FEVER
An early and widely used fact-checking benchmark dataset for claim verification based on Wikipedia text. It is the predecessor of FEVEROUS and focuses primarily on unstructured textual data.
Core Facts
ReAct significantly outperforms methods that use CoT or action alone on complex reasoning benchmarks such as HotpotQA and FEVER.
Through experiments on multi-hop reasoning benchmarks such as HotpotQA and FEVER, the ReAct paper demonstrated that compared to pure Chain-of-Thought reasoning, ReAct improved accuracy by approximately 10%–30% on tasks requiring external information retrieval.
Timeline (last 90 days)
The research team generated 470,400 responses on three datasets—FEVER, FEVEROUS, and SciFact—using two 8B-scale large language models (Qwen and Llama).
On the FEVER benchmark, an 8B-scale backbone model paired with the R2VC architecture achieved an accuracy improvement of 13.74% over the baseline.
After removing the candidate filtering component, FEVER accuracy dropped to 76.24%.
All Facts (6)
ReAct significantly outperforms methods that use CoT or action alone on complex reasoning benchmarks such as HotpotQA and FEVER.
80%VerifiedThrough experiments on multi-hop reasoning benchmarks such as HotpotQA and FEVER, the ReAct paper demonstrated that compared to pure Chain-of-Thought reasoning, ReAct improved accuracy by approximately 10%–30% on tasks requiring external information retrieval.
75%UnverifiedReAct论文在HotpotQA和FEVER等基准测试上证明ReAct模式相比纯推理或纯行动方法都有显著提升
85%UnverifiedThe research team generated 470,400 responses on three datasets—FEVER, FEVEROUS, and SciFact—using two 8B-scale large language models (Qwen and Llama).
50%UnverifiedOn the FEVER benchmark, an 8B-scale backbone model paired with the R2VC architecture achieved an accuracy improvement of 13.74% over the baseline.
50%UnverifiedAfter removing the candidate filtering component, FEVER accuracy dropped to 76.24%.
50%