List your company

Egregious Unfaithfulness on Aggregate of Seven Domains

5.8Egregious Unfaithfulness (%)Gemini 3 Flash
5.4287.93910.4512.961Feb 2, 2026

Evaluation results

MethodLinks
Gemini 3 Flash2026.025.8—
Gemma 3 27BParameters=27B2026.026.2—
Claude Opus 4.52026.026.4—
Qwen 3 32BParameters=32B2026.027.4—
GPT-5.22026.027.7—
Gemini 3 Pro2026.027.8—
GPT-5 Nano2026.028.4—
Claude Haiku 4.52026.028.8—
Gemma 3 1BParameters=1B2026.0212.9—
Qwen 3 0.6BParameters=0.6B2026.0215.1—