{"paper":{"title":"LLM+P: Empowering Large Language Models with Optimal Planning Proficiency","license":"http://creativecommons.org/licenses/by/4.0/","headline":"LLM+P lets language models generate optimal plans by routing problems through classical planners via PDDL translation.","cross_cats":["cs.RO"],"primary_cat":"cs.AI","authors_text":"Bo Liu, Joydeep Biswas, Peter Stone, Qiang Liu, Shiqi Zhang, Xiaohan Zhang, Yuqian Jiang","submitted_at":"2023-04-22T20:34:03Z","abstract_excerpt":"Large language models (LLMs) have demonstrated remarkable zero-shot generalization abilities: state-of-the-art chatbots can provide plausible answers to many common questions that arise in daily life. However, so far, LLMs cannot reliably solve long-horizon planning problems. By contrast, classical planners, once a problem is given in a formatted way, can use efficient search algorithms to quickly identify correct, or even optimal, plans. In an effort to get the best of both worlds, this paper introduces LLM+P, the first framework that incorporates the strengths of classical planners into LLMs"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"LLM+P is able to provide optimal solutions for most problems, while LLMs fail to provide even feasible plans for most problems.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The assumption that large language models can reliably and accurately translate natural language planning problems into syntactically and semantically correct PDDL without introducing errors that invalidate the planner's output.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LLM+P lets LLMs solve planning problems optimally by converting them to PDDL for classical planners and back to natural language.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LLM+P lets language models generate optimal plans by routing problems through classical planners via PDDL translation.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"dcbd98632e4f006e1f663741ecd0f24fd27e19b96485ba09bc11516271ec2c54"},"source":{"id":"2304.11477","kind":"arxiv","version":3},"verdict":{"id":"aa3bb118-bf73-4db8-8031-2d087bac6c9e","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-14T18:30:44.381182Z","strongest_claim":"LLM+P is able to provide optimal solutions for most problems, while LLMs fail to provide even feasible plans for most problems.","one_line_summary":"LLM+P lets LLMs solve planning problems optimally by converting them to PDDL for classical planners and back to natural language.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The assumption that large language models can reliably and accurately translate natural language planning problems into syntactically and semantically correct PDDL without introducing errors that invalidate the planner's output.","pith_extraction_headline":"LLM+P lets language models generate optimal plans by routing problems through classical planners via PDDL translation."},"references":{"count":68,"sample":[{"doi":"","year":1966,"title":"Eliza—a computer program for the study of natural language communication between man and machine,","work_id":"e968b6f0-dc67-4ba3-bbee-78c2dd5a50f5","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Gpt-4 technical report","work_id":"195de85d-887e-420c-ba80-6443af256f61","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"Chatgpt for robotics: Design principles and model abilities,","work_id":"5597eeb9-de68-43cb-a3df-a346ec30197c","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"arXiv preprint arXiv:2301.06627 , volume=","work_id":"9b09d8dc-9d65-4791-a90a-a6f71e2374d0","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":1909,"title":"Mixout: Effective regularization to finetune large-scale pretrained language models.arXiv preprint arXiv:1909.11299,","work_id":"16f75ae4-57ac-4285-a1ac-30c476fdfe71","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":68,"snapshot_sha256":"042e8486c61237d46a7d6fce10ca3989a4e02132ffdddc79d42bf3196964a5c9","internal_anchors":17},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}