{
  "schema": "lab-board/v0",
  "updated": "2026-05-25",
  "tagline": "Behavior Lab. Preregistered experiments, contracts, completions. Open in public.",
  "filters": ["all", "experiment", "trace pair", "contract", "completion analyzer", "benchmark", "scenario set", "workbench"],
  "cards": [
    {
      "id": "exp-006",
      "kind": "experiment",
      "title": "EXP-006 · Confirmatory E1+E2",
      "lede": "Preregistered 540-run replication of outcomes-only insufficiency and contract dominance across four canonical OPBR domains. Paper 2 v1.1.",
      "status": "converged",
      "version": "v1.1",
      "generation": 1,
      "updated": "2026-05-25",
      "license": "CC BY 4.0",
      "href": "exp/exp-006.html",
      "most_recent": true
    },
    {
      "id": "exp-004",
      "kind": "experiment",
      "title": "EXP-004 · Held-out repair",
      "lede": "On held-out OPBR scenarios, behavior contracts achieve F1 = 0.982 vs. best baseline F1 = 0.400. McNemar p ≈ 0. Paper 1 Study 5.",
      "status": "converged",
      "version": "v1.0",
      "generation": 1,
      "updated": "2026-05-10",
      "license": "CC BY 4.0",
      "href": "exp/exp-004.html",
      "signature": true
    },
    {
      "id": "exp-002",
      "kind": "experiment",
      "title": "EXP-002 · Semi-blind perturbation",
      "lede": "Under non-instructed perturbations across 4 domains, 91.4% of behavioral regressions produce passing outputs. 1,437 runs. Wilson CI [0.894, 0.932]. Paper 1 Study 2.",
      "status": "converged",
      "version": "v1.0",
      "generation": 1,
      "updated": "2026-04-28",
      "license": "CC BY 4.0",
      "href": "exp/exp-002.html"
    },
    {
      "id": "exp-003",
      "kind": "experiment",
      "title": "EXP-003 · Action-affordance merge",
      "lede": "Merging propose and commit into one ambiguous tool degrades every structural detector. 960 runs across 4 domains. Paper 1 Study 4.",
      "status": "converged",
      "version": "v1.0",
      "generation": 1,
      "updated": "2026-04-22",
      "license": "CC BY 4.0",
      "href": "exp/exp-003.html"
    },
    {
      "id": "exp-005",
      "kind": "experiment",
      "title": "EXP-005 · Completeness analyzer",
      "lede": "Heuristic: specification audit of Study 1 contracts. The Study 2 recall gap is enumerable specification work, not a flaw in the mechanism. Paper 1 Study 7.",
      "status": "converged",
      "version": "v1.0",
      "generation": 1,
      "updated": "2026-04-15",
      "license": "CC BY 4.0",
      "href": "exp/exp-005.html"
    },
    {
      "id": "exp-001",
      "kind": "experiment",
      "title": "EXP-001 · Controlled mechanism",
      "lede": "180/180 induced fast-lane runs pass output evaluation while behavior contracts fail. Drift paths also pass output+operational composite. 1,080 runs. Paper 1 Study 1.",
      "status": "converged",
      "version": "v1.0",
      "generation": 1,
      "updated": "2026-04-12",
      "license": "CC BY 4.0",
      "href": "exp/exp-001.html"
    },
    {
      "id": "exp-007",
      "kind": "experiment",
      "title": "EXP-007 · Cross-runtime (Codex CLI)",
      "lede": "Preregistered replication of EXP-004 under Codex CLI. Closes Gate E of Paper 1 acceptance. Runs pending Codex install + API key.",
      "status": "pending",
      "version": "v0.1",
      "generation": 1,
      "updated": "2026-05-25",
      "license": "CC BY 4.0",
      "href": "exp/exp-007.html"
    },
    {
      "id": "opbr-refund-commit",
      "kind": "trace pair",
      "title": "OPBR · refund commit",
      "lede": "Same correct refund. Two different paths. The canonical interactive case.",
      "status": "converged",
      "version": "v1.0",
      "generation": 6,
      "updated": "2026-05-24",
      "href": "opbr.html"
    },
    {
      "id": "citation-insertion",
      "kind": "trace pair",
      "title": "Citation insertion before evidence read",
      "lede": "Research agent inserts a citation before reading the source it cites.",
      "status": "running",
      "version": "v0.2",
      "generation": 2,
      "updated": "2026-05-24",
      "href": "../archive/checkpoints/round-05-c1-trace-diff.html"
    },
    {
      "id": "memory-write-before-resolve",
      "kind": "trace pair",
      "title": "Memory write before conflict resolve",
      "lede": "Agent commits a fact to memory while two retrieved sources contradict.",
      "status": "converged",
      "version": "v0.7",
      "generation": 4,
      "updated": "2026-05-22",
      "href": "../archive/checkpoints/round-05-c1-trace-diff.html"
    },
    {
      "id": "refund-commit-after-checks",
      "kind": "contract",
      "title": "refund.commit_after_checks",
      "lede": "Executable behavior contract. YAML schema + Python validator.",
      "status": "converged",
      "version": "v1.0",
      "generation": 3,
      "updated": "2026-05-20",
      "href": "../system/data/contracts/refund-commit-after-checks.json"
    },
    {
      "id": "memory-write-after-resolve",
      "kind": "contract",
      "title": "memory.write_after_resolve",
      "lede": "Contract draft pairing with the memory-write demo.",
      "status": "running",
      "version": "v0.3",
      "generation": 2,
      "updated": "2026-05-23",
      "href": "../rosetta-pages/index.html#contracts"
    },
    {
      "id": "research-cite-after-read",
      "kind": "contract",
      "title": "research.cite_after_read",
      "lede": "Companion contract to the citation-insertion demo.",
      "status": "converged",
      "version": "v0.9",
      "generation": 3,
      "updated": "2026-05-21",
      "href": "../rosetta-pages/index.html#contracts"
    },
    {
      "id": "output-passing-detector",
      "kind": "completion analyzer",
      "title": "Output-passing detector",
      "lede": "Static analysis: flag traces where output evals pass but contracts fail.",
      "status": "converged",
      "version": "v0.4",
      "generation": 3,
      "updated": "2026-05-18",
      "license": "MIT",
      "href": "../rosetta-pages/index.html#workbench"
    },
    {
      "id": "tool-use-sequence-drift",
      "kind": "trace pair",
      "title": "Tool-use sequence drift",
      "lede": "Agent reorders tool calls across runs. Skeleton only — expected to ship 06/14.",
      "status": "pending",
      "version": "v0.1",
      "generation": 1,
      "updated": "2026-05-22",
      "href": "../rosetta-pages/index.html#lab"
    },
    {
      "id": "paper1-study5",
      "kind": "scenario set",
      "title": "Paper 1, Study 5 · n=400",
      "lede": "Held-out scenario set the canonical demo is built from.",
      "status": "converged",
      "version": "locked",
      "generation": null,
      "updated": "2026-04-08",
      "href": "../rosetta-pages/papers/opbr-contracts.html"
    },
    {
      "id": "bracket-renderer",
      "kind": "workbench",
      "title": "Bracket renderer for trace-diff",
      "lede": "Visual primitive used in the OPBR demo. Pulling into the workbench.",
      "status": "pending",
      "version": "v0.1",
      "generation": 1,
      "updated": "2026-05-24",
      "href": "opbr.html"
    },
    {
      "id": "opbr-bench",
      "kind": "benchmark",
      "title": "OPBR-Bench",
      "lede": "50 scenarios. Reproducible. Public leaderboard.",
      "status": "converged",
      "version": "v1.0",
      "generation": null,
      "updated": "2026-05-10",
      "href": "../rosetta-pages/index.html#bench"
    }
  ]
}