Reasoning Quality Correlation Analysis on Synthetic Reasoning Traces Aggregate
0.2937Somers' DMarODE
Evaluation Results
| Method | Links | |
|---|---|---|
| MarODEBackbone=Qwen-3B-CoT2026.03 | 0.2937 | |
| MarODEBackbone=Deepseek-LLaMA-8B2026.03 | 0.2921 | |
| MarODEBackbone=Deepseek-Qwen-14B2026.03 | 0.2882 | |
| MarODEBackbone=GPT-OSS-20B2026.03 | 0.2634 | |
| MarODEBackbone=DeepSeek-Qwen-7B2026.03 | 0.2371 | |
| ROSCOE-SABackbone=Qwen-3B-CoT2026.03 | 0.1187 | |
| ROSCOE-SABackbone=Deepseek-LLaMA-8B2026.03 | 0.1117 | |
| ROSCOE-SABackbone=Deepseek-Qwen-14B2026.03 | 0.1111 | |
| ROSCOE-SABackbone=DeepSeek-Qwen-7B2026.03 | 0.1053 | |
| ROSCOE-SABackbone=GPT-OSS-20B2026.03 | 0.1048 |