Yann LeCun

LLM-JEPA: Large Language Models Meet Joint Embedding Predictive Architectures

From Tokens to Thoughts: How LLMs and Humans Trade Compression for Meaning

VL-JEPA: Joint Embedding Predictive Architecture for Vision-language

Attention Sinks and Compression Valleys in LLMs are Two Sides of the Same Coin

Cambrian-S: Towards Spatial Supersensing in Video

What drives success in physical planning with Joint-Embedding Predictive World Models?

Closing the Train-Test Gap in World Models for Gradient-Based Planning

LiveBench: A Challenging, Contamination-Limited LLM Benchmark

DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning

Whole-Body Conditioned Egocentric Video Prediction

Layer by Layer: Uncovering Hidden Representations in Language Models

Hierarchical World Models as Visual Whole-Body Humanoid Controllers

URLOST: Unsupervised Representation Learning without Stationarity or Topology

OSVI-WM: One-Shot Visual Imitation for Unseen Tasks using World-Model-Guided Trajectory Generation

Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models

Seq-VCR: Preventing Collapse in Intermediate Transformer Representations for Enhanced Reasoning

PooDLe🐩: Pooled and dense self-supervised learning from naturalistic videos

$\mathbb{X}$-Sample Contrastive Loss: Improving Contrastive Learning with Sample Similarity Graphs

DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning

Just How Flexible are Neural Networks in Practice?

Variance-Covariance Regularization Improves Representation Learning

Deep Neural Networks without Normalization