9 papers · 1 filter
Induction Heads Interpolate N-Grams
Francesco D'Angelo, Oguz Kaan Yuksel, Swathi Shree Narashiman +1
Induction heads are attention circuits believed to underlie in-context learning in transformers, yet a precise characterization of the estimators they implement remains elusive. We…
Learning What to Forget: Improving LLM Unlearning via Learned Token-Level Importance
Gizem Yüce, Giorgos Nikolaou, Nicolas Flammarion
Machine unlearning aims to remove targeted knowledge from a trained model while preserving its general capabilities. For autoregressive language models, not all tokens in a forget…
Transformers Learn Latent Mixture Models In-Context via Mirror Descent
Francesco D'Angelo, Nicolas Flammarion
Sequence modelling requires determining which past tokens are causally relevant from the context and their importance: a process inherent to the attention layers in transformers, y…
(How) Learning Rates Regulate Catastrophic Overtraining
Mark Rofin, Aditya Varre, Nicolas Flammarion
Supervised fine-tuning (SFT) is a common first stage of LLM post-training, teaching the model to follow instructions and shaping its behavior as a helpful assistant. At the same ti…
Gradient Flow Polarizes Softmax Outputs towards Low-Entropy Solutions
Aditya Varre, Mark Rofin, Nicolas Flammarion
Understanding the intricate non-convex training dynamics of softmax-based models is crucial for explaining the empirical success of transformers. In this article, we analyze the gr…
Incremental Learning of Sparse Attention Patterns in Transformers
Oğuz Kaan Yüksel, Rodrigo Alvarez Lucendo, Nicolas Flammarion
This paper studies simple transformers trained on a high-order Markov chain, where the model must incorporate information from multiple past positions, each with different statisti…
Exact Learning of Arithmetic with Differentiable Agents
Hristo Papazov, Francesco D'Angelo, Nicolas Flammarion
We explore the possibility of exact algorithmic learning with gradient-based methods and introduce a differentiable framework capable of strong length generalization on arithmetic…
Learning In-context n-grams with Transformers: Sub-n-grams Are Near-stationary Points
Aditya Varre, Gizem Yüce, Nicolas Flammarion
Motivated by empirical observations of prolonged plateaus and stage-wise progression during training, we investigate the loss landscape of transformer models trained on in-context…
Learning Algorithms in the Limit
Hristo Papazov, Nicolas Flammarion
This paper studies the problem of learning computable functions in the limit by extending Gold's inductive inference framework to incorporate \textit{computational observations} an…