Research map / Large language models / Agents, code and math
Interactive Plan Selection Using Linear Temporal Logic, Disjunctive Action Landmarks, and Natural Language Instruction
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Presents Lemming, a visualization tool that lets users interactively select plans using disjunctive action landmarks (SELECT, BUILD forward/backward modes) and natural-language-to-LTL constraints.
Problem
Users need to select a preferred plan from a large set of generated plans, but existing approaches show all plans as separate sequences or lack a human-in-the-loop selection interface, and preferences may not be known upfront.
Approach
Uses disjunctive action landmarks, computed via a greedy disambiguation graph partitioning the plan set, to surface the next choice points to the user, and integrates NL2LTL so users can add natural-language constraints that are compiled into new planning problems.
Key results
- Demonstrates four user experiences (SELECT, NL2LTL, BUILD forward, BUILD backward) for interactive plan selection; no quantitative results reported in the text.
Limitations
- Worst-case number of user choices is the same with or without landmarks
- Greedy disambiguation graph may miss the preferred plan, especially in BUILD mode
- Landmark-disambiguated plan collections cannot capture arbitrarily complex preferences not modeled in the domain
Experimental setup
Builds on: top-k planning NL2LTL LTL to PDDL compilation Reagraph
Closest papers by meaning
- Visual Planning: Let's Think Only with Images (ICLR 2026 · similarity 0.84)
- ACPBench: Reasoning About Action, Change, and Planning (AAAI 2025 · similarity 0.84)
- On the self-verification limitations of large language models on reasoning and planning tasks (ICLR 2025 · similarity 0.84)
- Chain of Thoughtlessness? An Analysis of CoT in Planning (NeurIPS 2024 · similarity 0.83)
- StrategyLLM: Large Language Models as Strategy Generators, Executors, Optimizers, and Evaluators for Problem Solving (NeurIPS 2024 · similarity 0.83)
- Learning Planning Abstractions from Language (ICLR 2024 · similarity 0.83)
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.