CodeCRDT: Observation-Driven Coordination for Multi-Agent LLM Code Generation

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autor principal: Pugachev, Sergey
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866909863104217088
author Pugachev, Sergey
author_facet Pugachev, Sergey
contents Multi-agent LLM systems fail to realize parallel speedups due to costly coordination. We present CodeCRDT, an observation-driven coordination pattern where agents coordinate by monitoring a shared state with observable updates and deterministic convergence, rather than explicit message passing. Using Conflict-Free Replicated Data Types (CRDTs), CodeCRDT enables lock-free, conflict-free concurrent code generation with strong eventual consistency. Evaluation across 600 trials (6 tasks, 50 runs per mode) shows both benefits and trade-offs: up to 21.1% speedup on some tasks, up to 39.4% slowdown on others, and 100% convergence with zero merge failures. The study formalizes observation-driven coordination for stochastic LLM agents, revealing semantic conflict rates (5-10%) and quality-performance tradeoffs, and provides empirical characterization of when parallel coordination succeeds versus fails based on task structure.
format Preprint
id arxiv_https___arxiv_org_abs_2510_18893
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CodeCRDT: Observation-Driven Coordination for Multi-Agent LLM Code Generation
Pugachev, Sergey
Distributed, Parallel, and Cluster Computing
Artificial Intelligence
Software Engineering
I.2.11; D.2.11
Multi-agent LLM systems fail to realize parallel speedups due to costly coordination. We present CodeCRDT, an observation-driven coordination pattern where agents coordinate by monitoring a shared state with observable updates and deterministic convergence, rather than explicit message passing. Using Conflict-Free Replicated Data Types (CRDTs), CodeCRDT enables lock-free, conflict-free concurrent code generation with strong eventual consistency. Evaluation across 600 trials (6 tasks, 50 runs per mode) shows both benefits and trade-offs: up to 21.1% speedup on some tasks, up to 39.4% slowdown on others, and 100% convergence with zero merge failures. The study formalizes observation-driven coordination for stochastic LLM agents, revealing semantic conflict rates (5-10%) and quality-performance tradeoffs, and provides empirical characterization of when parallel coordination succeeds versus fails based on task structure.
title CodeCRDT: Observation-Driven Coordination for Multi-Agent LLM Code Generation
topic Distributed, Parallel, and Cluster Computing
Artificial Intelligence
Software Engineering
I.2.11; D.2.11
url https://arxiv.org/abs/2510.18893