arXiv:2607. 15693v1 Announce Type: cross Abstract: We describe a model of perceptual inference in primary visual cortex (V1) equivalent to a minimal diffusion model whose function can be readily understood from its parameters.
arXiv:2610.13997v1 Announce Type: cross Abstract: Concept representations in diffusion models drift across network depth and denoising time, yet existing sparse autoencoders offer only per-block or p...
Diffusion Trajectory Modeling (DTM) treats the evolving feature maps of diffusion models as temporally structured trajectories rather than static snapshots. By interpreting each spatial patch’s progression across multiple timesteps as a trajectory, DTM captures semantic correspondence cues that prior methods miss. Experiments on SPair-71k, SPair-U, and AP-10K demonstrate that DTM achieves strong performance, highlighting the semantic value embedded in the diffusion process’s temporal axis.
arXiv:2606. 31699v1 Announce Type: cross Abstract: Sparse autoencoders (SAEs) have recently been proposed as interpretable tools for concept-level manipulation, under the assumption that isolated features can serve as controllable intervention points.
arXiv:2609.37974v1 Announce Type: cross Abstract: Masked diffusion models (MDMs) generate text by unmasking several tokens per step, but they are trained and sampled under different conditions. The m...
arXiv:2606. 15796v1 Announce Type: cross Abstract: Mechanistic interpretability seeks to explain neural network behavior by decomposing model computations into interpretable features and circuits.
arXiv:2512. 20963v3 Announce Type: replace Abstract: Diffusion models excel at generating high-quality, diverse samples, yet they risk memorizing training data when overfit to the training objective.
arXiv:2603.17825v2 Announce Type: replace Abstract: In this work, we study the role of Massive Activations (MAs), which are rare, high-magnitude spikes confined to a few fixed hidden dimensions in vi...
arXiv:2606. 05328v1 Announce Type: cross Abstract: Modern video diffusion models generate increasingly realistic and temporally coherent videos, motivating their use as candidate world simulators.
arXiv:2609.37080v1 Announce Type: new Abstract: Latent Diffusion Models (LDMs) typically adopt a two-stage pipeline: an auto-encoder (AE) is first pre-trained to define a latent space, then a diffusi...
EraseSAE introduces a surgical concept erasure method for text-to-video diffusion models, using sparse autoencoders to decompose activations into interpretable, monosemantic features. The framework employs a contrastive attribution mechanism to isolate concept-specific kernels and applies timestep-resolved masks during inference to remove target concepts while preserving unrelated content. Experiments show that EraseSAE achieves precise, robust concept removal with minimal quality loss, outperforming existing methods.
arXiv:2605.29858v3 Announce Type: replace Abstract: Temporal action localization (TAL) requires recognizing the target event and localizing its start and end times precisely in untrimmed videos. Rece...