Research map / Large language models / Retrieval and knowledge
Relink: Constructing Query-Driven Evidence Graph On-the-Fly for GraphRAG
Manzong Huang, Chenyang Bu, Yi He, Xingrui Zhuo, Xindong Wu
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 explicit factual KG backbone (Gb) (ablation) | Relink · HotpotQA | EM: 0.486 -> 0.558 | Improved |
| add latent relation pool with dynamic repair (Rc) (ablation) | Relink · HotpotQA | EM: 0.526 -> 0.558 | Improved |
| add latent relation pool with dynamic repair (Rc) (ablation) | Relink · 2WikiMultiHopQA | EM: 0.616 -> 0.628 | Improved |
| use trained query-driven ranker instead of embedding cosine similarity (ablation) | Relink · HotpotQA | EM: 0.450 -> 0.558 | Improved |
| use trained query-driven ranker (ablation) | Relink · 2WikiMultiHopQA | EM: 0.552 -> 0.628 | Improved |
| add contrastive alignment loss (ablation) | Relink · HotpotQA | EM: 0.518 -> 0.558 | Improved |
| add contrastive alignment loss (ablation) | Relink · 2WikiMultiHopQA | F1: 0.695 -> 0.722 | Improved |
| remove 90% of explicit graph edges, with dynamic repair (ablation) | Relink · 2WikiMultiHopQA | F1: 0.669 (F1 after removal) | No clear effect |
| remove 90% of explicit graph edges, without dynamic repair (ablation) | Relink w/o Rc · 2WikiMultiHopQA | F1: drops by 34.7% | Hurt |
| Relink vs HippoRAG | Relink · HotpotQA | EM: 0.498 -> 0.558 | Improved |
| Relink vs HippoRAG | Relink · MuSiQue-Full | EM: 0.190 -> 0.252 | Improved |
| Relink vs RAPTOR | Relink · MuSiQue-Ans | EM: 0.258 -> 0.304 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- HippoRAG — HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models (NeurIPS 2024) · better on 6 of 6 shared results
- RAPTOR — RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval (ICLR 2024) · better on 5 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Reasoning of Large Language Models over Knowledge Graphs with Super-Relations (ICLR 2025 · similarity 0.90)
- Simple is Effective: The Roles of Graphs and Large Language Models in Knowledge-Graph-Based Retrieval-Augmented Generation (ICLR 2025 · similarity 0.90)
- Think-on-Graph 2.0: Deep and Faithful Large Language Model Reasoning with Knowledge-guided Retrieval Augmented Generation (ICLR 2025 · similarity 0.90)
- Harnessing Large Language Models for Knowledge Graph Question Answering via Adaptive Multi-Aspect Retrieval-Augmentation (AAAI 2025 · similarity 0.89)
- Deliberation on Priors: Trustworthy Reasoning of Large Language Models on Knowledge Graphs (NeurIPS 2025 · similarity 0.89)
- Debate on Graph: A Flexible and Reliable Reasoning Framework for Large Language Models (AAAI 2025 · similarity 0.89)
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.
