Planning in the LLM Era: Building for Reliability and Efficiency

Michael Katz, Harsha Kokel, Kavitha Srinivas, Shirin Sohrabi

arXiv:2605.21902 · 2026-05-23 공개 · arXiv · PDF

inference-time resource-efficient verification symbolic-solvers reliable-planners llm-based-planning planner-generation intelligent-agents

Abstract

Growing attention to intelligent agents has put a spotlight on one of their central capabilities: planning. Early attempts to leverage large language models (LLMs) for planning relied on single-shot plan generation, followed by hybrid approaches that coupled LLMs with limited external search. These methods, unsound and incomplete by their very nature, often require substantial resources without yielding better solutions on unseen problems. As the limitations of LLMs become clearer, recent work has shifted toward using them at solution construction time -- generating symbolic solvers for a family of problems that can be verified and then used efficiently at inference time. This trend reflects the growing need for agents that are both reliable and resource-efficient. It also offers a path towards generating maintainable planners with minimal dependence on language models at inference time. In this paper, we argue that this shift reflects a broader realignment of the planning field in the LLM era. We examine three major categories of planner-generation methods, discuss their current limitations, and outline research steps towards a more reliable and efficient LLM-based generation of planners.

한국어 요약

한 줄 요약

**[Planning / Position Paper]** LLM 시대 planning은 single-shot 생성·hybrid search에서 solution construction time symbolic solver 생성으로 재정렬 중 — 3 planner-generation 카테고리·현재 한계·신뢰성·효율성 향상 연구 단계 제시.

핵심 기여도

핵심 아이디어

LLM planning 연구는 inference-time LLM 호출을 줄이고 LLM이 verified symbolic solver를 generate하도록 하는 방향으로 재정렬 중이며, 이것이 신뢰성·자원 효율성의 dual 요구를 동시에 만족하는 경로다.

기술적 접근법

주요 결과

의의 및 한계

**의의**: LLM·planning 통합 연구의 방향성 정리, symbolic solver generation 패러다임의 신뢰성·효율성 가치 제시, inference-time LLM 의존 최소화 비전. **한계**: Position paper로 신규 알고리즘·실험은 부재, 카테고리 경계가 일부 fuzzy, 실제 산업 채택 양상은 추가 검증 필요.

실용적 활용