List your company

Question Answering on SciTrek (Proxy Context)

88.5Exact MatchQwen3-4B-Instruct (ProxyCoT-RL)
16.32435.06253.872.538Apr 6, 2026

Evaluation results

MethodLinks
Qwen3-4B-Instruct (ProxyCoT-RL)Training Strategy=ProxyCoT-RL2026.0488.5
Qwen3-235B-ThinkingTraining Strategy=Zero-shot2026.0485.6
Qwen3-235B-InstructTraining Strategy=Zero-shot2026.0472.7
Gemma3-4B-IT (ProxyCoT-RL)Training Strategy=ProxyCoT-RL2026.0469.8
Qwen3-4B-Instruct (ProxyCoT-ZS)Training Strategy=ProxyCoT-ZS2026.0467.8
Qwen3-4B-InstructTraining Strategy=Zero-shot2026.0467.2
Qwen3-4B-InstructTraining Strategy=RLVR on C2026.0466.1
Gemma3-4B-IT (ProxyCoT-ZS)Training Strategy=ProxyCoT-ZS2026.0464.2
Gemma3-4B-ITTraining Strategy=SFT on C, CoT*2026.0453.1
Qwen3-4B-InstructTraining Strategy=SFT on C, CoT*2026.0445.8
Gemma3-4B-ITTraining Strategy=RLVR on C2026.0439.9
Qwen3-4B-InstructTraining Strategy=SFT on C2026.0439
Gemma3-4B-ITTraining Strategy=Zero-shot2026.0434.2
Gemma3-4B-ITTraining Strategy=SFT on C2026.0419.1