Research map / Reinforcement learning / Visual and representation RL
Dialog Policy Learning for Joint Clarification and Active Learning Queries
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 |
|---|---|---|---|
| learn all three policies instead of keeping clarification and active learning static (ablation) | Hierarchical dialog policy · iMaterialist policy_test, final batch | successful dialogs: 0.17 -> 0.33 | Helped |
| learn the active learning policy rather than use the static policy (ablation) | Q-Learning decision and A3C clarification policies · iMaterialist policy_test, final batch | successful dialogs: 0.15 -> 0.33 | Helped |
| learn the clarification policy rather than use the static policy (ablation) | Q-Learning decision and A3C active learning policies · iMaterialist policy_test, final batch | successful dialogs: 0.09 -> 0.33 | Helped |
| learn the decision policy rather than use the static policy (ablation) | A3C clarification and active learning policies · iMaterialist policy_test, final batch | successful dialogs: 0.27 -> 0.33 | No clear effect |
| learn the decision policy rather than use the static policy (ablation) | A3C clarification and active learning policies · iMaterialist policy_test, final batch | average dialog length: 20.00 -> 9.40 | Helped |
| learn the active learning policy rather than use the static policy (ablation) | Q-Learning decision and A3C clarification policies · iMaterialist policy_test, final batch | average dialog length: 14.16 -> 9.40 | Helped |
| learn the clarification policy rather than use the static policy (ablation) | Q-Learning decision and A3C active learning policies · iMaterialist policy_test, final batch | average dialog length: 1.00 -> 9.40 | Mixed |
| use learned policies instead of the static policy | Dialog policy in modified task setup · iMaterialist modified task, simulation | successful dialogs: 0.23 -> 0.65 | Helped |
| use learned policies instead of the static policy | Dialog policy in modified task setup · iMaterialist modified task, simulation | average dialog length: 20.0 -> 20.0 | No clear effect |
| use learned policies instead of the static policy | Dialog policy evaluated with crowdworkers · Amazon Mechanical Turk | successful dialogs: 0.06 -> 0.16 | Helped |
| use learned policies instead of the static policy | Dialog policy evaluated with crowdworkers · Amazon Mechanical Turk | average dialog length: 19.16 -> 18.86 | No clear effect |
| combine learned active learning and clarification policies | Fully learned dialog policy · iMaterialist policy_test, across test batches | successful dialogs | Helped |
Closest papers by meaning
- An Efficient Self-Learning Framework For Interactive Spoken Dialog Systems (ICML 2024 · similarity 0.84)
- TA&AT: Enhancing Task-Oriented Dialog with Turn-Level Auxiliary Tasks and Action-Tree Based Scheduled Sampling (AAAI 2024 · similarity 0.84)
- Clarification as Supervision: Reinforcement Learning for Vision-Language Interfaces (NeurIPS 2026 · similarity 0.84)
- Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training (ICLR 2025 · similarity 0.84)
- MMDU: A Multi-Turn Multi-Image Dialog Understanding Benchmark and Instruction-Tuning Dataset for LVLMs (NeurIPS 2024 · similarity 0.84)
- Memory-T1: Reinforcement Learning for Temporal Reasoning in Multi-session Agents (ICLR 2026 · similarity 0.83)
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.