List your company

Instruction Following on MT-Bench (hold-out)

9.6ScoreTRINITY
8.38328.69919.0159.3309Dec 4, 2025

Evaluation results

MethodLinks
TRINITYZero-shot=true2025.129.6
Gemini Pro 2.5Zero-shot=true2025.129.37
GPT-5Zero-shot=true2025.129.35
Claude-4-SonnetZero-shot=true2025.129.28
Qwen3-32BZero-shot=true, Evaluation Mode=direct2025.129.03
Qwen3-32BZero-shot=true, Evaluation Mode=reasoning2025.128.99
Gemma-3-27B-ITZero-shot=true2025.128.76
DeepSeek-R1-Qwen-32BZero-shot=true2025.128.43