Title: Efficient Finetuning of Large Language Models via Large-WidthAnalysis Abstract: Finetuning Large Language Models (LLMs) enhances their performance on downstream tasks — a desirable outcome if the model is used for a specific task.Parameter-efficient finetuning methods such as LoRA (Low-Rank Adaptation) are popular because they allow finetuning large models with relatively low cost. When using […]

Title: Learning normalized probability models with dual score matching Abstract: Learning probability models from data is at the heart of many learning tasks. We introduce a new framework for learning normalized energy (log probability) models inspired from diffusion generative models. The energy model is fitted to data by two "score matching" objectives: the first constrains […]

Title: What Does Adam Really Need? Understanding Second Moment Compressibility Through Training Dynamics Abstract: Adam is the go-to optimizer for training modern machine learning models, but it requires additional memory to maintain the moving averages of the gradients and their squares. Despite its success, a fundamental question remains unanswered: When can Adam's second moments be […]

Title: The Three P's of Modern Computer Vision: Pixels, Perception and Physics Abstract: For decades, computer vision has been defined by three grand challenges, elegantly summarized by Jitendra Malik et al as the "Three R's": Recognition (what is it?), Reconstruction (what is its 3D shape?), and Reorganization (what belongs together?). While this framework has guided […]

Title: Mitigating the Curse of Detail: Scaling Arguments for Feature Learning and Sample Complexity Abstract: Two pressing topics in the theory of deep learning are the interpretation of feature learning mechanisms and the determination of the implicit bias of networks in the rich regime. Current theories of rich feature learning effects revolve around networks with […]