HELIX: Model-Harness Co-evolution for Recursive Self-Improvement

2026-08-14

AI Summary

This study addresses the disconnection between agent frameworks and model capabilities by proposing a novel model-framework co-evolution paradigm. We construct a traceable evolutionary foundation that leverages typed port decomposition, atomic component reuse, and recipe orchestration to generate verified trajectories. These trajectories drive recursive model self-improvement and establish a closed-loop execution optimization mechanism. Experimental results on SWE-bench demonstrate that this approach increases code repair coverage by 4.0% and improves composite strategy performance by 58.0%. Furthermore, the method yields 438 high-quality training samples, effectively validating the superiority of the proposed co-evolutionary mechanism in bridging the gap between architectural design and model intelligence.

Abstract

Scaling agent capability has largely focused on improving the model, yet an interactive agent acts through a runtime harness that mediates context, tools, control flow, and stopping. The harness shapes both what a model can accomplish and the trajectories from which it learns. This coupling motivates model-harness co-evolution for recursive self-improvement: build harnesses for a fixed model, update the model from verified sibling trajectories, and rebuild the harnesses as model capabilities change. Realizing this loop requires a controlled way to evolve harnesses while preserving intervention identity and effect. We present HELIX, a source-traceable substrate for harness evolution. HELIX decomposes agent systems into typed ports, reusable atoms, recipes, product shells, and runtime policies. It makes interventions explicit and auditable while retaining trajectories, test outcomes, and provenance. Harness evolution thus serves two linked roles: improving fixed-model execution and producing matched successes, regressions, near misses, and alternative solutions as data for subsequent model improvement. We evaluate HELIX in one evolution round on code repair. A 65-candidate portfolio discovers a fixed harness that improves task coverage by 4.0% over Pi, while the full portfolio exposes up to 58.0% more verified coverage through complementary sibling behavior. Selected candidates are assessed with repeated runs and the SWE-bench evaluator. A 200-slot sibling slice yields 438 verified SFT, critic, filter, and preference records. These results show how harness, model, and data form a feedback system: harness evolution expands current capability and creates learning signal for the next model; model updates motivate the next round of harness evolution. HELIX provides an auditable interface for studying this recursive process. Code is available at https://github.com/HKUDS/HELIX.

Problem

• Model-Harness Co-evolution
• Recursive Self-Improvement
• Agent Harness
• Traceability

Innovation

• Model-Harness Co-evolution
• Recursive Self-Improvement
• Source-traceable Substrate
• Agent System Decomposition
• Feedback Loop