Get Started
Home
Topics
Search
Library
Research questionHow can recurring natural-language text functions become reusable local programs without runtime remote inference?Recurring text functions are often difficult to encode with reliable rules. Sending every input to a remote model adds cost, latency, and dependence on an external provider.
AI
Code Generation & Program Synthesis
Inference Optimization
Machine Learning
Natural Language Processing
Small / On-device Models
Technology
Latest papersRecent research connected to this question, newest first.Compile by Training: Turning Natural-Language Specifications into Local Neural FunctionsThis paper uses teacher models during compilation to generate examples for training an adapter around a compact local interpreter. It evaluates the approach on a difficult text-function benchmark subset and interactive demonstrations; teacher-model access is required for compilation but not runtime execution.research paper · Sep 3, 2026
Related questions
How can natural-language interfaces generate reliable SQL and GraphQL without runtime LLM inference?How can LLMs propagate local conversational revisions across dependent artifact components without excessive test-time computation?How can LLM coding agents construct persistent, editable platform-native data pipelines instead of one-off scripts?How can language models compress conversational and document context without reconstruction while preserving task-relevant information?