List your company

Multi-turn Dialogue on MT-Bench (Acceptance Length τ)

2.717Acceptance Length (τ)Llama 3.2-3B
1.800762.038632.27652.51437Jun 10, 2026

Evaluation results

MethodLinks
Llama 3.2-3Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0.3, αc (chain-loss coefficient)=402026.062.717
Llama 3.2-3Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0.3, αc (chain-loss coefficient)=02026.062.542
Llama 3.2-3Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0, αc (chain-loss coefficient)=02026.062.454
Llama 3.2-3Bγ (loss-decay constant)=None, αf (first-error focal coefficient)=0, αc (chain-loss coefficient)=02026.062.351
Llama-3-8Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0.3, αc (chain-loss coefficient)=402026.062.341
Llama-3-8Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0.3, αc (chain-loss coefficient)=02026.062.222
Qwen3-4Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0.3, αc (chain-loss coefficient)=402026.062.123
Llama-3-8Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0, αc (chain-loss coefficient)=02026.062.105
Qwen3-8Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0.3, αc (chain-loss coefficient)=402026.062.055
Qwen3-4Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0.3, αc (chain-loss coefficient)=02026.062
Qwen3-8Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0.3, αc (chain-loss coefficient)=02026.061.961
Qwen3-4Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0, αc (chain-loss coefficient)=02026.061.959
Llama-3-8Bγ (loss-decay constant)=None, αf (first-error focal coefficient)=0, αc (chain-loss coefficient)=02026.061.931
Qwen3-8Bγ (loss-decay constant)=10, αf (first-error focal coefficient)=0, αc (chain-loss coefficient)=02026.061.901
Qwen3-4Bγ (loss-decay constant)=None, αf (first-error focal coefficient)=0, αc (chain-loss coefficient)=02026.061.875
Qwen3-8Bγ (loss-decay constant)=None, αf (first-error focal coefficient)=0, αc (chain-loss coefficient)=02026.061.836