[论文] Program-as-Weights: A Programming Paradigm for Fuzzy Functions

论文概要

研究领域: NLP 作者: Wentao Zhang, Liliana Hotsko, Woojeong Kim 发布时间: 2026-07-04 arXiv: 2507.00480

中文摘要

许多日常编程任务难以用清晰的基于规则的实现来完成，例如对重要日志行发出警报、修复格式错误的JSON，或按意图对搜索结果排序，这些任务越来越多地外包给大语言模型API，代价是丧失了本地性、可复现性和经济性。我们提出了模糊函数编程：将此类函数从自然语言规范编译成紧凑、可在本地执行的神经制品。我们通过Program-as-Weights（PAW）实例化这一范式，其中一个在FuzzyBench（我们发布的1000万示例数据集）上训练的4B编译器，为冻结的轻量级解释器发射参数高效的适配器。执行PAW程序的0.6B Qwen3解释器与直接提示Qwen3-32B的性能相当，同时使用约五十分之一的推理内存，并在MacBook M3上以30 token/秒运行。PAW将基础模型从每次输入的问题求解器重新定义为工具构建者：每个函数定义调用一次，它产生一个小的可重用制品，其后续每次函数应用的调用既廉价又可离线运行。

原文摘要

Many everyday programming tasks resist clean rule-based implementation, such as alerting on important log lines, repairing malformed JSON, or ranking search results by intent, and are increasingly outsourced to large language model APIs at the cost of locality, reproducibility, and price. We propose fuzzy-function programming: compiling such a function from a natural-language specification into a compact, locally-executable neural artifact. We instantiate this paradigm with Program-as-Weights (PAW), in which a 4B compiler trained on FuzzyBench, a 10M-example dataset we release, emits parameter-efficient adapters for a frozen, lightweight interpreter. A 0.6B Qwen3 interpreter executing PAW programs matches the performance of direct prompting of Qwen3-32B, while using roughly one fiftieth of...

--- *自动采集于 2026-07-04*

#论文 #arXiv #NLP #小凯