[KongchangAI]
ConceptHadamard矩阵旋转

Hadamard Rotation

A data-independent fixed orthogonal transform based on the Hadamard matrix. It is computationally simple and requires no calibration data, but is insensitive to the activation distribution of a specific model; it is a special case within the SpectrumQuant spectral alignment framework.

Timeline (last 90 days)

Oct 15

Existing orthogonal transform methods fall roughly into two categories: data-independent fixed transforms (such as the Hadamard rotation) and rotation matrices learned during the calibration stage (such as SpinQuant and DartQuant).

Unverified50%

All Facts (1)

Source Articles