Anonymous code and data release for the ICLR 2027 submission.
- Project page: https://plan2pose.github.io
- Code and data documentation: this repository, tag
iclr2027-submission - Data archives and checkpoints: release assets of the tag, https://github.com/plan2pose/plan2pose-code/releases/tag/iclr2027-submission
| directory | contents |
|---|---|
code/ |
Plan2Pose (code/plan2pose/plan2pose_ref), the evaluation package, the ports of LINGO-Space, Diffusion-CCSP and StructDiffusion, the reference methods, and the table generator with the stored result files. Start with code/README.md: environment and one command per table. |
data/ |
Data card, split files of every pool and evaluation set, preparation scripts, and LARGE_FILES.tsv with the size and sha256 of every release asset. Start with data/README.md. |
Licences: code MIT (code/LICENSE); data and checkpoints CC BY 4.0 (data/LICENSE); the baseline
ports keep their upstream licences (code/THIRD_PARTY_LICENSES/). For licensing reasons, the
2,850 replay episodes built from GraspNet-1Billion object scans are not released; their identifiers
are listed in data/omitted_graspnet_episodes.txt.
The code and data were anonymised for review with the help of large language model coding agents and checked with automated scans.