New benchmark TestEvo-Bench evaluates AI agents on code and test co-evolution

Researchers have introduced TestEvo-Bench, a new benchmark designed to evaluate AI agents on their ability to co-evolve tests with code changes. The benchmark includes tasks for generating new tests and updating existing ones, grounded in real commit histories and executable environments. TestEvo-Bench aims to reduce data leakage by periodically mining new tasks, with the current snapshot containing over 1200 tasks from 152 open-source Java projects. AI

IMPACT This benchmark could drive improvements in AI agents' ability to understand and generate code and tests, leading to more robust software development tools.

RANK_REASON The cluster contains an academic paper introducing a new benchmark for AI research. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• arXiv
• Claude Code
• Claude Opus 4.7
• Gemini 3.1 Pro
• Gemini CLI
• Java
• SWE-Agent
• TestEvo-Bench

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →