![[2602.22067] Semantic Partial Grounding via LLMs](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.22067] Semantic Partial Grounding via LLMs
Summary
The paper proposes SPG-LLM, a novel approach for semantic partial grounding in AI planning that utilizes large language models (LLMs) to enhance efficiency and reduce computational bottlenecks.
Why It Matters
This research addresses a significant challenge in AI planning by improving grounding processes, which are crucial for effective task execution. By leveraging LLMs, the proposed method can potentially transform how AI systems handle complex planning tasks, making them faster and more efficient.
Key Takeaways
- SPG-LLM significantly reduces the size of grounded tasks by identifying irrelevant elements before grounding.
- The method achieves faster grounding times, often by orders of magnitude, compared to traditional approaches.
- It maintains or improves plan costs across various challenging benchmarks.
- The approach leverages textual and structural cues from PDDL descriptions, enhancing grounding efficiency.
- This research could lead to advancements in AI planning and operational efficiency in various applications.
Computer Science > Artificial Intelligence arXiv:2602.22067 (cs) [Submitted on 25 Feb 2026] Title:Semantic Partial Grounding via LLMs Authors:Giuseppe Canonaco, Alberto Pozanco, Daniel Borrajo View a PDF of the paper titled Semantic Partial Grounding via LLMs, by Giuseppe Canonaco and 2 other authors View PDF HTML (experimental) Abstract:Grounding is a critical step in classical planning, yet it often becomes a computational bottleneck due to the exponential growth in grounded actions and atoms as task size increases. Recent advances in partial grounding have addressed this challenge by incrementally grounding only the most promising operators, guided by predictive models. However, these approaches primarily rely on relational features or learned embeddings and do not leverage the textual and structural cues present in PDDL descriptions. We propose SPG-LLM, which uses LLMs to analyze the domain and problem files to heuristically identify potentially irrelevant objects, actions, and predicates prior to grounding, significantly reducing the size of the grounded task. Across seven hard-to-ground benchmarks, SPG-LLM achieves faster grounding-often by orders of magnitude-while delivering comparable or better plan costs in some domains. Subjects: Artificial Intelligence (cs.AI) Cite as: arXiv:2602.22067 [cs.AI] (or arXiv:2602.22067v1 [cs.AI] for this version) https://doi.org/10.48550/arXiv.2602.22067 Focus to learn more arXiv-issued DOI via DataCite (pending registration) Su...