Atlas

Benchmarks

← All benchmarks

ClaimBench - Missing evidence

Professional Work · 2026-10-20

ClaimBench claims copied from a different passage of the same filing with little wording in common with the excerpt (gold label insufficient evidence). Diagnostic split.

Top models (higher is better)

ModelScore
Claude Haiku 5.5100.0
Claude Opus 5.5100.0
Claude Sonnet 5.5100.0
GPT-6.1 Sol100.0
GPT-6 Luna100.0
Loading Atlas data…