Multi-hop Reasoning on StraQA
64.63AccuracyHSIR-SFT
Evaluation results
| Method | Links | |
|---|---|---|
| HSIR-SFTBackbone=Qwen2.5-1.5B, Training Iterations=12026.05 | 64.63 | |
| ReSTEMBackbone=Qwen2.5-1.5B, Training Iterations=12026.05 | 62.3 | |
| STaRBackbone=Qwen2.5-1.5B, Training Iterations=12026.05 | 61.28 | |
| SFT-OracleBackbone=Qwen2.5-1.5B, Training Iterations=12026.05 | 60.84 | |
| SFT-InitialBackbone=Qwen2.5-1.5B2026.05 | 59.1 |