Program-as-Weights：一种面向模糊函数的编程范式

AI 导读

Program-as-Weights (PAW) 提出模糊函数编程范式，将自然语言描述的函数编译为紧凑、可本地执行的神经制品。PAW 使用在 10M 示例数据集 FuzzyBench 上训练的 4B 编译器，为冻结的轻量级解释器输出参数高效适配器。0.6B 的 Qwen3 解释器执行 PAW 程序性能匹敌直接提示 Qwen3-32B，推理内存仅约五十分之一，在 MacBook M3 上达 30 tokens/s。该方法将基础模型从每次输入的求解器重新定义为工具构建器，一次函数定义后生成的制品可离线廉价复用。

推荐理由

这篇论文把模糊任务从调用大模型API变成了本地轻量函数，0.6B就能跑赢32B，省掉几十倍成本。对需要处理日志、JSON修复等经常性模糊任务的开发者是个真信号。

正文

View PDF HTML (experimental)

Abstract: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 the inference memory and running at 30 tokens/s on a MacBook M3. PAW reframes the foundation model from a per-input problem solver into a tool builder: invoked once per function definition, it produces a small reusable artifact whose subsequent calls per function application are cheap and offline.

Subjects:	Machine Learning (cs.LG); Artificial Intelligence (cs.AI); Computation and Language (cs.CL)
Cite as:	arXiv:2607.02512 [cs.LG]
	(or arXiv:2607.02512v1 [cs.LG] for this version)
	https://doi.org/10.48550/arXiv.2607.02512 arXiv-issued DOI via DataCite

Submission history

From: Yuntian Deng [view email]
[v1] Thu, 2 Jul 2026 17:59:50 UTC (1,727 KB)