Research map / Large language models / Retrieval and knowledge
VoG: Enhancing LLM Reasoning through Stepwise Verification on Knowledge Graphs
Wenxin Zhao, Jiachuan Wang, Yongqi Zhang, Shuangyin Li, Cheng Deng, Jun Wang, Lei 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 |
|---|---|---|---|
| add retrieval, stepwise verification and revision to the plan (ablation) | VoG · CWQ | Hits@1: 51.7 -> 64.7 | Improved |
| add retrieval, stepwise verification and revision to the plan (ablation) | VoG · WebQSP | Hits@1: 72.1 -> 83.2 | Improved |
| use bandit context selector instead of fixed Local strategy (ablation) | VoG · CWQ | Hits@1: 60.1 -> 64.7 | Improved |
| use bandit context selector instead of fixed Lookahead strategy (ablation) | VoG · CWQ | Hits@1: 63.6 -> 64.7 | Improved |
| use bandit context selector instead of fixed Global strategy (ablation) | VoG · WebQuestions | Hits@1: 60.6 -> 63.0 | Improved |
| add KG-aware prior and diversity bonuses to UCB (ablation) | VoG context selector · WebQSP | Hits@1: 74.7 -> 83.2 | Improved |
| add KG-aware prior and diversity bonuses to UCB (ablation) | VoG context selector · CWQ | Hits@1: 63.0 -> 64.7 | Improved |
| perturb UCB bonus and exploration weights by 20 percent (ablation) | VoG · CWQ/WebQSP | accuracy: fluctuation within 1.2% accuracy | No clear effect |
| use VoG instead of PoG as KG agent | GPT-3.5 agents · CWQ | Hits@1: 63.2 -> 64.7 | Improved |
| use VoG instead of PoG as KG agent | Qwen2.5-7B agents · CWQ | Hits@1: 46.0 -> 53.3 | Improved |
| use VoG instead of PoG | GPT-3.5 agents · CWQ | tokens per query: 8156.2 -> 6566.8 | Improved |
| use VoG (no fine-tuning) instead of fine-tuned GNN-RAG+RA | GPT-3.5 · CWQ | Hits@1: 68.7 -> 64.7 | Hurt |
Closest papers by meaning
- Think-on-Graph: Deep and Responsible Reasoning of Large Language Model on Knowledge Graph (ICLR 2024 · similarity 0.92)
- Reasoning on Graphs: Faithful and Interpretable Large Language Model Reasoning (ICLR 2024 · similarity 0.92)
- LightPROF: A Lightweight Reasoning Framework for Large Language Model on Knowledge Graph (AAAI 2025 · similarity 0.90)
- DARG: Dynamic Evaluation of Large Language Models via Adaptive Reasoning Graph (NeurIPS 2024 · similarity 0.90)
- Graph-constrained Reasoning: Faithful Reasoning on Knowledge Graphs with Large Language Models (ICML 2025 · similarity 0.90)
- A State-Transition Framework for Efficient LLM Reasoning (ICLR 2026 · 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.
