File size: 2,940 Bytes
c6a5d52
7adbe1b
 
 
c6a5d52
 
 
 
 
726c193
c6a5d52
726c193
 
 
c6a5d52
 
 
 
 
726c193
c6a5d52
 
 
726c193
c6a5d52
 
 
 
 
 
7adbe1b
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
---
license: other
license_name: nvidia-oneway-noncommercial-license
license_link: https://developer.download.nvidia.com/licenses/NVIDIA-OneWay-Noncommercial-License-22Mar2022.pdf
base_model: nvidia/GR00T-N1.6-3B
pipeline_tag: robotics
tags: [robotics, vla, gr00t, gr00t-n1.6, manipulation, maniguard, franka]
---

# GR00T-N1.6 - Sim2Real Clutter (task_0048) (joint, 2-cam)

NVIDIA Isaac **GR00T-N1.6-3B** fine-tuned on the ManiGuard **clutter-task0048** base task (sim Franka Panda).
Part of the ManiGuard VLA benchmark - GR00T vs pi0.5 on the same task families with
identical data, cameras, and controller.

## Model
- **Base:** [nvidia/GR00T-N1.6-3B](https://huggingface.co/nvidia/GR00T-N1.6-3B) - Eagle (nvidia/Eagle-Block2A-2B-v2) VLM + flow-matching DiT action head
- **Embodiment:** NEW_EMBODIMENT - Franka Panda, **8-D joint** state/action (7 arm joints + 1 gripper)
- **Cameras (2):** image_left (overview) + wrist (256x256)
- **Action:** arm = state-relative chunks (reconstructed to absolute at inference), gripper = absolute; 16-step horizon at 30 fps (0.53 s); NON_EEF (joint space)
- **Tuning:** GR00T-N1.6 default - VLM (LLM + visual) **frozen**, train projector + diffusion action head (**no LoRA**)

## Training
- single-card config, bf16, global batch 4, 50000 steps (~8 epochs over 23,904 frames), cosine LR (peak 2.5e-5, sqrt-scaled), warmup 0.05
- Data: [IDEAS-Lab-Northwestern/sim2real-clutter-task0048-sim-joint-5cam](https://huggingface.co/datasets/IDEAS-Lab-Northwestern/sim2real-clutter-task0048-sim-joint-5cam); videos decoded as H.264 for GR00T's torchcodec loader

## Usage
Load with `Gr00tPolicy` from [Isaac-GR00T (n1d6)](https://github.com/NVIDIA/Isaac-GR00T/tree/n1d6), `--embodiment-tag NEW_EMBODIMENT`. The included `processor/` carries the normalization stats + modality config.

> WARNING - Convention (must match at eval): joint-space JointController (absolute joint targets, NON_EEF) + 2 cameras (image_left overview + wrist). A mismatched controller or camera set silently feeds an out-of-distribution input.

## Paper & Citation

Part of **ManiGuard**: [paper (arXiv:2608.17386)](https://arxiv.org/abs/2608.17386) ·
[code](https://github.com/NU-IDEAS-Lab/ManiGuard) ·
[docs](https://nu-ideas-lab.github.io/ManiGuard/)

```bibtex
@misc{peng2026maniguard,
  title         = {{MANIGUARD}: A Benchmark and Data Suite for Specification-Grounded
                   Safety Evaluation and Improvement of Robotic Manipulation},
  author        = {Peng, Yiyan and Wang, Philip and Zhan, Simon Sinong and Lyu, Yiqi
                   and Ni, Zhenyang and Yan, Jixin and Wong, Fiorelli and Jiao, Ruochen
                   and Yin, Hang and Cao, Xinyu and Shao, Huajie and Li, Manling
                   and Zhang, Ruohan and Zhu, Qi},
  year          = {2026},
  eprint        = {2608.17386},
  archivePrefix = {arXiv},
  primaryClass  = {cs.RO},
  url           = {https://arxiv.org/abs/2608.17386},
}
```