EvolveNet: Collaborative Harness Evolution for Agent Self-Improvement

2026-08-05

AI Summary

This work addresses the limitation of existing large language model agent execution frameworks, which rely on centralized experience collection and thus struggle to adapt to real-world deployment environments that are distributed, heterogeneous, and subject to data isolation. The authors propose a collaborative framework evolution paradigm: a shared execution framework is deployed across local nodes, where it independently evolves under each node’s unique workload, and only program-level adaptation outcomes—not raw data—are uploaded for aggregation and redistribution. The core innovation lies in a novel program-level aggregation mechanism guided by scope types and evidence, enabling non-parametric, composable collaborative optimization. Evaluated across five domains—text-to-SQL, data science coding, competitive programming, software engineering, and agent workflows—the approach significantly enhances framework performance, particularly under heterogeneous workloads, demonstrating the efficacy of cross-agent adaptive composition.

Abstract

The capabilities of an LLM agent depend not only on its model but on the harness: the executable program that constructs context, invokes tools, verifies results, and recovers from failure. Recent work shows that evolving the harness yields persistent improvements without updating model weights. Existing approaches, however, assume that all execution experience can be routed to a single optimizer, which evolves one harness along a sequential trajectory. Real agent ecosystems violate that assumption: users, organizations, and environments generate isolated streams of experience that cannot be pooled, so the experience most worth learning from is exactly the experience that cannot be directly centralized. We introduce EvolveNet, a paradigm of collaborative harness evolution that moves experience extraction to the data. A shared harness is broadcast to data-local agent deployments, each of which evolves it on its own workload. Only the resulting program adaptations are composed into an updated shared harness and redistributed, so that every participating agent inherits operational experience discovered by the others. By shifting the aggregation boundary from raw workloads to learned adaptations, EvolveNet keeps workloads local and allows multiple evolutionary searches to proceed concurrently with reduced serial depth. Because independently modified programs cannot be averaged like model parameters and may conflict when composed, EvolveNet introduces scope-typed, evidence-guided program aggregation. Across five settings spanning text-to-SQL, data-science coding, competitive programming, software engineering, and agentic workflows, EvolveNet improves the shared harness in all five, with the largest gains under heterogeneous workloads, and ablations attribute the improvement to composition of adaptations from different agents rather than to selecting among them.

Problem

• harness evolution
• agent self-improvement
• decentralized experience
• collaborative learning
• program adaptation

Innovation

• collaborative evolution
• harness optimization
• program aggregation
• federated agent learning
• scope-typed composition