When Parallelism Pays Off: Cohesion-Aware Task Partitioning for Multi-Agent Coding

2026-05-30

AI Summary

This work addresses the challenge in multi-agent parallel code generation where communication overhead often offsets the gains from parallelization. To mitigate this, the authors propose a cohesion-aware multi-agent orchestration mechanism that constructs a dependency graph via static code analysis to identify structurally central files. Leveraging community detection and graph partitioning, the approach enables collaborative task decomposition that balances computational parallelism with communication costs, guided by a dependency-aware scheduler. Evaluated on 28 real-world tasks, the proposed method achieves up to a 14.0% improvement in task success rate, a 2.10× speedup, and a 35% reduction in API call costs compared to baseline approaches.

Abstract

Multi-agent Large Language Model (LLM) systems offer a way to decompose complex tasks, such as coding, through parallelization and context isolation. However, adding agents in practice introduces inter-agent communication overhead, which incurs extra cost and can sometimes offset the efficiency gains. We formalize multi-agent orchestration as a graph partitioning problem that captures the communication-to-computation trade-off: task decomposition can shorten critical-path computation, but cross-agent dependencies require costly context transfer. We instantiate this view in repository-level software engineering and present Cohesion-aware Coder (Co-Coder), which builds dependency graphs from static analysis, isolates structural hub files, partitions the graph via community detection, and executes the partition with a dependency-aware scheduler. Across 28 real-world tasks on DevEval and CodeProjectEval, Co-Coder advances the Pareto-frontier over sequential and file-based parallel baselines as well as Claude Code with Agent Teams, lifting pass rate by up to 14.0%, achieving up to a 2.10x wall-clock speedup, and reducing API cost by up to 35%, with the largest gains on the most dependency-dense projects. Co-coder demonstrates how cohesion-aware orchestration can make parallel coding agents both theoretically grounded and practically efficient, suggesting a broader design principle for multi-agent systems.

Problem

• multi-agent LLM
• task partitioning
• communication overhead
• parallel coding
• cohesion-aware orchestration

Innovation

• multi-agent LLM
• cohesion-aware partitioning
• graph partitioning
• dependency-aware scheduling
• parallel coding