Unverified50% confidenceFactExact time
In world models, the MDN-RNN typically uses an LSTM with 256 hidden units and 5 Gaussian mixture components to predict the distribution of the next frame's latent vector.
1
Sources
50%
Confidence
Long-term
Relevance
9/1/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLLM的每次调用可以被建模为伯努利试验，N次独立调用的通过次数服从二项分布B(N, p)67% similarUnverifiedBERT's MLM performs one-shot prediction and is primarily used for representation learning, whereas discrete diffusion models treat the masking ratio as a continuous variable from 0% to 100% to construct a complete Markov chain, achieving progressive generation through multi-step iteration.67% similarUnverifiedLLM自回归解码阶段每生成一个token都需从显存加载全量模型权重，批量较小时形成内存墙65% similarUnverifiedMTP adds multiple prediction heads to the model's output layer, enabling the model to simultaneously predict the probability distribution of the next N tokens in a single forward pass, achieving 2–4x inference acceleration without sacrificing generation quality.65% similarUnverifiedResearchers use graph neural networks (GNN) to learn neutronics simulations of breeding blankets, enabling sub-second prediction of TBR values for arbitrary combinations of geometric parameters with prediction errors controlled within 1%.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/838280API
curl https://kongchang.com/api/v1/knowledge/claims/838280MCP
get_claim(id=838280)