Research map / Reinforcement learning / Visual and representation RL
Goal-Oriented Dialogue Policy Learning from Failures
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 T-HER to augment DQN training with trimmed dialogue segments that achieve subgoals | DQN dialogue policy · Movie-booking dialogue simulator | learning rate | Helped |
| Use S-HER to augment DQN training by stitching dialogue segments | DQN dialogue policy · Movie-booking dialogue simulator | learning rate | Helped |
| Add prioritized experience replay to T-HER | T-HER · Movie-booking dialogue simulator | performance | Helped |
| Add prioritized experience replay to S-HER | S-HER · Movie-booking dialogue simulator | performance | Helped |
| Use T-HER instead of uniform experience replay or PER | DQN dialogue policy · Movie-booking dialogue simulator, cold start | success rate | Helped |
| Use S-HER instead of uniform experience replay or PER | DQN dialogue policy · Movie-booking dialogue simulator, cold start | success rate | Helped |
| Use S-HER instead of T-HER | DQN dialogue policy · Movie-booking dialogue simulator, cold start | success rate | Helped |
| Combine T-HER and S-HER | DQN dialogue policy · Movie-booking dialogue simulator, cold start | performance | Helped |
| Set the S-HER stitchability threshold too high (ε=2.0) (ablation) | S-HER · Movie-booking dialogue simulator | policy quality | Hurt |
| Set the S-HER stitchability threshold too low (<=0.1) (ablation) | S-HER · Movie-booking dialogue simulator | learning rate | Hurt |
| Remove supervised-learning warm start and reduce the replay pool from 100k to 1k (ablation) | DQN dialogue policy · Movie-booking dialogue simulator, cold start | success rate: 100k to 1k | Hurt |
Closest papers by meaning
- Curriculum-guided Hindsight Experience Replay (NeurIPS 2019 · similarity 0.89)
- From Outcomes to Actions: Leveraging Hindsight for Long-Horizon Language Agent Training (ICML 2026 · similarity 0.87)
- Revisiting Fundamentals of Experience Replay (ICML 2020 · similarity 0.87)
- Memory-T1: Reinforcement Learning for Temporal Reasoning in Multi-session Agents (ICLR 2026 · similarity 0.86)
- Simulation-Free Hierarchical Latent Policy Planning for Proactive Dialogues (AAAI 2025 · similarity 0.85)
- Cluster-Based Sampling in Hindsight Experience Replay for Robotic Tasks (Student Abstract) (AAAI 2024 · 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.