Hierarchical World Models as Visual Whole-Body Humanoid Controllers
机构 * UC San Diego（UC圣地亚哥大学） ； New York University（纽约大学） ； Meta AI
Comments Code and videos at https://nicklashansen.com/rlpuppeteer
作者
Yann LeCun
Deep Learning / Vision
RoboPEPP: Vision-Based Robot Pose and Joint Angle Estimation through Embedding Predictive Pre-Training
机构 * New York University Tandon School of Engineering（纽约大学Tandon工程学院） ； New York University Courant Institute of Mathematical Sciences（纽约大学Courant数学科学研究所） ； Meta-FAIR
PooDLe: Pooled and dense self-supervised learning from naturalistic videos
机构 * New York University（纽约大学） ； Meta
Comments Project page: https://agenticlearning.ai/poodle/
LiveBench: A Challenging, Contamination-Limited LLM Benchmark
Comments ICLR 2025 Spotlight
Navigation World Models
Comments CVPR 2025. Project page: https://www.amirbar.net/nwm/
Scaling Language-Free Visual Representation Learning
Comments Project page at https://davidfan.io/webssl/
URLOST: Unsupervised Representation Learning without Stationarity or Topology
Comments Accepted by ICLR 2025; Code will be available at this https://github.com/zeyuyun1/URLOST
Seq-VCR: Preventing Collapse in Intermediate Transformer Representations for Enhanced Reasoning
Improving Pre-trained Self-Supervised Embeddings Through Effective Entropy Maximization
Comments Published in Proceedings of the 28th International Conference on Artificial Intelligence and Statistics (AISTATS 2025). A preliminary version of this work also appeared in the NeurIPS 2024 Workshop on Self-Supervised Learning: Theory and Practice
Multi-modal AI for comprehensive breast cancer prognostication
Intuitive physics understanding emerges from self-supervised pretraining on natural videos
Comments 24 pages,14 figures, 5 tables
DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning
MetaMorph: Multimodal Understanding and Generation via Instruction Tuning
Comments Project page at tsb0601.github.io/metamorph
Video Representation Learning with Joint-Embedding Predictive Architectures
Does Representation Matter? Exploring Intermediate Layers in Large Language Models
Comments Accepted to 2024 NeurIPs Workshop on Machine Learning and Compression
Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs
Comments NeurIPS 2024 (Oral). Website at https://cambrian-mllm.github.io
Introduction to Latent Variable Energy-Based Models: A Path Towards Autonomous Machine Intelligence
Comments 23 pages + 1-page appendix, 11 figures. These notes follow the content of three lectures given by Yann LeCun during the Les Houches Summer School on Statistical Physics and Machine Learning in 2022. Feedback and comments are most welcome!
Journal ref J. Stat. Mech. (2024) 104011
Fine-Tuning Large Vision-Language Models as Decision-Making Agents via Reinforcement Learning
$\mathbb{X}$-Sample Contrastive Loss: Improving Contrastive Learning with Sample Similarity Graphs
Blockwise Self-Supervised Learning at Scale
Just How Flexible are Neural Networks in Practice?
Towards an Improved Understanding and Utilization of Maximum Manifold Capacity Representations
Light-weight probing of unsupervised representations for Reinforcement Learning
Comments To appear in the proceedings of the Reinforcement Learning Conference 2024
Towards a Framework for Openness in Foundation Models: Proceedings from the Columbia Convening on Openness in Artificial Intelligence
G-Retriever: Retrieval-Augmented Generation for Textual Graph Understanding and Question Answering
The Entropy Enigma: Success and Failure of Entropy Minimization
Advancing human-centric AI for robust X-ray analysis through holistic self-supervised learning
An Information-Theoretic Perspective on Variance-Invariance-Covariance Regularization
Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs
Comments Project page: https://tsb0601.github.io/mmvp_blog/
EgoPet: Egomotion and Interaction Data from an Animal's Perspective
Comments https://www.amirbar.net/egopet
