Research map / Learning theory and algorithms / Prediction and decision losses
Beyond Sequential Reranking: Reranker-Guided Search Improves Reasoning Intensive Retrieval
Haike Xu, Tong Chen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace sequential reranking with reranker-guided graph search (RGS) | Retrieve-and-Rerank, Reranker@100 · BRIGHT | NDCG@10: 25.3 -> 28.8 | Improved |
| replace sequential reranking with RGS | Retrieve-and-Rerank, Reranker@500 · BRIGHT | NDCG@10: 27.7 -> 33.0 | Improved |
| replace sequential reranking with RGS | Retrieve-and-Rerank, Reranker@100 · FollowIR | NDCG@10: 60.4 -> 63.3 | Improved |
| replace sequential reranking with RGS | Retrieve-and-Rerank, Reranker@500 · FollowIR | NDCG@10: 61.6 -> 64.2 | Improved |
| replace sequential reranking with RGS | Retrieve-and-Rerank, Reranker@100 · M-BEIR | NDCG@10: 25.9 -> 31.0 | Improved |
| replace sequential reranking with RGS | Retrieve-and-Rerank, Reranker@500 · M-BEIR | NDCG@10: 30.1 -> 38.3 | Improved |
| use RGS instead of SlideGAR | SlideGAR, Reranker@500 · BRIGHT | NDCG@10: 26.9 -> 33.0 | Improved |
| use RGS instead of sequential reranking on math/code | Retrieve-and-Rerank · BRIGHT AoPS and LeetCode | NDCG@10: 25.8 -> 22.1 | No clear effect |
| increase reranker budget from 100 to 500 | SlideGAR · FollowIR | NDCG@10: 62.5 -> 59.8 | Hurt |
| increase reranker budget from 100 to 500 | RGS · BRIGHT | NDCG@10: 28.8 -> 33.0 | Improved |
| start RGS from noisy start point or default ANNS start instead of exact closest document (ablation) | RGS · BRIGHT | NDCG@10 | No clear effect |
| run RGS on KNN graph instead of DiskANN graph (ablation) | RGS · BRIGHT | NDCG@10 | Hurt |
Closest papers by meaning
- Contrastive Retrieval Heads for Improved Attention-Based Reranking (NeurIPS 2026 · similarity 0.90)
- AcuRank: Uncertainty-Aware Adaptive Computation for Listwise Reranking (NeurIPS 2025 · similarity 0.90)
- DynamicRAG: Leveraging Outputs of Large Language Model as Feedback for Dynamic Reranking in Retrieval-Augmented Generation (NeurIPS 2025 · similarity 0.87)
- Attention in Large Language Models Yields Efficient Zero-Shot Re-Rankers (ICLR 2025 · similarity 0.86)
- Retro*: Optimizing LLMs for Reasoning-Intensive Document Retrieval (ICLR 2026 · similarity 0.85)
- Very Efficient Listwise Multimodal Reranking for Long Documents (ICML 2026 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
