Citation copied to clipboard
Authorship
Venue Type
Incremental Learning of Sparse Attention Patterns in Transformers
Incremental Learning of Sparse Attention Patterns in Transformers
Oğuz Kaan Yüksel, Rodrigo Alvarez Lucendo, Nicolas Flammarion
In ICML International Conference on Machine Learning , 2026
Training Dynamics
Paper arXiv Slides
Induction Heads Interpolate N-Grams
Induction Heads Interpolate N-Grams
Francesco D'Angelo*Equal contribution, Oğuz Kaan Yüksel*Equal contribution, Swathi Shree Narashiman, Nicolas Flammarion
In ICML International Conference on Machine Learning , 2026
Mechanistic Interpretability
Paper
Generalization Bounds for Autoregressive Processes and In-Context Learning
Generalization Bounds for Autoregressive Processes and In-Context Learning
Oğuz Kaan Yüksel, Nicolas Flammarion
In PriGM@EurIPS Workshop on Principles of Generative Modeling , 2025
Generalization
Paper
Incremental Learning of Sparse Attention Patterns in Transformers
Incremental Learning of Sparse Attention Patterns in Transformers
Oğuz Kaan Yüksel, Rodrigo Alvarez Lucendo, Nicolas Flammarion
In PriGM@EurIPS Workshop on Principles of Generative Modeling , 2025
Training Dynamics
Paper
On the Sample Complexity of Next-Token Prediction
On the Sample Complexity of Next-Token Prediction
Oğuz Kaan Yüksel, Nicolas Flammarion
In AISTATS Annual Conference on Artificial Intelligence and Statistics , 2025
Generalization
Paper
Long-Context Linear System Identification
Long-Context Linear System Identification
Oğuz Kaan Yüksel, Mathieu Even, Nicolas Flammarion
In ICLR International Conference on Learning Representations , 2025
Generalization
Paper
First-order ANIL provably learns representations despite overparametrization
First-order ANIL provably learns representations despite overparametrization
Oğuz Kaan Yüksel, Etienne Boursier, Nicolas Flammarion
In ICLR International Conference on Learning Representations , 2024
Meta Learning Training Dynamics
Paper
LatentCLR: A Contrastive Learning Approach for Unsupervised Discovery of Interpretable Directions
LatentCLR: A Contrastive Learning Approach for Unsupervised Discovery of Interpretable Directions
Oğuz Kaan Yüksel*Equal contribution, Enis Simsar*Equal contribution, Ezgi Gülperi Er, Pınar Yanardağ
In ICCV Proceedings of the IEEE International Conference on Computer Vision , 2021
Computer Vision
Paper Code Talk Website
Semantic Perturbations with Normalizing Flows for Improved Generalization
Semantic Perturbations with Normalizing Flows for Improved Generalization
Oğuz Kaan Yüksel, Sebastian Urban Stich, Martin Jaggi, Tatjana Chavdarova
In ICCV Proceedings of the IEEE International Conference on Computer Vision , 2021
Computer Vision
Paper Code Talk
Semantic Perturbations with Normalizing Flows for Improved Generalization
Semantic Perturbations with Normalizing Flows for Improved Generalization
Oğuz Kaan Yüksel, Sebastian Urban Stich, Martin Jaggi, Tatjana Chavdarova
In INNF+ Workshop, ICML ICML Workshop on Invertible Neural Networks, Normalizing Flows, and Explicit Likelihood Models , 2021
Computer Vision
Paper Code
[Re] Can gradient clipping mitigate label noise?
[Re] Can gradient clipping mitigate label noise?
David Mizrahi, Oğuz Kaan Yüksel, Aiday Marlen Kyzy
In ReScience C ReScience C , 2021
Regularization
Paper Code
Adversarial Training with Orthogonal Regularization
Adversarial Training with Orthogonal Regularization
Oğuz Kaan Yüksel, İnci Meliha Baytaş
In SIU Signal Processing and Communications Applications Conference , 2020
Adversarial Training Regularization
Paper
We propose orthogonal regularization during adversarial training to improve robustness. By encouraging weight matrices to stay orthogonal, the model learns more stable and transferable representations under attack.
No matching items