List your company

Instruction Following Evaluation on HelpSteer2 (val)

71.8QualitySEE
56.92860.78964.6568.511Jun 3, 2026

Evaluation results

MethodLinks
SEEJudge=Gemini 3.1 Flash-Lite2026.0671.870.9——
SEEunique examples=160, batch size=16, rounds=152026.0670.473.167.170
Adapted RLCRJudge=Gemini 3.1 Flash-Lite2026.0669.167——
Qwen 3-4BJudge=Gemini 3.1 Flash-Lite2026.0668.265.1——
Adapted RLCRunique examples=~5,000, batch size=482026.0666.267.55761.7
Qwen3-4Bstatus=base model2026.0664.463.25050
SEEJudge=Claude Sonnet 4.62026.0659.967——
Adapted RLCRJudge=Claude Sonnet 4.62026.0658.162.1——
Qwen 3-4BJudge=Claude Sonnet 4.62026.0657.558.5——