Gemini answered 2030-01-01; the verified outcome was 2058-02-04

Verified answer

✓2058-02-04 Stated in the maturity condition.

Gemini answered

✕2030-01-01 Estimated from context.

Decisive issue

Extraction

Evidence for the label

DOC-1#S3

Gemini cited

DOC-1#S1evidence score 0%

• Retrieval failure. None of the cited sections is among those that establish the answer; the relevant evidence was not retrieved.
• Wrong fact. Extracted 2030-01-01; record states 2058-02-04.
• Overconfident. Confidence 0.85 on an incorrect answer.

Classifications are produced by the failure analyzer against the verified label and the evidence map; the model never grades itself. Failure taxonomy →

• Mastermind✓Correctanswered 2058-02-04
• Claude✓Correctanswered 2058-02-04
• GPT✓Correctanswered 2058-02-04its failure page →

Open the full Case Replay with the documents →

• #1 Republic of Norland 2024Retrieval failure
• #21 Kingdom of Vestmark 2026Hallucinated evidence
• #31 Orinoco Retail Holdings plc 2027Retrieval failure
• #51 Brightwater Energy Corp 2024Hallucinated evidence
• #71 Meridian Telecom S.A. 2026Retrieval failure
• #90 Northgate Shipping Ltd 2028Retrieval failure

All Gemini extraction failures →

This page records Gemini’s answer from its latest official run on the frozen benchmark v0.1. If the model is re-run and the answer changes, the page shows the new result and the model page keeps the full run history. Corrections policy →