Online Learning of HTN Methods for integrated LLM-HTN Planning

Héctor Muñoz-Avila · 2025

Read Online Learning of HTN Methods for integrated LLM-HTN Planning by Héctor Muñoz-Avila (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2511-12901.txt.

Description

Yuesheng Xu and Hector Munoz-Avila teach a planning chatbot to stop asking for help with the same problems. Their system builds on ChatHTN, a planner that breaks tasks into subtasks using hierarchical task network methods, and calls ChatGPT when no stored method fits. The new work adds online learning: when the language model produces a decomposition, the planner generalizes and stores it as a method. Similar future tasks are solved without another call to the model.

Unlike simple memoization, the learned method applies to other instances of the same task, not just the case that triggered it. Experiments in two domains showed the learner cut the number of ChatGPT calls while solving at least as many problems, and sometimes more. The work appeared at a cognitive systems conference. It points toward hybrid planners that use a language model as a teacher whose lessons are kept, checked, and reused.

Catalog record

CREATORHéctor Muñoz-Avila
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE39 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2511-12901

Browse the full library · About the Tower of Babel