{
 "arm": {
  "box": "ollama-cloud via the laptop system ollama v0.32.14",
  "cards": [],
  "host": "http://127.0.0.1:withheld-2026-10-14",
  "mode": "generate",
  "model": "deepseek-v4-pro:0813-cloud",
  "name": "cloud-deepseek-v4-pro--0813-cloud",
  "note": "cloud reading 0: run.py generate path, verbatim body; no local GPU",
  "num_ctx": 131072,
  "power_cap_w": 0,
  "quant": "provider-served; see roster (not exposed by the daemon)",
  "runtime": "ollama",
  "seed": 20260921,
  "top_logprobs": 20
 },
 "ended": "2026-09-23T18:31:07Z",
 "first_row_utc": "2026-09-23T16:15:39Z",
 "kit": {
  "a_labels": {
   "no": 21,
   "yes": 42
  },
  "b_labelled": 1,
  "bank_sha256": "35d58866da02e3ac731814e63ef0ceac97a25ee8b224c1c32c29df6ddbcc3d1c",
  "built": null,
  "c_per_guest": 18,
  "c_roster": [
   {
    "id": "darwin",
    "name": "Charles Darwin"
   },
   {
    "id": "einstein",
    "name": "Albert Einstein"
   },
   {
    "id": "hypatia",
    "name": "Hypatia of Alexandria"
   },
   {
    "id": "ibn_sina",
    "name": "Ibn Sina"
   },
   {
    "id": "sagan",
    "name": "Carl Sagan"
   },
   {
    "id": "socrates",
    "name": "Socrates of Athens"
   }
  ],
  "counts": {
   "a": 63,
   "articles": 42,
   "b": 42,
   "c": 108
  },
  "docroot": "/workshop/s2s-research-hub/site",
  "seed": 20260921,
  "wall_sha256": "80484f8379a5977cf56033f335c27c8f54fac683c7adf3e800def5fa419e698c",
  "wall_url": "https://longtable.strata2signal.com/api/wall"
 },
 "last_row_utc": "2026-09-23T16:18:10Z",
 "protocol": {
  "body": {
   "logprobs": true,
   "options": {
    "num_ctx": 131072,
    "num_predict": 8,
    "seed": 20260921,
    "temperature": 0
   },
   "stream": false,
   "think": false,
   "top_logprobs": 20
  },
  "extraction": "first character of the reply that is one of the option letters (run.decide, generate branch)",
  "instructions": {
   "a": "Does the article in the state answer this reader's question? Question: %s",
   "c": "Six people are at a dinner table. Which of them said the line below?\nLine: %s",
   "d": "DOORMAN_SYSTEM v1.2 verbatim (kit sha checked)",
   "f": "Does the excerpt in the state answer this question, using only what the excerpt itself states? Question: %s"
  },
  "not_measured": "energy, residency, card: a cloud call has no local board",
  "noul": "tasks a and f apply run.py's yes/no calibration to the one-hot p, exactly as the gemma4 generate arm did; it never changes the pick",
  "prompt": "run.readout_prompt(state, instructions, options)",
  "source": "jev-2026-09-21/run.py call_ollama + decide (generate mode), run_addenda.py task_d / task_f; imported, not copied"
 },
 "raw_file": "/workshop/estate/bench/cost-of-intelligence-2026-09-23/cloud-reading-0/raw/deepseek-v4-pro--0813-cloud.a.raw.jsonl",
 "reading": "cost-of-intelligence 2026-09-23, cloud reading 0",
 "reasoning_mode": "exact: think=false, num_predict 8",
 "rows_file": "/workshop/estate/bench/cost-of-intelligence-2026-09-23/cloud-reading-0/rows/deepseek-v4-pro--0813-cloud.a.jsonl",
 "show": {
  "capabilities": [
   "completion",
   "tools",
   "thinking"
  ],
  "details": {
   "families": null,
   "family": "deepseek4",
   "format": "",
   "parameter_size": "1650497936906",
   "parent_model": "deepseek-v4-pro:0813",
   "quantization_level": "FP8"
  },
  "modified_at": "2026-08-13T15:00:00Z",
  "probe_file": "/workshop/estate/bench/cost-of-intelligence-2026-09-23/cloud-reading-0/probes/deepseek-v4-pro__0813-cloud.json",
  "remote_host": null,
  "remote_model": null
 },
 "started": "2026-09-23T16:15:39Z",
 "summary": {
  "accuracy": 0.9365079365079365,
  "accuracy_rule": "run.summarise: choice == label over reached rows; an unparsed reply is scored as run.decide leaves it (uniform probabilities, so its choice is the FIRST option)",
  "correct": 59,
  "done_reason_length": 6,
  "empty_replies": 0,
  "eval_tokens_total": 207,
  "failed_requests": 0,
  "format_failure_ids": [],
  "format_failures": 0,
  "http_429s": 0,
  "latency_what": "wall time of each successful POST to the local daemon, network + provider queue + compute included",
  "letter_first_correct": 59,
  "letter_first_rule": "correct AND the reply opens with its letter; unparsed and prose-first replies count wrong (a stricter READING of the same rows, never a different extraction)",
  "median_decode_tok_s_client": 20.264378402710793,
  "median_seconds": 2.28133948300092,
  "median_ttft_s": 2.1554783799947472,
  "n_items": 63,
  "n_labelled": 63,
  "n_reached": 63,
  "non_bare_replies": 7,
  "non_bare_rule": "a parsed reply that is not just one letter (brackets and punctuation allowed): the extraction read a letter out of prose",
  "p95_seconds": 4.2578164560036385,
  "prompt_tokens_max": 73041,
  "prompt_tokens_total": 762543,
  "prose_first_replies": 0,
  "reasoning_observed_rows": 0,
  "refused_ids": [],
  "refused_overlong": 0,
  "remote_models": [],
  "retried_rows": 0,
  "run_summarise": {
   "accuracy": 0.9365079365079365,
   "brier": 0.1254266934888739,
   "correct": 59,
   "errors": 0,
   "eval_tokens_total": 207,
   "floored_rows": 0,
   "letters_floored_total": 0,
   "median_seconds": 2.28133948300092,
   "n": 63,
   "n_labelled": 63,
   "n_reached": 63,
   "p95_seconds": 4.2578164560036385,
   "prompt_tokens_total": 762543,
   "refused": 0,
   "refused_by_reason": {},
   "reliability": [
    {
     "accuracy": null,
     "correct": 0,
     "gap": null,
     "hi": 0.1,
     "lo": 0.0,
     "mean_p": null,
     "n": 0,
     "sum_p": 0.0
    },
    {
     "accuracy": null,
     "correct": 0,
     "gap": null,
     "hi": 0.2,
     "lo": 0.1,
     "mean_p": null,
     "n": 0,
     "sum_p": 0.0
    },
    {
     "accuracy": null,
     "correct": 0,
     "gap": null,
     "hi": 0.3,
     "lo": 0.2,
     "mean_p": null,
     "n": 0,
     "sum_p": 0.0
    },
    {
     "accuracy": null,
     "correct": 0,
     "gap": null,
     "hi": 0.4,
     "lo": 0.3,
     "mean_p": null,
     "n": 0,
     "sum_p": 0.0
    },
    {
     "accuracy": null,
     "correct": 0,
     "gap": null,
     "hi": 0.5,
     "lo": 0.4,
     "mean_p": null,
     "n": 0,
     "sum_p": 0.0
    },
    {
     "accuracy": null,
     "correct": 0,
     "gap": null,
     "hi": 0.6,
     "lo": 0.5,
     "mean_p": null,
     "n": 0,
     "sum_p": 0.0
    },
    {
     "accuracy": null,
     "correct": 0,
     "gap": null,
     "hi": 0.7,
     "lo": 0.6,
     "mean_p": null,
     "n": 0,
     "sum_p": 0.0
    },
    {
     "accuracy": null,
     "correct": 0,
     "gap": null,
     "hi": 0.8,
     "lo": 0.7,
     "mean_p": null,
     "n": 0,
     "sum_p": 0.0
    },
    {
     "accuracy": null,
     "correct": 0,
     "gap": null,
     "hi": 0.9,
     "lo": 0.8,
     "mean_p": null,
     "n": 0,
     "sum_p": 0.0
    },
    {
     "accuracy": 0.9365079365079365,
     "correct": 59,
     "gap": 0.05703091599171084,
     "hi": 1.0,
     "lo": 0.9,
     "mean_p": 0.9935388524996474,
     "n": 63,
     "sum_p": 62.592947707477784
    }
   ],
   "unparsed": 0
  },
  "served_models": [
   "deepseek-v4-pro:0813"
  ],
  "streamed_rows": 63,
  "strict_accuracy": 0.9365079365079365,
  "strict_correct": 59,
  "strict_rule": "an unparsed reply counts wrong",
  "strict_wilson95": [
   0.8478082206274346,
   0.9750343821600765
  ],
  "thinking_rows": 0,
  "wilson95": [
   0.8478082206274346,
   0.9750343821600765
  ]
 },
 "tag": "deepseek-v4-pro:0813-cloud",
 "task": "a",
 "task_name": "docent gate (task a, 63)"
}