CoRL - Heungwoo/research GitHub Wiki
CoRL — Conference on Robot Learning
CoRL is the natural sibling venue to ICLR for VLA work — typically more focused on real-robot systems and end-to-end learning recipes.
Surveys hosted in this wiki
| Year | Location | Survey |
|---|---|---|
| CoRL 2026 🆕 | Austin, Nov 9–12, 2026 | [VLA & Manipulation Survey](/Heungwoo/research/wiki/CoRL-2026-VLA-Manipulation-Survey) (preliminary — 687 accepted; official list pending) |
| [CoRL 2025](/Heungwoo/research/wiki/CoRL-2025) | Seoul, Sept 27–30, 2025 | [VLA & Manipulation Survey](/Heungwoo/research/wiki/CoRL-2025-VLA-Manipulation-Survey) (live) |
Baseline / flagship papers at CoRL
- π0.5 (CoRL 2025 Oral) — the hierarchical VLA that seeds the current π series. See π series evolution for how it evolved into π0.6, π*0.6, and π0.7.
Foundational CoRL 2024 papers referenced across the wiki
- OpenVLA (CoRL 2024) — the open-source 7B autoregressive VLA (Prismatic VLM + discretized action tokens on Open X-Embodiment); the reference baseline most later VLAs report against.
- ReKep (CoRL 2024) — training-free manipulation via relational keypoint constraints written by a VLM and solved as real-time optimization.
← Back to Home