Program-as-Weights: A Programming Paradigm for Fuzzy Functions

2026-07-02

AI Summary

This work addresses the limitations of large model APIs in fuzzy programming tasks—such as log alerting and JSON repair—including poor locality, irreproducibility, and high computational cost. To overcome these challenges, the authors propose Fuzzy Function Programming (FFP), a novel paradigm that compiles natural language specifications into compact, locally executable neural modules. This approach reimagines foundation models not as step-by-step reasoners but as one-shot tool constructors, encoding program semantics directly into parameterized weights. Using a 4B-parameter compiler trained on FuzzyBench, task-specific adapters are generated and deployed on a frozen, lightweight 0.6B-parameter Qwen3 interpreter. The resulting Parameterized Adaptive Weight (PAW) programs match the performance of direct Qwen3-32B inference while reducing memory consumption to approximately 1/50th and achieving 30 tokens/s on a MacBook M3.

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.

Problem

• fuzzy functions
• programmability
• large language models
• local execution
• reproducibility

Innovation

• fuzzy-function programming
• Program-as-Weights
• parameter-efficient adapters
• local inference
• foundation model as tool builder