File size: 2,574 Bytes
a1323f4 dd641b1 a1323f4 dd641b1 a1323f4 4065962 a1323f4 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 | ---
license: cc-by-sa-4.0
library_name: n0vtla
tags:
- robotics
- vision-language-action
- tactile
- manipulation
---
# N0-VTLA - NeoSim Grasp Chip
Task policy for [N0-VTLA](https://github.com/neoteai/N0-VTLA), a vision-tactile-language-action
model that conditions a flow-matching action expert on predicted latent tactile tokens.
This is a **task policy, not a pretrained base**. For post-training on your own robot start from
[n0-vtla-base](https://huggingface.co/NeoteAI/n0-vtla-base).
| | |
|---|---|
| Config | `sim_single_arm_tactile` |
| Tactile pathway | enabled, `n_latent=5`, views `(tactile_a, tactile_b)` |
| Action space | 8-dim joint |
Single-task policy for the NeoSim `grasp_chip` task.
**100%**, first attempt.
### Serving
`action_horizon` is **16**, which differs from the config default, so it must be overridden:
```bash
VTLA_ASSET_ID=univtac_grasp_chip_joint_h16_norm \
python scripts/serve_zmq.py --config sim_single_arm_tactile \
--ckpt <this-dir> --action-horizon 16 --addr "tcp://127.0.0.1:5557"
```
Set `exec_horizon: 16` in the deploy YAML to match. Prompt: `grasp chip`
## Evaluation protocol
Measured on the UniVTAC simulator at commit `695a22d`
(branch `NeoSim` of [anlorla/UniVTAC](https://github.com/anlorla/UniVTAC)), on held-out seeds
starting at 100. Success criteria on three tasks were tightened after these numbers were
measured, so a success rate on this benchmark is not comparable without the simulator commit
beside it; see
[docs/EVAL.md](https://github.com/neoteai/N0-VTLA/blob/main/docs/EVAL.md) for the details and for
the full evaluation procedure.
## Caveat on the tactile pathway
This checkpoint carries the tactile pathway, but this benchmark cannot demonstrate that touch
contributes to the score. Object randomisation is +/-2-5 mm with no domain randomisation, so a
policy that ignores its cameras and its tactile sensors entirely can still score well. Use
`scripts/probe_z_tactile_dependence.py` to measure the causal contribution yourself.
## Citation
If you use this model in your research, please cite the [N0-VTLA paper](https://arxiv.org/abs/2607.23782):
```bibtex
@misc{neoteaiteam2026n0vtlascalingvisiontactilelanguageactionmodel,
title={$N_0$-VTLA: Scaling Vision-Tactile-Language-Action Model with Latent Tactile Tokens},
author={NeoteAI Team and Fudan TEAI Team},
year={2026},
eprint={2607.23782},
archivePrefix={arXiv},
primaryClass={cs.RO},
url={https://arxiv.org/abs/2607.23782},
}
```
## License
CC BY-SA 4.0, as the parent repository.
|