BAIR lists 33 Berkeley main-conference papers at CoRL 2026

Berkeley AI Research (BAIR) has put up a schedule page for the Conference on Robot Learning (CoRL), which takes place November 9 to 12, 2026, in Austin, Texas. The main conference runs November 9 to 11 and workshops fall on November 12. BAIR says its faculty, postdocs and students will present research at various sessions, including an Early Career Keynote from Professor Negar Mehr. The page covers papers co-authored with collaborators at other universities and companies. The inclusion rule: a paper is listed if at least one author is affiliated with UC Berkeley, and joint affiliations qualify. Faculty talks, session-chair roles and workshop participation are listed too. All times are CST (UTC−6), and the page was last updated November 8, 2026.
The summary tiles give the scale: 33 main-conference papers, 4 late-breaking results (LBR) posters, 10 orals and spotlights (3 orals and 7 spotlights), and 4 days of research. Monday, November 9 has 10 papers plus 3 LBR posters. Tuesday, November 10 has 11 papers plus 1 LBR poster. The late-breaking list is described as four posters "so far" and may grow as more are confirmed.
Monday opens with Spotlight S1, "Generalist policies, memory and transfer", from 9:00 to 9:30 AM at the Hilton Austin. It includes two papers. "MEM: Multi-Scale Embodied Memory for Vision Language Action Models" has 17 listed authors, among them Marcel Torne, Karl Pertsch, Sergey Levine, Chelsea Finn and Danny Driess. "Diagnosing Compositional Generalization in Sequential Robot Tasks" is by Yixiao Wang, Cheng-En Wu, Lingfeng Sun, Pengcheng Wang, Xiang Ji, Boyuan Liang, Guojian Zhan and Masayoshi Tomizuka. Both reappear at Poster Session I.
After the lunch break (12:00 to 1:30 PM), three late-breaking posters in an "Agentic Robotics" group are shown at Griffin, Level 2, in the JW Marriott Austin. They are "Scaling Online Exploration for Robot Agents" (board 119, LBR #15, with authors including Pieter Abbeel, Antonio Torralba, Dawn Song and Yilun Du); "Reconstruct, Practice, Go Real: Guided Self-Improvement for Embodied Agents" (board 120, LBR #19, with Abbeel and Nika Haghtalab among the authors, also at the Agentic Robotics workshop on Thursday, November 12, 4:15 to 5:55 PM); and "Flip Anything with Agentic Robotics Self-Learning" (board 131, LBR #54, by Shuangyu Xie, Kaiyuan Chen and Ken Goldberg, with a GitHub link).
At 2:00 to 2:30 PM at the Hilton Austin, Early Career Keynote II features Professor Negar Mehr with the talk "How Can Robots Learn to Interact and Coordinate with Others?" Ken Goldberg chairs the session.
Poster Session I runs 5:00 to 6:30 PM at the JW Marriott with nine posters. MEM is Poster 42. The others are "Large Video Planner Enables Generalizable Robot Control" (57), "How to Peel with a Knife: Aligning Fine-Grained Manipulation with Human Preference" (68), "HiRE: Hindsight Reward Editing for Policy Finetuning" (97), "Scalable Behavior Cloning with Open Data, Training, and Evaluation" (113), the compositional generalization paper (115), "AsyncVLA: An Asynchronous VLA for Fast and Robust Navigation on the Edge" (142), "GLST: Global-Local Spatio-Temporal Memory Decomposition Framework for Online Chained-Goal Navigation" (159), "ENPIRE: Agentic Robot Policy Self-Improvement in the Real World" (202) and "Energy-Based Policy Orchestration" (225).
The visible Tuesday schedule starts with Oral O4, "Reinforcement learning and motion reasoning", at 9:00 to 9:30 AM at the Hilton Austin. It features "Flow Policy Gradients for Robot Control" by Brent Yi, Hongsuk Choi, Himanshu Gaurav Singh and nine co-authors including Pieter Abbeel, Guanya Shi, Karen Liu and Angjoo Kanazawa. The paper also appears at Poster Session II, 5:00 to 6:30 PM, Poster 63, at the JW Marriott. The available text ends there, so Tuesday afternoon, Wednesday and the Thursday workshops are not covered here.
Key facts
- CoRL 2026 runs November 9 to 12 in Austin, Texas: main conference November 9 to 11, workshops November 12.
- BAIR's page counts 33 main-conference papers with at least one UC Berkeley-affiliated author, plus 4 late-breaking posters listed so far.
- Ten orals and spotlights are listed (3 orals, 7 spotlights); Monday has 10 papers plus 3 LBR posters, Tuesday 11 papers plus 1 LBR poster.
- Professor Negar Mehr gives Early Career Keynote II on Monday, November 9, 2:00 to 2:30 PM CST, titled "How Can Robots Learn to Interact and Coordinate with Others?"
- Monday's Spotlight S1 includes "MEM: Multi-Scale Embodied Memory for Vision Language Action Models"; Tuesday's Oral O4 includes "Flow Policy Gradients for Robot Control".
Why it matters
CoRL is a robot learning conference, and this page shows how much of its program involves Berkeley authors: 33 main-conference papers, 10 of them as orals or spotlights. The visible titles point to themes such as vision-language-action models with memory, policy learning from open data, reinforcement learning with flow policies, and agentic robotics where robots improve their own policies. The page is a schedule, not a report of findings, so it says little about what the papers achieve.
Who it affects
Attendees planning their days in Austin, and researchers following Berkeley groups in robot learning. The authors listed include Pieter Abbeel, Sergey Levine, Chelsea Finn, Jitendra Malik, Ken Goldberg and Angjoo Kanazawa, plus collaborators at other universities and companies. Anyone who wants to catch a specific paper or person can use the page to find the talk and poster slot.
How to use it
The page has a search box for a paper or person (a surname or shorter title works) and a filter by session type: keynote, orals, spotlights, posters, late-breaking results and workshops. Day tabs cover Monday to Thursday. Most entries link to the paper, the late-breaking program has its own link, and "Flip Anything with Agentic Robotics Self-Learning" links to GitHub. Times are CST (UTC−6). Papers appear in both their talk and poster sessions, so a spotlight or oral paper can be found again at the evening poster slot.
How solid is it
This is first-party: BAIR is listing its own members' sessions, and the page was updated November 8, 2026. The inclusion criterion is explicit (at least one UC Berkeley-affiliated author; joint affiliations qualify). The text available here is truncated: Tuesday afternoon, Wednesday and Thursday (workshops) sessions are not visible, so the 33-paper total could be checked here only against Monday's 10 and Tuesday's 11.
Risks and caveats
Times for oral and spotlight papers are session windows, and individual presentation order is not specified. The late-breaking list may grow as more posters are confirmed, so the count of 4 is provisional. The page gives no results or descriptions for the papers, nothing on awards, acceptance rates or attendance, and nothing about the Mehr keynote beyond its title. A count of papers with a Berkeley author says nothing about which work is most significant.
“BAIR faculty, postdocs and students will be presenting research at various sessions, including an Early Career Keynote from Professor Negar Mehr.”
— BAIR blog