List your company

Question Answering on NQ-Open Out-of-distribution (test)

49.93AccuracyBase
26.498832.581938.66544.7481May 15, 2026

Evaluation results

MethodLinks
BaseBackbone=Llama-3.1-8B-Instruct2026.0549.9310.0270.81
RLVRBackbone=Llama-3.1-8B-Instruct2026.0549.588.0369.14
CSRBackbone=Llama-3.1-8B-Instruct2026.0546.056.9276.67
RLVRBackbone=Mistral-7B-Instruct2026.0544.874.0172.28
BaseBackbone=Mistral-7B-Instruct2026.0544.566.3873.1
RLCRBackbone=Mistral-7B-Instruct2026.054414.4364.42
RLCRBackbone=Llama-3.1-8B-Instruct2026.0543.716.5561.27
CSRBackbone=Mistral-7B-Instruct2026.0543.353.5479.11
RDBackbone=Llama-3.1-8B-Instruct2026.0542.412.1873.65
RLCRBackbone=Qwen2.5-7B-Instruct2026.0537.521.7159.02
RLVRBackbone=Qwen2.5-7B-Instruct2026.0531.189.7277.91
BaseBackbone=Qwen2.5-7B-Instruct2026.0531.059.7678.49
CSRBackbone=Qwen2.5-7B-Instruct2026.0529.985.784.73
RDBackbone=Mistral-7B-Instruct2026.0529.723.2372.92
RDBackbone=Qwen2.5-7B-Instruct2026.0527.49.0766.94