Subgoal-Guided Policy Heuristic Search with Learned Subgoals
Jake Tuero, Michael Buro, Levi H. S. Lelis · 2025
Read Subgoal-Guided Policy Heuristic Search with Learned Subgoals by Jake Tuero, Michael Buro, Levi H. S. Lelis (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-2506-07255.txt.
Description
Jake Tuero, Michael Buro, and Levi Lelis improve how search policies learn from failed attempts in a 2025 paper. Policy-guided tree search solves problems by expanding a search tree steered by a learned policy, with guarantees tied to policy quality. Training normally needs complete solution paths, which are costly when problems are hard and early policies are random, so failed searches waste their data.
Their method learns subgoals, intermediate targets, plus policies aimed at each subgoal, from all the trees the search grows, failures included. That reuse makes learning far more sample efficient in the online setting. Tests show better learning of both the policy and the guiding heuristic than the standard approach, turning wasted attempts into training material.
Catalog record
| CREATOR | Jake Tuero, Michael Buro, Levi H. S. Lelis |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 65 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-07255