Type: url
Original: https://aclanthology.org/2026.findings-acl.1330/
Status: generated
Summary
The ACL Anthology landing page for “A Survey on Evaluation of LLM-based Agents” (Yehudai et al., Findings of ACL 2026), carrying the paper’s bibliographic metadata and its abstract. The paper surveys evaluation methods for LLM-based agents across five perspectives: core LLM capabilities for agentic workflows, application-specific benchmarks such as web and SWE agents, evaluation of generalist agents, the core dimensions of agent benchmarks, and evaluation frameworks and tools for agent developers. It reports a trend toward more realistic, continuously updated benchmarks and identifies gaps in assessing cost-efficiency, safety and robustness, and in fine-grained, scalable evaluation methods.