Imperial College London · robot AI model
Instant Policy
United Kingdom · 2024
Sources retrieved: 27 October 2026 (A, B, C)
3 of 7 key figures stated, as stated by the developer
- Model type
- Not stated
- What it controls
- Arms and handsA
- Inputs
- Depth Robot stateA
- Access
- Open weights (download)A
- Parameters
- Not stated
- Base model
- Not stated
- Licence type
- Not stated
Specifications
5/14 specifications disclosed 9 not stated
5 fields hidden 4 not-stated fields with a note stay ·
Not statedThe developer publishes nothing on it
Select a source marker for source details and notes.
Release1/1 disclosed
Released19 November 2024
Source and note: ReleasedB arxiv.org/abs/2411.12633 · Retrieved 27 October 2026Note. arXiv: “Submitted on 19 Nov 2024”.
Input and output2/4 disclosed
Model typeNot stated
Source and note: Model typeB arxiv.org/abs/2411.12633 · Retrieved 27 October 2026Note. The developers do not label it vision-language-action, world action model or embodied reasoning: “We introduce Instant Policy, which learns new tasks instantly (without further training) from just one or two demonstrations, achieving ICIL through two key components.”
InputsDepth Robot state
Source and note: InputsA github.com/vv19/instant_policy · Retrieved 27 October 2026Note. README: “the segmented point clouds expressed in the world frame”, “representing poses of the gripper in the world frame”.
What it controlsArms and hands
Source and note: What it controlsA github.com/vv19/instant_policy · Retrieved 27 October 2026Note. README: “Query the model to obtain end-effector actions”; “we provide examples of how to use Instant Policy for deployment on any robotic manipulator using parallel-jaw gripper”.
Degrees of freedom controlledNot stated
Size and speed0/2 disclosed
ParametersNot stated
Control rateNot stated
Where it runs and what it is built on0/2 disclosed
On robot or cloudNot stated
Base modelNot stated
Openness1/3 disclosed
AccessOpen weights (download)
Source and note: AccessA github.com/vv19/instant_policy · Retrieved 27 October 2026Note. README: “Download pre-trained weights.” The download script fetches them from a shared Google Drive folder: “gdown --folder https://drive.google.com/drive/folders/1hfyQ0DhZ8sCLrrH7dmE4WLMIibxZGVpI?usp=sharing”.
Licence documentNot stated
Source and note: Licence documentC github.com/vv19/instant_policy/blob/main/LICENSE · Retrieved 27 October 2026Note. The code repository carries the PolyForm Noncommercial licence (“SPDX-License-Identifier: PolyForm-Noncommercial-1.0.0”); the README and the download script name no licence for the pre-trained weights.
Licence typeNot stated
Source and note: Licence typeC github.com/vv19/instant_policy/blob/main/LICENSE · Retrieved 27 October 2026Note. The code repository carries the PolyForm Noncommercial licence (“SPDX-License-Identifier: PolyForm-Noncommercial-1.0.0”); the README and the download script name no licence for the pre-trained weights.
The developer’s own claims1/2 disclosed
Learns new tasksFrom a few examples
Source and note: Learns new tasksB arxiv.org/abs/2411.12633 · Retrieved 27 October 2026Note. Developer claim, arXiv abstract: “learns new tasks instantly (without further training) from just one or two demonstrations”.
Works across robotsNot stated
Source and note: Works across robotsA github.com/vv19/instant_policy · Retrieved 27 October 2026Note. The developers describe deployment on any manipulator with a parallel-jaw gripper. README: “we provide examples of how to use Instant Policy for deployment on any robotic manipulator using parallel-jaw gripper”. Abstract: “We also show how it can serve as a foundation for cross-embodiment and zero-shot transfer to language-defined tasks.”Yes only if the developer itself says the model runs on different robot models, for example robots from two manufacturers or two of one manufacturer’s models. Several units of the same robot do not count.
Notes
Instant Policy was introduced in the paper “Instant Policy: In-Context Imitation Learning via Graph Diffusion”, cited by its developers in the Proceedings of the International Conference on Learning Representations (ICLR).
The developers trained the model on simulated pseudo-demonstrations.
The model comes from the Robot Learning Lab at Imperial College London, led by Dr Edward Johns.
Sources
One line per source, not per figure: a source has one retrieval date.
- Agithub.com/vv19/instant_policyRetrieved 27 October 2026
- Barxiv.org/abs/2411.12633Retrieved 27 October 2026
- Cgithub.com/vv19/instant_policy/blob/main/LICENSERetrieved 27 October 2026