From Procedural Skills to Strategy Genes: Towards Experience-Driven Test-Time Evolution

TL;DR

This paper investigates how to effectively represent reusable experience for test-time control and iterative evolution, finding that compact Gene representations outperform documentation-oriented Skill packages in stability and effectiveness.

Contribution

It introduces and evaluates a Gene-based experience representation that is more stable, compact, and effective for control and evolution than traditional Skill packages.

Findings

Gene representation yields the strongest overall average control.

Gene remains competitive under structural perturbations.

Failure history in Gene improves iterative experience accumulation.

Abstract

This beta technical report asks how reusable experience should be represented so that it can function as effective test-time control and as a substrate for iterative evolution. We study this question in 4.590 controlled trials across 45 scientific code-solving scenarios. We find that documentation-oriented Skill packages provide unstable control: their useful signal is sparse, and expanding a compact experience object into a fuller documentation package often fails to help and can degrade the overall average. We further show that representation itself is a first-order factor. A compact Gene representation yields the strongest overall average, remains competitive under substantial structural perturbations, and outperforms matched-budget Skill fragments, while reattaching documentation-oriented material usually weakens rather than improves it. Beyond one-shot control, we show that Gene is also a better carrier for iterative experience accumulation: attached failure history is more effective in Gene than in Skill or freeform text, editable structure matters beyond content alone, and failure information is most useful when distilled into compact warnings rather than naively appended. On CritPt, gene-evolved systems improve over their paired base models from 9.1% to 18.57% and from 17.7% to 27.14%. These results suggest that the core problem in experience reuse is not how to supply more experience, but how to encode experience as a compact, control-oriented, evolution-ready object.