---
license: mit
base_model: openvla/openvla-7b
tags:
  - vla
  - robotics
  - lora
  - calvin
  - adversarial-robustness
---

# calvin-rdvla-lora-epoch0

**RD-VLA (OpenVLA-7B recurrent LoRA, k_iters=8) fine-tuned on CALVIN task_D_D — first post-epoch-0 checkpoint (step16151).**

Fine-tuned with LoRA on [CALVIN task_D_D](https://github.com/mees/calvin) for the paper:

> "Reasoning as a Double-Edged Sword: Vulnerability and Defense in VLA Pipelines"
> NeurIPS 2026 — University of Melbourne Physical AI Group

Code and probe results: [https://github.com/uom-physical-ai/calvin-vla-adversarial](https://github.com/uom-physical-ai/calvin-vla-adversarial)

## Checkpoint details

| Field | Value |
|-------|-------|
| Base model | `openvla/openvla-7b` |
| Fine-tune dataset | CALVIN task_D_D (512,077 episodes) |
| Step | 16151 |
| Epoch (training) | 0 |
| Adversarial probe step | 16151 |

## Adversarial probe results (Gaussian image noise, n=300, seed=42)

| σ | mean_RAS | p95_RAS | mean_RS |
|---|----------|---------|---------|
| 0.05 | 0.0158 | 0.0810 | 0.0354 |
| 0.10 | 0.0272 | 0.1611 | 0.0629 |
| 0.20 | 0.0673 | 0.2969 | 0.1227 |

RAS = Relative Action Shift (L2 deviation of predicted action, normalised). Lower = more robust.

## Usage

```python
from peft import PeftModel
from transformers import AutoModelForVision2Seq, AutoProcessor

base = AutoModelForVision2Seq.from_pretrained("openvla/openvla-7b")
model = PeftModel.from_pretrained(base, "uom-physical-ai/calvin-rdvla-lora-epoch0")
processor = AutoProcessor.from_pretrained("openvla/openvla-7b")
```

See the [GitHub repo](https://github.com/uom-physical-ai/calvin-vla-adversarial) for the full fine-tuning and probe pipeline.