Document Reranking on FEVER
100NDCG@5MonoT5-3B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MonoT5-3BThreshold (τ)=0.52026.06 | 100 | — | 99.9 | 91.5 | |
| Qwen3-Reranker-8BThreshold (τ)=0.52026.06 | 100 | — | 99.8 | 98.2 | |
| MonoT5-baseModel Parameters=0.2B, Threshold (τ)=0.52026.06 | 99.9 | — | 99.8 | 91.3 | |
| LLM2Vec-GenBase Model=Qwen3-8B, Threshold (τ)=02026.06 | 99.7 | — | 99.8 | 13.1 | |
| RSRankThreshold (τ)=02026.06 | 99.2 | — | 99.7 | 83.4 | |
| RankT5Retriever=Contriever, Category=Supervised Neural Reranker2026.05 | 81.556 | — | — | — | |
| RankT5 + CARRetriever=Contriever, Category=Supervised Neural Reranker, Confidence-Aware Reranking=Enabled2026.05 | 81.556 | 0 | — | — | |
| Cross-EncoderRetriever=Contriever, Category=Supervised Neural Reranker2026.05 | 78.457 | — | — | — | |
| Cross-Encoder + CARRetriever=Contriever, Category=Supervised Neural Reranker, Confidence-Aware Reranking=Enabled2026.05 | 78.457 | 0 | — | — | |
| ColBERTRetriever=Contriever, Category=Supervised Neural Reranker2026.05 | 76.169 | — | — | — | |
| ColBERT + CARRetriever=Contriever, Category=Supervised Neural Reranker, Confidence-Aware Reranking=Enabled2026.05 | 76.169 | 0 | — | — | |
| RankGPTLLM=Qwen2.5-7B-Instruct, Retriever=Contriever, Category=LLM-based Reranker2026.05 | 74.268 | — | — | — | |
| RankGPT + CARLLM=Qwen2.5-7B-Instruct, Retriever=Contriever, Category=LLM-based Reranker, Confidence-Aware Reranking=Enabled2026.05 | 74.268 | 0 | — | — | |
| ContrieverRetriever=Contriever, Category=Retriever Only2026.05 | 74.253 | — | — | — | |
| Contriever + CARRetriever=Contriever, Category=Retriever Only, Confidence-Aware Reranking=Enabled2026.05 | 74.253 | 0 | — | — | |
| QLM + CARLLM=Qwen2.5-7B-Instruct, Retriever=Contriever, Category=LLM-based Reranker, Confidence-Aware Reranking=Enabled2026.05 | 57.82 | 3.8 | — | — | |
| QLMLLM=Qwen2.5-7B-Instruct, Retriever=Contriever, Category=LLM-based Reranker2026.05 | 55.687 | — | — | — | |
| YesNo + CARLLM=Qwen2.5-7B-Instruct, Retriever=Contriever, Category=LLM-based Reranker, Confidence-Aware Reranking=Enabled2026.05 | 41.326 | 62.3 | — | — | |
| YesNoLLM=Qwen2.5-7B-Instruct, Retriever=Contriever, Category=LLM-based Reranker2026.05 | 25.462 | — | — | — |