{
 "accepted": 481,
 "api_errors": 0,
 "cost": {
  "input_tokens": 210702,
  "output_tokens": 52872,
  "total_usd": 0.047506,
  "usd_per_accepted": 9.9e-05
 },
 "domains_at_or_above_80pct": 9,
 "duration_s": 76.6,
 "generated_at": "2026-11-18T18:28:10.831849+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": 11,
 "latency_ms": {
  "p50": 839.0,
  "p95": 1817.4
 },
 "model": "claude-haiku-5-5",
 "model_label": "Claude Haiku 5 5",
 "n_scored": 500,
 "n_skipped": 0,
 "n_tasks": 500,
 "per_domain": {
  "business_decision": {
   "accepted": 35,
   "errors": 0,
   "invalid": 11,
   "n": 50,
   "rate": 0.7,
   "wilson95": [
    0.5625,
    0.809
   ]
  },
  "data_analysis": {
   "accepted": 48,
   "errors": 0,
   "invalid": 0,
   "n": 50,
   "rate": 0.96,
   "wilson95": [
    0.8654,
    0.989
   ]
  },
  "governance": {
   "accepted": 50,
   "errors": 0,
   "invalid": 0,
   "n": 50,
   "rate": 1.0,
   "wilson95": [
    0.9287,
    1.0
   ]
  },
  "memory_retrieval": {
   "accepted": 50,
   "errors": 0,
   "invalid": 0,
   "n": 50,
   "rate": 1.0,
   "wilson95": [
    0.9287,
    1.0
   ]
  },
  "planning": {
   "accepted": 50,
   "errors": 0,
   "invalid": 0,
   "n": 50,
   "rate": 1.0,
   "wilson95": [
    0.9287,
    1.0
   ]
  },
  "recovery": {
   "accepted": 50,
   "errors": 0,
   "invalid": 0,
   "n": 50,
   "rate": 1.0,
   "wilson95": [
    0.9287,
    1.0
   ]
  },
  "science_calculation": {
   "accepted": 48,
   "errors": 0,
   "invalid": 0,
   "n": 50,
   "rate": 0.96,
   "wilson95": [
    0.8654,
    0.989
   ]
  },
  "security_triage": {
   "accepted": 50,
   "errors": 0,
   "invalid": 0,
   "n": 50,
   "rate": 1.0,
   "wilson95": [
    0.9287,
    1.0
   ]
  },
  "software_repair": {
   "accepted": 50,
   "errors": 0,
   "invalid": 0,
   "n": 50,
   "rate": 1.0,
   "wilson95": [
    0.9287,
    1.0
   ]
  },
  "tool_orchestration": {
   "accepted": 50,
   "errors": 0,
   "invalid": 0,
   "n": 50,
   "rate": 1.0,
   "wilson95": [
    0.9287,
    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": 50,
  "policy_card_supplied": true,
  "prices_usd_per_mtok": {
   "input": 0.1,
   "output": 0.5
  },
  "prompt_sha256": "2dd37b92819b07348f5e12ef9d533a1bc91c69ffed8ac774e984fb4200e92531",
  "seed": 9402,
  "selection_seed": 9411,
  "task_ids_sha256": "7f11b7ef67523ba3d6eb6b9d6772dee1455d29648b271a861b3d1f8231486170",
  "tools": "none"
 },
 "rate": 0.962,
 "schema": "cain42.vagi_suite.model_arm.v1",
 "suite": "VAGI Behavior Suite v1",
 "wilson95": [
  0.9414,
  0.9755
 ]
}