CoRL 2026 — Conference on Robot Learning

CoRL 2026 — the Conference on Robot Learning — takes place from 9 to 12 November 2026 at the JW Marriott Austin in Austin, Texas, with workshops opening the week and the main single-track programme following. CoRL is the leading venue dedicated specifically to the intersection of machine learning and robotics: smaller and more tightly focused than IROS or ICRA, and for that reason one of the most closely watched conferences in AI right now. It is built for robot-learning and machine-learning researchers, PhD students, and the industry labs racing to turn learned policies into shipping products. Organised under the Robot Learning Foundation, the conference has grown quickly in influence as robotics has become one of the central arenas of modern AI.

The 2026 programme and research focus

CoRL concentrates on learning-based robotics, and its single-track format — a deliberate choice — keeps the whole community in one room around a tightly curated set of accepted papers, with oral presentations, interactive poster sessions and a strong slate of workshops. The research centre of gravity includes:

• Imitation learning and reinforcement learning for control and manipulation
• Vision-language-action (VLA) models and robot foundation models
• Learned perception and representation for embodied agents
• Sim-to-real transfer and large-scale, physics-based simulation
• Large-scale and cross-embodiment data collection that lets one model drive many different machines
• The manipulation and locomotion benchmarks that test whether learned policies actually work in the physical world

Because the proceedings are curated rather than sprawling, CoRL acceptances carry real signal, and the conference functions as an early-warning system for which robot-learning ideas the field is converging on. The papers, demos and debates here tend to anticipate the methods that show up across the wider robotics and AI literature a year or two later.

Who should attend

CoRL is essential for researchers and graduate students working at the machine-learning–robotics boundary — the people building and evaluating learned policies, datasets and architectures. Its scale and single-track design make it unusually good for sustained technical exchange: attendees genuinely see the same talks and argue over the same results, which is far harder at a multi-thousand-person, multi-track conference. The atmosphere is closer to a focused workshop than a trade show, and that concentration is precisely what regulars value. CoRL is also increasingly a recruiting ground. The robotics startups and frontier labs betting on general-purpose manipulation send teams to read the field first and to hire the students producing the work, so industry R&D staff and technical founders have strong reason to attend, not only academics. Many attendees come specifically to track which datasets, simulators and policy architectures are gaining traction before committing their own labs' resources to them.

Why it matters for AI and robotics in 2026

CoRL sits at the centre of the embodied-AI moment. The field's biggest current bet — that the recipe behind large language and vision models (scale, broad data, pretraining) can produce general-purpose robot policies — is being worked out here, through vision-language-action models, robot foundation models, and cross-embodiment datasets. With humanoids and general manipulation drawing enormous investment in 2026, CoRL is where the research underpinning those bets is presented, critiqued and stress-tested first. The conference's value is that it pairs ambition with evidence: papers must demonstrate learned behaviour on real or carefully simulated robots, which keeps the community honest about the gap between a promising result and a reliable, repeatable system. For the broader AI field, CoRL is the clearest annual read on whether the foundation-model playbook truly generalises from pixels and tokens to motors and grippers — the question on which a great deal of current robotics investment ultimately rests. Because the community is small enough to reach consensus, CoRL also tends to crystallise shared benchmarks and evaluation norms that the rest of the field then adopts, giving the conference an outsized influence on how progress in robot learning is measured.

Attending and registration

The official programme, accepted papers, workshop list and registration open on the conference site at corl.org. As at recent editions, expect reduced student rates and a strong emphasis on in-person participation, which the single-track format depends on. Researchers intending to submit should watch the abstract and full-paper deadlines closely, and anyone planning to attend should book Austin accommodation early: the compact venue and the field's surging interest tend to make space tight well before the November dates.