List your company

Reasoning Quality Correlation Analysis on Synthetic Reasoning Traces Aggregate

0.2937Somers' DMarODE
0.0972440.1482470.199250.250253Mar 2, 2026

Evaluation results

MethodLinks
MarODEBackbone=Qwen-3B-CoT2026.030.2937
MarODEBackbone=Deepseek-LLaMA-8B2026.030.2921
MarODEBackbone=Deepseek-Qwen-14B2026.030.2882
MarODEBackbone=GPT-OSS-20B2026.030.2634
MarODEBackbone=DeepSeek-Qwen-7B2026.030.2371
ROSCOE-SABackbone=Qwen-3B-CoT2026.030.1187
ROSCOE-SABackbone=Deepseek-LLaMA-8B2026.030.1117
ROSCOE-SABackbone=Deepseek-Qwen-14B2026.030.1111
ROSCOE-SABackbone=DeepSeek-Qwen-7B2026.030.1053
ROSCOE-SABackbone=GPT-OSS-20B2026.030.1048