List your company

Reasoning on GLOQA (test)

53.98AccuracyMultiRole-R1
31.765637.532843.349.0672Jul 27, 2025

Evaluation results

MethodLinks
MultiRole-R1Backbone=R1-Distill-Qwen-14B, Method Strategy=SelfConsis SFT+GRPO(RS)2025.0753.9891.32
SelfConsis SFT+GRPOBackbone=R1-Distill-Qwen-14B2025.0751.7390.33
MultiRole-R1Backbone=Qwen3-8B, Method Strategy=SelfConsis SFT+GRPO(RS)2025.0751.7289.88
SelfConsis SFT+GRPOBackbone=Qwen3-8B2025.0751.3787.74
SelfConsis SFT+DPOBackbone=R1-Distill-Qwen-14B2025.0751.3365.03
SelfConsis SFTBackbone=R1-Distill-Qwen-14B2025.0750.9881.04
SelfConsis SFT+DPOBackbone=Qwen3-8B2025.0750.6874.14
SelfConsis SFTBackbone=Qwen3-8B2025.0750.3277.07
Role-PlayBackbone=R1-Distill-Qwen-14B2025.0749.975.73
MultiRole-R1Backbone=R1-Distill-Qwen-7B, Method Strategy=SelfConsis SFT+GRPO(RS)2025.0749.189.67
MultiRole-R1Backbone=R1-Distill-Llama-8B, Method Strategy=SelfConsis SFT+GRPO(RS)2025.0749.0691.78
Self-RefineBackbone=R1-Distill-Qwen-14B2025.0749.0469.4
SelfConsis SFT+DPOBackbone=R1-Distill-Llama-8B2025.0748.8980.97
SelfConsis SFT+GRPOBackbone=R1-Distill-Llama-8B2025.0748.5589.36
SelfConsis SFTBackbone=R1-Distill-Llama-8B2025.0748.1787.26
Role-PlayBackbone=Qwen3-8B2025.0747.5754.89
SelfConsis SFT+GRPOBackbone=R1-Distill-Qwen-7B2025.0747.2287.46
Self-RefineBackbone=Qwen3-8B2025.0745.2585
Role-PlayBackbone=R1-Distill-Llama-8B2025.0744.8783.02
SelfConsis SFT+DPOBackbone=R1-Distill-Qwen-7B2025.0744.261.17
More thinkBackbone=R1-Distill-Llama-8B2025.0744.0487.19
More thinkBackbone=Qwen3-8B2025.0743.3978.98
Self-RefineBackbone=R1-Distill-Llama-8B2025.0743.1981.11
Self-RefineBackbone=R1-Distill-Qwen-7B2025.0743.1359.88
SelfConsis SFTBackbone=R1-Distill-Qwen-7B2025.0743.1385.58
Zero-shot CoTBackbone=Qwen3-8B2025.0742.1360.99
Role-PlayBackbone=R1-Distill-Qwen-7B2025.0741.6777.75
More thinkBackbone=R1-Distill-Qwen-14B2025.0741.684.04
Zero-shot CoTBackbone=R1-Distill-Llama-8B2025.0738.4187.07
Zero-shot CoTBackbone=R1-Distill-Qwen-14B2025.0736.8279.18
More thinkBackbone=R1-Distill-Qwen-7B2025.0736.4286.9
Zero-shot CoTBackbone=R1-Distill-Qwen-7B2025.0732.6265.88