Newest papers

VeriEvolve: Evaluator-First Evolution for Tool-Using Agents

Yueshen Li, Kamer Ali Yüksel, Gökhan Tür, Dilek Z. Hakkani-Tur, João Pedro Maia · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.29005 · cited by 0 works (OpenAlex) · open-access PDF

Indonesian

Salah satu hambatan utama bagi agen self-improving adalah kurangnya sinyal kebugaran yang dapat diandalkan untuk tugas-tugas yang realistis. Kami memperkenalkan \textsc{VeriEvolve}, kerangka kerja evolusi evaluator pertama yang membangun lapisan evaluasi yang hilang ini sebelum mengoptimalkan agen. Dari instruksi tugas, VeriEvolve mensintesis keadaan tantangan berbasis instruksi dan evaluator yang dapat dieksekusi yang memeriksa mereka. Selama setiap putaran evolusi alur kerja, evaluator dibekukan dan berfungsi sebagai sinyal kebugaran. Di antara putaran, VeriEvolve meninjau evaluator dengan mengkonversi keadaan akhir dari limo-limo yang dihasilkan oleh eksekutor. Evaluator untuk penelitian eksekutif yang digunakan untuk pencarian dapat mengevaluasi kembali semua 30 tugas MCPM Filarkes, VeriEvolve membangun 30

Machine translation (Meta M2M100); the English original follows.

Abstract

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 instruction-grounded challenge states and an executable evaluator that checks them. During each round of workflow evolution, the evaluator is frozen and serves as the fitness signal. Between rounds, VeriEvolve revises the evaluator by converting the end states of strong and weak runs into a pair of counterfactual test cases. Across all 30 MCPMark Filesystem tasks, VeriEvolve constructs 30 evaluators with 141 challenge states, achieving 141/141 agreement on the generated suites and 113/122 agreement on held-out states produced by executor agents. Evaluator-guided evolution increases externally verified success from 6/30 to 18/30 for GLM-5.2 and from 27/30 to 30/30 for Claude; refinement removes all three false positives that the weak executor exposes. Cross-environment studies on PostgreSQL, Playwright, Notion, and GitHub further expose complementary limits from observability, state representation, and search/execution capacity. These results show that executable evaluators can be synthesized and used to guide search when task success is observable and an external verifier is available for calibration.

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Executor, Executable, Computer science, Task (project management), Counterfactual thinking, Stateful firewall

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange