{
 "round": "two-frontiers-cove-r1",
 "gate": "G-CALIBRATE",
 "verdict": "PASS",
 "detail": "every one of the 6 carried seats filed all 6 anchor cells, and each publishes its own anchor mean and its offset from the panel's over all 6",
 "started_utc": "2026-09-06T19:05:00Z",
 "finished_utc": "2026-09-06T19:05:00Z",
 "evidence": {
  "seats": [
   {
    "seat": "deepseek-v4-pro",
    "family": "deepseek",
    "anchor_cells": 6,
    "anchor_mean": 6.0,
    "offset_from_panel": -0.361
   },
   {
    "seat": "gemma4-31b",
    "family": "google",
    "anchor_cells": 6,
    "anchor_mean": 6.667,
    "offset_from_panel": 0.306
   },
   {
    "seat": "kimi-k3",
    "family": "moonshot",
    "anchor_cells": 6,
    "anchor_mean": 5.5,
    "offset_from_panel": -0.861
   },
   {
    "seat": "mistral-large-3-675b",
    "family": "mistral",
    "anchor_cells": 6,
    "anchor_mean": 7.333,
    "offset_from_panel": 0.972
   },
   {
    "seat": "nemotron-3-ultra",
    "family": "nvidia",
    "anchor_cells": 6,
    "anchor_mean": 6.25,
    "offset_from_panel": -0.111
   },
   {
    "seat": "qwen3.5-397b",
    "family": "alibaba",
    "anchor_cells": 6,
    "anchor_mean": 6.417,
    "offset_from_panel": 0.056
   }
  ],
  "cells": 36,
  "cells_expected": 36,
  "cells_expected_full_panel": 42,
  "complete_basis": "6 asks x 6 seats that sat = 36. The registered table's 42 is 6 x 7 seats; the difference is a seat RETIRED at its audition, not an anchor cell a seat failed to file, and those are different findings.",
  "by_ask_spread": {
   "S1-ask-A": 3.0,
   "S1-ask-B": 3.0,
   "S2-ask-A": 5.5,
   "S2-ask-B": 3.0,
   "S3-ask-A": 2.0,
   "S3-ask-B": 4.5
  },
  "a_wide_spread_is_a_finding": "a wide spread is a finding, not a failure — no arm is refused for it and no threshold exists — and a narrow one is not a calibration (§8).",
  "per_ask_between_seat_spread": "NOT-RUN as a SCORER field — `anchor_report` groups by SCENARIO and this round has two asks per scenario, so its `by_scenario[*]` reports 14 seats for a seven-seat panel and its spread pools two DIFFERENT anchor replies. The report layer below re-groups the scorer's own anchor cells by (seat, ask) so all 36 that EXIST are kept — six asks × the six seats that sat — against the 42 a seven-seat panel would have filed; the difference is the retired chair (A6) and not a missing reading. The scorer's own grouping prints beside it, labelled MIXED.",
  "problems": []
 },
 "arms": {},
 "records": [],
 "withheld": {
  "rule": "a receipt publishes in the kit, so these fields are removed here rather than at kit time — and named, because a field quietly absent from a receipt is indistinguishable from a gate that never read it."
 },
 "note": "written by harness/cove/gates_cove.py. Every generation call rides the same send() the scored run uses; a gate that hand-rolled its own call would receipt a field the run never sent."
}