{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "on-01", "task": "ka", "slug": "three-new-voices-at-the-narrators-chair", "label": "yes", "kind": "on-page", "state_sha256": "d944ffba7d7a97d70ffb305031ae8069d7e1bf47491eb08abdaea5d81af2ffe1", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:48Z"}
{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "on-02", "task": "ka", "slug": "two-new-frontier-models-at-the-rules-desk", "label": "yes", "kind": "on-page", "state_sha256": "b5c4d7119044bfb86fe877d20dd455d38788622a8cb8fd3da273cf881e554504", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:48Z"}
{"mode": "kev-noul", "prompt_sha256": "f8ebce04dd689673c8f6656b0c067eeaf2f9c2930c89dd7a8d480047c0c9db60", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.995929862284104}}, "usage": {"input_tokens": 2914, "output_tokens": 1}, "latency_ms": 797.7842260006582, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 2914, "compiled_prompt_sha256": "1bd9a502439c06a3395f832fc06f80152b118ed49f3ef7f26c2f261c451ccac3", "compiled_ids_sha256": "645742c30d9ba70edd95010cb21bb88c0de8a5062b2f504984a66e71250addd0", "candidate_token_ids": [32, 33], "logits": [19.0, 13.5], "runtime_probabilities": [0.9959298968315125, 0.004070138093084097], "decide_seconds": 0.7682689419998496, "probabilities_f64_minus_runtime_max_abs": 3.4547408489515874e-08}}}, "kev_noul": 0.995929862284104, "choice": "yes", "probs": {"yes": 0.995929862284104, "no": 0.004070137715896038}, "confidence": 0.9918597245682079, "seconds": 0.7979991840002185, "prompt_tokens": 2914, "eval_tokens": 1, "server_latency_ms": 797.7842260006582, "prompt_eval_ms": null, "total_ms": 797.9991840002185, "done_reason": null, "id": "on-03", "task": "ka", "slug": "six-worlds-painted-large", "label": "yes", "kind": "on-page", "state_sha256": "4902422ba42eec05800d9a72fd69c58559d0cb098878daec40505c90f50184d4", "request_sha256": "f8ebce04dd689673c8f6656b0c067eeaf2f9c2930c89dd7a8d480047c0c9db60", "compiled_tokens": 2914, "compiled_prompt_sha256": "1bd9a502439c06a3395f832fc06f80152b118ed49f3ef7f26c2f261c451ccac3", "compiled_ids_sha256": "645742c30d9ba70edd95010cb21bb88c0de8a5062b2f504984a66e71250addd0", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:49Z"}
{"mode": "kev-noul", "prompt_sha256": "4b35dd2369d7faeb0ff675f5ae0784719f669b50156491251f05bd9838f208f9", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9981897889737974}}, "usage": {"input_tokens": 4655, "output_tokens": 1}, "latency_ms": 1224.1151199996239, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 4655, "compiled_prompt_sha256": "edbe397bec6b71309e063cfbfb1325282875effcd62ba4f7f2bdf3d61c9a56d5", "compiled_ids_sha256": "b54721457b99b6118df9689c7df0e3898dbea1c3f00f12adc9d5098e352d34c9", "candidate_token_ids": [32, 33], "logits": [18.75, 12.4375], "runtime_probabilities": [0.9981897473335266, 0.0018102108733728528], "decide_seconds": 1.2009979079994082, "probabilities_f64_minus_runtime_max_abs": 4.164027078701338e-08}}}, "kev_noul": 0.9981897889737974, "choice": "yes", "probs": {"yes": 0.9981897889737974, "no": 0.0018102110262026017}, "confidence": 0.9963795779475948, "seconds": 1.224305888001254, "prompt_tokens": 4655, "eval_tokens": 1, "server_latency_ms": 1224.1151199996239, "prompt_eval_ms": null, "total_ms": 1224.305888001254, "done_reason": null, "id": "on-04", "task": "ka", "slug": "the-beat-lab-asked-and-answered", "label": "yes", "kind": "on-page", "state_sha256": "423a54ae5c1707a69cc67454a7c1fb261af05e3661a12dd44a3ac170ae6fdbea", "request_sha256": "4b35dd2369d7faeb0ff675f5ae0784719f669b50156491251f05bd9838f208f9", "compiled_tokens": 4655, "compiled_prompt_sha256": "edbe397bec6b71309e063cfbfb1325282875effcd62ba4f7f2bdf3d61c9a56d5", "compiled_ids_sha256": "b54721457b99b6118df9689c7df0e3898dbea1c3f00f12adc9d5098e352d34c9", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:50Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-05", "task": "ka", "slug": "ten-minutes-with-living-artists", "label": "yes", "kind": "on-page", "state_sha256": "3d3d19395679e12b9871c93853d3d888cfff4352b0fff0d5a432a99980d72e14", "request_sha256": "544f0ee64df7e0df092e1d7b112006bcef1ba90b07629cd639c6c83585ded309", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:50Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-06", "task": "ka", "slug": "listen-for-yourself", "label": "yes", "kind": "on-page", "state_sha256": "ff4ba7e7329efa416398cff233e4d8fbc43838da73f017c9081eeeaaf26349cc", "request_sha256": "f7a46135ce9830b095ff8a908ea65af2ec1853bcda47ecc7e72e53a9ce0dfb2c", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:50Z"}
{"mode": "kev-noul", "prompt_sha256": "af029828977102ecd6e92bc895faae7cfeb96442254aca83b40d201ab59db54a", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9984021545059827}}, "usage": {"input_tokens": 5463, "output_tokens": 1}, "latency_ms": 1427.5929689974873, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5463, "compiled_prompt_sha256": "114a98c17ad96ebd0809a80d1095a04395b78290895f79a8e803b5960e068aae", "compiled_ids_sha256": "b3527ead527e354324c5d715ace90f4b2cf11002473b337ffeca47007097e90f", "candidate_token_ids": [32, 33], "logits": [18.875, 12.4375], "runtime_probabilities": [0.9984021782875061, 0.001597845577634871], "decide_seconds": 1.4013059999997495, "probabilities_f64_minus_runtime_max_abs": 2.378152341719897e-08}}}, "kev_noul": 0.9984021545059827, "choice": "yes", "probs": {"yes": 0.9984021545059827, "no": 0.0015978454940173137}, "confidence": 0.9968043090119654, "seconds": 1.427795978001086, "prompt_tokens": 5463, "eval_tokens": 1, "server_latency_ms": 1427.5929689974873, "prompt_eval_ms": null, "total_ms": 1427.795978001086, "done_reason": null, "id": "on-07", "task": "ka", "slug": "half-an-hour-with-dead-composers", "label": "yes", "kind": "on-page", "state_sha256": "21d3e92b73064718cb5ca0a3d5fd514b31f0b656626e5a7e8e2baeff97735bfa", "request_sha256": "af029828977102ecd6e92bc895faae7cfeb96442254aca83b40d201ab59db54a", "compiled_tokens": 5463, "compiled_prompt_sha256": "114a98c17ad96ebd0809a80d1095a04395b78290895f79a8e803b5960e068aae", "compiled_ids_sha256": "b3527ead527e354324c5d715ace90f4b2cf11002473b337ffeca47007097e90f", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:52Z"}
{"mode": "kev-noul", "prompt_sha256": "7159206ba651b0038e881478a3249949339ca1db0ab69b0e0fa8c77920eb0595", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9988304897349445}}, "usage": {"input_tokens": 3575, "output_tokens": 1}, "latency_ms": 941.9020840032317, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 3575, "compiled_prompt_sha256": "3f41dbda1c02846c30408a09b01edb260ab6015d11a8c9f0b4e6c4c5b586357f", "compiled_ids_sha256": "236342a1ddaeb34f15a2f948bb7eb41d37ae888eff688ece0aa557bbd8be36c4", "candidate_token_ids": [32, 33], "logits": [18.875, 12.125], "runtime_probabilities": [0.9988304972648621, 0.0011695101857185364], "decide_seconds": 0.923019869998825, "probabilities_f64_minus_runtime_max_abs": 7.529917578352752e-09}}}, "kev_noul": 0.9988304897349445, "choice": "yes", "probs": {"yes": 0.9988304897349445, "no": 0.0011695102650555178}, "confidence": 0.997660979469889, "seconds": 0.9420499930020014, "prompt_tokens": 3575, "eval_tokens": 1, "server_latency_ms": 941.9020840032317, "prompt_eval_ms": null, "total_ms": 942.0499930020014, "done_reason": null, "id": "on-08", "task": "ka", "slug": "chopin-in-five-minutes", "label": "yes", "kind": "on-page", "state_sha256": "ce76bf2f182f387a020db77e9380690ce1522227e2f85e1345b2d33cba49c186", "request_sha256": "7159206ba651b0038e881478a3249949339ca1db0ab69b0e0fa8c77920eb0595", "compiled_tokens": 3575, "compiled_prompt_sha256": "3f41dbda1c02846c30408a09b01edb260ab6015d11a8c9f0b4e6c4c5b586357f", "compiled_ids_sha256": "236342a1ddaeb34f15a2f948bb7eb41d37ae888eff688ece0aa557bbd8be36c4", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:53Z"}
{"mode": "kev-noul", "prompt_sha256": "adf96d957a673f6825ed15571fe62c5b1c6c91a7d9bd2cbf6792820a261295ac", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9986749775827868}}, "usage": {"input_tokens": 6258, "output_tokens": 1}, "latency_ms": 1615.4603979994135, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 6258, "compiled_prompt_sha256": "237337dd9de5776e4b659f54ad37930734391599004df06817bf46041d90e878", "compiled_ids_sha256": "9497c108708c1203d67d9538fe3772bbbd47b0878fc7b14ac009c73bf9a59f04", "candidate_token_ids": [32, 33], "logits": [18.875, 12.25], "runtime_probabilities": [0.9986749291419983, 0.001325022429227829], "decide_seconds": 1.5843785659999412, "probabilities_f64_minus_runtime_max_abs": 4.844078849153277e-08}}}, "kev_noul": 0.9986749775827868, "choice": "yes", "probs": {"yes": 0.9986749775827868, "no": 0.0013250224172132175}, "confidence": 0.9973499551655736, "seconds": 1.6156492560003244, "prompt_tokens": 6258, "eval_tokens": 1, "server_latency_ms": 1615.4603979994135, "prompt_eval_ms": null, "total_ms": 1615.6492560003244, "done_reason": null, "id": "on-09", "task": "ka", "slug": "the-same-sixteen", "label": "yes", "kind": "on-page", "state_sha256": "830caf9405ece8c6ff598d3adfbdaf8da136fd260537d81f576a5059e5a17953", "request_sha256": "adf96d957a673f6825ed15571fe62c5b1c6c91a7d9bd2cbf6792820a261295ac", "compiled_tokens": 6258, "compiled_prompt_sha256": "237337dd9de5776e4b659f54ad37930734391599004df06817bf46041d90e878", "compiled_ids_sha256": "9497c108708c1203d67d9538fe3772bbbd47b0878fc7b14ac009c73bf9a59f04", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:54Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-10", "task": "ka", "slug": "two-hours-on-battery", "label": "yes", "kind": "on-page", "state_sha256": "c4c4db33a200bbafecb6bc46874d79161edbca9f5ee4e5451aac9225077b64f1", "request_sha256": "d9c1aa5ce8328386bf5b822eab474fe8debc8954f00773d30f75876f46b67d9e", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:54Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-11", "task": "ka", "slug": "the-instrument-travels", "label": "yes", "kind": "on-page", "state_sha256": "f316ed88d19f7f94e1870200edec052a28e370f3c5c57589b7c66a8182758693", "request_sha256": "23af0206dfe7b32d92912604f44c2146cd59ed7ff0bd53a231c1b935feb4f573", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:54Z"}
{"mode": "kev-noul", "prompt_sha256": "cc21c98043fdd32ece00bb9deeaef0bb80eda78cd24c4ae19e7df71d59b8600d", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9989677689632452}}, "usage": {"input_tokens": 1824, "output_tokens": 1}, "latency_ms": 496.284133001609, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 1824, "compiled_prompt_sha256": "1abfbe90a5a151a421c88de314fcfaf9440b64c378bbf6981c702ba3dbf25c32", "compiled_ids_sha256": "9a6bc56cc9a5ef132b4e7b3a61c99f5dbe120dd341fc8f15d46e405ee2c0245f", "candidate_token_ids": [32, 33], "logits": [19.25, 12.375], "runtime_probabilities": [0.9989677667617798, 0.0010322310263291001], "decide_seconds": 0.48757245000160765, "probabilities_f64_minus_runtime_max_abs": 2.2014654543767165e-09}}}, "kev_noul": 0.9989677689632452, "choice": "yes", "probs": {"yes": 0.9989677689632452, "no": 0.0010322310367547605}, "confidence": 0.9979355379264905, "seconds": 0.49635799300085637, "prompt_tokens": 1824, "eval_tokens": 1, "server_latency_ms": 496.284133001609, "prompt_eval_ms": null, "total_ms": 496.35799300085637, "done_reason": null, "id": "on-12", "task": "ka", "slug": "nine-worlds-one-dog", "label": "yes", "kind": "on-page", "state_sha256": "84c842a71a59b037055f5448244f732c94578d47c68a907c19c9a4a0324a0095", "request_sha256": "cc21c98043fdd32ece00bb9deeaef0bb80eda78cd24c4ae19e7df71d59b8600d", "compiled_tokens": 1824, "compiled_prompt_sha256": "1abfbe90a5a151a421c88de314fcfaf9440b64c378bbf6981c702ba3dbf25c32", "compiled_ids_sha256": "9a6bc56cc9a5ef132b4e7b3a61c99f5dbe120dd341fc8f15d46e405ee2c0245f", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:55Z"}
{"mode": "kev-noul", "prompt_sha256": "aadaa76601faa41b91b4e9a4253f0d299bde20b1be555f0c0f12c1ca5486e92a", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9988304897349445}}, "usage": {"input_tokens": 6966, "output_tokens": 1}, "latency_ms": 1810.0553349977417, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 6966, "compiled_prompt_sha256": "8cb2a234150912a5700c6e4d0cd11cf421ea25af29c4277d76ec257a2b8208cb", "compiled_ids_sha256": "fcbbaa3e3ea1444d3c017c20b59689a4dbb42b587a72c01fb4aade641f7e4e39", "candidate_token_ids": [32, 33], "logits": [18.75, 12.0], "runtime_probabilities": [0.9988304972648621, 0.0011695101857185364], "decide_seconds": 1.77745155500088, "probabilities_f64_minus_runtime_max_abs": 7.529917578352752e-09}}}, "kev_noul": 0.9988304897349445, "choice": "yes", "probs": {"yes": 0.9988304897349445, "no": 0.0011695102650555178}, "confidence": 0.997660979469889, "seconds": 1.8102593639996485, "prompt_tokens": 6966, "eval_tokens": 1, "server_latency_ms": 1810.0553349977417, "prompt_eval_ms": null, "total_ms": 1810.2593639996485, "done_reason": null, "id": "on-13", "task": "ka", "slug": "what-150-watts-buys", "label": "yes", "kind": "on-page", "state_sha256": "0fd39e7ea02081243919348e7a76a2acba2433694142679ddfd8fbaa29a3d00d", "request_sha256": "aadaa76601faa41b91b4e9a4253f0d299bde20b1be555f0c0f12c1ca5486e92a", "compiled_tokens": 6966, "compiled_prompt_sha256": "8cb2a234150912a5700c6e4d0cd11cf421ea25af29c4277d76ec257a2b8208cb", "compiled_ids_sha256": "fcbbaa3e3ea1444d3c017c20b59689a4dbb42b587a72c01fb4aade641f7e4e39", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:57Z"}
{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "on-14", "task": "ka", "slug": "how-the-beat-lab-works", "label": "yes", "kind": "on-page", "state_sha256": "fdb6d4bf89ebedc44a2d887309bf63ba59598c898a6eca00760115d49bc2133d", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:57Z"}
{"mode": "kev-noul", "prompt_sha256": "2bcf96723063667f7a5e82dc4bea15e8c566c6a67130c7d3cbc4e458657eab24", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.998498817743263}}, "usage": {"input_tokens": 5231, "output_tokens": 1}, "latency_ms": 1345.3129200024705, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5231, "compiled_prompt_sha256": "ac82da4918dfdba0291a940a7c2cc2b743971cd8d6cb2a07fe24bd22d31d0804", "compiled_ids_sha256": "783cd17bee5fee78a008afed881ca1810ba4ef81f844eae55489a70cf3024cf7", "candidate_token_ids": [32, 33], "logits": [19.0, 12.5], "runtime_probabilities": [0.998498797416687, 0.0015011822106316686], "decide_seconds": 1.3187212640004873, "probabilities_f64_minus_runtime_max_abs": 2.032657597794696e-08}}}, "kev_noul": 0.998498817743263, "choice": "yes", "probs": {"yes": 0.998498817743263, "no": 0.0015011822567370103}, "confidence": 0.996997635486526, "seconds": 1.345475988000544, "prompt_tokens": 5231, "eval_tokens": 1, "server_latency_ms": 1345.3129200024705, "prompt_eval_ms": null, "total_ms": 1345.475988000544, "done_reason": null, "id": "on-15", "task": "ka", "slug": "the-typist-and-the-developer", "label": "yes", "kind": "on-page", "state_sha256": "2d33747167f415355e5637f99962534c215404a547456ccde9c80e2d7591cf5d", "request_sha256": "2bcf96723063667f7a5e82dc4bea15e8c566c6a67130c7d3cbc4e458657eab24", "compiled_tokens": 5231, "compiled_prompt_sha256": "ac82da4918dfdba0291a940a7c2cc2b743971cd8d6cb2a07fe24bd22d31d0804", "compiled_ids_sha256": "783cd17bee5fee78a008afed881ca1810ba4ef81f844eae55489a70cf3024cf7", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:58Z"}
{"mode": "kev-noul", "prompt_sha256": "74c848205284da7c2e3107feda1f2348604771c9f8c84dc7e02efae8787e5f79", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9978172835546547}}, "usage": {"input_tokens": 4452, "output_tokens": 1}, "latency_ms": 1175.4359129990917, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 4452, "compiled_prompt_sha256": "d16c5504d0a9269ad05b803d893a01b26b3196bb7c2154e0837fb5aebe9289c1", "compiled_ids_sha256": "072e551272f3ebe0eff1b91b6906d8c644e9a5ede31d205d64e163ef8fe77631", "candidate_token_ids": [32, 33], "logits": [19.0, 12.875], "runtime_probabilities": [0.9978172779083252, 0.002182716503739357], "decide_seconds": 1.1534882619998825, "probabilities_f64_minus_runtime_max_abs": 5.646329515052173e-09}}}, "kev_noul": 0.9978172835546547, "choice": "yes", "probs": {"yes": 0.9978172835546547, "no": 0.0021827164453452896}, "confidence": 0.9956345671093094, "seconds": 1.1755746019989601, "prompt_tokens": 4452, "eval_tokens": 1, "server_latency_ms": 1175.4359129990917, "prompt_eval_ms": null, "total_ms": 1175.5746019989601, "done_reason": null, "id": "on-16", "task": "ka", "slug": "a-rig-your-friend-already-owns", "label": "yes", "kind": "on-page", "state_sha256": "9d22ebe41036f74e0c8501e4631982f4d610eebb419288b4535c8a5e9f9b65d4", "request_sha256": "74c848205284da7c2e3107feda1f2348604771c9f8c84dc7e02efae8787e5f79", "compiled_tokens": 4452, "compiled_prompt_sha256": "d16c5504d0a9269ad05b803d893a01b26b3196bb7c2154e0837fb5aebe9289c1", "compiled_ids_sha256": "072e551272f3ebe0eff1b91b6906d8c644e9a5ede31d205d64e163ef8fe77631", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:59Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-17", "task": "ka", "slug": "how-a-vision-model-sees", "label": "yes", "kind": "on-page", "state_sha256": "6a885683413930cdad4c104dc4ede2330b45bb9903d54621ace0d4f518a89fa9", "request_sha256": "f7e2a4ea6ecd51b7f9f233118addb1fdd88d344a05bdcd77161f004a73b3bab1", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:45:59Z"}
{"mode": "kev-noul", "prompt_sha256": "9ed4b6bd4df4dbfce12a9a35ffc8a8e6cdc67795590fc668d5806c02f879bfb4", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9989012689270753}}, "usage": {"input_tokens": 5126, "output_tokens": 1}, "latency_ms": 1334.5873520011082, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5126, "compiled_prompt_sha256": "bb1962fc7394d3cfa5e3ba5e2faba08bb918d42732a89ac0f4a753dc0c7c2036", "compiled_ids_sha256": "f2c17ad71e2d07fef432f096382eaa55cc394b00c5c69246bc79d32d84d1b378", "candidate_token_ids": [32, 33], "logits": [19.25, 12.4375], "runtime_probabilities": [0.9989012479782104, 0.001098731067031622], "decide_seconds": 1.3105850770007237, "probabilities_f64_minus_runtime_max_abs": 2.0948864865033556e-08}}}, "kev_noul": 0.9989012689270753, "choice": "yes", "probs": {"yes": 0.9989012689270753, "no": 0.0010987310729246857}, "confidence": 0.9978025378541506, "seconds": 1.3347427109983983, "prompt_tokens": 5126, "eval_tokens": 1, "server_latency_ms": 1334.5873520011082, "prompt_eval_ms": null, "total_ms": 1334.7427109983983, "done_reason": null, "id": "on-18", "task": "ka", "slug": "fifty-seven-milliseconds", "label": "yes", "kind": "on-page", "state_sha256": "bec5d08ff12a12a4d87ca9d1d31e3bec848dfe329b6a47e9b8616e071da4ba5c", "request_sha256": "9ed4b6bd4df4dbfce12a9a35ffc8a8e6cdc67795590fc668d5806c02f879bfb4", "compiled_tokens": 5126, "compiled_prompt_sha256": "bb1962fc7394d3cfa5e3ba5e2faba08bb918d42732a89ac0f4a753dc0c7c2036", "compiled_ids_sha256": "f2c17ad71e2d07fef432f096382eaa55cc394b00c5c69246bc79d32d84d1b378", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:00Z"}
{"mode": "kev-noul", "prompt_sha256": "edb28ef33c63af998e62baa172266a3b22df050244e0d4c044638c421e7785c0", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9990889488055994}}, "usage": {"input_tokens": 5834, "output_tokens": 1}, "latency_ms": 1527.435673000582, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5834, "compiled_prompt_sha256": "5841ae17fd03368af3da0e3d76eae5b3570c9000eb719447d8ad3e0358fc17b5", "compiled_ids_sha256": "3c93b3ffe360c2dfd77bab27a915d4569e0573cda648da831d5734fbe10e0ebc", "candidate_token_ids": [32, 33], "logits": [19.125, 12.125], "runtime_probabilities": [0.9990890026092529, 0.0009110512328334153], "decide_seconds": 1.4982010470012028, "probabilities_f64_minus_runtime_max_abs": 5.38036535324693e-08}}}, "kev_noul": 0.9990889488055994, "choice": "yes", "probs": {"yes": 0.9990889488055994, "no": 0.0009110511944006028}, "confidence": 0.9981778976111988, "seconds": 1.5276224820008792, "prompt_tokens": 5834, "eval_tokens": 1, "server_latency_ms": 1527.435673000582, "prompt_eval_ms": null, "total_ms": 1527.6224820008792, "done_reason": null, "id": "on-19", "task": "ka", "slug": "the-dog-the-dice-and-the-painter", "label": "yes", "kind": "on-page", "state_sha256": "a136b61fadbc1ff5eded7327a0b3f4818be7eb2a2122cfd25b64de203f98cb0f", "request_sha256": "edb28ef33c63af998e62baa172266a3b22df050244e0d4c044638c421e7785c0", "compiled_tokens": 5834, "compiled_prompt_sha256": "5841ae17fd03368af3da0e3d76eae5b3570c9000eb719447d8ad3e0358fc17b5", "compiled_ids_sha256": "3c93b3ffe360c2dfd77bab27a915d4569e0573cda648da831d5734fbe10e0ebc", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:02Z"}
{"mode": "kev-noul", "prompt_sha256": "1d5ab0e80fcbfc522cbbe4dba6522120f03b714d21695880515d101660dfdb04", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9988304897349445}}, "usage": {"input_tokens": 5259, "output_tokens": 1}, "latency_ms": 1408.2205170016096, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5259, "compiled_prompt_sha256": "fa14595e009a37dde7b2c0f25a9f3ce0ced432a2f174bc67b7b50188c0b89dbd", "compiled_ids_sha256": "0b6f2a4a2c1222a63cdd024647268780a660fe7c2e889778e98cda5762656f12", "candidate_token_ids": [32, 33], "logits": [18.875, 12.125], "runtime_probabilities": [0.9988304972648621, 0.0011695101857185364], "decide_seconds": 1.3799535850012035, "probabilities_f64_minus_runtime_max_abs": 7.529917578352752e-09}}}, "kev_noul": 0.9988304897349445, "choice": "yes", "probs": {"yes": 0.9988304897349445, "no": 0.0011695102650555178}, "confidence": 0.997660979469889, "seconds": 1.4084005059994524, "prompt_tokens": 5259, "eval_tokens": 1, "server_latency_ms": 1408.2205170016096, "prompt_eval_ms": null, "total_ms": 1408.4005059994524, "done_reason": null, "id": "on-20", "task": "ka", "slug": "where-new-knowledge-comes-from", "label": "yes", "kind": "on-page", "state_sha256": "26b4da101ba323b5f8e9999ee8b445f148b5cf9757bf21acaaa991f748bd4225", "request_sha256": "1d5ab0e80fcbfc522cbbe4dba6522120f03b714d21695880515d101660dfdb04", "compiled_tokens": 5259, "compiled_prompt_sha256": "fa14595e009a37dde7b2c0f25a9f3ce0ced432a2f174bc67b7b50188c0b89dbd", "compiled_ids_sha256": "0b6f2a4a2c1222a63cdd024647268780a660fe7c2e889778e98cda5762656f12", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:03Z"}
{"mode": "kev-noul", "prompt_sha256": "cd046120ec7849c3fe98cc45aa86a5fe535bea84f2adb6e91abe6fedcd6a34ba", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9953904278206259}}, "usage": {"input_tokens": 5207, "output_tokens": 1}, "latency_ms": 1350.0995030008198, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5207, "compiled_prompt_sha256": "f9f3a98179d84c8db5d4fcf5f65998efc1f9cdcbcbf7950d44bb663d18563c01", "compiled_ids_sha256": "1c13e6f85308d775bcf9310347a0f84a104f3564d1790749c82aacb6765e39d0", "candidate_token_ids": [32, 33], "logits": [18.75, 13.375], "runtime_probabilities": [0.9953904151916504, 0.004609572701156139], "decide_seconds": 1.3224720859980152, "probabilities_f64_minus_runtime_max_abs": 1.262897553200304e-08}}}, "kev_noul": 0.9953904278206259, "choice": "yes", "probs": {"yes": 0.9953904278206259, "no": 0.004609572179374077}, "confidence": 0.9907808556412518, "seconds": 1.3502772920001007, "prompt_tokens": 5207, "eval_tokens": 1, "server_latency_ms": 1350.0995030008198, "prompt_eval_ms": null, "total_ms": 1350.2772920001007, "done_reason": null, "id": "on-21", "task": "ka", "slug": "compression-is-the-objective", "label": "yes", "kind": "on-page", "state_sha256": "68bc7c0890d5d05afefc1bbd6178315fddd18fb35775c738f50544096a982f7f", "request_sha256": "cd046120ec7849c3fe98cc45aa86a5fe535bea84f2adb6e91abe6fedcd6a34ba", "compiled_tokens": 5207, "compiled_prompt_sha256": "f9f3a98179d84c8db5d4fcf5f65998efc1f9cdcbcbf7950d44bb663d18563c01", "compiled_ids_sha256": "1c13e6f85308d775bcf9310347a0f84a104f3564d1790749c82aacb6765e39d0", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:05Z"}
{"mode": "kev-noul", "prompt_sha256": "284c6b4eaa96f868169bb2e6f9693a52a1b617f6046fa04c5f1cbc436ef1f25c", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9987551566979116}}, "usage": {"input_tokens": 7148, "output_tokens": 1}, "latency_ms": 1857.0029949987656, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 7148, "compiled_prompt_sha256": "eb1cc1612dac8ae3d4b5e1ad9fefbe91447df45934a713b19bcce79f67c5d650", "compiled_ids_sha256": "a0e1b29fe7526ddc5fa90920fe9712d9a74579bfce719050ca45014dd9036ad6", "candidate_token_ids": [32, 33], "logits": [18.75, 12.0625], "runtime_probabilities": [0.9987550973892212, 0.0012448433553799987], "decide_seconds": 1.8211884499978623, "probabilities_f64_minus_runtime_max_abs": 5.930869040327025e-08}}}, "kev_noul": 0.9987551566979116, "choice": "yes", "probs": {"yes": 0.9987551566979116, "no": 0.0012448433020884053}, "confidence": 0.9975103133958232, "seconds": 1.8572139440002502, "prompt_tokens": 7148, "eval_tokens": 1, "server_latency_ms": 1857.0029949987656, "prompt_eval_ms": null, "total_ms": 1857.2139440002502, "done_reason": null, "id": "on-22", "task": "ka", "slug": "the-free-speed-wasnt-free", "label": "yes", "kind": "on-page", "state_sha256": "3ae132432376b190cec501235414ae89382270e55515119fc3f0b4d51c4d6de3", "request_sha256": "284c6b4eaa96f868169bb2e6f9693a52a1b617f6046fa04c5f1cbc436ef1f25c", "compiled_tokens": 7148, "compiled_prompt_sha256": "eb1cc1612dac8ae3d4b5e1ad9fefbe91447df45934a713b19bcce79f67c5d650", "compiled_ids_sha256": "a0e1b29fe7526ddc5fa90920fe9712d9a74579bfce719050ca45014dd9036ad6", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:07Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-23", "task": "ka", "slug": "the-compressed-photograph", "label": "yes", "kind": "on-page", "state_sha256": "30efd8ed8a4dcfd536f819d785eb18027a4f37e47597ae0e18294a5ede65b01a", "request_sha256": "330898f43055b8d1b8e0c8f31843f1ba31ca73a1132e96da4f147f9c99ab302c", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:07Z"}
{"mode": "kev-noul", "prompt_sha256": "dc069ad09e93b282d9e3b744ce733f164c25b6c16e916e03f86f23dcb553c53c", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9985896415033819}}, "usage": {"input_tokens": 7304, "output_tokens": 1}, "latency_ms": 1940.7150830011233, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 7304, "compiled_prompt_sha256": "8673351de400a9ed7e8afda494dd8c479d188ccab8bf744e23f178c5218d2a18", "compiled_ids_sha256": "8e65854d97de35c28392d5f3632b30722879d24419a8fea5f2c7ea91293b0af5", "candidate_token_ids": [32, 33], "logits": [19.0, 12.4375], "runtime_probabilities": [0.9985895752906799, 0.0014103584690019488], "decide_seconds": 1.902530385999853, "probabilities_f64_minus_runtime_max_abs": 6.621270198792217e-08}}}, "kev_noul": 0.9985896415033819, "choice": "yes", "probs": {"yes": 0.9985896415033819, "no": 0.0014103584966180804}, "confidence": 0.9971792830067638, "seconds": 1.9409355809984845, "prompt_tokens": 7304, "eval_tokens": 1, "server_latency_ms": 1940.7150830011233, "prompt_eval_ms": null, "total_ms": 1940.9355809984845, "done_reason": null, "id": "on-24", "task": "ka", "slug": "three-at-the-table", "label": "yes", "kind": "on-page", "state_sha256": "5ee526a9690f5511741a41a6eb1e91418702e2cd8e49d1c2db9e03802a7bde5c", "request_sha256": "dc069ad09e93b282d9e3b744ce733f164c25b6c16e916e03f86f23dcb553c53c", "compiled_tokens": 7304, "compiled_prompt_sha256": "8673351de400a9ed7e8afda494dd8c479d188ccab8bf744e23f178c5218d2a18", "compiled_ids_sha256": "8e65854d97de35c28392d5f3632b30722879d24419a8fea5f2c7ea91293b0af5", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:08Z"}
{"mode": "kev-noul", "prompt_sha256": "f1ea881a09acefb481e491b99d0fa7892b9a7e558ca155a01fa58e2619b119b2", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9989012689270753}}, "usage": {"input_tokens": 5221, "output_tokens": 1}, "latency_ms": 1353.663655998389, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5221, "compiled_prompt_sha256": "b984ea421cdf0d50830e309547d784530c1080453d8e875ad33cbcb942fb1e57", "compiled_ids_sha256": "593b050c83c688cebd226712842dac0c58f12fa2a8e5eafc6d2efe5b4ac2c690", "candidate_token_ids": [32, 33], "logits": [18.75, 11.9375], "runtime_probabilities": [0.9989012479782104, 0.001098731067031622], "decide_seconds": 1.3255198020015087, "probabilities_f64_minus_runtime_max_abs": 2.0948864865033556e-08}}}, "kev_noul": 0.9989012689270753, "choice": "yes", "probs": {"yes": 0.9989012689270753, "no": 0.0010987310729246857}, "confidence": 0.9978025378541506, "seconds": 1.3538328649992764, "prompt_tokens": 5221, "eval_tokens": 1, "server_latency_ms": 1353.663655998389, "prompt_eval_ms": null, "total_ms": 1353.8328649992764, "done_reason": null, "id": "on-25", "task": "ka", "slug": "reading-is-fast", "label": "yes", "kind": "on-page", "state_sha256": "79b9c9e1be3ac3cdc484fec1f5eb9a83bf909612d6eb1c3c7c83bb469125ff2f", "request_sha256": "f1ea881a09acefb481e491b99d0fa7892b9a7e558ca155a01fa58e2619b119b2", "compiled_tokens": 5221, "compiled_prompt_sha256": "b984ea421cdf0d50830e309547d784530c1080453d8e875ad33cbcb942fb1e57", "compiled_ids_sha256": "593b050c83c688cebd226712842dac0c58f12fa2a8e5eafc6d2efe5b4ac2c690", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:10Z"}
{"mode": "kev-noul", "prompt_sha256": "db3855f1aa358bda6553b365078e77ce8df6b68b5e99a51f512c13e97d5e9ce3", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9986749775827868}}, "usage": {"input_tokens": 5608, "output_tokens": 1}, "latency_ms": 1482.2279100008018, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5608, "compiled_prompt_sha256": "80304fbc22a730018429dba667947be2fa0716e71b7c3dd7b8165d0652bfa5d4", "compiled_ids_sha256": "47d89d0b710075af09468ecca4c9e4d97c1a416d0b5507e6baef4a909fb22e97", "candidate_token_ids": [32, 33], "logits": [19.0, 12.375], "runtime_probabilities": [0.9986749291419983, 0.001325022429227829], "decide_seconds": 1.4508873300001142, "probabilities_f64_minus_runtime_max_abs": 4.844078849153277e-08}}}, "kev_noul": 0.9986749775827868, "choice": "yes", "probs": {"yes": 0.9986749775827868, "no": 0.0013250224172132175}, "confidence": 0.9973499551655736, "seconds": 1.4824119990007603, "prompt_tokens": 5608, "eval_tokens": 1, "server_latency_ms": 1482.2279100008018, "prompt_eval_ms": null, "total_ms": 1482.4119990007603, "done_reason": null, "id": "on-26", "task": "ka", "slug": "three-librarians", "label": "yes", "kind": "on-page", "state_sha256": "8883de675623f6941e01030f0e305b0a8f73197ce266d599ff002b148a135167", "request_sha256": "db3855f1aa358bda6553b365078e77ce8df6b68b5e99a51f512c13e97d5e9ce3", "compiled_tokens": 5608, "compiled_prompt_sha256": "80304fbc22a730018429dba667947be2fa0716e71b7c3dd7b8165d0652bfa5d4", "compiled_ids_sha256": "47d89d0b710075af09468ecca4c9e4d97c1a416d0b5507e6baef4a909fb22e97", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:11Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-27", "task": "ka", "slug": "the-new-kid", "label": "yes", "kind": "on-page", "state_sha256": "676acfa90c7de336adc467067853954373df2cc40b45fb728501238f7b963b79", "request_sha256": "f1eeaf9128a6de26f5c7be831b1b8f7c4d6fd25c99f9c01f5dde951ec2e73c10", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:11Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-28", "task": "ka", "slug": "llms-txt", "label": "yes", "kind": "on-page", "state_sha256": "96cab65488f086e43a1b03248954f7a7abadb4ae49c8591b267586825532ff41", "request_sha256": "8da3fb4865799703851d3c144a54788423c49155fd44bff34831e0882ad2f2a7", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:11Z"}
{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "on-29", "task": "ka", "slug": "the-open-call", "label": "yes", "kind": "on-page", "state_sha256": "dd50fd254da5ef89e4284d245102a3ca9f9f7dd4e41f90c7687adb01223798a4", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:11Z"}
{"mode": "kev-noul", "prompt_sha256": "c0ddf9d73aa3cefab86b3e69e03b8e812a3a5fa698bdee051b5d28197024b673", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9990302480322174}}, "usage": {"input_tokens": 4707, "output_tokens": 1}, "latency_ms": 1246.323499999562, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 4707, "compiled_prompt_sha256": "8991c891ad0648a9bdd74c5d37aae5dda176c5bb3b83870e432c229980e99f9a", "compiled_ids_sha256": "00272c479ee9666e368b7babd5cb441e5dac03bbcdc34d81b7aa4e43f77d3d57", "candidate_token_ids": [32, 33], "logits": [19.375, 12.4375], "runtime_probabilities": [0.9990302324295044, 0.0009697519126348197], "decide_seconds": 1.2206623160018353, "probabilities_f64_minus_runtime_max_abs": 1.5602713010487435e-08}}}, "kev_noul": 0.9990302480322174, "choice": "yes", "probs": {"yes": 0.9990302480322174, "no": 0.000969751967782595}, "confidence": 0.9980604960644348, "seconds": 1.2464838390005752, "prompt_tokens": 4707, "eval_tokens": 1, "server_latency_ms": 1246.323499999562, "prompt_eval_ms": null, "total_ms": 1246.4838390005752, "done_reason": null, "id": "on-30", "task": "ka", "slug": "where-your-question-goes", "label": "yes", "kind": "on-page", "state_sha256": "cdc24e58a19354d77487d8df5e8782b2c15b4483afe7405f34478bf769d44b11", "request_sha256": "c0ddf9d73aa3cefab86b3e69e03b8e812a3a5fa698bdee051b5d28197024b673", "compiled_tokens": 4707, "compiled_prompt_sha256": "8991c891ad0648a9bdd74c5d37aae5dda176c5bb3b83870e432c229980e99f9a", "compiled_ids_sha256": "00272c479ee9666e368b7babd5cb441e5dac03bbcdc34d81b7aa4e43f77d3d57", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:13Z"}
{"mode": "kev-noul", "prompt_sha256": "7a56fea6c96e13f42ed4a44e374ab48cb8f33b63e0a508f72d864481b95a69d3", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9984021545059827}}, "usage": {"input_tokens": 4928, "output_tokens": 1}, "latency_ms": 1293.4542280017922, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 4928, "compiled_prompt_sha256": "0426a35073f7b39e707b5422234ce10e50617dbf612f232f17ff4ec2062bbb51", "compiled_ids_sha256": "ef4de9f70f9d399b9404b12bcdbf31ef0b36a6364294e62a3eb29b28409bed91", "candidate_token_ids": [32, 33], "logits": [18.875, 12.4375], "runtime_probabilities": [0.9984021782875061, 0.001597845577634871], "decide_seconds": 1.2657692200009478, "probabilities_f64_minus_runtime_max_abs": 2.378152341719897e-08}}}, "kev_noul": 0.9984021545059827, "choice": "yes", "probs": {"yes": 0.9984021545059827, "no": 0.0015978454940173137}, "confidence": 0.9968043090119654, "seconds": 1.2936272070000996, "prompt_tokens": 4928, "eval_tokens": 1, "server_latency_ms": 1293.4542280017922, "prompt_eval_ms": null, "total_ms": 1293.6272070000996, "done_reason": null, "id": "on-31", "task": "ka", "slug": "how-we-work", "label": "yes", "kind": "on-page", "state_sha256": "77d23d89a5adf6c0d561223d2c37308b5118a80f1f07f35f78bb1b0d08276e99", "request_sha256": "7a56fea6c96e13f42ed4a44e374ab48cb8f33b63e0a508f72d864481b95a69d3", "compiled_tokens": 4928, "compiled_prompt_sha256": "0426a35073f7b39e707b5422234ce10e50617dbf612f232f17ff4ec2062bbb51", "compiled_ids_sha256": "ef4de9f70f9d399b9404b12bcdbf31ef0b36a6364294e62a3eb29b28409bed91", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:14Z"}
{"mode": "kev-noul", "prompt_sha256": "2b8132cb0cfe51926d74fdca96be58607c86cbb4cf971432cd0fd6a73e19f2a5", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9993736658418905}}, "usage": {"input_tokens": 5832, "output_tokens": 1}, "latency_ms": 1533.774823998101, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5832, "compiled_prompt_sha256": "52e9469037d5e6d0bc838e1c8289e6fa42a3fc7d6a763b781a4779d71d27fc7f", "compiled_ids_sha256": "8f47b3a3903b41929a0f32b78846cd9d02556100a969c1e103e95c0de047dc0c", "candidate_token_ids": [32, 33], "logits": [19.25, 11.875], "runtime_probabilities": [0.999373733997345, 0.000626334163825959], "decide_seconds": 1.504097061999346, "probabilities_f64_minus_runtime_max_abs": 6.815545450233884e-08}}}, "kev_noul": 0.9993736658418905, "choice": "yes", "probs": {"yes": 0.9993736658418905, "no": 0.0006263341581095316}, "confidence": 0.9987473316837809, "seconds": 1.5339438829978462, "prompt_tokens": 5832, "eval_tokens": 1, "server_latency_ms": 1533.774823998101, "prompt_eval_ms": null, "total_ms": 1533.9438829978462, "done_reason": null, "id": "on-32", "task": "ka", "slug": "the-move", "label": "yes", "kind": "on-page", "state_sha256": "3d1d2fe049acbdd5e286fbc87c126a77f0410d05b6a51694ef22f6658b409e9f", "request_sha256": "2b8132cb0cfe51926d74fdca96be58607c86cbb4cf971432cd0fd6a73e19f2a5", "compiled_tokens": 5832, "compiled_prompt_sha256": "52e9469037d5e6d0bc838e1c8289e6fa42a3fc7d6a763b781a4779d71d27fc7f", "compiled_ids_sha256": "8f47b3a3903b41929a0f32b78846cd9d02556100a969c1e103e95c0de047dc0c", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:15Z"}
{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "on-33", "task": "ka", "slug": "chair-trials", "label": "yes", "kind": "on-page", "state_sha256": "33688798032704bdb7db628ffa9c2c557245b03bcc2611f357853ee101698402", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:15Z"}
{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "on-34", "task": "ka", "slug": "cove-voice-head-to-head", "label": "yes", "kind": "on-page", "state_sha256": "53932c8b9bd51ad2a1b2e91fef30dd9dfc61b5c1e531fb3ee4818db6fbf2f5ee", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:15Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-35", "task": "ka", "slug": "outside-judges", "label": "yes", "kind": "on-page", "state_sha256": "fc7235d0fabb2221f559392fc0236d981f9898d96425df0d7355f2d6f11d17d6", "request_sha256": "75cf89ce8145ee0fe72c39c2235cdad4cfb0757e13eddc7b65203c9c6fd8980a", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:15Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-36", "task": "ka", "slug": "august-arrivals", "label": "yes", "kind": "on-page", "state_sha256": "4e79d2cabc185d243c590fe6100c9ba3cd0a34262da8edab751ca5859dfa0d9b", "request_sha256": "caa9227270616296639b0112ed1ca9b6cd49e8bb08ad5a39b6a5ec589f57fbad", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:15Z"}
{"mode": "kev-noul", "prompt_sha256": "d521a044697a9469f7dee687aa84b4ab159e1eb698c4756fb4c4c4bea9aef755", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9988304897349445}}, "usage": {"input_tokens": 7010, "output_tokens": 1}, "latency_ms": 1832.297794000624, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 7010, "compiled_prompt_sha256": "29d18bb15e86d042468b353c22bfd91201e4d4ba387b1c83ceb01e9230a2234d", "compiled_ids_sha256": "70de4bd9a66a14d01b0394ffe7e4c7e8362a0fc62339b6e4bc1b8d57806a201d", "candidate_token_ids": [32, 33], "logits": [18.75, 12.0], "runtime_probabilities": [0.9988304972648621, 0.0011695101857185364], "decide_seconds": 1.8000197719993594, "probabilities_f64_minus_runtime_max_abs": 7.529917578352752e-09}}}, "kev_noul": 0.9988304897349445, "choice": "yes", "probs": {"yes": 0.9988304897349445, "no": 0.0011695102650555178}, "confidence": 0.997660979469889, "seconds": 1.8324859230015136, "prompt_tokens": 7010, "eval_tokens": 1, "server_latency_ms": 1832.297794000624, "prompt_eval_ms": null, "total_ms": 1832.4859230015136, "done_reason": null, "id": "on-37", "task": "ka", "slug": "seat-trials", "label": "yes", "kind": "on-page", "state_sha256": "c9374a27b5e06de1440ffed7d81385fa68f7572790c0f553f8f2948c0218eea1", "request_sha256": "d521a044697a9469f7dee687aa84b4ab159e1eb698c4756fb4c4c4bea9aef755", "compiled_tokens": 7010, "compiled_prompt_sha256": "29d18bb15e86d042468b353c22bfd91201e4d4ba387b1c83ceb01e9230a2234d", "compiled_ids_sha256": "70de4bd9a66a14d01b0394ffe7e4c7e8362a0fc62339b6e4bc1b8d57806a201d", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:17Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-38", "task": "ka", "slug": "voice-trials", "label": "yes", "kind": "on-page", "state_sha256": "34230a4e79273a8665f9112f22ddbdd39df22fdcb0190066ae293f8a9c7c34dc", "request_sha256": "90c5a1d337de9dfabce1ad0966b52b7bac9b65938417d5534d5f2fcf62e95501", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:17Z"}
{"mode": "kev-noul", "prompt_sha256": "cfd0106359e94c981d662777f27744b65dbd2675d8e3da4f2f252fa40e2fec35", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.9964063974185798}}, "usage": {"input_tokens": 8187, "output_tokens": 1}, "latency_ms": 2142.029607999575, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 8187, "compiled_prompt_sha256": "12ad24e118a10273ebd25074ced3fed63df066b719655dfb47c2a9ef78e840f0", "compiled_ids_sha256": "159f7c62e7b55a3bd6697645e168f75a9b822547bf119e267a9b78c8fe24558f", "candidate_token_ids": [32, 33], "logits": [18.75, 13.125], "runtime_probabilities": [0.9964063763618469, 0.0035936026833951473], "decide_seconds": 2.100225810001575, "probabilities_f64_minus_runtime_max_abs": 2.105673291286081e-08}}}, "kev_noul": 0.9964063974185798, "choice": "yes", "probs": {"yes": 0.9964063974185798, "no": 0.0035936025814201633}, "confidence": 0.9928127948371597, "seconds": 2.1422624370025005, "prompt_tokens": 8187, "eval_tokens": 1, "server_latency_ms": 2142.029607999575, "prompt_eval_ms": null, "total_ms": 2142.2624370025005, "done_reason": null, "id": "on-39", "task": "ka", "slug": "licences", "label": "yes", "kind": "on-page", "state_sha256": "cce674dd43eecdff402378e65d4ae1f1f64bb6db5702c2694bed4c5733c37090", "request_sha256": "cfd0106359e94c981d662777f27744b65dbd2675d8e3da4f2f252fa40e2fec35", "compiled_tokens": 8187, "compiled_prompt_sha256": "12ad24e118a10273ebd25074ced3fed63df066b719655dfb47c2a9ef78e840f0", "compiled_ids_sha256": "159f7c62e7b55a3bd6697645e168f75a9b822547bf119e267a9b78c8fe24558f", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:19Z"}
{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "on-40", "task": "ka", "slug": "hear-it-for-yourself", "label": "yes", "kind": "on-page", "state_sha256": "679607114cbafcf5a158729b859012472889e4ec49b0c3ae61df66a3fafc6827", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:19Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "on-41", "task": "ka", "slug": "the-ceiling-is-not-the-corpus", "label": "yes", "kind": "on-page", "state_sha256": "f4bb84be11307889eba78ab0a1e5af28c73807a48012d5c6850a848cece69cb8", "request_sha256": "9b5a7fd2e981cc720b5a06acec3c5f820cd06bb034afab220cbcc050e5a7c0f1", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:19Z"}
{"mode": "kev-noul", "prompt_sha256": "95467f4791fac437dacefc0e3b46bc5cb14a90a6a1946e9d21a675d2f575eaf6", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.995929862284104}}, "usage": {"input_tokens": 6634, "output_tokens": 1}, "latency_ms": 1754.7663189980085, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 6634, "compiled_prompt_sha256": "5c6d5150ebd3a3bd0318b43c76687f69ecfcb21acbe8bd55640cd0520b294006", "compiled_ids_sha256": "673446875e774e62507bc37eeeaf5e5d3cd09f65a523ee2c26a18b7328719648", "candidate_token_ids": [32, 33], "logits": [18.875, 13.375], "runtime_probabilities": [0.9959298968315125, 0.004070138093084097], "decide_seconds": 1.7162460550025571, "probabilities_f64_minus_runtime_max_abs": 3.4547408489515874e-08}}}, "kev_noul": 0.995929862284104, "choice": "yes", "probs": {"yes": 0.995929862284104, "no": 0.004070137715896038}, "confidence": 0.9918597245682079, "seconds": 1.7549895480005944, "prompt_tokens": 6634, "eval_tokens": 1, "server_latency_ms": 1754.7663189980085, "prompt_eval_ms": null, "total_ms": 1754.9895480005944, "done_reason": null, "id": "on-42", "task": "ka", "slug": "the-indifferent-prediction", "label": "yes", "kind": "on-page", "state_sha256": "de97acfa81114f48f10f062ee956f9beb81fce1b56537b3ca2a7ad894da658b6", "request_sha256": "95467f4791fac437dacefc0e3b46bc5cb14a90a6a1946e9d21a675d2f575eaf6", "compiled_tokens": 6634, "compiled_prompt_sha256": "5c6d5150ebd3a3bd0318b43c76687f69ecfcb21acbe8bd55640cd0520b294006", "compiled_ids_sha256": "673446875e774e62507bc37eeeaf5e5d3cd09f65a523ee2c26a18b7328719648", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:21Z"}
{"mode": "kev-noul", "prompt_sha256": "8d600d4f9e1d55c8f6ae4216cc3c3edd312644ed4004b2522ff7dfdc90a4b634", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0013250224172131609}}, "usage": {"input_tokens": 6956, "output_tokens": 1}, "latency_ms": 1824.3986449997465, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 6956, "compiled_prompt_sha256": "b156e7c2044906052f9569e06999999ddb9fc4e27ee21586b69c805a5a1aa161", "compiled_ids_sha256": "faf00be32fc92a946a57764ba0cc531b9cb340e59f3e6d2ca0c73c19e998c20d", "candidate_token_ids": [32, 33], "logits": [12.375, 19.0], "runtime_probabilities": [0.001325022429227829, 0.9986749291419983], "decide_seconds": 1.7918414949999715, "probabilities_f64_minus_runtime_max_abs": 4.844078849153277e-08}}}, "kev_noul": 0.0013250224172131609, "choice": "no", "probs": {"yes": 0.0013250224172131609, "no": 0.9986749775827869}, "confidence": 0.9973499551655738, "seconds": 1.8246027029999823, "prompt_tokens": 6956, "eval_tokens": 1, "server_latency_ms": 1824.3986449997465, "prompt_eval_ms": null, "total_ms": 1824.6027029999823, "done_reason": null, "id": "off-01", "task": "ka", "slug": "what-150-watts-buys", "label": "no", "kind": "off-page", "state_sha256": "0fd39e7ea02081243919348e7a76a2acba2433694142679ddfd8fbaa29a3d00d", "request_sha256": "8d600d4f9e1d55c8f6ae4216cc3c3edd312644ed4004b2522ff7dfdc90a4b634", "compiled_tokens": 6956, "compiled_prompt_sha256": "b156e7c2044906052f9569e06999999ddb9fc4e27ee21586b69c805a5a1aa161", "compiled_ids_sha256": "faf00be32fc92a946a57764ba0cc531b9cb340e59f3e6d2ca0c73c19e998c20d", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:23Z"}
{"mode": "kev-noul", "prompt_sha256": "552e46aa2217dd8fdb55e81b1f8782bdbc1c7fb1ef2a23af452d4bb037cbea32", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0010987310729246428}}, "usage": {"input_tokens": 4925, "output_tokens": 1}, "latency_ms": 1302.5149280001642, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 4925, "compiled_prompt_sha256": "4a6b0d4b6980f14bef7526e9432c0b43fd867a8f2e6c83e0a49d5a586beb651b", "compiled_ids_sha256": "accd3f579f223ef8c9e893a30ce53e124f52a23f80e143b1aae85905aaff4023", "candidate_token_ids": [32, 33], "logits": [12.4375, 19.25], "runtime_probabilities": [0.001098731067031622, 0.9989012479782104], "decide_seconds": 1.2747247620027338, "probabilities_f64_minus_runtime_max_abs": 2.0948864865033556e-08}}}, "kev_noul": 0.0010987310729246428, "choice": "no", "probs": {"yes": 0.0010987310729246428, "no": 0.9989012689270753}, "confidence": 0.9978025378541506, "seconds": 1.302682706998894, "prompt_tokens": 4925, "eval_tokens": 1, "server_latency_ms": 1302.5149280001642, "prompt_eval_ms": null, "total_ms": 1302.682706998894, "done_reason": null, "id": "off-02", "task": "ka", "slug": "how-we-work", "label": "no", "kind": "off-page", "state_sha256": "77d23d89a5adf6c0d561223d2c37308b5118a80f1f07f35f78bb1b0d08276e99", "request_sha256": "552e46aa2217dd8fdb55e81b1f8782bdbc1c7fb1ef2a23af452d4bb037cbea32", "compiled_tokens": 4925, "compiled_prompt_sha256": "4a6b0d4b6980f14bef7526e9432c0b43fd867a8f2e6c83e0a49d5a586beb651b", "compiled_ids_sha256": "accd3f579f223ef8c9e893a30ce53e124f52a23f80e143b1aae85905aaff4023", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:24Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "off-03", "task": "ka", "slug": "llms-txt", "label": "no", "kind": "off-page", "state_sha256": "96cab65488f086e43a1b03248954f7a7abadb4ae49c8591b267586825532ff41", "request_sha256": "92846aa94ff8202ba46abcdcff4cc2605ea586391ee0ccb8ca610ce4dfdd29bf", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:24Z"}
{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "off-04", "task": "ka", "slug": "three-new-voices-at-the-narrators-chair", "label": "no", "kind": "off-page", "state_sha256": "d944ffba7d7a97d70ffb305031ae8069d7e1bf47491eb08abdaea5d81af2ffe1", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:24Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "off-05", "task": "ka", "slug": "two-hours-on-battery", "label": "no", "kind": "off-page", "state_sha256": "c4c4db33a200bbafecb6bc46874d79161edbca9f5ee4e5451aac9225077b64f1", "request_sha256": "20e3610dcf34195439f1acc62acc6349f52980cd8e9b8c61155d5632aa3e2edf", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:24Z"}
{"mode": "kev-noul", "prompt_sha256": "9b62a1b35b3173c70b56c9acb97460e2a7e69d86a505e5032218cb9966259975", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0006263341581094201}}, "usage": {"input_tokens": 5219, "output_tokens": 1}, "latency_ms": 1357.8174139984185, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5219, "compiled_prompt_sha256": "33782e20773aff604d83e65a9d36788cdbf9a25c0a609f8d444287984c137b25", "compiled_ids_sha256": "d84f54ad8417204873581c8cb5de15670c6ecd8b47030b4f4a681033532db81a", "candidate_token_ids": [32, 33], "logits": [11.75, 19.125], "runtime_probabilities": [0.000626334163825959, 0.999373733997345], "decide_seconds": 1.3304090149977128, "probabilities_f64_minus_runtime_max_abs": 6.815545450233884e-08}}}, "kev_noul": 0.0006263341581094201, "choice": "no", "probs": {"yes": 0.0006263341581094201, "no": 0.9993736658418906}, "confidence": 0.9987473316837812, "seconds": 1.3579840029997285, "prompt_tokens": 5219, "eval_tokens": 1, "server_latency_ms": 1357.8174139984185, "prompt_eval_ms": null, "total_ms": 1357.9840029997285, "done_reason": null, "id": "off-06", "task": "ka", "slug": "reading-is-fast", "label": "no", "kind": "off-page", "state_sha256": "79b9c9e1be3ac3cdc484fec1f5eb9a83bf909612d6eb1c3c7c83bb469125ff2f", "request_sha256": "9b62a1b35b3173c70b56c9acb97460e2a7e69d86a505e5032218cb9966259975", "compiled_tokens": 5219, "compiled_prompt_sha256": "33782e20773aff604d83e65a9d36788cdbf9a25c0a609f8d444287984c137b25", "compiled_ids_sha256": "d84f54ad8417204873581c8cb5de15670c6ecd8b47030b4f4a681033532db81a", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:26Z"}
{"mode": "kev-noul", "prompt_sha256": "84543fe40f7be55f38cedd0c0f4e77d5730b7aeb7551a6fd8294abf7816ebddf", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.001410358496618023}}, "usage": {"input_tokens": 4443, "output_tokens": 1}, "latency_ms": 1183.2615029998124, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 4443, "compiled_prompt_sha256": "a743325d5b143657a18f4d393c95d8b476ed0a93800f03b12351bd2e8b69fe24", "compiled_ids_sha256": "d5c972444b078ff5ce4995e0e1b35e47d283ab49e07917dd9f503245f2a542b6", "candidate_token_ids": [32, 33], "logits": [12.4375, 19.0], "runtime_probabilities": [0.0014103584690019488, 0.9985895752906799], "decide_seconds": 1.1614542300012545, "probabilities_f64_minus_runtime_max_abs": 6.621270198792217e-08}}}, "kev_noul": 0.001410358496618023, "choice": "no", "probs": {"yes": 0.001410358496618023, "no": 0.998589641503382}, "confidence": 0.9971792830067641, "seconds": 1.1834123020016705, "prompt_tokens": 4443, "eval_tokens": 1, "server_latency_ms": 1183.2615029998124, "prompt_eval_ms": null, "total_ms": 1183.4123020016705, "done_reason": null, "id": "off-07", "task": "ka", "slug": "a-rig-your-friend-already-owns", "label": "no", "kind": "off-page", "state_sha256": "9d22ebe41036f74e0c8501e4631982f4d610eebb419288b4535c8a5e9f9b65d4", "request_sha256": "84543fe40f7be55f38cedd0c0f4e77d5730b7aeb7551a6fd8294abf7816ebddf", "compiled_tokens": 4443, "compiled_prompt_sha256": "a743325d5b143657a18f4d393c95d8b476ed0a93800f03b12351bd2e8b69fe24", "compiled_ids_sha256": "d5c972444b078ff5ce4995e0e1b35e47d283ab49e07917dd9f503245f2a542b6", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:27Z"}
{"mode": "kev-noul", "prompt_sha256": "713661cd57354fb28ebff6d4bfea3fe674b6a1f304124e9af12717cf96e1acbe", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.005911068856243796}}, "usage": {"input_tokens": 5831, "output_tokens": 1}, "latency_ms": 1536.7539519975253, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5831, "compiled_prompt_sha256": "177ba09734eac6b4ccb994bdaa1effe5e07ba36a995cb0aa2f8ddf05660eeb4b", "compiled_ids_sha256": "7c230617a8bf884091a29671a844c96a13f0bb0fbb4945ce575a0913f5e540f6", "candidate_token_ids": [32, 33], "logits": [13.5, 18.625], "runtime_probabilities": [0.005911068990826607, 0.9940890073776245], "decide_seconds": 1.5069622699993488, "probabilities_f64_minus_runtime_max_abs": 7.623386832911194e-08}}}, "kev_noul": 0.005911068856243796, "choice": "no", "probs": {"yes": 0.005911068856243796, "no": 0.9940889311437562}, "confidence": 0.9881778622875124, "seconds": 1.5369232410012046, "prompt_tokens": 5831, "eval_tokens": 1, "server_latency_ms": 1536.7539519975253, "prompt_eval_ms": null, "total_ms": 1536.9232410012046, "done_reason": null, "id": "off-08", "task": "ka", "slug": "the-move", "label": "no", "kind": "off-page", "state_sha256": "3d1d2fe049acbdd5e286fbc87c126a77f0410d05b6a51694ef22f6658b409e9f", "request_sha256": "713661cd57354fb28ebff6d4bfea3fe674b6a1f304124e9af12717cf96e1acbe", "compiled_tokens": 5831, "compiled_prompt_sha256": "177ba09734eac6b4ccb994bdaa1effe5e07ba36a995cb0aa2f8ddf05660eeb4b", "compiled_ids_sha256": "7c230617a8bf884091a29671a844c96a13f0bb0fbb4945ce575a0913f5e540f6", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:28Z"}
{"mode": "kev-noul", "prompt_sha256": "2a2695e3876d1b582ebab9d92a35db85ae1f131c90e682872ae735ead287cd2b", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0007554056327295276}}, "usage": {"input_tokens": 5126, "output_tokens": 1}, "latency_ms": 1342.4753120016248, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5126, "compiled_prompt_sha256": "1720bd220750603c9e498108019ba6dcc02cff0dc58ebf6343e5f0f78d60e8bc", "compiled_ids_sha256": "63644f172bffe644284fdc0a92f50997103fd60b8b9bf91d836a5a6c41ebd4d4", "candidate_token_ids": [32, 33], "logits": [12.4375, 19.625], "runtime_probabilities": [0.000755405577365309, 0.9992445707321167], "decide_seconds": 1.317619482997543, "probabilities_f64_minus_runtime_max_abs": 2.3635153834078437e-08}}}, "kev_noul": 0.0007554056327295276, "choice": "no", "probs": {"yes": 0.0007554056327295276, "no": 0.9992445943672704}, "confidence": 0.9984891887345408, "seconds": 1.3426335900003323, "prompt_tokens": 5126, "eval_tokens": 1, "server_latency_ms": 1342.4753120016248, "prompt_eval_ms": null, "total_ms": 1342.6335900003323, "done_reason": null, "id": "off-09", "task": "ka", "slug": "fifty-seven-milliseconds", "label": "no", "kind": "off-page", "state_sha256": "bec5d08ff12a12a4d87ca9d1d31e3bec848dfe329b6a47e9b8616e071da4ba5c", "request_sha256": "2a2695e3876d1b582ebab9d92a35db85ae1f131c90e682872ae735ead287cd2b", "compiled_tokens": 5126, "compiled_prompt_sha256": "1720bd220750603c9e498108019ba6dcc02cff0dc58ebf6343e5f0f78d60e8bc", "compiled_ids_sha256": "63644f172bffe644284fdc0a92f50997103fd60b8b9bf91d836a5a6c41ebd4d4", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:30Z"}
{"mode": "kev-noul", "prompt_sha256": "9b0b371518ff7abb1bcdef463b5d83200d64ad49e561aa94ba78125be9665ff7", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0012448433020883743}}, "usage": {"input_tokens": 2913, "output_tokens": 1}, "latency_ms": 775.2713930021855, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 2913, "compiled_prompt_sha256": "83b3a336eba6b8faeb18def0c78ea042a820fa485ae67a2d7d3a5fc57fc9e3e4", "compiled_ids_sha256": "95615075893ffeb4db3647d9312eaaaa7ef8166e8ef2d0b567a15ef98aea3485", "candidate_token_ids": [32, 33], "logits": [12.4375, 19.125], "runtime_probabilities": [0.0012448433553799987, 0.9987550973892212], "decide_seconds": 0.7605676550010685, "probabilities_f64_minus_runtime_max_abs": 5.930869040327025e-08}}}, "kev_noul": 0.0012448433020883743, "choice": "no", "probs": {"yes": 0.0012448433020883743, "no": 0.9987551566979116}, "confidence": 0.9975103133958232, "seconds": 0.7753780620005273, "prompt_tokens": 2913, "eval_tokens": 1, "server_latency_ms": 775.2713930021855, "prompt_eval_ms": null, "total_ms": 775.3780620005273, "done_reason": null, "id": "off-10", "task": "ka", "slug": "six-worlds-painted-large", "label": "no", "kind": "off-page", "state_sha256": "4902422ba42eec05800d9a72fd69c58559d0cb098878daec40505c90f50184d4", "request_sha256": "9b0b371518ff7abb1bcdef463b5d83200d64ad49e561aa94ba78125be9665ff7", "compiled_tokens": 2913, "compiled_prompt_sha256": "83b3a336eba6b8faeb18def0c78ea042a820fa485ae67a2d7d3a5fc57fc9e3e4", "compiled_ids_sha256": "95615075893ffeb4db3647d9312eaaaa7ef8166e8ef2d0b567a15ef98aea3485", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:31Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "off-11", "task": "ka", "slug": "outside-judges", "label": "no", "kind": "off-page", "state_sha256": "fc7235d0fabb2221f559392fc0236d981f9898d96425df0d7355f2d6f11d17d6", "request_sha256": "19d9e4f9f0d5b96afe38a6413aa876999067343b558556fca9a50d2df81ec057", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:31Z"}
{"mode": "kev-noul", "prompt_sha256": "f509dfbc3ebb640055c777670874e04b146aa31d974e7416edf8428ae2de8b83", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0006263341581094201}}, "usage": {"input_tokens": 5228, "output_tokens": 1}, "latency_ms": 1355.0236650007719, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5228, "compiled_prompt_sha256": "99ab9888957fcb030c783278c7b37103c6b48ec6b062400aed4051c7e97bbfb9", "compiled_ids_sha256": "4094aaf71d8eb66c3fd663751f4449553694e02378119439e7e501ac1984016d", "candidate_token_ids": [32, 33], "logits": [12.0, 19.375], "runtime_probabilities": [0.000626334163825959, 0.999373733997345], "decide_seconds": 1.3292053430013766, "probabilities_f64_minus_runtime_max_abs": 6.815545450233884e-08}}}, "kev_noul": 0.0006263341581094201, "choice": "no", "probs": {"yes": 0.0006263341581094201, "no": 0.9993736658418906}, "confidence": 0.9987473316837812, "seconds": 1.3551960540025902, "prompt_tokens": 5228, "eval_tokens": 1, "server_latency_ms": 1355.0236650007719, "prompt_eval_ms": null, "total_ms": 1355.1960540025902, "done_reason": null, "id": "off-12", "task": "ka", "slug": "the-typist-and-the-developer", "label": "no", "kind": "off-page", "state_sha256": "2d33747167f415355e5637f99962534c215404a547456ccde9c80e2d7591cf5d", "request_sha256": "f509dfbc3ebb640055c777670874e04b146aa31d974e7416edf8428ae2de8b83", "compiled_tokens": 5228, "compiled_prompt_sha256": "99ab9888957fcb030c783278c7b37103c6b48ec6b062400aed4051c7e97bbfb9", "compiled_ids_sha256": "4094aaf71d8eb66c3fd663751f4449553694e02378119439e7e501ac1984016d", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:32Z"}
{"mode": "kev-noul", "prompt_sha256": "e9a0fe7c450713b43dc95f2c9b5af7257cf5c7de9b680dbaa37f90bbc64fe0f8", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0021827164453451808}}, "usage": {"input_tokens": 5198, "output_tokens": 1}, "latency_ms": 1352.538415001618, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 5198, "compiled_prompt_sha256": "a87dba0180cb1b5060b6860296d117704f106262dc0aefedfa758b8937185492", "compiled_ids_sha256": "6517fffa0d330d494b72d60ad610d24c927d966254ec9699a38ae65b6c0ffa7f", "candidate_token_ids": [32, 33], "logits": [13.0, 19.125], "runtime_probabilities": [0.002182716503739357, 0.9978172779083252], "decide_seconds": 1.3252826440002536, "probabilities_f64_minus_runtime_max_abs": 5.646329515052173e-09}}}, "kev_noul": 0.0021827164453451808, "choice": "no", "probs": {"yes": 0.0021827164453451808, "no": 0.9978172835546548}, "confidence": 0.9956345671093096, "seconds": 1.3527194140006031, "prompt_tokens": 5198, "eval_tokens": 1, "server_latency_ms": 1352.538415001618, "prompt_eval_ms": null, "total_ms": 1352.7194140006031, "done_reason": null, "id": "off-13", "task": "ka", "slug": "compression-is-the-objective", "label": "no", "kind": "off-page", "state_sha256": "68bc7c0890d5d05afefc1bbd6178315fddd18fb35775c738f50544096a982f7f", "request_sha256": "e9a0fe7c450713b43dc95f2c9b5af7257cf5c7de9b680dbaa37f90bbc64fe0f8", "compiled_tokens": 5198, "compiled_prompt_sha256": "a87dba0180cb1b5060b6860296d117704f106262dc0aefedfa758b8937185492", "compiled_ids_sha256": "6517fffa0d330d494b72d60ad610d24c927d966254ec9699a38ae65b6c0ffa7f", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:33Z"}
{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "off-14", "task": "ka", "slug": "the-open-call", "label": "no", "kind": "off-page", "state_sha256": "dd50fd254da5ef89e4284d245102a3ca9f9f7dd4e41f90c7687adb01223798a4", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:33Z"}
{"refused": "kev-422", "choice": null, "probs": null, "seconds": null, "server_message": "HTTP 422: invalid_request: Compiled prompt exceeds token budget; no truncation is allowed", "n_options": 2, "mode": "kev-noul", "id": "off-15", "task": "ka", "slug": "the-new-kid", "label": "no", "kind": "off-page", "state_sha256": "676acfa90c7de336adc467067853954373df2cc40b45fb728501238f7b963b79", "request_sha256": "afd2a4577ff2913d759649def5f3b2644e5e08e6a1c1eb230628a8f33099cf68", "bridge_refused": {"status": 422, "code": "invalid_request", "message": "Compiled prompt exceeds token budget; no truncation is allowed"}, "prompt_sha256_matches_kev9b": null, "kev9b_refused": "kev-422", "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:33Z"}
{"mode": "kev-noul", "prompt_sha256": "b20457a7455573562b8b154b5d9aca5a954b6a2434349474b2a591f6214d0f38", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0010987310729246428}}, "usage": {"input_tokens": 4705, "output_tokens": 1}, "latency_ms": 1250.2743090008153, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 4705, "compiled_prompt_sha256": "1578694f810a08bf721dbf37586bb36cf5bfb979fc3f4862db6e10e3691ad8d9", "compiled_ids_sha256": "51e867960290f6d2de032f439359d1832e0beee26288c25d3b9c5365c7dd8f66", "candidate_token_ids": [32, 33], "logits": [12.4375, 19.25], "runtime_probabilities": [0.001098731067031622, 0.9989012479782104], "decide_seconds": 1.2246761850001349, "probabilities_f64_minus_runtime_max_abs": 2.0948864865033556e-08}}}, "kev_noul": 0.0010987310729246428, "choice": "no", "probs": {"yes": 0.0010987310729246428, "no": 0.9989012689270753}, "confidence": 0.9978025378541506, "seconds": 1.2504587980001816, "prompt_tokens": 4705, "eval_tokens": 1, "server_latency_ms": 1250.2743090008153, "prompt_eval_ms": null, "total_ms": 1250.4587980001816, "done_reason": null, "id": "off-16", "task": "ka", "slug": "where-your-question-goes", "label": "no", "kind": "off-page", "state_sha256": "cdc24e58a19354d77487d8df5e8782b2c15b4483afe7405f34478bf769d44b11", "request_sha256": "b20457a7455573562b8b154b5d9aca5a954b6a2434349474b2a591f6214d0f38", "compiled_tokens": 4705, "compiled_prompt_sha256": "1578694f810a08bf721dbf37586bb36cf5bfb979fc3f4862db6e10e3691ad8d9", "compiled_ids_sha256": "51e867960290f6d2de032f439359d1832e0beee26288c25d3b9c5365c7dd8f66", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:35Z"}
{"mode": "kev-noul", "prompt_sha256": "8f17af6313832b7a0ccf7340d2d3e7ccbbf39825e0127d3fca2cf9fda26773bf", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0011695102650555148}}, "usage": {"input_tokens": 1825, "output_tokens": 1}, "latency_ms": 502.27899699893896, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 1825, "compiled_prompt_sha256": "c5d9185b828324dbe9810ef231080f603ada63ea5e7cea516e3f42a7fc760328", "compiled_ids_sha256": "fc0923ce3a96f7960b95c8c9bbcfacfeac7d854f512ca13d65e51e8ee4a7dafb", "candidate_token_ids": [32, 33], "logits": [12.375, 19.125], "runtime_probabilities": [0.0011695101857185364, 0.9988304972648621], "decide_seconds": 0.4930216879984073, "probabilities_f64_minus_runtime_max_abs": 7.529917578352752e-09}}}, "kev_noul": 0.0011695102650555148, "choice": "no", "probs": {"yes": 0.0011695102650555148, "no": 0.9988304897349445}, "confidence": 0.997660979469889, "seconds": 0.5023540760012111, "prompt_tokens": 1825, "eval_tokens": 1, "server_latency_ms": 502.27899699893896, "prompt_eval_ms": null, "total_ms": 502.35407600121107, "done_reason": null, "id": "off-17", "task": "ka", "slug": "nine-worlds-one-dog", "label": "no", "kind": "off-page", "state_sha256": "84c842a71a59b037055f5448244f732c94578d47c68a907c19c9a4a0324a0095", "request_sha256": "8f17af6313832b7a0ccf7340d2d3e7ccbbf39825e0127d3fca2cf9fda26773bf", "compiled_tokens": 1825, "compiled_prompt_sha256": "c5d9185b828324dbe9810ef231080f603ada63ea5e7cea516e3f42a7fc760328", "compiled_ids_sha256": "fc0923ce3a96f7960b95c8c9bbcfacfeac7d854f512ca13d65e51e8ee4a7dafb", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:35Z"}
{"refused": "overlong", "choice": null, "probs": null, "seconds": null, "why": "inherited from arm 2's own refusal rows so the denominator matches (GAPS-CARD.md, arm 12)", "id": "off-18", "task": "ka", "slug": "how-the-beat-lab-works", "label": "no", "kind": "off-page", "state_sha256": "fdb6d4bf89ebedc44a2d887309bf63ba59598c898a6eca00760115d49bc2133d", "prompt_sha256_matches_kev9b": null, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:35Z"}
{"mode": "kev-noul", "prompt_sha256": "aca3d93ad50e4f4eaf9e80c4f149926870065ff34d47a39b4f05dd7116ffb923", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0013250224172131609}}, "usage": {"input_tokens": 7291, "output_tokens": 1}, "latency_ms": 1920.8767849995638, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 7291, "compiled_prompt_sha256": "275d2689568ae23cda11b437697bc0c4c93192e6cca1c9a24348f7a1852ab590", "compiled_ids_sha256": "bc1a005d1429bb0e9d8ae4c1e1c6792b7d5f95036f2d7a57af8e24ef6d0c4411", "candidate_token_ids": [32, 33], "logits": [12.75, 19.375], "runtime_probabilities": [0.001325022429227829, 0.9986749291419983], "decide_seconds": 1.8822033209980873, "probabilities_f64_minus_runtime_max_abs": 4.844078849153277e-08}}}, "kev_noul": 0.0013250224172131609, "choice": "no", "probs": {"yes": 0.0013250224172131609, "no": 0.9986749775827869}, "confidence": 0.9973499551655738, "seconds": 1.9210954629998014, "prompt_tokens": 7291, "eval_tokens": 1, "server_latency_ms": 1920.8767849995638, "prompt_eval_ms": null, "total_ms": 1921.0954629998014, "done_reason": null, "id": "off-19", "task": "ka", "slug": "three-at-the-table", "label": "no", "kind": "off-page", "state_sha256": "5ee526a9690f5511741a41a6eb1e91418702e2cd8e49d1c2db9e03802a7bde5c", "request_sha256": "aca3d93ad50e4f4eaf9e80c4f149926870065ff34d47a39b4f05dd7116ffb923", "compiled_tokens": 7291, "compiled_prompt_sha256": "275d2689568ae23cda11b437697bc0c4c93192e6cca1c9a24348f7a1852ab590", "compiled_ids_sha256": "bc1a005d1429bb0e9d8ae4c1e1c6792b7d5f95036f2d7a57af8e24ef6d0c4411", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:37Z"}
{"mode": "kev-noul", "prompt_sha256": "9b582945b20a9d0a581a3907b8efa5f8c9be40c164523b0b971b01b4a8fc46f5", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0031726828424851893}}, "usage": {"input_tokens": 6251, "output_tokens": 1}, "latency_ms": 1633.9450260020385, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 6251, "compiled_prompt_sha256": "7605b289f127cd843596e77c2ddedcc2af9b36bdffa26a8c2ae2fa2b1dbc6f3f", "compiled_ids_sha256": "c4b36fee5baeba7261302317f84218601e1cdc225de1c23ce41f7d8fdc48e5aa", "candidate_token_ids": [32, 33], "logits": [13.125, 18.875], "runtime_probabilities": [0.0031726830638945103, 0.9968273043632507], "decide_seconds": 1.6027853550003783, "probabilities_f64_minus_runtime_max_abs": 1.27942640926193e-08}}}, "kev_noul": 0.0031726828424851893, "choice": "no", "probs": {"yes": 0.0031726828424851893, "no": 0.9968273171575148}, "confidence": 0.9936546343150297, "seconds": 1.6341345849978097, "prompt_tokens": 6251, "eval_tokens": 1, "server_latency_ms": 1633.9450260020385, "prompt_eval_ms": null, "total_ms": 1634.1345849978097, "done_reason": null, "id": "off-20", "task": "ka", "slug": "the-same-sixteen", "label": "no", "kind": "off-page", "state_sha256": "830caf9405ece8c6ff598d3adfbdaf8da136fd260537d81f576a5059e5a17953", "request_sha256": "9b582945b20a9d0a581a3907b8efa5f8c9be40c164523b0b971b01b4a8fc46f5", "compiled_tokens": 6251, "compiled_prompt_sha256": "7605b289f127cd843596e77c2ddedcc2af9b36bdffa26a8c2ae2fa2b1dbc6f3f", "compiled_ids_sha256": "c4b36fee5baeba7261302317f84218601e1cdc225de1c23ce41f7d8fdc48e5aa", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:39Z"}
{"mode": "kev-noul", "prompt_sha256": "5da80e7cb08a3aa664717d70834de08a1c2bb7cf6e2d33e45b04ab19c2672493", "n_options": 2, "kev_response": {"model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "answers": {"q": {"type": "noul", "noul": 0.0024726231566347748}}, "usage": {"input_tokens": 7006, "output_tokens": 1}, "latency_ms": 1842.6254149999295, "apus": {"q": {"effort": "high", "executed_layers": 32, "projection": "full_head", "calibrated": false, "primitive": "noul", "compiled_tokens": 7006, "compiled_prompt_sha256": "e1a2c44e954cdb213f4baa50323be36d4dd90c6e16dc90de7f8d16e758d00c55", "compiled_ids_sha256": "900508349ff6a88a850dfbbdc51e59cbdc1896250966867f488d071bc844f06e", "candidate_token_ids": [32, 33], "logits": [12.75, 18.75], "runtime_probabilities": [0.0024726232513785362, 0.9975274205207825], "decide_seconds": 1.8095107289991574, "probabilities_f64_minus_runtime_max_abs": 4.3677417127518936e-08}}}, "kev_noul": 0.0024726231566347748, "choice": "no", "probs": {"yes": 0.0024726231566347748, "no": 0.9975273768433652}, "confidence": 0.9950547536867305, "seconds": 1.842804232997878, "prompt_tokens": 7006, "eval_tokens": 1, "server_latency_ms": 1842.6254149999295, "prompt_eval_ms": null, "total_ms": 1842.804232997878, "done_reason": null, "id": "off-21", "task": "ka", "slug": "seat-trials", "label": "no", "kind": "off-page", "state_sha256": "c9374a27b5e06de1440ffed7d81385fa68f7572790c0f553f8f2948c0218eea1", "request_sha256": "5da80e7cb08a3aa664717d70834de08a1c2bb7cf6e2d33e45b04ab19c2672493", "compiled_tokens": 7006, "compiled_prompt_sha256": "e1a2c44e954cdb213f4baa50323be36d4dd90c6e16dc90de7f8d16e758d00c55", "compiled_ids_sha256": "900508349ff6a88a850dfbbdc51e59cbdc1896250966867f488d071bc844f06e", "prompt_sha256_matches_kev9b": true, "card": "RTX 3090 (fp 33a0b4acb8e8)", "box": "benchbox", "os": {"pretty_name": "Ubuntu 26.04 LTS", "id": "ubuntu", "version_id": "26.04", "kernel": "7.0.0-31-generic", "arch": "x86_64", "source": "/etc/os-release"}, "dry": false, "preflight": false, "effort": "high", "runtime": "openjet_runtime (APUS-OpenJev-v1-4B@422b3741) in-process via deployment/contracts.py (APUS-OpenJev-v1@af276da5), one question per request", "model_revision": "422b3741f8b5c092eeefef847c1ca89d78337d45", "bridge_revision": "af276da53a2e7a6d0fbdbafc9280bf9982b2d3a6", "torch": "2.14.0+cu130", "torch_cuda": "13.0", "transformers": "5.16.1", "tokenizers": "0.23.2", "dtype": "bfloat16", "kernels": "reference (no flash-linear-attention), sdpa", "calibrated": false, "arm": "apus-4b-high", "model": "apus-ailab/APUS-OpenJev-v1-4B@422b3741", "quant": "bf16 (merged LoRA r=8; full weights incl. the vision tower; tied embeddings)", "card_uuid": "GPU-fp:33a0b4acb8e8", "num_ctx": 8191, "power_cap_w": 300, "seed": 20260921, "stamp": "2026-09-28T16:46:40Z"}