Research map / Large language models / Agents, code and math
Self-CriTeach: LLM Self-Teaching and Self-Critiquing for Improving Robotic Planning via Automated Domain Generation
Jinbang Huang, Zhiyuan Li, Yuanzhao Hu, Zhanguang Zhang, Mark Coates, Xingyue Quan, Yingxue Zhang
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 (CPO) with self-generated domain reward after SFT (ablation) | SCT SFT-only · Overall | planning success rate: 0.43 -> 0.46 | Improved |
| add RL (CPO) with self-generated domain reward after SFT (ablation) | SCT SFT-only · Overall | progress score: 0.67 -> 0.76 | Improved |
| add SFT before RL (ablation) | SCT CPO-only · Overall | planning success rate: 0.31 -> 0.46 | Improved |
| add symbolic-to-CoT transformation to SFT data (ablation) | SCT Symbol · Overall | planning success rate: 0.38 -> 0.46 | Improved |
| add symbolic-to-CoT transformation to SFT data (ablation) | SCT Symbol · Prepare Experiment (unseen) | planning success rate: 0.16 -> 0.45 | Improved |
| add symbolic-to-CoT transformation to SFT data (ablation) | SCT Symbol · BW Align (seen) | planning success rate: 0.84 -> 0.75 | Hurt |
| use CPO instead of DPO for RL (ablation) | SCT DPO-only · Overall | planning success rate: 0.29 -> 0.31 | Improved |
| use LCCS instead of CPO for RL (ablation) | SCT CPO-only · Overall | progress score: 0.69 -> 0.71 | Improved |
| apply Self-CriTeach to Qwen3-4B | Qwen3-4B · Overall | planning success rate: 0.26 -> 0.46 | Improved |
| apply Self-CriTeach to Qwen3-4B | Qwen3-4B · BW Hard (seen) | planning success rate: 0.24 -> 0.45 | Improved |
| apply Self-CriTeach to Llama-3.1-8B | Llama-3.1-8B · Overall | planning success rate: 0.003 -> 0.277 | Improved |
| apply Self-CriTeach to Qwen3-8B | Qwen3-8B · Overall | planning success rate: 0.35 -> 0.49 | Improved |
Closest papers by meaning
- Plan-Seq-Learn: Language Model Guided RL for Solving Long Horizon Robotics Tasks (ICLR 2024 · similarity 0.89)
- Closed-Loop Long-Horizon Robotic Planning via Equilibrium Sequence Modeling (ICML 2025 · similarity 0.88)
- Hierarchical Planning for Complex Tasks with Knowledge Graph-RAG and Symbolic Verification (ICML 2025 · similarity 0.88)
- UniDomain: Pretraining a Unified PDDL Domain from Real-World Demonstrations for Generalizable Robot Task Planning (NeurIPS 2025 · similarity 0.87)
- Stabilizing Contrastive RL: Techniques for Robotic Goal Reaching from Offline Data (ICLR 2024 · similarity 0.87)
- On the self-verification limitations of large language models on reasoning and planning tasks (ICLR 2025 · similarity 0.87)
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.
