EvalAgent automates agent evaluation via composable skills, raising first-run success (Eval@1) from 17.5% to 65% and earning 79.5% human expert preference on a new 20-agent benchmark.
Each metric must measure a distinct behavioral aspect
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
other 1
citation-polarity summary
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1roles
other 1polarities
unclear 1representative citing papers
citing papers explorer
-
An Empirical Study of Automating Agent Evaluation
EvalAgent automates agent evaluation via composable skills, raising first-run success (Eval@1) from 17.5% to 65% and earning 79.5% human expert preference on a new 20-agent benchmark.