Research map / Large language models / Retrieval and knowledge
PathMind: A Retrieve-Prioritize-Reason Framework for Knowledge Graph Reasoning with Large Language Models
Yu Liu, Xixun Lin, Yanmin Shang, Yangxi Li, Shi Wang, Yanan Cao
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 path prioritization module (ablation) | PathMind · CWQ | Hits@1: 0.643 -> 0.707 | Improved |
| add path prioritization module (ablation) | PathMind · WebQSP | Hits@1: 0.840 -> 0.895 | Improved |
| add path-wise preference alignment (DPO) (ablation) | PathMind · CWQ | Hits@1: 0.672 -> 0.707 | Improved |
| add path-wise preference alignment (DPO) (ablation) | PathMind · WebQSP | Hits@1: 0.871 -> 0.895 | Improved |
| add two-phase training (SFT + DPO) (ablation) | PathMind · CWQ | F1: 0.274 -> 0.614 | Improved |
| use important (prioritized) paths instead of shortest paths (ablation) | PathMind · CWQ | Hits@1: 0.662 -> 0.707 | Improved |
| use important paths instead of random paths (ablation) | PathMind · WebQSP | Hits@1: 0.356 -> 0.895 | Improved |
| increase top-K node selection beyond 3 (ablation) | PathMind · WebQSP and CWQ | F1 | Hurt |
| switch backbone from Llama3.1-8B to Llama2-7B (ablation) | PathMind · WebQSP | Hits@1: 0.895 -> 0.864 | Hurt |
| switch backbone from Llama3.1-8B to Qwen2-7B (ablation) | PathMind · CWQ | Hits@1: 0.707 -> 0.665 | Hurt |
| single-pass reasoning over prioritized paths vs iterative synergy methods | PathMind · WebQSP | # input tokens: 7,069 -> 216 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- EPERM — EPERM: An Evidence Path Enhanced Reasoning Model for Knowledge Graph Question and Answering (AAAI 2025) · better on 4 of 4 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 on Graphs: Faithful and Interpretable Large Language Model Reasoning (ICLR 2024 · similarity 0.93)
- Graph-constrained Reasoning: Faithful Reasoning on Knowledge Graphs with Large Language Models (ICML 2025 · similarity 0.92)
- Plan-on-Graph: Self-Correcting Adaptive Planning of Large Language Model on Knowledge Graphs (NeurIPS 2024 · similarity 0.91)
- Debate on Graph: A Flexible and Reliable Reasoning Framework for Large Language Models (AAAI 2025 · similarity 0.91)
- Think-on-Graph: Deep and Responsible Reasoning of Large Language Model on Knowledge Graph (ICLR 2024 · similarity 0.91)
- EPERM: An Evidence Path Enhanced Reasoning Model for Knowledge Graph Question and Answering (AAAI 2025 · similarity 0.91)
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.
