Learning and Enforcing Context-Sensitive Control for LLMs

arXiv cs.CL / 4/14/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper proposes a framework to automatically learn context-sensitive control constraints for LLM outputs, addressing the manual-specification burden of prior approaches.
It uses a two-phase pipeline: syntactic exploration to collect diverse model outputs for learning, then constraint exploitation to enforce the learned rules during generation.
Experiments indicate the method achieves perfect constraint adherence even with small 1B-parameter LLMs, while reportedly outperforming larger models and some state-of-the-art reasoning systems.
The authors claim a first integration of context-sensitive grammar learning directly with LLM generation, aiming to preserve generation validity without hand-crafted constraints.

Abstract

Controlling the output of Large Language Models (LLMs) through context-sensitive constraints has emerged as a promising approach to overcome the limitations of Context-Free Grammars (CFGs) in guaranteeing generation validity. However, such constraints typically require manual specification -- a significant barrier demanding specialized expertise. We introduce a framework that automatically learns context-sensitive constraints from LLM interactions through a two-phase process: syntactic exploration to gather diverse outputs for constraint learning, followed by constraint exploitation to enforce these learned rules during generation. Experiments demonstrate that our method enables even small LLMs (1B parameters) to learn and generate with perfect constraint adherence, outperforming larger counterparts and state-of-the-art reasoning models. This work represents the first integration of context-sensitive grammar learning with LLM generation, eliminating manual specification while maintaining generation validity.