Short papers
Up to 4 pages (excluding references), describing completed research, ongoing work, or benchmark datasets.
Machine Learning/CORE Unranked
pretrain-to-adapt 2026
CoRL 2026 Workshop on Pretrain to Adapt
The Pretrain to Adapt workshop at CoRL 2026 explores what makes a pretrained robot policy an effective foundation for downstream adaptation, bridging research in pretraining generalist policies and adaptation methods such as fine-tuning, in-context learning, and test-time steering. The workshop aims to identify properties that enhance adaptability, develop metrics for readiness to adapt, and align pretraining objectives with adaptation goals.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Full papers
Up to 8 pages (excluding references), describing completed research, ongoing work, or benchmark datasets.
Research areas in scope
01
Topics of Interest
Pretraining objectives for adaptable policies (behavior cloning, return-conditioned, world-model, value-based, in-context objectives)Composing and curating heterogeneous pretraining data (demonstrations, play data, suboptimal/failure trajectories, RL rollouts) for downstream adaptabilityCoupling pretraining recipes with adaptation methods: RL finetuning, in-context learning, and test-time steeringSteerability and plasticity of pretrained policiesAdapting pretrained policies to new tasks, embodiments, and environmentsMetrics and benchmarks for measuring readiness to adaptWorld models as pretraining objectives, evaluation tools, or RL environmentsReward modeling and learned rewards for continual policy improvementPosition papers, work-in-progress, and retrospectives on any of the above
Policies worth checking twice
- Submissions will be double-anonymized.
- Dual submissions are welcome if under review or recently published at relevant venues.
- The workshop is non-archival; accepted papers will be posted on the workshop website.
- Accepted papers must be presented in person.
- Page limits: short papers up to 4 pages, full papers up to 8 pages (both excluding references).
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026