Search papers, labs, and topics across Lattice.
This paper introduces a self-evolving agent-harness framework that enhances the performance of a frozen language model by improving the surrounding harness rather than the model weights themselves. By separating the proposal of changes from their evaluation, the framework ensures that credited improvements are reliable, utilizing a gated, categorical quality-diversity archive to address specific pathologies. The results demonstrate significant performance gains across seven domains, with improvements ranging from +9 to +15.5 percentage points, indicating effective generalization rather than mere overfitting.
A self-evolving framework achieves up to 15.5 percentage points in performance gains by intelligently refining the agent's harness without altering the underlying model.
An LLM agent's real-task performance is shaped as much by the harness around its model as by the frozen model itself: its prompts, injected knowledge, runtime control, and configuration. In deployment the harness is often the only lever available, so improving it automatically is the natural way to raise performance without touching the weights. The hard part is not generating changes but knowing which one truly helped. Self-generated feedback is noisy, and an apparent gain can be a measurement artifact or an edit that merely overfits the tasks it was tuned on. We present a self-evolving agent-harness framework that separates proposing changes from crediting them: a language model diagnoses failures and proposes patches, while all sampling, measurement, and significance testing are owned by deterministic code, so every credited improvement is trustworthy by construction. Patches populate a gated, categorical quality-diversity archive (GSME) keyed on the (WHERE x WHY) pathology an edit addresses rather than the tasks it fixes, an anti-overfitting inductive bias; generalization is measured on a sealed test scored only after evolution. Across seven domains with a frozen open-weight model, the harness is train-selected and scored once on a sealed test; its credited gains there are +9 to +15.5pp and retain 86-147% of the training gain, evidence they generalize rather than overfit. The winning patch tracks the model's dominant pathology, not its size or family: changing the model can change the pathology and the patch, while the same pathology-to-patch match recurs across two model families. What transfers is the diagnose-and-credit loop, not any specific harness.
