List your company

General Utility Evaluation on MT-Bench (MTB Score)

7.33MTB ScoreDPO mix50/1000/aug
5.10445.68226.266.8378May 1, 2026

Evaluation results

MethodLinks
DPO mix50/1000/augModel=Gemma-3-4B2026.057.33
DPO mix50/1000/augModel=Qwen3-4B2026.057.19
baselineModel=Gemma-3-4B2026.057.1
baselineModel=Qwen3-4B2026.057.04
baselineModel=Qwen2.5-7B2026.057.03
SFT mix0/200/cleanModel=Qwen2.5-7B2026.057.03
DPO mix50/1000/cleanModel=Qwen2.5-7B2026.057
SFT mix50/1000/augModel=Qwen3-4B2026.056.9
baselineModel=Llama-3.1-8B2026.056.82
DPO mix50/1000/augModel=Llama-3.1-8B2026.056.72
SFT mix50/200/augModel=Llama-3.1-8B2026.056.59
SFT mix50/1000/augModel=Gemma-3-4B2026.056.47
baselineModel=Mistral-7B2026.055.94
DPO mix50/1000/augModel=Mistral-7B2026.055.88
SFT mix50/1000/augModel=Mistral-7B2026.055.19