4 papers
RL Post-Training Builds Compositional Reasoning Strategies
Azwar Abdulsalam, Nishil Patel, Andrew Saxe
Does RL post-training merely amplify primitive skills already latent in a base model, or can it compose primitive skills into new higher-level strategies? We study this question in…
A Theory of Initialisation's Impact on Specialisation
Devon Jarvis, Sebastian Lee, Clémentine Carla Juliette Dominé +2
Prior work has demonstrated a consistent tendency in neural networks engaged in continual learning tasks, wherein intermediate task similarity results in the highest levels of cata…
The RL Perceptron: Generalisation Dynamics of Policy Learning in High Dimensions
Nishil Patel, Sebastian Lee, Stefano Sarao Mannelli +2
Reinforcement learning (RL) algorithms have proven transformative in a range of domains. To tackle real-world domains, these systems often use neural networks to learn policies dir…
Tilting the Odds at the Lottery: the Interplay of Overparameterisation and Curricula in Neural Networks
Stefano Sarao Mannelli, Yaraslau Ivashynka, Andrew Saxe +1
A wide range of empirical and theoretical works have shown that overparameterisation can amplify the performance of neural networks. According to the lottery ticket hypothesis, ove…