Unverified50% confidenceConceptExact time
Distribution Shift refers to the fact that the behavior an Agent exhibits during training and controlled testing may not represent its behavior in a real deployment environment.
1
Sources
50%
Confidence
Long-term
Relevance
10/16/2026
First Seen
Sources
The AI Agent Security Crisis: More Than Just a Cybersecurity Problem
twitteryoshua_bengio10/5/2026
Related Entities
Related Claims
VerifiedBehavioral cloning suffers from the distribution shift problem: during training the policy only encounters the state distribution covered by expert trajectories, and during deployment the accumulation of small deviations causes the actual state distribution visited to gradually diverge from the training distribution.79% similarUnverifiedAsking the demonstrator to adjust their posture may seem like a low-cost compromise, but it quietly changes the distribution of the training data.65% similarUnverifiedEvaluating within the training distribution easily overestimates model capability, while out-of-distribution benchmarks require the model to generalize to unseen scene and interaction combinations, better exposing the gap between 'looking right' and 'being physically right'.65% similarUnverifiedDomain Shift refers to the problem where a model performs well on training data but degrades significantly after being deployed to a new environment65% similarUnverified针对特定工具做专项强化训练的副作用是对非训练分布工具的泛化能力下降，学术上称为分布外泛化失败63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1048992API
curl https://kongchang.com/api/v1/knowledge/claims/1048992MCP
get_claim(id=1048992)