1

Navigating the Latent Space Dynamics of Neural Models
Neural networks transform high-dimensional data into compact, structured representations, often modeled as elements of a lower …
Cite arXiv ICLR 2026 Oral
Navigating the Latent Space Dynamics of Neural Models
Language Models are Injective and Hence Invertible
Transformer components such as non-linear activations and normalization are inherently non-injective, suggesting that different inputs …
Cite arXiv Thread GitHub
Language Models are Injective and Hence Invertible
Implicit Inversion turns CLIP into a Decoder
CLIP is a discriminative model trained to align images and text in a shared embedding space. Due to its multimodal structure, it serves …
Cite arXiv GitHub
Implicit Inversion turns CLIP into a Decoder
MASS: MoErging through Adaptive Subspace Selection
Model merging has recently emerged as a lightweight alternative to ensembling, combining multiple fine-tuned models into a single set …
Cite arXiv
MASS: MoErging through Adaptive Subspace Selection
Video Unlearning via Low-Rank Refusal Vector
Video generative models democratize the creation of visual content through intuitive instruction following, but they also inherit the …
Cite arXiv GitHub
Video Unlearning via Low-Rank Refusal Vector
EuleroDec: A Complex-Valued RVQ-VAE for Efficient and Robust Audio Coding
Audio codecs power discrete music generative modelling, music streaming and immersive media by shrinking PCM audio to …
Cite arXiv
EuleroDec: A Complex-Valued RVQ-VAE for Efficient and Robust Audio Coding
Model Merging Improves Zero-Shot Generalization in Bioacoustic Foundation Models
Foundation models capable of generalizing across species and tasks represent a promising new frontier in bioacoustics, with NatureLM …
Cite arXiv URL GitHub Spotlight (top 15%)
Model Merging Improves Zero-Shot Generalization in Bioacoustic Foundation Models
Two-Scale Latent Dynamics for Recurrent-Depth Transformers
Recurrent-depth transformers scale test-time compute by iterating latent computations before emitting tokens. We study the geometry of …
Cite arXiv
Two-Scale Latent Dynamics for Recurrent-Depth Transformers
Membership and Dataset Inference Attacks on Large Audio Generative Models
Generative audio models, based on diffusion and autoregressive architectures, have advanced rapidly in both quality and expressiveness. …
Cite
Membership and Dataset Inference Attacks on Large Audio Generative Models
Implicit-ARAP: Efficient Handle-Guided Neural Field Deformation via Local Patch Meshing
Neural fields have emerged as a powerful representation for 3D geometry, enabling compact and continuous modeling of complex shapes. …
Cite arXiv GitHub
Implicit-ARAP: Efficient Handle-Guided Neural Field Deformation via Local Patch Meshing