5 papers
A Scalable Measure of Loss Landscape Curvature for Analyzing the Training Dynamics of LLMs
Dayal Singh Kalra, Jean-Christophe Gagnon-Audet, Andrey Gromov +4
Understanding the curvature evolution of the loss landscape is fundamental to analyzing the training dynamics of neural networks. The most commonly studied measure, Hessian sharpne…
On the origin of neural scaling laws: from random graphs to natural language
Maissam Barkeshli, Alberto Alfarano, Andrey Gromov
Scaling laws have played a major role in the modern AI revolution, providing practitioners predictive power over how the model performance will improve with increasing data, comput…
PARQ: Piecewise-Affine Regularized Quantization
Lisa Jin, Jianhao Ma, Zechun Liu +3
We develop a principled method for quantization-aware training (QAT) of large-scale machine learning models. Specifically, we show that convex, piecewise-affine regularization (PAR…
The Unreasonable Ineffectiveness of the Deeper Layers
Andrey Gromov, Kushal Tirumala, Hassan Shapourian +2
How is knowledge stored in an LLM's weights? We study this via layer pruning: if removing a certain layer does not affect model performance in common question-answering benchmarks,…
Spectral Journey: How Transformers Predict the Shortest Path
Andrew Cohen, Andrey Gromov, Kaiyu Yang +1
Decoder-only transformers lead to a step-change in capability of large language models. However, opinions are mixed as to whether they are really planning or reasoning. A path to m…