RuleChef: Making AI Rules Editable by Humans
Researchers introduced RuleChef, a system that uses LLMs to generate executable rules for NLP tasks like text classification, NER, and relation extraction. Rules are created from task descriptions and labeled examples, then iteratively improved via human feedback or additional examples. The system can also bootstrap rules from any existing model's input-output pairs.
Researchers unveiled RuleChef, a framework that uses large language models (LLMs) to create editable, executable rules for NLP tasks. These tasks include text classification, named entity recognition (NER), and relation extraction. The rules are generated from a task description and a set of labeled examples, then iteratively improved based on additional examples and human feedback on existing rules.
This matters because it makes AI more accessible. Normally, editing AI rules requires coding skills, but RuleChef lets anyone tweak the rules easily. For example, a small business owner could adjust how their AI categorizes customer emails without needing a developer. The framework can also bootstrap rules using observed input-output pairs from any existing model for a given task, meaning it can learn from models already in use.
Crucially, LLMs are used only at learning time — once the rules are generated, they run without requiring an LLM, making them lightweight and interpretable. The rules are human-editable, allowing non-experts to refine AI behavior directly.