Harness-Search: Guiding Long-Horizon Search through Multi-Agent Coordination

2026-10-04

AI Summary

This study addresses the challenges of policy drift, evidence omission, and premature termination in long-horizon search caused by history accumulation in single-agent systems. To this end, it proposes Harness-Search, a multi-agent framework that innovatively decouples a monolithic policy into three privilege-restricted collaborative roles: a retrieval proposer, a state updater, and a summary auditor. By establishing a propose-submit-audit closed-loop mechanism integrated with persistent state management and explicit evidence sufficiency verification, the framework effectively mitigates local error propagation. Experimental results demonstrate that Harness-Search significantly enhances performance across seven benchmarks, improving recall by 4.60–27.92 points and final answer recall by 12.34–30.13 points, while effectively reducing redundant retrievals as the search deepens.

Abstract

Long-horizon search requires agents to gather evidence across multiple steps and synthesize it into well-supported answers. The recent agent harnesses provide a natural and promising framework to support such long-running search processes. As interaction histories grow, one single agent in harnesses might get stuck and cause the policy to lose track of unresolved questions, overlook useful evidence, or terminate before sufficient support has been collected. One of promising way is to decouple three distinct responsibilities of proposing retrieval actions, updating persistent state, and deciding when to stop rather than concentrating them within a single policy. Targeted at it, we introduce Harness-Search, a multi-agent search harness to reduce the local errors propagating across subsequent exploration, evidence curation, and termination decisions. In particular, Harness-Search assigns these responsibilities to three permission-bounded authorities: a Retrieval Policy that proposes search operations, a Memory Operator that validates and commits persistent-state updates, and a Summary Auditor that accepts or rejects termination based on the sufficiency of the curated evidence. Together, these roles form a Propose-Commit-Audit loop in which actions are proposed, persistent evidence is selectively committed, and stopping decisions are subjected to an explicit sufficiency check. Across seven long-horizon search benchmarks, Harness-Search improves both retrieval and answer generation under the same policy backbone, increasing Recall by 4.60-27.92 points and Final-Answer Recall by 12.34-30.13 points over the strongest harness-based baseline on each evidence-retrieval benchmark. Moreover, trajectory-level analyses show that Harness-Search continues to accumulate useful evidence and expand evidence coverage with less redundant retrieval as the search history grows.

Problem

• long-horizon search
• multi-agent coordination
• evidence retrieval
• agent harness
• error propagation

Innovation

• Multi-Agent Coordination
• Long-Horizon Search
• Propose-Commit-Audit Loop
• Agent Harness
• Evidence Curation