| title | Compile by Training: Turning Natural-Language Specifications into Local Neural Functions |
|---|---|
| authors | Yuntian Deng, Pengyu Nie, Stuart Shieber |
| date | 2026-09-03 |
| venue | arXiv (preprint, not peer reviewed) |
| id | arXiv:2609.04199 |
| affiliation | University of Waterloo |
| upvotes | 55 (Hugging Face Daily Papers, 2026-11-01) |
| links | abs pdf hf code |
| abstract | Many recurring text functions are easy to describe but difficult to implement with rules, while calling a large remote model for every input introduces repeated cost, latency, and dependency on a provider. We present compile by training, which turns a natural-language specification into a reusable neural function. At compile time, teacher models generate task-specific examples that are used to train a small adapter for a compact interpreter. The resulting function runs without the teachers and can be stored, versioned, and composed like ordinary software. On FuzzyBench-Hard, a subset on which the Program-as-Weights fast compiler produced no exact matches, compile by training reaches 83.6% semantic accuracy. This higher accuracy comes with a higher compile-time cost: roughly a minute rather than seconds for the fast compiler. We deploy the compiler in a public interactive service and demonstrate compiled functions in a multi-site website helper, a language-controlled 3D avatar, and a bidirectional English-Claudish translator. |
cite
@misc{deng2026compile,
title={Compile by Training: Turning Natural-Language Specifications into Local Neural Functions},
author={Yuntian Deng and Pengyu Nie and Stuart Shieber},
year={2026},
eprint={2609.04199},
archivePrefix={arXiv},
url={https://arxiv.org/abs/2609.04199}
}