List your company

General Question Answering on TriviaQA OOD

67.5Exact Match AccuracyRECON
16.5429.774356.23Oct 12, 2025

Evaluation results

MethodLinks
RECONBackbone=Qwen2.5-7B-Base, Training Protocol=PPO, RECON Stage=Stage 1 + Stage 22025.1067.5
Search-R1Backbone=Qwen2.5-7B-Base, Training Protocol=PPO2025.1063.8
RECONBackbone=Qwen2.5-3B-Base, Training Protocol=PPO, RECON Stage=Stage 1 + Stage 22025.1061.2
Rejection Sampling2025.1059.2
Search-R1Backbone=Qwen2.5-3B-Base, Training Protocol=PPO2025.1058.7
RAG2025.1058.5
RECONBackbone=Qwen2.5-3B-Base, Training Protocol=PPO, RECON Stage=Stage 2 Distillation-only2025.1057.5
RECONBackbone=Qwen2.5-7B-Base, Training Protocol=PPO, RECON Stage=Stage 2 Distillation-only2025.1055.6
R1-base2025.1053.9
R1-instruct2025.1053.7
IRCoT2025.1047.8
Search-o12025.1044.3
Direct Inference2025.1040.8
SFT2025.1035.4
CoT2025.1018.5