List your company
ResearchFormerly WizWand

Preference Modeling on HH-RLHF

61.4AccuracyLLM-Judge
44.55248.92653.357.674May 26, 2026
Updated 5mo ago

Evaluation results

MethodLinks
LLM-JudgeBackbone Model=GPT-OSS-20B, Budget=N=42026.0561.4
UABBackbone Model=Gemma3-27B, Budget=N=42026.0560.8
UABBackbone Model=Cohere, Budget=N=42026.0560.8
UABBackbone Model=GPT-OSS-20B, Budget=N=42026.0560.6
LLM-JudgeBackbone Model=Gemma3-27B, Budget=N=42026.0560.2
UniformBackbone Model=GPT-OSS-20B, Budget=N=42026.0560.1
N=1Backbone Model=Cohere, Budget=N=12026.0560
UniformBackbone Model=Cohere, Budget=N=42026.0559.7
N=1Backbone Model=GPT-OSS-20B, Budget=N=12026.0559.4
UniformBackbone Model=Gemma3-27B, Budget=N=42026.0559.4
N=1Backbone Model=Gemma3-27B, Budget=N=12026.0559.1
LLM-JudgeBackbone Model=Cohere, Budget=N=42026.0558.7
UABModel=Llama3.2-3B, N=42026.0554.3
UABModel=Qwen2.5-7B, N=42026.0553.2
LengthModel=Qwen2.5-7B, N=42026.0553
LLM-JudgeModel=Qwen2.5-7B, N=42026.0553
N=1Model=Qwen2.5-7B, N=12026.0552.8
UniformModel=Llama3.2-3B, N=42026.0552.2
UABModel=Qwen2.5-1.5B, N=42026.0551.7
LengthModel=Llama3.2-3B, N=42026.0551.3
RandomModel=Qwen2.5-7B, N=42026.0551.3
N=1Model=Qwen2.5-1.5B, N=12026.0550.9
RandomModel=Llama3.2-3B, N=42026.0550.5
UniformModel=Qwen2.5-7B, N=42026.0550.5
LengthModel=Qwen2.5-1.5B, N=42026.0550.4
LLM-JudgeModel=Qwen2.5-1.5B, N=42026.0550.2
LLM-JudgeModel=Llama3.2-3B, N=42026.0549.2
UniformModel=Qwen2.5-1.5B, N=42026.0548.2
RandomModel=Qwen2.5-1.5B, N=42026.0546.6
N=1Model=Llama3.2-3B, N=12026.0545.2