Mathematical Reasoning on MathQA (Exact Match Accuracy and Generation Tokens)
60.1Exact Match AccuracyTILR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| TILRBackbone=Qwen2.5-Math-1.5B, LoRA rank=16, Internalization curriculum=Stepwise, Reasoning method=TILR2026.06 | 60.1 | 9 | |
| RefinementBackbone=Qwen2.5-Math-1.5B, LoRA rank=16, Internalization curriculum=Stepwise, Reasoning method=Refinement2026.06 | 59.6 | 9 | |
| CoconutBackbone=Qwen2.5-Math-1.5B, LoRA rank=16, Internalization curriculum=Stepwise, Reasoning method=Coconut2026.06 | 57.4 | 9 | |
| AdaAnchorBackbone=Qwen2.5-Math-1.5B, LoRA rank=16, Internalization curriculum=Stepwise, Reasoning method=AdaAnchor2026.06 | 53.1 | 9 | |
| CoTBackbone=Qwen2.5-Math-1.5B, LoRA rank=16, Internalization curriculum=Stepwise, Reasoning method=CoT2026.06 | 52.4 | 68 | |
| No-CoTBackbone=Qwen2.5-Math-1.5B, LoRA rank=16, Internalization curriculum=Stepwise, Reasoning method=No-CoT2026.06 | 42.2 | 5 |