Probabilistic Programs of Thought
Poorva Garg, Renato Lui Geh, Daniel Israel, Todd Millstein, Kyle Richardson, Guy Van den Broeck · 2026
Read Probabilistic Programs of Thought by Poorva Garg, Renato Lui Geh, Daniel Israel, Todd Millstein, Kyle Richardson, Guy Van den Broeck (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-2604-17290.txt.
Description
Garg, Geh, Israel, Millstein, Richardson, and Van den Broeck rethink how language models sample programs, a process that normally burns one costly generation per candidate. Their probabilistic programs of thought fold a model's next-token odds into a single generated program, compactly standing for exponentially many variants that can then be sampled cheaply. Across code and math benchmarks, the approach improved scores while needing fewer generations from the model.
Catalog record
| CREATOR | Poorva Garg, Renato Lui Geh, Daniel Israel, Todd Millstein, Kyle Richardson, Guy Van den Broeck |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 82592 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.17290