AUG 18, 2026 · PREPRINT
On the Fragility of Self-Improving Agents: Variance, Task Order, and Underspecification
arXiv
A re-evaluation study of existing methods identifying robustness issues through variance and task-order testing, without proposing a validated solution; raises important methodological concerns but lacks the definitive evidence needed for practice change.
Study details
DesignEmpirical re-evaluation with variance quant…
InterventionIncorporation of detailed rubrics and envir…
ComparatorBaseline memory construction without enhanc…