Razvan Pascanu

A fresh approach improves connections between images and their captions.

Examining ways to maintain skills in RL during fine-tuning.

Research reveals strategies to improve adaptability of neural networks in dynamic conditions.

Hawk and Griffin set a new standard in efficient language processing.

Examining the role of representation in PPO and its impact on agent performance.

An analysis of Transformers' struggles with counting and copying tasks.

Explore the role of attention mechanisms in machine learning.

A new approach improves learning efficiency in reinforcement learning environments.

Investigating how transformers learn and generalize from compositional tasks.

A new approach helps neural networks learn from shifting data without forgetting past knowledge.