Guillaume Lajoie

MesaNet: Sequence Modeling by Locally Optimal Test-Time Training

Iterative Amortized Inference: Unifying In-Context Learning and Learned Optimizers

Recursive Self-Aggregation Unlocks Deep Thinking in Large Language Models

Generalizable, real-time neural decoding with hybrid state-space models

Towards a formal theory of compositionality

Multi-agent cooperation through learning-aware policy gradients

Expressivity of Neural Networks with Random Weights and Learned Biases

Tracing the Representation Geometry of Language Models from Pretraining to Post-training

Towards a Formal Theory of Representational Compositionality

Does learning the right latent variables necessarily improve in-context learning?

Accelerating Training with Neuron Interaction and Nowcasting Networks

Does learning the right latent variables necessarily improve in-context learning?

In-context learning and Occam's razor

In-Context Learning and Occam's Razor

Memory-Pruning Algorithm for Bayesian Optimization with Strict Computational Cost Guarantees