StructuredAgent : Planning with AND/OR Trees for Long-Horizon Web Tasks Work done while at Amazon
Elita A. Lobo, Xu Chen, Jingjing Meng, Nan Xi, Yang Jiao, Chirag Agarwal, Yair Zick, Yan Gao · 2026
Read StructuredAgent : Planning with AND/OR Trees for Long-Horizon Web Tasks Work done while at Amazon by Elita A. Lobo, Xu Chen, Jingjing Meng, Nan Xi, Yang Jiao, Chirag Agarwal, Yair Zick, Yan Gao (2026) 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-2603-05294.txt.
Description
ELita Lobo and team offer STRUCTUREDAGENT, hierarchical planning interleaving planning and execution via AND/OR trees for long-horizon web tasks. Current LLM web agents falter on complex long-horizon tasks through limited memory, weak planning, and greedy premature termination. The method separates shape planning from LLM reasoning, letting error recovery through node repair, exploration via OR nodes, and modular plans for human intervention. On 630 WebArena tasks it reaches roughly 53% success versus 46% for AgentOccam, and gains reach 10% over the strongest older model on multi-constraint Amazon shopping tasks.
Catalog record
| CREATOR | Elita A. Lobo, Xu Chen, Jingjing Meng, Nan Xi, Yang Jiao, Chirag Agarwal, Yair Zick, Yan Gao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 148244 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.05294