# BAIR Presents Advanced Robotics Research at CoRL 2026 Berkeley researchers showcase breakthroughs in embodied AI and robot learning at CoRL 2026. - **Source:** [Berkeley AI Research Blog](https://bair.berkeley.edu/blog) - **Published:** 2026-11-09 - **Topic:** [technology](https://brifing.si/c/technology) - **Original article:** - **On Brifing:** https://brifing.si/s/c2fe1866-ce0f-4bc0-9a42-3605ab907be0 ## CoRL 2026 Conference Overview The Conference on Robot Learning (CoRL) is held from November 9 to 12, 2026, in Austin, Texas. This premier event gathers global experts to discuss the latest advancements in robotic systems. The Berkeley Artificial Intelligence Research (BAIR) team is participating extensively, presenting numerous papers and hosting sessions throughout the four-day program. ## Early Career Keynote Speaker Professor Negar Mehr will deliver an Early Career Keynote address during the conference. Her presentation, titled 'How Can Robots Learn to Interact and Coordinate with Others?', focuses on collaborative robotics. This session is chaired by Ken Goldberg and highlights the critical role of social interaction in modern autonomous robot development and deployment. ## Focus on Embodied Memory Researchers presented 'MEM: Multi-Scale Embodied Memory for Vision Language Action Models' during the Monday spotlight session. This work explores how robots utilize multi-scale memory structures to improve performance in vision-language tasks. The project involves a large collaborative team, including experts from various institutions, aiming to enhance generalist policy capabilities in robotics. ## Advancing Dexterous Manipulation The 'EgoScale' project, featured on Tuesday, addresses the challenge of scaling dexterous manipulation using diverse egocentric human data. By leveraging large-scale human interaction datasets, the researchers aim to improve how robots handle complex assembly tasks. This research is a significant step toward creating more versatile and capable robotic manipulation systems. ## Humanoid Control Breakthroughs The 'SuperMimic' paper introduces a method for scaling humanoid whole-body control using a differentiable world model. Presented by Qiayuan Liao and colleagues, this research emphasizes the integration of sophisticated control mechanisms with learned world models. The work is designed to enhance the fluidity and precision of humanoid movements in dynamic environments. ## Late-Breaking Results Posters The conference includes a dedicated track for late-breaking results, showcasing emerging ideas that push the boundaries of current robotics. Current posters from Berkeley authors cover topics like agentic robotics self-learning and real-to-sim toolkits. These sessions provide a platform for rapid dissemination of novel findings that are still in active development stages. ## Generalist Policy Research Several papers, such as 'Large Video Planner Enables Generalizable Robot Control', investigate the development of generalist policies. These models aim to enable robots to perform a wide variety of tasks without needing task-specific training. By aligning fine-grained manipulation with human preferences, researchers are bridging the gap between theoretical models and practical applications. ## Collaborative Research Efforts Many of the presented papers are co-authored with collaborators from various universities and industry partners. This collaborative approach reflects the interdisciplinary nature of modern robotics research. By combining expertise from different fields, the BAIR team ensures their work addresses both fundamental scientific challenges and real-world implementation hurdles in robotic systems. ## Poster Session Dynamics Poster sessions serve as a primary venue for detailed technical discussions between researchers and attendees. With sessions scheduled throughout the conference, participants can engage directly with authors regarding specific methodologies, such as hindsight reward editing or asynchronous VLA navigation. These interactions are essential for refining algorithms and fostering future research collaborations. ## Technical Session Structure The conference program is organized into thematic sessions including reinforcement learning, motion reasoning, and dexterous learning. Each session provides a structured environment for presenting oral papers and spotlight presentations. This organization helps attendees navigate the dense schedule of technical content, ensuring that specialized research reaches the most relevant academic and industrial audiences. ## Methodological Innovations Research presented at CoRL 2026 highlights a shift toward data-driven approaches in robot learning. Techniques like flow policy gradients and discrete diffusion policies demonstrate the move toward more robust, asynchronous execution in robotic agents. These methodological advancements are critical for improving the reliability of robots operating in unpredictable, real-world environments. ## Future Research Implications The work presented by BAIR at CoRL 2026 underscores the rapid evolution of embodied AI. By focusing on memory, generalist policies, and human-robot coordination, these researchers are setting the agenda for the next generation of autonomous machines. The findings shared here will likely influence future developments in both industrial automation and personal robotics. ## Sources covering this story - http://bair.berkeley.edu/blog/2026/11/09/bair-at-corl-2026/ — --- _Summary written by Brifing from the original reporting. Copyright in the original article belongs to its publisher; read the full piece at the link above._