TestSuiteBench: Coding Agents Fail to Write Bug-Catching Tests

ETH researchers released TestSuiteBench, a benchmark where coding agents generate test suites from specs of 165 real Rust/Go/C/C++ apps. Results show frontier agents miss over half of logic defects, revealing a major weakness in test generation.

2026-10-12 ~ 2026-10-13 · 2 related posts

• TestSuiteBench: coding agents' tests catch fewer than 55% of logical bugs — OfirPress · 2026-10-12
• TestSuiteBench: coding agents kill only ~55% of bugs with generated test suites — lbeurerkellner · 2026-10-13