Which Self-Improvements Should We Trust? Reliable Self-Improvement When Agents Reuse Their Benchmarks
Abstract
As recursive self-improvement (RSI) rapidly advances, reliable evaluation becomes critical for guiding adaptive search. RSI typically relies on finite evaluation resources, such as fixed benchmarks, to determine which modifications are retained and what is proposed next. However, when these finite resources are repeatedly reused, new candidates are proposed based on feedback from the same evaluation set, so the search trajectory can adaptively overfit and empirical improvement may not reflect genuine population improvement on the underlying task distribution. Some existing methods account for multiple comparisons but assume that candidates are chosen independently of the evaluation set, and therefore do not control this adaptive dependence. To address this, we propose REUSE (Risk-controlled Evaluation Under Sequential Evolution), a certified evaluation and promotion framework that allows a fixed evaluation set to support repeated adaptive decisions while providing statistical guarantees. For a user-specified error level α, with probability at least 1−α, every promoted modification is a genuine population improvement on the underlying task distribution. REUSE achieves this by strictly limiting the evaluation feedback returned to the search process and accounting for possible promotion histories within the error budget. We develop detailed statistical theory for RSI evaluation in this setting, including simultaneous error control, valid lower bounds on cumulative improvement, and a characterization of the fundamental limits of adaptive evaluation reuse. In live self-improvement experiments, REUSE commits substantially fewer false promotions than evaluation frameworks from current RSI systems and error-controlled baselines, reducing the proportion of false promotions from up to 20.7% to 0%, while achieving final true population performance comparable to the best baselines.
Related
- Spectral Improvements of Geometric Inequalities on Closed Kähler Manifolds
- Almost-orthogonal Strichartz estimates and radial improvements
- Distribution-Dependent Improvements of McDiarmid's Inequality via Rényi Divergence
- Improvements of Classical Eigenvalue Inequalities I: Quantitative Hersch-deficit and the Joint Spectral Range on the Two-Sphere
- Optimal Non-Homogeneous Improvements of the Anisotropic Hardy Inequality
BibTeX
@misc{sun2026which,
title = {{Which Self-Improvements Should We Trust? Reliable Self-Improvement When Agents Reuse Their Benchmarks}},
author = {Xiaojing Sun and Yuhan Zeng and Zihua She and Xiao Wang},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.33180},
eprint = {2609.33180},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.33180},
}
RIS
TY - UNPB
TI - Which Self-Improvements Should We Trust? Reliable Self-Improvement When Agents Reuse Their Benchmarks
AU - Xiaojing Sun
AU - Yuhan Zeng
AU - Zihua She
AU - Xiao Wang
PY - 2026
DA - 2026/09/29
PB - arXiv
DO - 10.48550/arxiv.2609.33180
UR - https://arxiv.org/abs/2609.33180
AB - As recursive self-improvement (RSI) rapidly advances, reliable evaluation becomes critical for guiding adaptive search. RSI typically relies on finite evaluation resources, such as fixed benchmarks, to determine which modifications are retained and what is proposed next. However, when these finite resources are repeatedly reused, new candidates are proposed based on feedback from the same evaluation set, so the search trajectory can adaptively overfit and empirical improvement may not reflect genuine population improvement on the underlying task distribution. Some existing methods account for multiple comparisons but assume that candidates are chosen independently of the evaluation set, and therefore do not control this adaptive dependence. To address this, we propose REUSE (Risk-controlled Evaluation Under Sequential Evolution), a certified evaluation and promotion framework that allows a fixed evaluation set to support repeated adaptive decisions while providing statistical guarantees. For a user-specified error level α, with probability at least 1−α, every promoted modification is a genuine population improvement on the underlying task distribution. REUSE achieves this by strictly limiting the evaluation feedback returned to the search process and accounting for possible promotion histories within the error budget. We develop detailed statistical theory for RSI evaluation in this setting, including simultaneous error control, valid lower bounds on cumulative improvement, and a characterization of the fundamental limits of adaptive evaluation reuse. In live self-improvement experiments, REUSE commits substantially fewer false promotions than evaluation frameworks from current RSI systems and error-controlled baselines, reducing the proportion of false promotions from up to 20.7% to 0%, while achieving final true population performance comparable to the best baselines.
ER -