6 papers
Data-Aware Random Feature Kernel for Transformers
Amirhossein Farzam, Hossein Mobahi, Nolan Andrew Miller +1
Transformers excel across domains, yet their quadratic attention complexity poses a barrier to scaling. Random-feature attention, as in Performers, can reduce this cost to linear i…
How new data permeates LLM knowledge and how to dilute it
Chen Sun, Renat Aksitov, Andrey Zhmoginov +5
Large language models learn and continually learn through the accumulation of gradient-based updates, but how individual pieces of new information affect existing knowledge, leadin…
Long Context In-Context Compression by Getting to the Gist of Gisting
Aleksandar Petrov, Mark Sandler, Andrey Zhmoginov +2
Long context processing is critical for the adoption of LLMs, but existing methods often introduce architectural complexity that hinders their practical adoption. Gisting, an in-co…
Learning and Unlearning of Fabricated Knowledge in Language Models
Chen Sun, Nolan Andrew Miller, Andrey Zhmoginov +2
What happens when a new piece of knowledge is introduced into the training data and how long does it last while a large language model (LM) continues to train? We investigate this…
Uncovering mesa-optimization algorithms in Transformers
Johannes von Oswald, Maximilian Schlegel, Alexander Meulemans +10
Some autoregressive models exhibit in-context learning capabilities: being able to learn as an input sequence is processed, without undergoing any parameter changes, and without be…
Narrowing the Focus: Learned Optimizers for Pretrained Models
Gus Kristiansen, Mark Sandler, Andrey Zhmoginov +4
In modern deep learning, the models are learned by applying gradient updates using an optimizer, which transforms the updates based on various statistics. Optimizers are often hand…