Unverified50% confidenceTradeoffExact time
Using generative model rollouts as a data source carries a risk of distribution shift: the generative model's own physical priors may contain biases or simplifications, making the data unable to fully reflect the complex dynamics of the real world.
1
Sources
50%
Confidence
Long-term
Relevance
10/13/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTraditional autoregressive models typically require additional processing or approximation methods to obtain reliable probability distributions, giving diffusion models a structural advantage in this regard.74% similarUnverified扩散模型通过全局并行生成和迭代修正，能从根本上规避自回归模型的「曝光偏差」（Exposure Bias）累积错误问题73% similarUnverifiedMainstream generative models such as diffusion models and flow matching models are essentially probabilistic generators that produce new samples by sampling from data distributions.73% similarUnverifiedDomain Adaptation shows that a medium-sized model thoroughly fine-tuned on target distribution data can often outperform a large general-purpose model that has not seen such data on test sets from that distribution.72% similarUnverified领先模型若被用于生成训练数据，可能将自身偏差系统性地传递给下一代，即'模型坍塌'风险72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1013384API
curl https://kongchang.com/api/v1/knowledge/claims/1013384MCP
get_claim(id=1013384)