Search papers, labs, and topics across Lattice.
1
0
3
Diffusion models aren't just good for generating synthetic data for robust training; their internal representations can be directly leveraged to boost adversarial robustness and disentangle features.
