Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Thought Cloning: Learning to Think while Acting by Imitating Human Thinking
Shengran Hu, Jeff Clune
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 the Thought Cloning loss (imitate human thoughts) to the same architecture (ablation) | Thought Cloning agent · BabyAI BossLevel | success rate (%): 65.5 -> 96.2 | Improved |
| train with thought imitation instead of behavioral cloning | BabyAI agent · BabyAI BossLevel | success rate (%): 91.2 -> 96.2 | Improved |
| give BC twice as much action data (ablation) | Behavioral Cloning · BabyAI BossLevel | success rate (%): 91.2 -> 91.4 | No clear effect |
| scale BC architecture to match TC parameters (ablation) | Behavioral Cloning · BabyAI BossLevel | success rate (%): 91.2 -> 91.9 | No clear effect |
| scale BC to matched parameters and 2x data (ablation) | Behavioral Cloning · BabyAI BossLevel | success rate (%): 91.2 -> 92.7 | No clear effect |
| gradually decay teacher-forcing rate from 100% to 0% instead of constant teacher forcing or abrupt switch (ablation) | Thought Cloning training · BabyAI BossLevel | success rate | Improved |
| Thought Cloning vs BC on out-of-distribution environments (zero-shot) | Thought Cloning agent · OOD BabyAI (Behavioral/Cognitive difficulty) | success rate | Improved |
| Thought Cloning vs BC when fine-tuning on OOD environments | Thought Cloning agent · OOD BabyAI fine-tuning | success rate | Improved |
| Precrime Intervention: halt agent when unsafe thought detected | Thought Cloning agent · BabyAI unsafe-behavior tests | fraction of episodes with unsafe behavior | Improved |
| inject oracle high-level thoughts into the TC agent | Thought Cloning agent · OOD BabyAI | success rate | Improved |
Closest papers by meaning
- When Can Model-Free Reinforcement Learning be Enough for Thinking? (NeurIPS 2025 · similarity 0.87)
- Hypothetical Minds: Scaffolding Theory of Mind for Multi-Agent Tasks with Large Language Models (ICLR 2025 · similarity 0.86)
- Thought Communication in Multiagent Collaboration (NeurIPS 2025 · similarity 0.86)
- A Smooth Sea Never Made a Skilled SAILOR: Robust Imitation via Learning to Search (NeurIPS 2025 · similarity 0.85)
- Think or Not? Selective Reasoning via Reinforcement Learning for Vision-Language Models (NeurIPS 2025 · similarity 0.85)
- Theory of Mind and Persuasion Beyond Conversation: Assessing the Capacity of LLMs to Induce Belief States via Planning and Action (NeurIPS 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.
