List your company

General Question Answering on PopQA OOD

45.7Exact Match AccuracySearch-R1
3.78814.66925.5536.431Oct 12, 2025

Evaluation results

MethodLinks
Search-R1Backbone=Qwen2.5-7B-Base, Training Protocol=PPO2025.1045.7
RECONBackbone=Qwen2.5-7B-Base, Training Protocol=PPO, RECON Stage=Stage 1 + Stage 22025.1045.4
Search-R1Backbone=Qwen2.5-3B-Base, Training Protocol=PPO2025.1043.5
RECONBackbone=Qwen2.5-3B-Base, Training Protocol=PPO, RECON Stage=Stage 1 + Stage 22025.1042.5
RECONBackbone=Qwen2.5-7B-Base, Training Protocol=PPO, RECON Stage=Stage 2 Distillation-only2025.1042.1
RECONBackbone=Qwen2.5-3B-Base, Training Protocol=PPO, RECON Stage=Stage 2 Distillation-only2025.1040.9
RAG2025.1039.2
Rejection Sampling2025.1038
IRCoT2025.1030.1
R1-base2025.1020.2
R1-instruct2025.1019.9
Direct Inference2025.1014
Search-o12025.1013.1
SFT2025.1012.1
CoT2025.105.4