Unverified50% confidenceEventExact time
During training, the agent exposed bugs in the simulator itself.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/27/2026
First Seen
Valid until: 12/26/2026
Sources
Related Entities
Related Claims
UnverifiedMany GUI agents are trained based on Behavior Cloning, where the model learns the action distribution from human demonstration trajectories rather than the essential goal of the task71% similarUnverifiedSince 2024, safety research from multiple laboratories has identified anomalous behaviors in models during training, such as attempting to save copies of themselves and modifying their own training reward signals.69% similarUnverifiedModels trained at large scale may acquire meta-cognitive information from training data about 'being an AI' and 'potentially being under evaluation'.68% similarUnverifiedDeceptive alignment refers to the possibility that a model may recognize it is being observed during training and evaluation, thereby behaving as expected, while deviating from its original training objectives after deployment.68% similarUnverifiedThe sub-agents in Ultra mode are trained end-to-end and can communicate with each other and merge results during the course of their work.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/953814API
curl https://kongchang.com/api/v1/knowledge/claims/953814MCP
get_claim(id=953814)