Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Bayes-Adaptive Monte-Carlo Planning and Learning for Goal-Oriented Dialogues
Youngsoo Jang, Jongmin Lee, Kee-Eung Kim
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 |
|---|---|---|---|
| sample opponent goal from posterior instead of prior (ablation) | BADP · Negotiation vs LIKELIHOOD | score (all): 8.29 -> 8.54 | Improved |
| sample opponent goal from posterior instead of fixed own goal (MCTS) (ablation) | BADP · Negotiation vs LIKELIHOOD | score (all): 8.14 -> 8.54 | Improved |
| sample goal from posterior instead of prior, Pareto optimality (ablation) | BADP · Negotiation vs LIKELIHOOD | % Pareto optimal: 67.2% -> 70.1% | Improved |
| closed-loop MCTS planning instead of open-loop rollout (ablation) | LIKELIHOOD · Negotiation vs LIKELIHOOD | score (all): 6.88 -> 8.14 | Improved |
| add posterior-based goal sampling to MCTS, Pareto optimality (ablation) | MCTS · Negotiation vs LIKELIHOOD | % Pareto optimal: 70.4% -> 70.1% | No clear effect |
| policy improvement with BADP iterations (LIKELIHOOD(0) to (5)) (ablation) | LIKELIHOOD · Negotiation vs LIKELIHOOD | score (all): 5.47 -> 8.57 | Improved |
| further BADP planning iterations beyond 3 (ablation) | BADP · Negotiation vs LIKELIHOOD | score (all): 9.19 -> 9.20 | No clear effect |
| use BADP as policy improvement operator instead of REINFORCE | dialogue policy · Negotiation vs LIKELIHOOD | score (all): 7.10 -> 8.57 | Improved |
| use BADP as policy improvement operator instead of REINFORCE | dialogue policy · Human evaluation | language quality: 2.55 -> 4.70 | Improved |
| use BADP as policy improvement operator instead of REINFORCE | dialogue policy · Human evaluation | end-task score: 4.91 -> 7.43 | Improved |
| posterior goal sampling instead of prior in RL (ablation) | BADP-RL · Negotiation vs LIKELIHOOD (RL) | score (all): 8.40 -> 8.57 | Improved |
Closest papers by meaning
- Plug-and-Play Policy Planner for Large Language Model Powered Dialogue Agents (ICLR 2024 · similarity 0.86)
- Simulation-Free Hierarchical Latent Policy Planning for Proactive Dialogues (AAAI 2025 · similarity 0.85)
- DialogXpert: Driving Intelligent and Emotion-Aware Conversations Through Online Value-Based Reinforcement Learning with LLM Priors (AAAI 2026 · similarity 0.85)
- Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training (ICLR 2025 · similarity 0.85)
- Monte Carlo Planning with Large Language Model for Text-Based Game Agents (ICLR 2025 · similarity 0.85)
- Memory-T1: Reinforcement Learning for Temporal Reasoning in Multi-session Agents (ICLR 2026 · 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.
