Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Natural Language PDDL (NL-PDDL) for Open-world Goal-oriented Commonsense Regression Planning in Embodied AI
Xiaotian Liu, Armin Toroghi, Jiazhou Liang, David Courtis, Ruiwen Li, Ali Pesaranghader, Jaehong Kim, Tanmana Sadhu, Hyejeong Jeon, Scott Sanner
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 |
|---|---|---|---|
| use NL-PDDL lifted regression instead of ReAct (w examples) | ALFWorld Text planner · ALFWorld Text | success rate (%): 81% -> 94% | Improved |
| use NL-PDDL instead of ReAct in token cost | ALFWorld Text planner · ALFWorld Text | # tokens: 5,507,266 -> 443,407 | Improved |
| use NL-PDDL instead of fine-tuned EMMA-10 | ALFWorld Vision planner · ALFWorld Vision | success rate (%): 82% -> 84% | Improved |
| use NL-PDDL instead of direct GPT-4o | ALFWorld Vision planner · ALFWorld Vision | success rate (%): 8% -> 84% | Improved |
| use NL-PDDL instead of direct LLM planners on Mystery Blocksworld | Blocksworld planner · Mystery Blocksworld | success rate (%): 0% -> 70% | Improved |
| use classical Fast Downward instead of NL-PDDL on aligned Blocksworld | Blocksworld planner · Blocksworld | success rate (%): 70% -> 100% | Hurt |
| introduce goal-action misalignment | NL-PDDL · ALFWorld Text | success rate (%): 94% -> 91% | Hurt |
| introduce goal-action misalignment | NL-PDDL · ALFWorld Vision | success rate (%): 84% -> 80% | Hurt |
| introduce goal-action misalignment | NL-PDDL · Blocksworld | success rate (%): 70% -> 70% | No clear effect |
| introduce goal-action misalignment | ReAct (w examples) · ALFWorld Text | success rate (%): 81% -> 79% | Hurt |
| introduce goal-action misalignment | Fast Downward · Blocksworld | success rate (%): 100% -> 0% | Hurt |
| increase optimal plan depth beyond 6 | NL-PDDL · Blocksworld | success rate (%): 100% -> 84% | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Reflexion-3 — Reflexion: language agents with verbal reinforcement learning (NeurIPS 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Embodied Task Planning via Graph-Informed Action Generation with Large Language Models (ICML 2026 · similarity 0.89)
- Leveraging Environment Interaction for Automated PDDL Translation and Planning with Large Language Models (NeurIPS 2024 · similarity 0.89)
- World-aware Planning Narratives Enhance Large Vision-Language Model Planner (NeurIPS 2025 · similarity 0.88)
- Multi-Modal Grounded Planning and Efficient Replanning for Learning Embodied Agents with a Few Examples (AAAI 2025 · similarity 0.88)
- R2C: Mapping Room to Chessboard to Unlock LLM As Low-Level Action Planner (CVPR 2025 · similarity 0.88)
- Agent Planning with World Knowledge Model (NeurIPS 2024 · similarity 0.88)
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.
