latent spaces
Latent spaces are abstract representations of data learned by models, where similar items are positioned closer together. In generative models like VAEs or GANs, these spaces help capture the essence of the data, enabling tasks like interpolation and generation of new samples.
- $\boldsymbol{\lambda}$-Orthogonality Regularization for Compatible Representation Learning
- Bridging Brains and Concepts: Interpretable Visual Decoding from fMRI with Semantic Bottlenecks
- Counterfactual Image Editing with Disentangled Causal Latent Space
- Latent Space Factorization in LoRA
- LeapFactual: Reliable Visual Counterfactual Explanation Using Conditional Flow Matching
- TADA: Improved Diffusion Sampling with Training-free Augmented DynAmics
- TimeWak: Temporal Chained-Hashing Watermark for Time Series Data
- Towards Unified and Lossless Latent Space for 3D Molecular Latent Diffusion Modeling