Research map / Graph learning / Knowledge and temporal graphs
HiTKG: Towards Goal-Oriented Conversations via Multi-Hierarchy Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Use the hierarchical transformer graph walker with multiscale inputs and MetaPath representations | HiTKG · OpenDialKG test | path@1 recall: 23.72 -> 25.99 | Helped |
| Use the hierarchical transformer graph walker with multiscale inputs and MetaPath representations | HiTKG · OpenDialKG test | tgt@25 recall: 78.79 -> 86.09 | Helped |
| Use 2-entity MetaPaths instead of the full method’s MetaPath (ablation) | HiTKG · OpenDialKG test | path@1 recall: 25.99 -> 24.16 | Hurt |
| Swap the KG-path-history and dialogue-history attention layers (ablation) | HiTKG · OpenDialKG test | path@1 recall: 25.99 -> 24.55 | Hurt |
| Omit speaker 2 utterances (ablation) | HiTKG · OpenDialKG test | path@1 recall: 25.99 -> 25.82 | Hurt |
| Remove the Output-level Length Head (ablation) | HiTKG · OpenDialKG test | path@1 recall: 25.99 -> 24.49 | Hurt |
| Predict entity and relation paths separately instead of using MetaPath (ablation) | HiTKG · OpenDialKG test | path@1 recall: 25.99 -> 23.98 | Hurt |
| Add the global-goal reinforcement-learning stage | HiTKG · Global-goal success, distance 3 | success rate (%): 2 -> 66 | Helped |
| Add the global-goal reinforcement-learning stage | HiTKG · Global-goal success, distance 5 | success rate (%): 0 -> 27 | Helped |
| Add the global-goal reinforcement-learning stage | HiTKG · Global-goal success, distance 7 | success rate (%): 0 -> 11 | Helped |
| Add distance embeddings to the MetaPath for stage-2 learning | HiTKG · OpenDialKG test | tgt@3 recall: 46.29 -> 46.67 | Helped |
| Add distance embeddings to the MetaPath for stage-2 learning | HiTKG · OpenDialKG test | path@1 recall: 25.99 -> 25.12 | Hurt |
Closest papers by meaning
- A Goal Interaction Graph Planning Framework for Conversational Recommendation (AAAI 2024 · similarity 0.86)
- GOALNET: Interleaving Neural Goal Predicate Inference with Classical Planning for Generalization in Robot Instruction Following (AAAI 2024 · similarity 0.86)
- Planning without Search: Refining Frontier LLMs with Offline Goal-Conditioned RL (NeurIPS 2025 · similarity 0.86)
- Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training (ICLR 2025 · similarity 0.86)
- TA&AT: Enhancing Task-Oriented Dialog with Turn-Level Auxiliary Tasks and Action-Tree Based Scheduled Sampling (AAAI 2024 · similarity 0.86)
- MA-RLHF: Reinforcement Learning from Human Feedback with Macro Actions (ICLR 2025 · similarity 0.85)
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.