Optimizing Attention with Mirror Descent: Generalized Max-Margin Token Selectionby Addison Kristanto Julistiono, Davoud Ataee Tarzanagh,…
Fine-Tuning Discrete Diffusion Models via Reward Optimization with Applications to DNA and Protein Designby Chenyu…
Improving Discrete Optimisation Via Decoupled Straight-Through Gumbel-Softmaxby Rushi Shah, Mingyuan Yan, Michael Curtis Mozer, Dianbo…
AERO: Softmax-Only LLMs for Efficient Private Inferenceby Nandan Kumar Jha, Brandon ReagenFirst submitted to arxiv…
The Bayesian Confidence (BACON) Estimator for Deep Neural Networksby Patrick D. Kee, Max J. Brown,…
Beyond Linear Approximations: A Novel Pruning Approach for Attention Matrixby Yingyu Liang, Jiangxuan Long, Zhenmei…
Quadratic Gating Functions in Mixture of Experts: A Statistical Insightby Pedram Akbarian, Huy Nguyen, Xing…
TABCF: Counterfactual Explanations for Tabular Data Using a Transformer-Based VAEby Emmanouil Panagiotou, Manuel Heurich, Tim…
LoLCATs: On Low-Rank Linearizing of Large Language Modelsby Michael Zhang, Simran Arora, Rahul Chalamala, Alan…
Training Dynamics of Transformers to Recognize Word Co-occurrence via Gradient Flow Analysisby Hongru Yang, Bhavya…