Rob Fergus

Cambrian-S: Towards Spatial Supersensing in Video

KL-Regularized Reinforcement Learning for Generative Modelling is Designed to Mode Collapse

Understanding outer learning rates in Local SGD

Efficient Exploration and Discriminative World Model Learning with an Object-Centric Abstraction

Language Agents Mirror Human Causal Reasoning Biases. How Can We Help Them Think Like Scientists?

Training Language Models on Synthetic Edit Sequences Improves Code Synthesis

BALROG: Benchmarking Agentic LLM and VLM Reasoning On Games

D3: A Dataset for Training Code LMs to Act Diff-by-Diff