Research map / Large language models / Retrieval and knowledge
LiR3AG: A Lightweight Rerank Reasoning Strategy Framework for Retrieval-Augmented Generation
Guo Chen, Junjie Huang, Huaijin Xie, Fei Sun, Tao Jia
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 |
|---|---|---|---|
| add Retriever (context retrieval) to LiR3AG (ablation) | LiR3AG · MuSiQue | F1: 0.293 -> 0.464 | Improved |
| add Reranker (relevance filtering and ordering) (ablation) | LiR3AG · MuSiQue | F1: 0.233 -> 0.464 | Improved |
| add Reasoning Constructor (structured reasoning steps) (ablation) | LiR3AG · MuSiQue | F1: 0.238 -> 0.464 | Improved |
| add Reranker (ablation) | LiR3AG · MultiHop-RAG | F1: 0.619 -> 0.673 | Improved |
| increase top-k from 1 to 10 (ablation) | LiR3AG · MultiHop-RAG | F1: 0.597 -> 0.707 | Improved |
| increase top-k from 1 to 10 (more input tokens) (ablation) | LiR3AG · MuSiQue | F1: 0.356 -> 0.497 | Improved |
| use larger reranker/constructor model (GPT-4o-mini instead of Qwen3-8B) (ablation) | LiR3AG · MuSiQue | F1: 0.344 -> 0.464 | Improved |
| scale reranker/constructor from Qwen3-8B to Qwen3-32B (ablation) | LiR3AG · MultiHop-RAG | F1: 0.634 -> 0.635 | No clear effect |
| enable thinking mode in Qwen3 under Vanilla RAG | Qwen3 Vanilla RAG · MultiHop-RAG | F1: 0.611 -> 0.613 (8B), 0.668 -> 0.654 (32B) | Hurt |
| enable thinking mode in Qwen3 under Vanilla RAG | Qwen3 Vanilla RAG · MuSiQue | F1: 0.239 -> 0.379 (8B) | Improved |
| add retrieved context to direct output | Qwen3-8B no-think · HotpotQA | F1: 0.257 -> 0.403 | Improved |
Closest papers by meaning
- Retrieval is Not Enough: Enhancing RAG through Test-Time Critique and Optimization (NeurIPS 2025 · similarity 0.91)
- From Complex to Atomic: Enhancing Augmented Generation via Knowledge-Aware Dual Rewriting and Reasoning (ICML 2025 · similarity 0.91)
- Simple is Effective: The Roles of Graphs and Large Language Models in Knowledge-Graph-Based Retrieval-Augmented Generation (ICLR 2025 · similarity 0.90)
- Cooperative Retrieval-Augmented Generation for Question Answering: Mutual Information Exchange and Ranking by Contrasting Layers (NeurIPS 2025 · similarity 0.90)
- StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization (ICLR 2025 · similarity 0.90)
- DynamicRAG: Leveraging Outputs of Large Language Model as Feedback for Dynamic Reranking in Retrieval-Augmented Generation (NeurIPS 2025 · similarity 0.90)
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.
