Introduction
Dexterous manipulation remains one of robotics' most difficult challenges. Although researchers have made progress over several decades, robots still struggle to perform tasks that humans accomplish with ease. Addressing this problem will require advances across mechanical design, motion planning, control, and machine learning. This workshop will bring together experts from these areas to discuss the key gaps, emerging approaches, and future directions for the field.
Rather than serving only as a venue for presenting recent results, the workshop will be structured around viewpoints, evidence, and reflection. Invited speakers will articulate distinct perspectives on the future of dexterous manipulation, while curated posters and live demonstrations from academia and industry will provide concrete examples of the current state of the art.
Through this format, the workshop aims to help the community identify where different approaches agree, where they diverge, and what evidence would be needed to change current assumptions. By connecting high-level perspectives with real systems and open discussion, we hope to clarify shared bottlenecks and define concrete priorities for building robust, generalizable, and deployable dexterous robotic systems.
Core Challenges & Research Questions
The workshop is organized around several competing but complementary hypotheses for how dexterous manipulation may scale. Rather than treating hardware, perception, control, and learning as separate topics, we ask how these components interact and which bottlenecks must be resolved for dexterous manipulation to move beyond isolated demonstrations.
Evaluation and Evidence- How should the community decide what is working? The field currently contains many impressive demonstrations, but lacks shared standards for comparing progress across platforms and approaches.
- What should count as evidence of general-purpose dexterity? Can the community develop benchmarks and reporting practices that make it easier to compare hardware-first, control-first, and learning-first approaches?
- How can robotic systems extract useful information when the hand itself occludes the object or scene?
- How can policies remain robust across changes in lighting, clutter, viewpoints, and unseen objects?
- What role should tactile sensing play: can touch become as informative as vision for manipulation, or is it primarily a complementary signal? What tactile sensor designs and representations are most useful for scalable manipulation?
- What innovations are needed to make robotic hands more compact, efficient, durable, and capable?
- How can future hand designs be smarter rather than simply stronger, and how can we improve hardware longevity while reducing maintenance and cost?
- Will dexterous manipulation be driven by general-purpose foundation models, or by task-specific models tailored to particular objects, embodiments, and environments?
- How can policies adapt to dynamic, forceful, or high-speed interactions?
- Can the community standardize the collection and use of human hand data to better train and evaluate robotic manipulation policies?
- How can policies generalize from limited, noisy, or biased demonstrations?
- What infrastructure and algorithms are needed to scale reinforcement learning across hundreds of manipulation tasks with minimal intervention?
Invited Speakers
Workshop Schedule
Tentative half-day afternoon schedule (subject to change):
| Time | Section |
|---|---|
| 2:00 PM – 2:20 PM | Invited Speaker 1 |
| 2:20 PM – 2:40 PM | Invited Speaker 2 |
| 2:40 PM – 3:00 PM | Invited Speaker 3 |
| 3:00 PM – 3:30 PM | Spotlight Session |
| 3:30 PM – 4:00 PM | Poster Session |
| 4:00 PM – 4:40 PM | Live Demo Session |
| 4:40 PM – 5:00 PM | Invited Speaker 4 |
| 5:00 PM – 5:55 PM | Panel Discussion |
| 5:55 PM – 6:00 PM | Closing Remarks & Award Announcement |
Call for Papers and Demos
We welcome submissions on both algorithmic and hardware advances in dexterous manipulation. Outstanding submissions will receive awards, ensuring recognition for high-quality work. We especially encourage submissions that clarify assumptions, report failure modes, compare alternative approaches, or provide evidence for what is needed to move dexterous manipulation beyond isolated demonstrations.
Each accepted short paper will be eligible for a poster presentation, and selected papers will be invited to give a short spotlight talk. Note: poster and spotlight presentations must be given in person.
We are particularly excited to provide a platform for showcasing real-world robotic systems. Even without a formal paper submission, we encourage submissions of videos demonstrating your robots in action. For those attending in person, there will be dedicated opportunities to showcase your robots live at the workshop!
We encourage researchers to submit work in the following areas (the list is not exhaustive):
- Evaluation and Evidence: benchmarks, reporting standards, and shared protocols that compare hardware-first, control-first, and learning-first approaches to dexterity.
- Perception — Seeing & Feeling the World: robust visual perception under occlusion and clutter, tactile sensor designs and representations, and multi-modal fusion for manipulation.
- Hardware — Redesigning Dexterous Hands: compact, efficient, durable, and capable hand designs; soft and compliant mechanisms; durability and maintainability.
- Scaling Robot Skill Learning: general-purpose vs. task-specific policies, foundation models for manipulation, dynamic / forceful / high-speed interactions, and standardized human-hand data.
- Reinforcement Learning and Demonstration Data: generalizing from limited or biased demonstrations, infrastructure for scaling RL across many tasks with minimal intervention.
- Sim-to-real and Data Generation: simulation for contact-rich tasks, low-cost teleoperation, and cross-embodiment transfer.
- Any additional related topics in dexterous manipulation 😄
Submission Guidelines
- Submission Portal for Papers and Demos: OpenReview link — coming soon
- Paper Submission Guidelines:
- Double-Blind Review: The review process is strictly double-blind. Please ensure your submission is fully anonymized. Do not include author names, affiliations, or identifying acknowledgements in the main paper, appendix, or supplementary materials.
- Paper Length: Submissions should be up to 4 pages maximum, excluding references, acknowledgements, and appendices.
- Format:
- Please follow the standard CoRL 2026 main conference format. Include references and appendices in the same PDF as the main paper, and submit any optional videos as a zip file in the supplementary material section.
- The maximum file size is 100MB. More information on format can be found here.
- Review Process:
- All submissions will undergo a peer-review process and will be evaluated based on their relevance and contribution to the workshop's topics.
- Accepted submissions will be featured at the workshop through poster spotlight talks and poster sessions.
- Dual Submissions:
- Papers in preparation or under review at other major venues are welcome.
- Previously published works are also permitted, provided their publication status is explicitly stated at the time of submission.
- Note: per CoRL 2026 policy, full papers accepted to the main conference are discouraged from also being submitted as workshop contributions.
- Demo Submission Guidelines:
- Summary Document: Please upload a summary document as a PDF. There are no strict page limits or formatting requirements for this document.
- Demo Video: Please upload demo videos as a single zip file under the supplementary material field in OpenReview.
- Visibility: Submissions and reviews will remain private. Only accepted papers will be made public.
Timeline
- Submission Portal Open: TBD
- Submission Deadline: TBD
- Notification: TBD
- Workshop Date: TBD
