XGrammar-2: Efficient Dynamic Structured Generation Engine for Agentic LLMs

TL;DR

XGrammar-2 is a novel structured generation engine that significantly improves efficiency for dynamic agentic workloads in LLMs by supporting flexible structure switching and cache reuse.

Contribution

It introduces TagDispatch and Cross-Grammar Cache for dynamic structure management and reuse, enabling faster and more efficient structured generation.

Findings

Over 6x faster compilation than prior engines

Near-zero end-to-end overhead in LLM serving systems

Effective support for dynamic structure switching and reuse

Abstract

Modern LLM agents increasingly rely on dynamic structured generation, such as tool calling and response protocols. Unlike traditional structured generation with static structures, these workloads vary both across requests and within a request, posing new challenges to existing engines. We present XGrammar-2, a structured generation engine for dynamic agentic workloads. Our design is based on two key ideas: first-class support for tag-triggered structure switching, and fine-grained reuse across requests with different output structures. Concretely, XGrammar-2 introduces TagDispatch for dynamic structural dispatching and Cross-Grammar Cache for substructure-level cache reuse across grammars. It further improves efficiency with an Earley-based adaptive token mask cache, just-in-time compilation, and repetition state compression. Experiments show that XGrammar-2 achieves over 6x faster compilation than prior structured generation engines, and incurs near-zero end-to-end overhead in modern LLM serving systems.