Most day-to-day text processing tasks—intent routing, JSON repair, log triage, fuzzy search—are "fuzzy": they resist brittle rule systems but don't need a 30B model call per input. The paper's core insight is to turn a foundation model from a per-input problem solver into a tool builder: compile once into a small neural artifact and run that artifact locally and cheaply.
Program-as-Weights: A Programming Paradigm for Fuzzy Functions
Compiles natural-language function specifications into compact, locally-executable neural programs (PAW) that run on a small frozen interpreter; a 4B compiler emits LoRA adapters for a 0.6B runtime to provide offline, low-memory fuzzy text functions.
Introduction
Information
- Websitearxiv.org
- OrganizationsUniversity of Waterloo, Cornell University, Harvard University
- AuthorsWentao Zhang, Liliana Hotsko, Woojeong Kim, Pengyu Nie, Stuart Shieber, Yuntian Deng
- Published date2026/07/02
More Items
Evaluates fast-decision models across 2,919 tasks in 11 domains, measuring both capability and reliability through option-order changes and cases with no valid answer.
Studies how to simplify Hyper-Connections and reduce their training overhead. SimpleHC removes residual mixing and shares state access, matching mHC performance on a 30B MoE model.
Lets AI agents read task-relevant evidence from external content instead of receiving entire documents by default. Intent tracking, adaptive readers, and an exposure budget reduce token use and latency while retaining a faithful full-view fallback.