Skip to content
RobothecaBeta
Log in

PRTS

  • AI model
  • Humanoids and robot learning

PRTS is a vision-language-action foundation model built on a Qwen3-VL backbone, pre-trained with reward-label-free contrastive RL to judge how close a state is to a language goal.

  • Non-commercial
  • Non-commercial weights

Made by TeleHuman (TeleAI), China (checked on github.com/TeleHuman (opens in a new tab)).

Source on GitHub: github.com/TeleHuman/PRTS (opens in a new tab). Facts read at commit 15cd007 (opens in a new tab).

Licenses

Code
CC-BY-NC-4.0 (opens in a new tab)

Read by Robotheca from LICENSE (opens in a new tab) (CC-BY-NC-4.0) at commit 15cd007 (opens in a new tab), on 16 Oct 2026.

Commercial use
No

The code and weights licenses do not allow commercial use.

Labels

  • Non-commercialThe code license (CC-BY-NC-4.0) does not allow commercial use.
  • Non-commercial weightsThe weights license (CC-BY-NC-4.0) does not allow commercial use.

What it does

  • Adds a contrastive value head trained alongside the action loss, using only the temporal structure of demonstrations.
  • Ships a pre-trained PRTS-4B checkpoint on Hugging Face, plus a LIBERO post-trained checkpoint.
  • Provides fine-tuning code driven by YAML configs over LeRobot v2.1 datasets, with a DeepSpeed launch script.
  • Includes a websocket policy server with an EnvMode registry for serving fine-tuned checkpoints.
  • Reports results on LIBERO, LIBERO-Plus, LIBERO-Pro, SimplerEnv and a MolmoSpaces leaderboard entry.

Written by Robotheca from the project's README (opens in a new tab) at commit 15cd007.

Getting started

Difficulty
Advanced

Maintenance

Status
Maintained: a commit in the last 12 months
Newest commit
2 Jun 2026 on main
Latest release
None published on GitHub

Checked on GitHub on 16 Oct 2026.

Links

Robots on Robotheca that use it

No robot on Robotheca is linked to it yet. Links come from a robot's robot.yaml, from its dependency files read as text, or from Robotheca's editors.

Robotheca lists AI models as a reference: nothing is hosted or run here, and every fact links to where it was read.