Research map / Large language models / Agents, code and math
ToMAP: Training Opponent-Aware LLM Persuaders with Theory of Mind
Peixuan Han, Zijia Liu, Jiaxuan You
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 |
|---|---|---|---|
| add RL training | Qwen2.5-3B · Table 1 avg | agreement shift: 11.76 -> 14.60 | Improved |
| add SFT on successful conversations | Qwen2.5-3B · Table 1 avg | agreement shift: 11.76 -> 12.49 | No clear effect |
| add ToM modules on top of RL | Qwen2.5-3B · Table 1 avg | agreement shift: 14.60 -> 17.48 | Improved |
| add ToM modules on top of RL | LLaMa3.2-3B · Table 1 avg | agreement shift: 21.76 -> 24.35 | Improved |
| add ToM info to untrained base model (ablation) | Qwen2.5-3B · Qwen2.5-7B persuadee, 3 datasets | agreement shift: 8.44 -> 8.18 | No clear effect |
| add ToM info zero-shot to larger models (ablation) | Gemma-3-27B · Qwen2.5-7B persuadee, 3 datasets | agreement shift: 11.92 -> 13.26 | Improved |
| add ToM info to RL-trained model (ablation) | Qwen2.5-3B+ToMAP · Qwen2.5-7B persuadee, 3 datasets | agreement shift: 14.47 -> 19.19 | Improved |
| add long-CoT prompt (ablation) | Qwen2.5-3B · CMV | agreement shift: 12.75 -> 14.25 | No clear effect |
| add attitude predictor to counterclaim predictor (ablation) | ToMAP · CMV | agreement shift: 15.57 -> 23.01 | Improved |
| add attitude predictor to counterclaim predictor (ablation) | ToMAP · Anthropic | agreement shift: 9.55 -> 15.82 | Improved |
| use ground-truth attitudes instead of predicted (ablation) | ToMAP · CMV | agreement shift: 23.01 -> 23.57 | No clear effect |
| add ToM modules to RL | Qwen2.5-3B · LLaMa3.1-8B persuadee, CMV (OOD) | agreement shift: 12.15 -> 10.77 | Hurt |
Closest papers by meaning
- Theory of Mind and Persuasion Beyond Conversation: Assessing the Capacity of LLMs to Induce Belief States via Planning and Action (NeurIPS 2026 · similarity 0.89)
- Adaptive Theory of Mind for LLM-based Multi-Agent Coordination (AAAI 2026 · similarity 0.86)
- AutoToM: Scaling Model-based Mental Inference via Automated Agent Modeling (NeurIPS 2025 · similarity 0.85)
- Should we be going MAD? A Look at Multi-Agent Debate Strategies for LLMs (ICML 2024 · similarity 0.85)
- OmniToM: Benchmarking Theory of Mind in LLMs via Explicit Belief Modeling (NeurIPS 2026 · similarity 0.85)
- ToMATO: Verbalizing the Mental States of Role-Playing LLMs for Benchmarking Theory of Mind (AAAI 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.
