List your company

Open-ended instruction following on Arena-Hard-Auto v2.0

51.8Win Rate (Response)SEE
49.92850.41450.951.386Jun 3, 2026

Evaluation results

MethodLinks
SEE2026.0651.857.560.9
Adapted RLCR2026.0650.256.954.7
Qwen3-4B2026.065056.851.7