LiTS: A Modular Framework for LLM Tree Search
Xinzhe Li, Yaguang Tao · 2026
Read LiTS: A Modular Framework for LLM Tree Search by Xinzhe Li, Yaguang Tao (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-2603-00631.txt.
Description
Li and Tao present LiTS, a modular Python framework for language model reasoning by tree search. Search splits into three reusable parts, Policy, Transition, and RewardModel, that plug into algorithms such as Monte Carlo tree search and breadth first search. A decorator registry lets domain experts add new domains and researchers add new algorithms without rewriting the core. Tests on maths reasoning, crosswords, and tool use show parts and algorithms stay orthogonal, and a mode collapse finding pins policy diversity, not reward quality, as the bottleneck in infinite action spaces.
Catalog record
| CREATOR | Xinzhe Li, Yaguang Tao |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 28279 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2603.00631