-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathnora.html
292 lines (264 loc) · 13.3 KB
/
nora.html
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
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="generator" content="Hugo 0.66.0" />
<meta name="viewport" content="width=device-width, initial-scale=1">
<link href="https://fonts.googleapis.com/css?family=Roboto:300,400,600" rel="stylesheet" type="text/css">
<link rel="stylesheet" href="https://cdnjs.cloudflare.com/ajax/libs/highlight.js/8.4/styles/github.min.css">
<!-- <link rel="stylesheet" href="../css/normalize.css"> -->
<!-- <link rel="stylesheet" href="../css/skeleton.css"> -->
<!-- <link rel="stylesheet" href="../css/custom.css"> -->
<link rel="stylesheet" href="https://cdn.jsdelivr.net/npm/[email protected]/dist/css/bootstrap.min.css"
integrity="sha384-Gn5384xqQ1aoWXA+058RXPxPg6fy4IWvTNh0E263XmFcJlSAwiGgFAW/dAiS6JXm" crossorigin="anonymous">
<link rel="alternate" href="index.xml" type="application/rss+xml" title="Speech Research">
<!-- <link rel="shortcut icon" href="favicon.png" type="image/x-icon" /> -->
<title>NORA: A Small Open-Sourced Generalist Vision Language Action Model for Embodied Tasks</title>
<style>
figcaption {
color: inherit;
font-size: inherit;
font-family: inherit;
}
table {
font-family: inherit;
font-size: inherit;
}
thead {
background-color: inherit;
border-bottom: inherit;
}
h2 {
font-size: 1.5rem;
}
.container {
color: #061E61;
}
</style>
</head>
<font size="5">
<div class="container">
<header role="banner">
</header>
<main role="main">
<article itemscope itemtype="https://schema.org/BlogPosting">
<br>
<h1 itemprop="headline" align="center">
<font color="000093" size="7">NORA: A Small Open-Sourced Generalist Vision Language Action Model for Embodied Tasks</font>
</h1>
<br>
<p style="line-height:1" align="center"><b>
<font color="061E61">Chia-Yu Hung<sup>1</sup>, Qi Sun<sup>1</sup>, Pengfei Hong<sup>1</sup></font>
</b></p>
<p style="line-height:1" align="center"><b>
<font color="061E61">Amir Zadeh<sup>2</sup>, Chuan Li<sup>2</sup></font>
</b></p>
<p style="line-height:1" align="center"><b>
<font color="061E61">U-Xuan Tan<sup>1</sup>, Navonil Majumder<sup>1</sup>, Soujanya Poria<sup>1</sup></font>
</b></p>
<p style="line-height:0.6" align="center">
<font color="061E61"><sup>1</sup>DeCLaRe Lab, Singapore University of Technology and Design, Singapore</font>
</p>
<p style="line-height:0.6" align="center">
<font color="061E61"><sup>2</sup>Lambda Labs</font>
</p>
<!-- <p style="line-height:0.6" align="center">
<font color="061E61"> <sup>2</sup>Department of EEE, Imperial College London, London, UK</font>
</p>
<p style="line-height:1" align="center">
<font color="061E61"><sup>*</sup>Equal Contribution</font>
</p> -->
<section itemprop="entry-text">
<br>
<div class="container">
<center>
<p><a href="https://arxiv.org/abs/2504.19854">[Paper on ArXiv]</a> <a href="https://github.com/declare-lab/nora">[Code on GitHub]</a> <a href="https://huggingface.co/collections/declare-lab/nora-6811ba3e820ef362d9eca281">[Hugging Face]</a></p>
</center>
</div>
<h2 id="abstract">
<font color="000093">Abstract</font>
</h2>
<p style="text-align: justify;">
<font color="061E61">Existing Visual-Language-Action (VLA) models have shown promising performance in zero-shot scenarios, demonstrating impressive task execution and reasoning capabilities. However, a significant challenge arises from the limitations of visual encoding, which can result in failures during tasks such as object grasping. Moreover, these models typically suffer from high computational overhead due to their large sizes, often exceeding 7B parameters. While these models excel in reasoning and task planning, the substantial computational overhead they incur makes them impractical for real-time robotic environments, where speed and efficiency are paramount. Given the common practice of fine-tuning VLA models for specific tasks, there is a clear need for a smaller, more efficient model that can be fine-tuned on consumer-grade GPUs. To address the limitations of existing VLA models, we propose NORA, a 3B-parameter model designed to reduce computational overhead while maintaining strong task performance. NORA adopts the Qwen-2.5-VL-3B multimodal model as the backbone, leveraging its superior visual-semantic understanding to enhance visual reasoning and action grounding. Additionally, our NORA is trained on 970k real-world robot demonstrations and equipped with the FAST+ tokenizer for efficient action sequence generation. Experimental results demonstrate that NORA outperforms existing large-scale VLA models, achieving better task performance with significantly reduced computational overhead, making it a more practical solution for real-time robotic autonomy.</font>
</p>
<!-- <h2 id="note"> -->
<!-- <font color="000093">Note</font> -->
<!-- </h2> -->
<!-- <li> -->
<!-- <font color="061E61"> TANGO generates text-conditional sound effects, including human speech, and music.</font> -->
<!-- </li> -->
<!-- <li> -->
<!-- <font color="061E61"> The LDM is trained on a four A6000 GPUs, with text supervision from instruction tuned LLM -- FLAN-T5. </font> -->
<!-- </li> -->
<!-- <li> -->
<!-- <font color="061E61"> Despite training TANGO's LDM with 63x less data, it manages to produce superior sound quality to the baselines</font> -->
<!-- </li> -->
<br>
<figure>
<p align="center"><img src="../NORA.png" width="100%" class="center" /></p>
<figcaption>
<p style="text-align: justify">
<font color="061E61"><b>Figure 1:</b> NORA, as depicted in this figure, has three major components: (i) image encoder, (ii) vision language model, and (iii) FAST+ action tokenizer. The image encoder encodes the current state of the environment. Subsequently, the VLM predicts the next action in order to accomplish the input goal, given the current state. Thereafter, FAST+ decodes the VLM output tokens into actionable robot tokens.</font>
</p>
</figcaption>
</figure>
<br>
<p id="demos"><b>
<font color="061E61">Examples of NORA in action:</font>
</b></p>
<table class="table" align="center" style="table-layout: fixed;word-break:break-word">
<tr>
<td scope="col" width="33%">
<font color="061E61"><i>Task Instruction</i></font>
</td>
<td scope="col" width="66%">
<font color="061E61"><i>Policy Execution</i></font>
</td>
</tr>
<tbody>
<tr>
<td scope="row">Put the blue cube on the plate</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/vhRcvLRs5lI"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
<tr>
<td scope="row">Put the corn and carrot in pan</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/k3seqSaGftU"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
<tr>
<td scope="row">Put banana and carrot in pot</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/8RzuMsOkw2o"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
<tr>
<td scope="row">Move the banana close to the pan</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/AFZP3-1MEvY"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
<tr>
<td scope="row">Put carrot in the pot</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/L9eydFNtEWI"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
<tr>
<td scope="row">Put the pink toy at the right corner</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/N0ZTQq3G37A"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
<tr>
<td scope="row">Put the carrot and hotdog in pot</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/QigpozIXbQ0"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
<tr>
<td scope="row">Put the blue cube on the plate</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/Xajw9eaB940"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
<tr>
<td scope="row">Put banana in pot</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/YF6RH24j5nI"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
<tr>
<td scope="row">Put the red bottle and the hamburger in the pan</td>
<td>
<iframe width="640" height="360"
src="https://www.youtube.com/embed/ftHc9aISAQw"
frameborder="0" allowfullscreen>
</iframe>
</td>
</tr>
</tbody>
</table>
<h2 id="Acknowledgement">
<font color="000093">Acknowledgement</font>
</h2>
<div class="container">
<p>
This website is created based on <a href = "https://github.com/AudioLDM/AudioLDM.github.io."> https://github.com/AudioLDM/AudioLDM.github.io </a>
</p>
</div>
</section>
</article>
</main>
</div>
</font>
<script>
(function (i, s, o, g, r, a, m) {
i['GoogleAnalyticsObject'] = r; i[r] = i[r] || function () {
(i[r].q = i[r].q || []).push(arguments)
}, i[r].l = 1 * new Date(); a = s.createElement(o),
m = s.getElementsByTagName(o)[0]; a.async = 1; a.src = g; m.parentNode.insertBefore(a, m)
})(window, document, 'script', '//www.google-analytics.com/analytics.js', 'ga');
ga('create', 'UA-139981676-1', 'auto');
ga('send', 'pageview');
</script>
<script src="//cdnjs.cloudflare.com/ajax/libs/highlight.js/8.4/highlight.min.js"></script>
<script>hljs.initHighlightingOnLoad();</script>
<script type="text/x-mathjax-config">
MathJax.Hub.Config({
HTML: ["input/TeX","output/HTML-CSS"],
TeX: {
Macros: {
bm: ["\\boldsymbol{#1}", 1],
argmax: ["\\mathop{\\rm arg\\,max}\\limits"],
argmin: ["\\mathop{\\rm arg\\,min}\\limits"]},
extensions: ["AMSmath.js","AMSsymbols.js"],
equationNumbers: { autoNumber: "AMS" } },
extensions: ["tex2jax.js"],
jax: ["input/TeX","output/HTML-CSS"],
tex2jax: { inlineMath: [ ['$','$'], ["\\(","\\)"] ],
displayMath: [ ['$$','$$'], ["\\[","\\]"] ],
processEscapes: true },
"HTML-CSS": { availableFonts: ["TeX"],
linebreaks: { automatic: true } }
});
</script>
<script type="text/x-mathjax-config">
MathJax.Hub.Config({
tex2jax: {
skipTags: ['script', 'noscript', 'style', 'textarea', 'pre', 'code']
}
});
</script>
<script type="text/javascript" async
src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.4/MathJax.js?config=TeX-MML-AM_CHTML">
</script>
</html>