Volkan Cevher

Matching Multiple Experts: On the Exploitability of Multi-Agent Imitation Learning

Diffusion-based Cumulative Adversarial Purification for Vision Language Models

Multi-modal Data Mixtures for Vision-Language Model Training

Space Filling Curves as Spatial Priors for Small or Data-Scarce Vision Transformers

Selective Rotary Position Embedding

ESLM: Risk-Averse Selective Language Modeling with Hierarchical Batch Selection

Alignment from Ranking and Rating Information

Learning Equilibria from Data: Provably Efficient Multi-Agent Imitation Learning

Ascent Fails to Forget

Generalization of noisy SGD in unbounded non-convex settings

Certified Robustness Under Bounded Levenshtein Distance

Efficient Large Language Model Inference with Neural Block Linearization

Linear Attention for Efficient Bidirectional Sequence Modeling

IL-SOAR : Imitation Learning with Soft Optimistic Actor cRitic

Addressing Label Shift in Distributed Learning via Entropy Regularization​

Generalized Gradient Norm Clipping & Non-Euclidean $(L_0,L_1)$-Smoothness

Generalization of noisy SGD under isoperimetry

Continuous-Time Analysis of Heavy Ball Momentum in Min-Max Games

Best of Both Worlds: Regret Minimization versus Minimax Play

Training Deep Learning Models with Norm-Constrained LMOs

Efficient Interpolation between Extragradient and Proximal Methods for Weak MVIs

Faster Inference of Flow-Based Generative Models via Improved Data-Noise Coupling

How Gradient descent balances features: A dynamical analysis for two-layer neural networks

Quantum-PEFT: Ultra parameter-efficient fine-tuning

Robustness in Both Domains: CLIP Needs a Robust Text Encoder

iFedDR: Auto-Tuning Local Computation with Inexact Douglas-Rachford Splitting in Federated Learning

Chameleon: A Flexible Data-mixing Framework for Language Model Pretraining and Finetuning

Quantized Optimistic Dual Averaging with Adaptive Layer-wise Compression

A Proximal Operator for Inducing 2:4-Sparsity

Multi-Step Preference Optimization via Two-Player Markov Games

Layer-wise Quantization for Quantized Optimistic Dual Averaging

LION: A bidirectional framework that trains like a Transformer and infers like an RNN

Imbalance-Regularized LoRA: A Plug-and-Play Method for Improving Fine-Tuning of Foundation Models

Adversarial Training for Defense Against Label Poisoning Attacks

A Phase Transition Induces Catastrophic Overfitting in Adversarial Training

Accelerating Spectral Clustering under Fairness Constraints

Space-time self-attention for graph signal processing

Leveraging the Context through Multi-Round Interactions for Jailbreaking Attacks

HeNCler: Node Clustering in Heterophilous Graphs via Learned Asymmetric Similarity

CONTINUAL FINITE-SUM MINIMIZATION UNDER THE POLYAK-ŁOJASIEWICZ CONDITION