List your company

Question Answering on NQ Noisy (test)

37.71F1 ScoreOpenDecoder
14.320420.392726.46532.5373Jan 13, 2026

Evaluation results

MethodLinks
OpenDecoderBackbone=Qwen-2.5-3B-Instruct2026.0137.7133.82
RbFTBackbone=Qwen-2.5-3B-Instruct2026.0135.530.7
Vanilla SFTBackbone=Qwen-2.5-3B-Instruct2026.0134.9832.83
AstuteRAGBackbone=Qwen-2.5-3B-Instruct2026.0132.3629
InstructRAGBackbone=Qwen-2.5-3B-Instruct2026.0128.0929.33
RobustRAGBackbone=Qwen-2.5-3B-Instruct2026.0125.2130.2
Vanilla RAGBackbone=Qwen-2.5-3B-Instruct2026.0115.2232.7