8 papers
Spatially Grounded Concept-Based Image Classification
Ran Eisenberg, Amit Rozner, Ethan Fetaya +1
Deep neural networks can achieve high accuracy while relying on evidence that is hard to inspect or misaligned with the intended task. Concept Bottleneck Models (CBMs) expose human…
Learning Permutation from Structure Without Supervision
Ran Eisenberg, Ofir Lindenbaum
Many learning problems require uncovering a hidden ordering that reveals structure in unordered data, such as monotonicity in sorting or spatial continuity in jigsaw reconstruction…
Train Less, Infer Faster: Efficient Model Finetuning and Compression via Structured Sparsity
Jonathan Svirsky, Yehonathan Refael, Ofir Lindenbaum
Fully finetuning foundation language models (LMs) with billions of parameters is often impractical due to high computational costs, memory requirements, and the risk of overfitting…
Unsupervised Feature Selection Through Group Discovery
Shira Lifshitz, Ofir Lindenbaum, Gal Mishne +2
Unsupervised feature selection (FS) is essential for high-dimensional learning tasks where labels are not available. It helps reduce noise, improve generalization, and enhance inte…
SUMO: Subspace-Aware Moment-Orthogonalization for Accelerating Memory-Efficient LLM Training
Yehonathan Refael, Guy Smorodinsky, Tom Tirer +1
Low-rank gradient-based optimization methods have significantly improved memory efficiency during the training of large language models (LLMs), enabling operations within constrain…
No Prior, No Leakage: Revisiting Reconstruction Attacks in Trained Neural Networks
Yehonatan Refael, Guy Smorodinsky, Ofir Lindenbaum +1
The memorization of training data by neural networks raises pressing concerns for privacy and security. Recent work has shown that, under certain conditions, portions of the traini…