Sergey Levine
- A Stable Whitening Optimizer for Efficient Neural Network Training
- Compute-Optimal Scaling for Value-Based Deep RL
- Consistently Simulating Human Personas with Multi-Turn Reinforcement Learning
- Derivative-Free Guidance in Continuous and Discrete Diffusion Models with Soft Value-based Decoding
- Horizon Reduction Makes RL Scalable
- Knowledge Insulating Vision-Language-Action Models: Train Fast, Run Fast, Generalize Better
- Offline Goal-conditioned Reinforcement Learning with Quasimetric Representations
- Planning without Search: Refining Frontier LLMs with Offline Goal-Conditioned RL
- Real-Time Execution of Action Chunking Flow Policies
- Reinforcement Learning with Action Chunking
- Self-Challenging Language Model Agents
- Temporal Representation Alignment: Successor Features Enable Emergent Compositionality in Robot Instruction Following