google/gemini-3.1-pro-preview · Model identity and integrity report

Family

google

V3 sub-model reference

Available ✓

Checks

知識截止 · 能力探針 · 拒絕模板 · 拒絕溫度參數

Family-fingerprint coverage

17 weighted rules · 5 feature categories

Concrete reference answers, refusal templates, and individual signal weights are not published because disclosure would help spoofing endpoints evade detection.

124 runs · 74 hosts · 75 anomalies

Family-fingerprint consistency

  google          98  ████████████████████░░░░░   79% (expected)
  unknown         12  ███░░░░░░░░░░░░░░░░░░░░░░   10% (suspicious)
  anthropic        8  ██░░░░░░░░░░░░░░░░░░░░░░░    6% (suspicious)
  openai           6  █░░░░░░░░░░░░░░░░░░░░░░░░    5% (suspicious)

Observed V3 sub-model mapping

  → google/gemini-3.1-pro-preview           55 (44%) ✓ correct sim=0.87
  → abstain                                 40 (32%) — low confidence
  → unknown                                 20 (16%) ⚠ mismatch
  → google/gemini-3.1-flash-lite-preview     6 ( 5%) ⚠ mismatch sim=0.91
  → openai/gpt-5.4                           2 ( 2%) ⚠ mismatch sim=1.00
  → anthropic/claude-sonnet-4.6              1 ( 1%) ⚠ mismatch sim=0.70

Capability score distribution (auxiliary)

  0-49     ██████████████░░░░░░░░░░░   71
  50-69    ██░░░░░░░░░░░░░░░░░░░░░░░    9
  70-89    ████████░░░░░░░░░░░░░░░░░   40
  90-100   █░░░░░░░░░░░░░░░░░░░░░░░░    4

BazaarLink Probe combines a Bayesian V2 family classifier across nine fingerprint categories with V3 sub-model references to verify endpoints claiming google/gemini-3.1-pro-preview. Low family consistency, a different V3 match, or divergence from selfClaim is flagged as anomalous. BazaarLink Probe home.