{
 "accepted": 96,
 "api_errors": 0,
 "cost": {
  "input_tokens": 41577,
  "output_tokens": 6233,
  "total_usd": 0.581936,
  "usd_per_accepted": 0.006062
 },
 "domains_at_or_above_80pct": 9,
 "duration_s": 31.2,
 "generated_at": "2026-11-18T18:26:13.509181+00:00",
 "honest_limits": [
  "This is a model baseline on a constructed suite, not a general-intelligence benchmark.",
  "The prompt supplies the same policy card the library codifies; it is a policy-informed baseline.",
  "The model has no tools, no memory, no retries, and no execution sandbox beyond the suite checker.",
  "The library arm is hand-written for these families; comparing it to a raw model is a ceiling comparison, not a capability claim.",
  "No independent verifier and no witness. Not AGI, not VAGI, not production."
 ],
 "invalid_json": 4,
 "latency_ms": {
  "p50": 926.8,
  "p95": 4156.6
 },
 "model": "claude-fable-5-2",
 "model_label": "Claude Fable 5 2",
 "n_scored": 100,
 "n_skipped": 0,
 "n_tasks": 100,
 "per_domain": {
  "business_decision": {
   "accepted": 10,
   "errors": 0,
   "invalid": 0,
   "n": 10,
   "rate": 1.0,
   "wilson95": [
    0.7225,
    1.0
   ]
  },
  "data_analysis": {
   "accepted": 9,
   "errors": 0,
   "invalid": 1,
   "n": 10,
   "rate": 0.9,
   "wilson95": [
    0.5958,
    0.9821
   ]
  },
  "governance": {
   "accepted": 10,
   "errors": 0,
   "invalid": 0,
   "n": 10,
   "rate": 1.0,
   "wilson95": [
    0.7225,
    1.0
   ]
  },
  "memory_retrieval": {
   "accepted": 10,
   "errors": 0,
   "invalid": 0,
   "n": 10,
   "rate": 1.0,
   "wilson95": [
    0.7225,
    1.0
   ]
  },
  "planning": {
   "accepted": 10,
   "errors": 0,
   "invalid": 0,
   "n": 10,
   "rate": 1.0,
   "wilson95": [
    0.7225,
    1.0
   ]
  },
  "recovery": {
   "accepted": 10,
   "errors": 0,
   "invalid": 0,
   "n": 10,
   "rate": 1.0,
   "wilson95": [
    0.7225,
    1.0
   ]
  },
  "science_calculation": {
   "accepted": 10,
   "errors": 0,
   "invalid": 0,
   "n": 10,
   "rate": 1.0,
   "wilson95": [
    0.7225,
    1.0
   ]
  },
  "security_triage": {
   "accepted": 10,
   "errors": 0,
   "invalid": 0,
   "n": 10,
   "rate": 1.0,
   "wilson95": [
    0.7225,
    1.0
   ]
  },
  "software_repair": {
   "accepted": 7,
   "errors": 0,
   "invalid": 3,
   "n": 10,
   "rate": 0.7,
   "wilson95": [
    0.3968,
    0.8922
   ]
  },
  "tool_orchestration": {
   "accepted": 10,
   "errors": 0,
   "invalid": 0,
   "n": 10,
   "rate": 1.0,
   "wilson95": [
    0.7225,
    1.0
   ]
  }
 },
 "protocol": {
  "attempts_per_task": 1,
  "budget_cap_usd": 1.5,
  "invalid_json_is_incorrect": true,
  "max_tokens": 400,
  "memory": "none",
  "n_per_domain": 10,
  "policy_card_supplied": true,
  "prices_usd_per_mtok": {
   "input": 8.0,
   "output": 40.0
  },
  "prompt_sha256": "4fa1e16298107de3552ef1be9a0c65b933c06e7b119cc15b13e85474c2eedfa6",
  "seed": 9402,
  "selection_seed": 9411,
  "task_ids_sha256": "bb28cebc24dde7d992b68f650c65e09e1f571f2008c85b872aa1284fda8b0968",
  "tools": "none"
 },
 "rate": 0.96,
 "schema": "cain42.vagi_suite.model_arm.v1",
 "suite": "VAGI Behavior Suite v1",
 "wilson95": [
  0.9016,
  0.9843
 ]
}