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

Hindi

स्वयं सुधार एजेंटों के लिए एक केंद्रीय बाधा वास्तविक राज्य कार्यों के लिए एक विश्वसनीय फिटनेस सिग्नल की अनुपस्थिति है। हम \textsc{VeriEvolve} को पेश करते हैं, एक कार्य-प्रथम विकास ढांचा जो एजेंट को अनुकूलित करने से पहले इस गायब मूल्यांकन परत का निर्माण करता है। कार्य निर्देश से, VeriEvolve निर्देश-आधारित चुनौती राज्यों और एक निष्पादित मूल्यांकक को संश्लेषित करता है जो उन्हें चेक करता है। कार्य प्रवाह के प्रत्येक दौर के विकास के दौरान, मूल्यांकक को फ्रीज किया जाता है और फिटनेस सिग्नल के रूप में कार्य करता है। दौरों के बीच, VeriEvolve मूल्यांकक को मजबूत और कमजोर कार्यों के अंत राज्यों को एक जोड़े में बदल

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