Research map / Multi-agent systems and games / LLM agents
ESCA: An Emotional Support Conversation Agent for Enhancing Reasonable Strategy Planning and Effective Expression
Jing Li, Yanxin Luo, Donghong Han, Yimeng Zhan, Xiaoming Fu, Baiyou Qiao, Gang Wu
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes ESCA, an LLM-based emotional support conversation agent aimed at improving strategy planning and the expression of support strategies in responses.
Problem
Emotional Support Conversation (ESC) requires planning reasonable support strategies over multi-turn dialogue and expressing them effectively in responses; existing methods struggle with both.
Approach
An LLM-based agent (ESCA) designed to plan support strategies more reasonably and express them more effectively in generated responses; specific mechanism not stated in the abstract provided.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add seeker state information (emotion, trust, behavior, stage) to the strategy planner (ablation) | ESCA · ESConv goal achievement | SR@10 (%): 37.31 -> 47.44 | Improved |
| add external knowledge to the prompt generator (ablation) | ESCA · ESConv goal achievement | SR@10 (%): 40.33 -> 47.44 | Improved |
| add the trained prompt generator instead of inserting retrieved knowledge directly (ablation) | ESCA · ESConv goal achievement | SR@10 (%): 32.31 -> 47.44 | Improved |
| add online RL (PPO) optimization of the prompt generator (ablation) | ESCA · ESConv goal achievement | SR@10 (%): 44.61 -> 47.44 | Improved |
| add online RL (PPO) optimization of the prompt generator (ablation) | ESCA · ESConv response generation | Distinct-1/2/3: D-1 6.32 -> 6.15; D-2 31.73 -> 30.51 | Hurt |
| use ESCA (LLaMA2-7b) instead of DPDP (LLaMA-13b) as the goal-achievement agent | strategy planning agent · ESConv goal achievement | SR@10 (%): 32.31 -> 47.44 | Improved |
| use ESCA instead of D2RCU on response quality | response generator · ESConv response generation | BLEU-4 / ROUGE-L: B-4 2.68 -> 2.50; R-L 18.12 -> 17.87; D-2 28.89 -> 30.51 | Mixed |
Closest papers by meaning
- SocialSim: Towards Socialized Simulation of Emotional Support Conversation (AAAI 2025 · similarity 0.86)
- DialogXpert: Driving Intelligent and Emotion-Aware Conversations Through Online Value-Based Reinforcement Learning with LLM Priors (AAAI 2026 · similarity 0.86)
- RLVER: Reinforcement Learning with Verifiable Emotion Rewards for Empathetic Agents (ICLR 2026 · similarity 0.85)
- Plug-and-Play Policy Planner for Large Language Model Powered Dialogue Agents (ICLR 2024 · similarity 0.84)
- Simulation-Free Hierarchical Latent Policy Planning for Proactive Dialogues (AAAI 2025 · similarity 0.83)
- Learning to Clarify: Multi-turn Conversations with Action-Based Contrastive Self-Training (ICLR 2025 · 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.
