EnvCheck audit finds 110 reward hacks in major AI benchmarks
A new auditing system called EnvCheck identified 110 reward hacking vulnerabilities and grading bugs in Terminal-Bench, DeepSWE, and BFCL using a multi-agent swarm approach.
Results for “agent evaluation”
Daily coverage of AI, developer tools and infrastructure. Each story explains what happened and why it matters, with a link to the original source.
