{
 "accepted": 81,
 "cost": {
  "input_tokens": 79530,
  "output_tokens": 10121,
  "total_usd": 0.26027
 },
 "duration_s": 31.3,
 "generated_at": "2026-11-18T20:29:30.320074+00:00",
 "honest_limits": [
  "The shift probe is deliberately adversarial to the hand-written library; a low model score is a gap map, not a general-intelligence result.",
  "The policy card supplies the intended rules; the software task's hidden target (x % 3) is not inferable from the public tests, so failure there is expected.",
  "Constructed tasks; no tools, no memory, one attempt; no external reviewer.",
  "Not AGI, not VAGI, not production."
 ],
 "invalid_json": 2,
 "latency_ms": {
  "p50": 1550.4,
  "p95": 4167.6
 },
 "library_reference": 5,
 "model": "claude-sonnet-5-5",
 "n_scored": 100,
 "n_skipped": 0,
 "n_tasks": 100,
 "per_domain": {
  "business_decision": {
   "accepted": 10,
   "invalid": 0,
   "n": 10,
   "rate": 1.0
  },
  "data_analysis": {
   "accepted": 10,
   "invalid": 0,
   "n": 10,
   "rate": 1.0
  },
  "governance": {
   "accepted": 10,
   "invalid": 0,
   "n": 10,
   "rate": 1.0
  },
  "memory_retrieval": {
   "accepted": 10,
   "invalid": 0,
   "n": 10,
   "rate": 1.0
  },
  "planning": {
   "accepted": 6,
   "invalid": 2,
   "n": 10,
   "rate": 0.6
  },
  "recovery": {
   "accepted": 10,
   "invalid": 0,
   "n": 10,
   "rate": 1.0
  },
  "science_calculation": {
   "accepted": 10,
   "invalid": 0,
   "n": 10,
   "rate": 1.0
  },
  "security_triage": {
   "accepted": 10,
   "invalid": 0,
   "n": 10,
   "rate": 1.0
  },
  "software_repair": {
   "accepted": 5,
   "invalid": 0,
   "n": 10,
   "rate": 0.5
  },
  "tool_orchestration": {
   "accepted": 0,
   "invalid": 0,
   "n": 10,
   "rate": 0.0
  }
 },
 "rate": 0.81,
 "schema": "cain42.vagi_suite.shift_model_arm.v1",
 "suite": "VAGI Behavior Suite v1 \u2014 distribution shift",
 "suite_seed": 9502,
 "wilson95": [
  0.7222,
  0.8749
 ]
}