List your company

Reasoning on CALI OOD (test)

77.95AccuracyMultiRole-R1
49.19456.659564.12571.5905Jul 27, 2025

Evaluation results

MethodLinks
MultiRole-R1Backbone=Qwen3-8B, Method Strategy=SelfConsis SFT+GRPO(RS)2025.0777.9583.84
SelfConsis SFT+GRPOBackbone=Qwen3-8B2025.0777.8380.36
MultiRole-R1Backbone=R1-Distill-Qwen-14B, Method Strategy=SelfConsis SFT+GRPO(RS)2025.0776.589.08
SelfConsis SFT+DPOBackbone=Qwen3-8B2025.0776.169.58
SelfConsis SFTBackbone=R1-Distill-Qwen-14B2025.0776.0873.65
SelfConsis SFTBackbone=Qwen3-8B2025.0775.9672.78
More thinkBackbone=R1-Distill-Qwen-14B2025.0775.976.81
SelfConsis SFT+DPOBackbone=R1-Distill-Qwen-14B2025.0775.8271.71
SelfConsis SFT+GRPOBackbone=R1-Distill-Qwen-14B2025.0775.6584.92
More thinkBackbone=Qwen3-8B2025.0775.168.44
Zero-shot CoTBackbone=R1-Distill-Qwen-14B2025.0775.0571.83
Zero-shot CoTBackbone=Qwen3-8B2025.0773.457.43
MultiRole-R1Backbone=R1-Distill-Llama-8B, Method Strategy=SelfConsis SFT+GRPO(RS)2025.0771.4890.55
Self-RefineBackbone=R1-Distill-Qwen-14B2025.0771.2878.22
MultiRole-R1Backbone=R1-Distill-Qwen-7B, Method Strategy=SelfConsis SFT+GRPO(RS)2025.0770.8583.31
SelfConsis SFT+GRPOBackbone=R1-Distill-Qwen-7B2025.0770.8382.15
Role-PlayBackbone=Qwen3-8B2025.0770.2650.49
SelfConsis SFTBackbone=R1-Distill-Llama-8B2025.0770.0579.77
SelfConsis SFT+DPOBackbone=R1-Distill-Llama-8B2025.0769.8176.62
Self-RefineBackbone=Qwen3-8B2025.0769.448.16
SelfConsis SFT+GRPOBackbone=R1-Distill-Llama-8B2025.0769.2683.37
SelfConsis SFT+DPOBackbone=R1-Distill-Qwen-7B2025.0768.1964.09
Role-PlayBackbone=R1-Distill-Qwen-14B2025.0767.4170.74
SelfConsis SFTBackbone=R1-Distill-Qwen-7B2025.0767.3578.94
More thinkBackbone=R1-Distill-Llama-8B2025.0764.4180.3
Role-PlayBackbone=R1-Distill-Llama-8B2025.0762.777.32
Self-RefineBackbone=R1-Distill-Llama-8B2025.0761.9578.87
Zero-shot CoTBackbone=R1-Distill-Llama-8B2025.0760.8473.98
More thinkBackbone=R1-Distill-Qwen-7B2025.0760.4578.82
Role-PlayBackbone=R1-Distill-Qwen-7B2025.0752.6967.43
Self-RefineBackbone=R1-Distill-Qwen-7B2025.0750.7666.09
Zero-shot CoTBackbone=R1-Distill-Qwen-7B2025.0750.352.22