AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

Rethinking the Reranker: Boundary-Aware Evidence Selection for Robust Retrieval-Augmented Generation

Jiashuo Sun, Pengcheng Jiang, Saizhuo Wang, Jiajun Fan, Heng Wang, Siru Ouyang, Ming Zhong, Yizhu Jiao, Chengsong Huang, Xueqiang Xu, Pengrui Han, Peiran Li and 4 more

ICML 2026 · Large language models › Retrieval and knowledge · approach: RAG robustness and evaluation · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
remove boundary reward from selector (ablation)BAR-RAG · NQ/PopQA/HotpotQA avgexact match: 40.2 -> 44.2Improved
add relevance reward to selector (ablation)BAR-RAG · NQ/PopQA/HotpotQA avgexact match: 43.5 -> 44.2Improved
add citation reward to generator (ablation)BAR-RAG · NQ/PopQA/HotpotQA avgexact match: 42.7 -> 44.2Improved
add Stage 2 generator fine-tuning on selected evidence (ablation)BAR-RAG · NQ/PopQA/HotpotQA avgexact match: 36.8 -> 44.2Improved
add Stage 1 selector training (ablation)BAR-RAG · NQ/PopQA/HotpotQA avgexact match: 42.2 -> 44.2Improved
add Stage 1 selector training (ablation)BAR-RAG · HotpotQAexact match: 41.9 -> 38.8Hurt
add training-time filtering of trivial and unanswerable queries (ablation)BAR-RAG selector · NQ/PopQA/HotpotQA avgexact match: 35.6 -> 44.2Improved
increase iterations from 1 to 2 (ablation)BAR-RAG · 7-benchmark avg, Qwen-2.5-7Bexact match: 37.2 -> 38.8Improved
increase iterations from 2 to 3 (ablation)BAR-RAG · 7-benchmark avg, Qwen-2.5-3Bexact match: 34.3 -> 34.2No clear effect
add a neural reranker to standard RAGRAG · 7-benchmark avg, Qwen-2.5-7Bexact match: 26.9 -> 28.0Improved
use BAR-RAG instead of RAG SFTQwen-2.5-7B-Instruct · 7-benchmark avgexact match: 31.8 -> 39.1Improved
use IRCoT instead of standard RAGQwen-2.5-7B-Instruct · 7-benchmark avgexact match: 26.9 -> 22.6Hurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy