ProcTHOR
ProcTHOR (ProcTHOR-10K, ArchitecTHOR)
Generator of procedurally built interactive houses in AI2-THOR, with a 10,000-house training set and a 10-house test set.1
Comparisons with real robots
Not checked2
Details
No real-robot experiments in the paper; datasheet lists Sim2Real only as a possible use.
Details
About
- What it is
- Simulator Inferred1
More
Classified by the Atlas from how the authors describe and distribute it.
- Built by
- PRIOR team, Allen Institute for AI5
More
Checked 2026-10-10.
- Released
- 2022-06 (arXiv v1 2022-06-14, only version)1
More
Checked 2026-10-10.
- Version
- procthor package 0.0.1 (2022-08-21); ProcTHOR-10K 0.1.2 (2022-07-11)4
More
GitHub releases. Checked 2026-10-10.
- Last update
- procthor release 0.0.1 on 2022-08-21; procthor-10k release 0.1.2 on 2022-07-11; last pushes 2023-04-07 (procthor) and 2022-12-14 (procthor-10k)4
More
GitHub API. Checked 2026-10-10.
- Status
- Dormant Inferred4
More
No releases since 2022, last push 2023-04. Checked 2026-10-10.
Setup
- Runs in
- Simulation2
More
Checked 2026-10-10.
- Robot
- Wheels only, Arm on wheels2
More
ObjectNav trained with a simulated LoCoBot agent; arm agents for manipulation. Checked 2026-10-10.
- Setting
- Whole home5
More
Floorplans with 1-10 rooms; custom scene types such as classrooms, libraries, offices also possible. Checked 2026-10-10.
Scoring and access
- Scored by
- Success rate, Path efficiency2
More
ArchitecTHOR ObjectNav reported as success rate and SPL (31.4% SR, 0.195 SPL). Checked 2026-10-10.
- Leaderboard
- None1
More
Checked 2026-10-10.
- Code licence
- Apache-2.04
More
Read the LICENSE file. Checked 2026-10-10.
- Data licence
- Apache-2.0 (procthor-10k LICENSE); paper datasheet: houses, asset database and code under Apache 2.0, ArchitecTHOR under Apache 2.0, usable for commercial and non-commercial purposes6
More
Checked 2026-10-10.
- Commercial use
- Allowed2
More
Datasheet sentence states commercial and non-commercial use. Checked 2026-10-10.
- Published at
- NeurIPS 2022 (Advances in Neural Information Processing Systems 35, Main Conference Track); NeurIPS virtual site labels it 'Outstanding Paper'7
More
Award label: https://neurips.cc/virtual/2022/poster/54832 Checked 2026-10-10.
Sources 7
- 1ProcTHOR: Large-Scale Embodied AI Using Procedural GenerationPaper · Jun 2022 · checked 10 Oct 2026
- 2ProcTHOR: Large-Scale Embodied AI Using Procedural Generation (full text)Paper · Jun 2022 · checked 10 Oct 2026
- 3Semantic Scholar API recordIndex · checked 10 Oct 2026
- 4allenai/procthor on GitHub (repository)Repository · checked 10 Oct 2026
- 5https://procthor.allenai.org/Official site · checked 10 Oct 2026
- 6allenai/procthor-10k on GitHub (repository)Repository · checked 10 Oct 2026
- 7🏘️ ProcTHOR: Large-Scale Embodied AI Using Procedural GenerationPaper · checked 10 Oct 2026
Change history
- Created as a basic entry: identity facts checked at primary sources (phase 1 re-verification).