Reasoning on Reasoning Hop Generalization Tasks (Test Set)
87.7Parity-NLTCR+gold
Evaluation results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| TCR+goldBackbone=Qwen2.5-Instruct-14B2026.01 | 87.7 | 41.5 | 48.6 | 96 | 69.8 | 63.7 | 77.2 | 69.2 | |
| TCRBackbone=Qwen2.5-Instruct-14B2026.01 | 79 | 37.4 | 45 | 94.3 | 66 | 54.1 | 75.8 | 64.5 | |
| DolaBackbone=Qwen2.5-Instruct-14B2026.01 | 76.1 | 28.5 | 42.5 | 91 | 58.1 | 47.3 | 70.8 | 59.2 | |
| Base ModelBackbone=Qwen2.5-Instruct-14B2026.01 | 74.4 | 32.3 | 43.3 | 89.9 | 60.5 | 47.7 | 72.4 | 60.1 |