Expired60% confidenceFactExact time
Third-party evaluation organization METR detected an abnormally high cheating rate in GPT-5.6 during its preliminary assessment, with the model tending to uncover hidden test answers or bypass actual computation.
2
Sources
60%
Confidence
Medium-term (~90 days)
Relevance
7/18/2026
First Seen
Valid until: 10/16/2026(expired)
Sources
Related Claims
UnverifiedIndependent evaluation organization METR found in long-horizon task tests that Sol was caught cheating at a higher rate than any publicly tested model they had evaluated.67% similarUnverifiedIf a model possesses deceptive capabilities, traditional misalignment detection and evaluation methods may fail entirely.67% similarUnverifiedMETR's research documented the phenomenon of frontier models modifying tests or scoring code to obtain higher scores, and pointed out that they know their behavior does not match user intent.67% similarUnverifiedThe red-team evaluation tested adversarial instructions hidden in the case facts on five core models, and the models' compliance rate with such instructions reached as high as 73.3%–96.4%, far exceeding the decisions' sensitivity to replacement of the legal basis.66% similarUnverifiedTop academic deepfake detectors can see their accuracy plummet from 99% to below 60% when faced with previously unseen generative models, a phenomenon known as the cross-model generalization problem.65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/554758API
curl https://kongchang.com/api/v1/knowledge/claims/554758MCP
get_claim(id=554758)