BAIR at CoRL 2026

Overview of BAIR Participation at CoRL 2026

The Berkeley Artificial Intelligence Research (BAIR) lab is participating in the 2026 Conference on Robot Learning (CoRL), held November 9–12 in Austin, Texas. The lab's presence includes 33 main-conference papers, 10 oral and spotlight presentations, and four late-breaking results posters. BAIR faculty, students, and postdocs are also contributing to workshop organization and keynote sessions throughout the four-day event.

Key Research Themes

The research presented by BAIR members at CoRL 2026 highlights several core trends in modern robotics, including:

• Embodied AI and Vision-Language-Action (VLA) Models: Research focuses on memory-augmented models, such as MEM: Multi-Scale Embodied Memory, and asynchronous VLA architectures for edge navigation.
• Humanoid Control and Loco-Manipulation: Several papers address whole-body control, including SuperMimic, which utilizes differentiable world models, and TANGO, a VLA-based navigation model for cluttered environments.
• Agentic Robotics and Self-Improvement: A significant portion of the work explores autonomous agentic behavior, including self-learning frameworks like ENPIRE and GaP (Graph-as-Policy).
• Dexterous Manipulation: New approaches to fine-grained manipulation are explored through tactile-reactive policies (T-Rex) and the use of diverse egocentric human data (EgoScale).

Featured Presentations and Keynotes

Professor Negar Mehr is scheduled to deliver an Early Career Keynote titled "How Can Robots Learn to Interact and Coordinate with Others?" on November 9. Additionally, BAIR members are heavily involved in workshop organization, covering topics such as sample-efficient reinforcement learning (R2RL), bimanual dexterous tool use, and sim-to-real-to-field transfer.

Schedule Highlights

• Monday, Nov 9: Focuses on generalist policies, memory, and agentic robotics, featuring the Early Career Keynote by Professor Negar Mehr.
• Tuesday, Nov 10: Covers reinforcement learning, motion reasoning, and humanoid skills, with oral presentations on Flow Policy Gradients and EgoScale.
• Wednesday, Nov 11: Highlights optimization and humanoid locomotion, including spotlights on Playful Agentic Robot Learning and LadderMan.
• Thursday, Nov 12: Dedicated to workshops, including sessions on sample-efficient RL, bimanual manipulation, and agentic robotics demos.