Publications

ZipIt!: Multitask Model Merging without Training
Cite OpenReview
What Mechanisms Does Knowledge Distillation Distill?
Cite OpenReview PMLR
WavSpA: Wavelet Space Attention for Boosting Transformers' Long Sequence Learning Ability
Cite OpenReview PMLR
Visual Expertise Explains Image Inversion Effects
Cite OpenReview PMLR
Variational Classification
Cite OpenReview
Unsupervised learning on spontaneous retinal activity leads to efficient neural representation geometry
Cite OpenReview PMLR
Universality of intrinsic dimension of latent representations across models
Cite OpenReview
UniCat: Crafting a Stronger Fusion Baseline for Multimodal Re-Identification
Cite OpenReview
Understanding Mode Connectivity via Parameter Space Symmetry
Cite OpenReview
Understanding Learning Dynamics of Neural Representations via Feature Visualization at Scale
Cite OpenReview
Towards Measuring Representational Similarity of Large Language Models
Cite OpenReview
Testing Assumptions Underlying a Unified Theory for the Origin of Grid Cells
Cite OpenReview
Supervising Variational Autoencoder Latent Representations with Language
Cite OpenReview PMLR
Sufficient conditions for offline reactivation in recurrent neural networks
Cite OpenReview
Subjective Randomness and In-Context Learning
Cite OpenReview
Soft Matching Distance: A metric on neural representations that captures single-neuron tuning
Cite OpenReview PMLR
SimVAE: Narrowing the gap between Discriminative & Generative Representation Learning
Cite OpenReview
SHARCS: Shared Concept Space forExplainable Multimodal Learning
Cite OpenReview
Semi-Ensemble: A Simple Approach Over-parameterize Model Interpolation
Cite OpenReview PMLR
SAM-CLIP: Merging Vision Foundation Models towards Semantic and Spatial Understanding
Cite OpenReview
Role Taxonomy of Units in Deep Neural Networks
Cite OpenReview PMLR
ReWaRD: Retinal Waves for Pre-Training Artificial Neural Networks Mimicking Real Prenatal Development
Cite OpenReview PMLR
Revisiting Supervision for Continual Representation Learning
Cite OpenReview
Representational constraints underlying similarity between task-optimized neural systems
Cite OpenReview
Randomly Weighted Neuromodulation in Neural Networks Facilitates Learning of Manifolds Common Across Tasks
Cite OpenReview PMLR
Predictive variational autoencoder for learning robust representations of time-series data
Cite OpenReview
On Transferring Expert Knowledge from Tabular Data to Images
Cite OpenReview PMLR
On the universality of neural codes in vision
Cite OpenReview
On the Robustness of Neural Collapse and the Neural Collapse of Robustness
Cite OpenReview
On the Direct Alignment of Latent Spaces
Cite OpenReview PMLR
On Feature Learning of Recursive Feature Machines and Automatic Relevance Determination
Cite OpenReview
On consequences of finetuning on data with highly discriminative features
Cite OpenReview
Object-Centric Semantic Vector Quantization
Cite OpenReview PMLR
NoPose-NeuS: Jointly Optimizing Camera Poses with Neural Implicit Surfaces for Multi-view Reconstruction
Cite OpenReview PMLR
NEUCORE: Neural Concept Reasoning for Composed Image Retrieval
Cite OpenReview PMLR
Multimodal decoding of human brain activity into images and text
Cite OpenReview PMLR
Multi-timescale reinforcement learning in the brain
Cite OpenReview
Model Merging by Gradient Matching
Cite OpenReview
Mixture of Multimodal Interaction Experts
Cite OpenReview
MeSa: Masked, Geometric, and Supervised Pre-training for Monocular Depth Estimation
Cite OpenReview PMLR
Linearly Structured World Representations in Maze-Solving Transformers
Cite OpenReview PMLR
Linear Mode Connectivity in Sparse Neural Networks
Cite OpenReview
Leveraging Self-Consistency for Data-Efficient Amortized Bayesian Inference
Cite OpenReview
Inverted-Attention Transformers can Learn Object Representations: Insights from Slot Attention
Cite OpenReview
Instruction-tuned LLMs with World Knowledge are More Aligned to the Human Brain
Cite OpenReview
Increasing Brain-LLM Alignment via Information-Theoretic Compression
Cite OpenReview
Implicit Representations for Image Segmentation
Cite OpenReview
Hybrid Early Fusion for Multi-Modal Biomedical Representations
Cite OpenReview
How Good is a Single Basin?
Cite OpenReview
How does fine-tuning affect your model? Mechanistic analysis on procedural tasks
Cite OpenReview
Growing Brains: Co-emergence of Anatomical and Functional Modularity in Recurrent Neural Networks
Cite OpenReview
Grokking as Simplification: A Nonlinear Complexity Perspective
Cite OpenReview
Event-Based Contrastive Learning for Medical Time Series
Cite OpenReview
Evaluation of Representational Similarity Scores Across Human Visual Cortex
Cite OpenReview
Estimating shape distances on neural representations with limited samples
Cite OpenReview
Efficient Multimodal Alignment: To Freeze or Not to Freeze?
Cite OpenReview
Ecological data and objectives align deep neural network representations with humans
Cite OpenReview
Duality of Bures and Shape Distances with Implications for Comparing Neural Representations
Cite OpenReview PMLR
Distributional Reinforcement Learning in the Mammalian Brain
Cite OpenReview
Disentangling Linear Mode Connectivity
Cite OpenReview
DisCoV: Disentangling Time Series Representations via Contrastive based $l$-Variational Inference
Cite OpenReview PMLR
Degradation and plasticity in convolutional neural networks: An investigation of internal representations
Cite OpenReview
Deep Multimodal Emotion Recognition using Modality Aware Attention Network for Unifying Representations in Neural Models
Cite OpenReview
Comparing Representational and Functional Similarity in Small Transformer Language Models
Cite OpenReview
Comparing neural models using their perceptual discriminability predictions
Cite OpenReview PMLR
CLIP meets Model Zoo Experts: Pseudo-Supervision for Visual Enhancement
Cite OpenReview
Characterizing pre-trained and task-adapted molecular representations
Cite OpenReview
Channel Vision Transformers: An Image Is Worth C x 16 x 16 Words
Cite OpenReview
Bio-inspired parameter reuse: Exploiting inter-frame representation similarity with recurrence for accelerating temporal visual processing
Cite OpenReview PMLR
Are Spiking Neural Networks more expressive than Artificial Neural Networks?
Cite OpenReview
An Information-Theoretic Understanding of Maximum Manifold Capacity Representations
Cite OpenReview
A sparse null code emerges in deep neural networks
Cite OpenReview PMLR
A General Method for Testing Bayesian Models using Neural Data
Cite OpenReview PMLR
A Compact Representation for Bayesian Neural Networks By Removing Permutation Symmetry
Cite OpenReview