Preprint
Oct 2026
VeriEvolve: Evaluator-First Evolution for Tool-Using Agents
A central obstacle to self-improving agents is the absence of a reliable fitness signal for realistic stateful tasks. We introduce \textsc{VeriEvolve}, an evaluator-first evolution framework that constructs this missing evaluation layer before optimizing the agent. From task instruction, VeriEvolve synthesizes instruct...
Yue-Shen Li, K. Yüksel, Gokhan Tur et al.
· 0 citations