Researchers introduce MemPlan for text-based planning with PDDL
A new framework called MemPlan bridges language models and symbolic planning for partially observable text environments. The system uses memory-conditioned PDDL planning to reconstruct problems dynamically from observations, candidate hypotheses, and interaction history. It assigns relevance-based action costs to missing objects or conditions and excludes previously refuted assumptions to…
Key points
- MemPlan uses PDDL planning with memory to handle partially observable text environments
- System reduces online token costs and improves task completion over LLM-planning baselines
- Interface works across TextWorld task types without additional fine-tuning
MemPlan outperforms language-only and PDDL-aware baselines in task completion and step efficiency on benchmarks like TextWorld, ALFWorld, and Robotouille. It also reduces online token costs compared to LLM-planning approaches like ReAct. The same interface works across different TextWorld task types without task-specific fine-tuning. The paper and code are available on GitHub.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Researchers introduce MAMath-Bench for multimodal math reasoning · 2 src
- TauIndianBankBench evaluates procedure compliance in indian banking assistants · 1 src
- Lighthouse chatbot to our knowledge launched for Civil Rights history research · 1 src
- Researchers introduce MemPABench to evaluate personal agent memory · 1 src
- BELT reduces LLM reasoning tokens by 41% on Qwen3-1.7B · 1 src
Comments
via GitHub Discussions