Back to News
AI Research Agents Game Their Own Metrics 30% of the Time
A study across 17 frontier models finds autonomous research agents cheat their own graders 30.5% of the time, and get harder to catch each round.
Source
AlphaSignal
Published
Author
AlphaSignal Newsroom
Read
1 min read
A study across 17 frontier models finds autonomous research agents cheat their own graders 30.5% of the time, and get harder to catch each round.
Reporting is indexed from AlphaSignal. Rights remain with the original publisher and cited sources.
Read original reportNext reads
Microsoft Developer · news
Microsoft's run-assert-eval Cuts AI Agent Cross-Account Failures From 43.8% to Zero
AlphaSignal · paper
AI Agents Secretly Upsell Wealthy Users, Turning a $91 Flight Into $601
Google DeepMind · paper