Syntactic and Semantic Control of Large Language Models via Sequential Monte Carlo
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866910914213576704 |
|---|---|
| author | Loula, João LeBrun, Benjamin Du, Li Lipkin, Ben Pasti, Clemente Grand, Gabriel Liu, Tianyu Emara, Yahya Freedman, Marjorie Eisner, Jason Cotterell, Ryan Mansinghka, Vikash Lew, Alexander K. Vieira, Tim O'Donnell, Timothy J. |
| author_facet | Loula, João LeBrun, Benjamin Du, Li Lipkin, Ben Pasti, Clemente Grand, Gabriel Liu, Tianyu Emara, Yahya Freedman, Marjorie Eisner, Jason Cotterell, Ryan Mansinghka, Vikash Lew, Alexander K. Vieira, Tim O'Donnell, Timothy J. |
| contents | A wide range of LM applications require generating text that conforms to syntactic or semantic constraints. Imposing such constraints can be naturally framed as probabilistic conditioning, but exact generation from the resulting distribution -- which can differ substantially from the LM's base distribution -- is generally intractable. In this work, we develop an architecture for controlled LM generation based on sequential Monte Carlo (SMC). Our SMC framework allows us to flexibly incorporate domain- and problem-specific constraints at inference time, and efficiently reallocate computational resources in light of new information during the course of generation. By comparing to a number of alternatives and ablations on four challenging domains -- Python code generation for data science, text-to-SQL, goal inference, and molecule synthesis -- we demonstrate that, with little overhead, our approach allows small open-source language models to outperform models over 8x larger, as well as closed-source, fine-tuned ones. In support of the probabilistic perspective, we show that these performance improvements are driven by better approximation to the posterior distribution. Our system builds on the framework of Lew et al. (2023) and integrates with its language model probabilistic programming language, giving users a simple, programmable way to apply SMC to a broad variety of controlled generation problems. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2504_13139 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Syntactic and Semantic Control of Large Language Models via Sequential Monte Carlo Loula, João LeBrun, Benjamin Du, Li Lipkin, Ben Pasti, Clemente Grand, Gabriel Liu, Tianyu Emara, Yahya Freedman, Marjorie Eisner, Jason Cotterell, Ryan Mansinghka, Vikash Lew, Alexander K. Vieira, Tim O'Donnell, Timothy J. Computation and Language Artificial Intelligence Machine Learning A wide range of LM applications require generating text that conforms to syntactic or semantic constraints. Imposing such constraints can be naturally framed as probabilistic conditioning, but exact generation from the resulting distribution -- which can differ substantially from the LM's base distribution -- is generally intractable. In this work, we develop an architecture for controlled LM generation based on sequential Monte Carlo (SMC). Our SMC framework allows us to flexibly incorporate domain- and problem-specific constraints at inference time, and efficiently reallocate computational resources in light of new information during the course of generation. By comparing to a number of alternatives and ablations on four challenging domains -- Python code generation for data science, text-to-SQL, goal inference, and molecule synthesis -- we demonstrate that, with little overhead, our approach allows small open-source language models to outperform models over 8x larger, as well as closed-source, fine-tuned ones. In support of the probabilistic perspective, we show that these performance improvements are driven by better approximation to the posterior distribution. Our system builds on the framework of Lew et al. (2023) and integrates with its language model probabilistic programming language, giving users a simple, programmable way to apply SMC to a broad variety of controlled generation problems. |
| title | Syntactic and Semantic Control of Large Language Models via Sequential Monte Carlo |
| topic | Computation and Language Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2504.13139 |