List your company

Knowledge-intensive Question Answering on CRAG

42.3TruthfulnessGPT-5
-55.252-29.926-4.620.726Sep 30, 2025

Evaluation results

MethodLinks
GPT-52025.0942.324.7—
TruthRLBackbone=Llama3.3-70B-Instruct2025.0940.617.3—
TruthRLRetrieval Setting=With Retrieval, Backbone=Llama3.1-8B-Inst2025.0937.219.456.6
OpenAI o32025.0933.233.1—
TruthRLRetrieval Setting=With Retrieval, Backbone=Qwen2.5-7B-Inst2025.0933.117.350.4
PromptingBackbone=Llama3.3-70B-Instruct2025.0932.126.4—
RLKFRetrieval Setting=With Retrieval, Backbone=Qwen2.5-7B-Inst2025.0925.221.947.1
RFTRetrieval Setting=With Retrieval, Backbone=Qwen2.5-7B-Inst2025.0922.631.454
TruthRLRetrieval Setting=Without Retrieval, Backbone=Llama3.1-8B-Inst2025.0922.416.338.7
TruthRL-BinaryRetrieval Setting=With Retrieval, Backbone=Llama3.1-8B-Inst2025.0920.839.560.3
RLKFRetrieval Setting=With Retrieval, Backbone=Llama3.1-8B-Inst2025.0918.930.649.5
RLHFRetrieval Setting=With Retrieval, Backbone=Qwen2.5-7B-Inst2025.0916.535.952.4
TruthRLRetrieval Setting=Without Retrieval, Backbone=Qwen2.5-7B-Inst2025.0916.28.724.9
R-TuningRetrieval Setting=With Retrieval, Backbone=Llama3.1-8B-Inst2025.0915.233.148.4
R-TuningRetrieval Setting=With Retrieval, Backbone=Qwen2.5-7B-Inst2025.0913.43548.4
RLHFRetrieval Setting=With Retrieval, Backbone=Llama3.1-8B-Inst2025.0913.139.752.7
PromptingRetrieval Setting=With Retrieval, Backbone=Qwen2.5-7B-Inst2025.0910.638.449
TruthRL-BinaryRetrieval Setting=With Retrieval, Backbone=Qwen2.5-7B-Inst2025.098.445.353.7
PromptingRetrieval Setting=With Retrieval, Backbone=Llama3.1-8B-Inst2025.095.343.548.8
RLHFRetrieval Setting=Without Retrieval, Backbone=Llama3.1-8B-Inst2025.091.435.336.7
SFTRetrieval Setting=With Retrieval, Backbone=Llama3.1-8B-Inst2025.091.449.350.7
RLKFRetrieval Setting=Without Retrieval, Backbone=Llama3.1-8B-Inst2025.09-0.632.832.2
SFTRetrieval Setting=With Retrieval, Backbone=Qwen2.5-7B-Inst2025.09-2.351.248.8
RFTRetrieval Setting=With Retrieval, Backbone=Llama3.1-8B-Inst2025.09-3.748.845.1
PromptingRetrieval Setting=Without Retrieval, Backbone=Llama3.1-8B-Inst2025.09-4.444.540.1
RLKFRetrieval Setting=Without Retrieval, Backbone=Qwen2.5-7B-Inst2025.09-6.128.122.1
R-TuningRetrieval Setting=Without Retrieval, Backbone=Qwen2.5-7B-Inst2025.09-7.521.914.5
RFTRetrieval Setting=Without Retrieval, Backbone=Llama3.1-8B-Inst2025.09-7.648.140.4
R-TuningRetrieval Setting=Without Retrieval, Backbone=Llama3.1-8B-Inst2025.09-13.739.525.8
TruthRL-BinaryRetrieval Setting=Without Retrieval, Backbone=Llama3.1-8B-Inst2025.09-14.557.242.8
RFTRetrieval Setting=Without Retrieval, Backbone=Qwen2.5-7B-Inst2025.09-16.846.729.9
PromptingRetrieval Setting=Without Retrieval, Backbone=Qwen2.5-7B-Inst2025.09-17.448.130.6
RLHFRetrieval Setting=Without Retrieval, Backbone=Qwen2.5-7B-Inst2025.09-17.445.728.3
TruthRL-BinaryRetrieval Setting=Without Retrieval, Backbone=Qwen2.5-7B-Inst2025.09-29.264.535.3
SFTRetrieval Setting=Without Retrieval, Backbone=Llama3.1-8B-Inst2025.09-42.171.128.9
SFTRetrieval Setting=Without Retrieval, Backbone=Qwen2.5-7B-Inst2025.09-51.575.724.3