Pieter Abbeel

Rodrigues Network for Learning Robot Actions

Relative Entropy Pathwise Policy Optimization

SARM: Stage-Aware Reward Modeling for Long Horizon Robot Manipulation

D-REX: Differentiable Real-to-Sim-to-Real Engine for Learning Dexterous Grasping

Diffusion Guidance Is a Controllable Policy Improvement Operator

GaussGym: An open-source real-to-sim framework for learning locomotion from pixels

Flow Policy Gradients for Legged Robots

ViTacFormer: Learning Cross-Modal Representation for Visuo-Tactile Dexterous Manipulation

What really matters in matrix whitening optimizers?

One Step Diffusion via Shortcut Models

Prioritized Generative Replay

DexGarmentLab: Dexterous Garment Manipulation Environment with Generalizable Policy

Object-centric 3D Motion Field for Robot Learning from Human Videos

Compute-Optimal Scaling for Value-Based Deep RL

Bigger, Regularized, Categorical: High-Capacity Value Functions are Efficient Multi-Task Learners

Protein Language Model Fitness is a Matter of Preference

Coarse-to-fine Q-Network with Action Sequence for Data-Efficient Reinforcement Learning

MaxInfoRL: Boosting exploration in reinforcement learning through information gain maximization

A Stable Whitening Optimizer for Efficient Neural Network Training

SEMDICE: Off-policy State Entropy Maximization via Stationary Distribution Correction Estimation

Textbook Consistency Weighted Internet Improves Efficiency Twofold

ElasticTok: Adaptive Tokenization for Image and Video

SOMBRL: Scalable and Optimistic Model-Based RL

World Model on Million-Length Video And Language With Blockwise RingAttention

Cliqueformer: Model-Based Optimization With Structured Transformers

Value-Based Deep RL Scales Predictably

Video2Policy: Scaling up Manipulation Tasks in Simulation through Internet Videos

Generating All-Atom Protein Structure from Sequence-Only Training Data

Reinforcement Learning with Action Sequence for Data-Efficient Robot Learning

Chip Placement with Diffusion Models

OTTER: A Vision-Language-Action Model with Text-Aware Visual Feature Extraction

Early Fusion Helps Vision Language Action Models Generalize Better