Unverified50% confidenceFactExact time
The backpropagation implemented by VIDRAFT on the NPU is standard reverse-mode gradient computation, not a gradient-free alternative, but the technical mechanism has not been disclosed.
1
Sources
50%
Confidence
Long-term
Relevance
10/14/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe project does not use a spiking mechanism and still operates within the standard backpropagation/weight update framework, and is therefore distinct from neuromorphic computing at the hardware level.68% similarUnverifiedVIDRAFT completed full backpropagation on an inference-specific NPU, pretrained a language model from scratch, and released it as VKORE-46M-NPU.66% similarVerifiedBackpropagation is essentially a dynamic programming algorithm for efficiently computing gradients, which reduces the computational complexity of gradient calculation by propagating error signals layer by layer from the output layer to the input layer and reusing intermediate computation results.66% similarVerified神经网络的反向传播算法是链式法则的直接应用，梯度下降依赖微积分中的偏导数概念63% similarUnverifiedNormalizing Trajectory Models (NTM) model each step in the reverse process as an expressive conditional normalizing flow, rather than Gaussian denoising63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1032398API
curl https://kongchang.com/api/v1/knowledge/claims/1032398MCP
get_claim(id=1032398)