Math Reasoning on ThmQA
34.3Multi@5 AccuracyDRIFT-5turn
Evaluation Results
| Method | Links | |
|---|---|---|
| DRIFT-5turnBackbone=Qwen2.5-3B-Instruct, Protocol=Ours, Max turns=52026.05 | 34.3 | |
| UFO-5turnBackbone=Qwen2.5-3B-Instruct, Protocol=Multi-turn baselines (Online), Max turns=52026.05 | 33.5 | |
| SFT-5turnBackbone=Qwen2.5-3B-Instruct, Protocol=Multi-turn baselines (Offline), Max turns=52026.05 | 33.1 | |
| SCoRe-2turnBackbone=Qwen2.5-3B-Instruct, Protocol=Multi-turn baselines (Online), Max turns=22026.05 | 31.8 | |
| UFO-5turnBackbone=Llama3.1-8B-Instruct, Protocol=Multi-turn baselines (Online), Max turns=52026.05 | 30.3 | |
| PPOBackbone=Llama3.1-8B-Instruct, Protocol=Single-turn baselines2026.05 | 30.1 | |
| DRIFT-5turnBackbone=Llama3.1-8B-Instruct, Protocol=Ours, Max turns=52026.05 | 29.9 | |
| SFTBackbone=Qwen2.5-3B-Instruct, Protocol=Single-turn baselines2026.05 | 29.8 | |
| STaR-2turnBackbone=Qwen2.5-3B-Instruct, Protocol=Multi-turn baselines (Offline), Max turns=22026.05 | 29.5 | |
| PPOBackbone=Qwen2.5-3B-Instruct, Protocol=Single-turn baselines2026.05 | 29 | |
| STaR-2turnBackbone=Llama3.1-8B-Instruct, Protocol=Multi-turn baselines (Offline), Max turns=22026.05 | 26.2 | |
| SCoRe-2turnBackbone=Llama3.1-8B-Instruct, Protocol=Multi-turn baselines (Online), Max turns=22026.05 | 26.1 | |
| BaseBackbone=Qwen2.5-3B-Instruct, Protocol=Single-turn baselines2026.05 | 26 | |
| SFTBackbone=Llama3.1-8B-Instruct, Protocol=Single-turn baselines2026.05 | 24.3 | |
| SFT-5turnBackbone=Llama3.1-8B-Instruct, Protocol=Multi-turn baselines (Offline), Max turns=52026.05 | 23.3 | |
| BaseBackbone=Llama3.1-8B-Instruct, Protocol=Single-turn baselines2026.05 | 22 |