List your company

Open-domain task on Arena-Hard

5.17Error (%)C-PAC
5.01926.03717.0558.0729Jan 30, 2026

Evaluation results

MethodLinks
C-PACUncertainty=Verbalized, Clustering type=Joint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.015.17010.3
C-PACUncertainty=Verbalized, Clustering type=Disjoint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.015.27010.91
C-PACUncertainty=Logits, Clustering type=Disjoint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.54047.01
PACUncertainty=Verbalized, Clustering type=Disjoint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.633.6830.07
PACUncertainty=Verbalized, Clustering type=Joint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.674.0430.15
C-PACUncertainty=Logits, Clustering type=Joint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.69049.32
C-PACUncertainty=Router, Clustering type=Joint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.77042.02
PACUncertainty=Logits, Clustering type=Disjoint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.84051.77
C-PACUncertainty=Router, Clustering type=Disjoint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.84042.34
PACUncertainty=Router, Clustering type=Disjoint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.94.4433.07
PACUncertainty=Logits, Clustering type=Joint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.94052.25
PACUncertainty=Router, Clustering type=Joint clustering, Backbone=Qwen LLMs, Epsilon (ε)=0.12026.018.944.6933.39