Research map / Reinforcement learning / Visual and representation RL
GaussianPath:A Bayesian Multi-Hop Reasoning Framework for Knowledge Graph Reasoning
Guojia Wan, Bo Du
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 Bayesian representation and Q-function modeling (ablation) | Q-path · FB15K237 | MRR: 0.247 -> 0.423 | Improved |
| Add Bayesian representation and Q-function modeling (ablation) | Q-path · WN18RR | MRR: 0.336 -> 0.458 | Improved |
| Add Bayesian representation and Q-function modeling (ablation) | Q-path · NELL995 | MRR: 0.645 -> 0.748 | Improved |
| Add Thompson sampling for action execution (ablation) | GaussianPath without Thompson Sampling · FB15K237 | MRR: 0.338 -> 0.423 | Improved |
| Add Thompson sampling for action execution (ablation) | GaussianPath without Thompson Sampling · WN18RR | MRR: 0.435 -> 0.458 | Improved |
| Add Thompson sampling for action execution (ablation) | GaussianPath without Thompson Sampling · NELL995 | MRR: 0.693 -> 0.748 | Improved |
| Initialize entity/relation Gaussian priors with pre-trained KG2E embeddings | GaussianPath · NELL995 | MRR: 0.748 -> 0.756 | No clear effect |
| Use Bayesian uncertainty and Thompson sampling instead of one-point RL representations | RL-based multi-hop reasoning · NELL995 | MRR: 0.725 -> 0.748 | Improved |
| Use GaussianPath rather than Multihop-KG | Knowledge graph reasoning · Kinship | MRR: 0.865 -> 0.874 | Improved |
| Use GaussianPath rather than Multihop-KG | Knowledge graph reasoning · UMLS | MRR: 0.940 -> 0.884 | Hurt |
| Use GaussianPath rather than MINERVA | Relation link prediction · NELL995 | MAP: 88.4 -> 91.3 | Improved |
| Use GaussianPath rather than MINERVA | Relation link prediction · FB15K-237 | MAP: 42.3 -> 55.2 | 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.
- M-walk — M-Walk: Learning to Walk over Graphs using Monte Carlo Tree Search (NeurIPS 2018) · better on 2 of 3 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
- Walk Wisely on Graph: Knowledge Graph Reasoning with Dual Agents via Efficient Guidance-Exploration (AAAI 2025 · similarity 0.90)
- MMhops-R1: Multimodal Multi-hop Reasoning (AAAI 2026 · similarity 0.87)
- LoongRL: Reinforcement Learning for Advanced Reasoning over Long Contexts (ICLR 2026 · similarity 0.86)
- Flow of Reasoning: Training LLMs for Divergent Reasoning with Minimal Examples (ICML 2025 · similarity 0.86)
- $\texttt{G1}$: Teaching LLMs to Reason on Graphs with Reinforcement Learning (NeurIPS 2025 · similarity 0.86)
- Grounding Multi-Hop Reasoning in Structural Causal Models via Group Relative Policy Optimization (ICML 2026 · similarity 0.86)
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.
