List your company

General Reasoning on GPQA (multi@5 accuracy)

72.7multi@5 AccuracyDRIFT-5turn
43.78851.29458.866.306May 29, 2026

Evaluation results

MethodLinks
DRIFT-5turnBackbone=Qwen2.5-3B-Instruct, Protocol=Ours, Max turns=52026.0572.7
UFO-5turnBackbone=Qwen2.5-3B-Instruct, Protocol=Multi-turn baselines (Online), Max turns=52026.0571.2
SCoRe-2turnBackbone=Qwen2.5-3B-Instruct, Protocol=Multi-turn baselines (Online), Max turns=22026.0563.2
UFO-5turnBackbone=Llama3.1-8B-Instruct, Protocol=Multi-turn baselines (Online), Max turns=52026.0561.6
DRIFT-5turnBackbone=Llama3.1-8B-Instruct, Protocol=Ours, Max turns=52026.0560.1
STaR-2turnBackbone=Qwen2.5-3B-Instruct, Protocol=Multi-turn baselines (Offline), Max turns=22026.0558.6
SFT-5turnBackbone=Qwen2.5-3B-Instruct, Protocol=Multi-turn baselines (Offline), Max turns=52026.0558
STaR-2turnBackbone=Llama3.1-8B-Instruct, Protocol=Multi-turn baselines (Offline), Max turns=22026.0555.7
SCoRe-2turnBackbone=Llama3.1-8B-Instruct, Protocol=Multi-turn baselines (Online), Max turns=22026.0554.8
SFTBackbone=Qwen2.5-3B-Instruct, Protocol=Single-turn baselines2026.0551
PPOBackbone=Llama3.1-8B-Instruct, Protocol=Single-turn baselines2026.0550.2
SFT-5turnBackbone=Llama3.1-8B-Instruct, Protocol=Multi-turn baselines (Offline), Max turns=52026.0548.5
BaseBackbone=Qwen2.5-3B-Instruct, Protocol=Single-turn baselines2026.0547.9
SFTBackbone=Llama3.1-8B-Instruct, Protocol=Single-turn baselines2026.0546.4
PPOBackbone=Qwen2.5-3B-Instruct, Protocol=Single-turn baselines2026.0545.1
BaseBackbone=Llama3.1-8B-Instruct, Protocol=Single-turn baselines2026.0544.9