Update pipeline tag and add paper link

README.md

@@ -1,11 +1,11 @@
 ---
-license: mit
-language:
-- en
 base_model:
 - IPEC-COMMUNITY/spatialvla-4b-224-pt
-pipeline_tag: image-text-to-text
+language:
+- en
 library_name: transformers
+license: mit
+pipeline_tag: robotics
 tags:
 - VLA
 - Foundation Vision-language-action Model
@@ -13,6 +13,16 @@ tags:
 - robotics
 ---
 
+# Paper title and link
+
+The model was presented in the paper [From Intention to Execution: Probing the Generalization Boundaries of Vision-Language-Action Models](https://huggingface.co/papers/2506.09930).
+
+# Paper abstract
+
+The abstract of the paper is the following:
+
+One promise that Vision-Language-Action (VLA) models hold over traditional imitation learning for robotics is to leverage the broad generalization capabilities of large Vision-Language Models (VLMs) to produce versatile, "generalist" robot policies. However, current evaluations of VLAs remain insufficient. Traditional imitation learning benchmarks are unsuitable due to the lack of language instructions. Emerging benchmarks for VLAs that incorporate language often come with limited evaluation tasks and do not intend to investigate how much VLM pretraining truly contributes to the generalization capabilities of the downstream robotic policy. Meanwhile, much research relies on real-world robot setups designed in isolation by different institutions, which creates a barrier for reproducibility and accessibility. To address this gap, we introduce a unified probing suite of 50 simulation-based tasks across 10 subcategories spanning language instruction, vision, and objects. We systematically evaluate several state-of-the-art VLA architectures on this suite to understand their generalization capability. Our results show that while VLM backbones endow VLAs with robust perceptual understanding and high level planning, which we refer to as good intentions, this does not reliably translate into precise motor execution: when faced with out-of-distribution observations, policies often exhibit coherent intentions, but falter in action execution. Moreover, finetuning on action data can erode the original VLM's generalist reasoning abilities. We release our task suite and evaluation code to serve as a standardized benchmark for future VLAs and to drive research on closing the perception-to-action gap. More information, including the source code, can be found at this https URL
+
 # SpatialVLA Fine-Tuned on fractal & bridge
 
 This model was produced by fine-tuning the [SpatialVLA model](IPEC-COMMUNITY/spatialvla-4b-224-pt) on the **bridge dataset** for Simpler-env benchmark.
@@ -30,6 +40,8 @@ This model was produced by fine-tuning the [SpatialVLA model](IPEC-COMMUNITY/spa
 - **Repository:** [https://github.com/SpatialVLA/SpatialVLA](https://github.com/SpatialVLA/SpatialVLA)
 - **Paper:** [SpatialVLA: Exploring Spatial Representations for Visual-Language-Action Model](https://arxiv.org/abs/2501.15830)
 - **Project Page & Videos:** [https://spatialvla.github.io/](https://spatialvla.github.io/)
+- **Project Page (INT-ACT):** [https://ai4ce.github.io/INT-ACT/](https://ai4ce.github.io/INT-ACT/)
+
 
 ## Uses
 
@@ -107,302 +119,19 @@ bash scripts/spatialvla_4b_finetune/finetune_lora.sh
 
 - SimplerEnv evaluation on Google Robot tasks.
 
-<table border="1" class="dataframe">
-  <thead>
-    <tr style="text-align: center;">
-      <th rowspan="2">Model</th>
-      <th colspan="4">Visual Matching</th>
-      <th colspan="4">Variant Aggregation</th>
-    </tr>
-    <tr style="text-align: center;">
-      <th>Pick Coke Can</th>
-      <th>Move Near</th>
-      <th>Open/Close Drawer</th>
-      <th>#Average</th>
-      <th>Pick Coke Can</th>
-      <th>Move Near</th>
-      <th>Open/Close Drawer</th>
-      <th>#Average</th>
-    </tr>
-  </thead>
-  <tbody>
-    <tr>
-      <td>RT-1 (Begin)</td>
-      <td>2.7%</td>
-      <td>5.0%</td>
-      <td>13.9%</td>
-      <td>6.8%</td>
-      <td>2.2%</td>
-      <td>4.0%</td>
-      <td>6.9%</td>
-      <td>4.2%</td>
-    </tr>
-    <tr>
-      <td>RT-1 (15%)</td>
-      <td>71.0%</td>
-      <td>35.4%</td>
-      <td>56.5%</td>
-      <td>60.2%</td>
-      <td>81.3%</td>
-      <td>44.6%</td>
-      <td>26.7%</td>
-      <td>56.2%</td>
-    </tr>
-    <tr>
-      <td>RT-1 (Converged)</td>
-      <td>85.7%</td>
-      <td>44.2%</td>
-      <td>73.0%</td>
-      <td>74.6%</td>
-      <td>89.8%</td>
-      <td>50.0%</td>
-      <td>32.3%</td>
-      <td>63.3%</td>
-    </tr>
-    <tr>
-      <td>HPT</td>
-      <td>56.0%</td>
-      <td>60.0%</td>
-      <td>24.0%</td>
-      <td>46.0%</td>
-      <td>--</td>
-      <td>--</td>
-      <td>31.0%</td>
-      <td>45.0%</td>
-    </tr>
-    <tr>
-      <td>TraceVLA</td>
-      <td>28.0%</td>
-      <td>53.7%</td>
-      <td>57.0%</td>
-      <td>42.0%</td>
-      <td>60.0%</td>
-      <td>56.4%</td>
-      <td>29.4%</td>
-      <td>39.6%</td>
-    </tr>
-    <tr>
-      <td>RT-1-X</td>
-      <td>56.7%</td>
-      <td>31.7%</td>
-      <td>59.7%</td>
-      <td>53.4%</td>
-      <td>49.0%</td>
-      <td>32.3%</td>
-      <td>35.3%</td>
-      <td>64.3%</td>
-    </tr>
-    <tr>
-      <td>RT-2-X</td>
-      <td>78.7%</td>
-      <td>77.9%</td>
-      <td>25.0%</td>
-      <td>60.7%</td>
-      <td>82.3%</td>
-      <td>79.2%</td>
-      <td>--</td>
-      <td>--</td>
-    </tr>
-  <tr>
-      <td>Octo-Base</td>
-      <td>17.0%</td>
-      <td>4.2%</td>
-      <td>22.7%</td>
-      <td>16.8%</td>
-      <td>0.6%</td>
-      <td>3.1%</td>
-      <td>1.1%</td>
-      <td>1.1%</td>
-    </tr>
-    <tr>
-      <td>OpenVLA</td>
-      <td>16.3%</td>
-      <td>46.2%</td>
-      <td>35.6%</td>
-      <td>27.7%</td>
-      <td>54.5%</td>
-      <td>47.7%</td>
-      <td>17.7%</td>
-      <td>39.8%</td>
-    </tr>
-    <tr>
-      <td>RoboVLM (zero-shot)</td>
-      <td>72.7%</td>
-      <td>66.3%</td>
-      <td>26.8%</td>
-      <td>56.3%</td>
-      <td>68.3%</td>
-      <td>56.0%</td>
-      <td>8.5%</td>
-      <td>46.3%</td>
-    </tr>
-    <tr>
-      <td>RoboVLM (fine-tuning)</td>
-      <td>77.3%</td>
-      <td>61.7%</td>
-      <td>43.5%</td>
-      <td>63.4%</td>
-      <td>75.6%</td>
-      <td>60.0%</td>
-      <td>10.6%</td>
-      <td>51.3%</td>
-    </tr>
-    <tr>
-      <td>SpatialVLA (zero-shot)</td>
-      <td><b>81.0%</b></td>
-      <td><b>69.6%</b></td>
-      <td><b>59.3%</b></td>
-      <td><b>71.9%</b></td>
-      <td><b>89.5%</b></td>
-      <td><b>71.7%</b></td>
-      <td>36.2%</td>
-      <td><b>68.8%</b></td>
-    </tr>
-    <tr>
-      <td>SpatialVLA (fine-tuning)</td>
-      <td><b>86.0%</b></td>
-      <td><b>77.9%</b></td>
-      <td>57.4%</td>
-      <td><b>75.1%</b></td>
-      <td>88.0%</td>
-      <td>72.7%</td>
-      <td>41.8%</td>
-      <td><b>70.7%</b></td>
-    </tr>
-  </tbody>
-</table>
-
+[Table 1]
 
 - SimplerEnv evaluation on WidowX Robot tasks.
 
-<table border="1" class="dataframe">
-  <thead>
-    <tr style="text-align: center;">
-      <th rowspan="2">Model</th>
-      <th colspan="2">Put Spoon on Towel</th>
-      <th colspan="2">Put Carrot on Plate</th>
-      <th colspan="2">Stack Green Block on Yellow Block</th>
-      <th colspan="2">Put Eggplant in Yellow Basket</th>
-      <th rowspan="2">#Overall Average</th>
-    </tr>
-    <tr style="text-align: center;">
-      <th>Grasp Spoon</th>
-      <th>Success</th>
-      <th>Grasp Carrot</th>
-      <th>Success</th>
-      <th>Grasp Green Block</th>
-      <th>Success</th>
-      <th>Grasp Eggplant</th>
-      <th>Success</th>
-    </tr>
-  </thead>
-  <tbody>
-    <tr>
-      <td>RT-1-X</td>
-      <td>16.7%</td>
-      <td>0.0%</td>
-      <td>20.8%</td>
-      <td>4.2%</td>
-      <td>8.3%</td>
-      <td>0.0%</td>
-      <td>0.0%</td>
-      <td>0.0%</td>
-      <td>1.1%</td>
-    </tr>
-    <tr>
-      <td>Octo-Base</td>
-      <td>34.7%</td>
-      <td>12.5%</td>
-      <td>52.8%</td>
-      <td>8.3%</td>
-      <td>31.9%</td>
-      <td>0.0%</td>
-      <td>66.7%</td>
-      <td>43.1%</td>
-      <td>16.0%</td>
-    </tr>
-    <tr>
-      <td>Octo-Small</td>
-      <td>77.8%</td>
-      <td>47.2%</td>
-      <td>27.8%</td>
-      <td>9.7%</td>
-      <td>40.3%</td>
-      <td>4.2%</td>
-      <td>87.5%</td>
-      <td>56.9%</td>
-      <td>30.0%</td>
-    </tr>
-    <tr>
-      <td>OpenVLA</td>
-      <td>4.1%</td>
-      <td>0.0%</td>
-      <td>33.3%</td>
-      <td>0.0%</td>
-      <td>12.5%</td>
-      <td>0.0%</td>
-      <td>8.3%</td>
-      <td>4.1%</td>
-      <td>1.0%</td>
-    </tr>
-    <tr>
-      <td>RoboVLM (zero-shot)</td>
-      <td>37.5%</td>
-      <td>20.8%</td>
-      <td>33.3%</td>
-      <td>25.0%</td>
-      <td>8.3%</td>
-      <td>8.3%</td>
-      <td>0.0%</td>
-      <td>0.0%</td>
-      <td>13.5%</td>
-    </tr>
-    <tr>
-      <td>RoboVLM (fine-tuning)</td>
-      <td>54.2%</td>
-      <td>29.2%</td>
-      <td>25.0%</td>
-      <td>25.0%</td>
-      <td>45.8%</td>
-      <td>12.5%</td>
-      <td>58.3%</td>
-      <td>58.3%</td>
-      <td>31.3%</td>
-    </tr>
-    <tr>
-      <td>SpatialVLA (zero-shot)</td>
-      <td><b>25.0%</b></td>
-      <td><b>20.8%</b></td>
-      <td><b>41.7%</b></td>
-      <td>20.8%</td>
-      <td><b>58.3%</b></td>
-      <td>25.0%</td>
-      <td><b>79.2%</b></td>
-      <td>70.8%</td>
-      <td><b>34.4%</b></td>
-    </tr>
-    <tr>
-      <td>SpatialVLA (fine-tuning)</td>
-      <td><b>20.8%</b></td>
-      <td>16.7%</td>
-      <td>29.2%</td>
-      <td>25.0%</td>
-      <td><b>62.5%</b></td>
-      <td>29.2%</td>
-      <td><b>100.0%</b></td>
-      <td><b>100.0%</b></td>
-      <td><b>42.7%</b></td>
-    </tr>
-  </tbody>
-</table>
+[Table 2]
 
 - Zero-shot Robot Control Evaluation on WidowX Robot.
 
-<img src="https://cdn-uploads.huggingface.co/production/uploads/6535045a910b844786a6642f/SUPyXwcdfnWranO04tulL.png" alt="perform">
+[Image 1]
 
 - Spatial Understanding Capability Evaluation.
 
-<img src="https://cdn-uploads.huggingface.co/production/uploads/6535045a910b844786a6642f/g-EfM-6M7iM9IYryUTwLA.png" alt="perform">
+[Image 2]
 
 
 ## Citation
@@ -419,4 +148,6 @@ bash scripts/spatialvla_4b_finetune/finetune_lora.sh
       primaryClass={cs.RO},
       url={https://arxiv.org/abs/2501.15830}, 
 }
-```
+```
+
+**Note:**  [Table 1] and [Table 2] refer to the tables present in the original model card. [Image 1] and [Image 2] refer to the images.  These are not recreated here due to their length and complexity.