-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
447 lines (390 loc) · 19.8 KB
/
index.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
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<!-- Meta tags for social media banners, these should be filled in appropriatly as they are your "business card" -->
<!-- Replace the content tag with appropriate information -->
<meta name="description" content="Project website for the exploratory search for learning representations project">
<meta property="og:title" content="CLEA: Contrastive Learning from Exploratory Actions"/>
<meta property="og:description" content="Project website for the exploratory search for learning representations project"/>
<meta property="og:url" content="https://interaction-lab.github.io/CLEA/"/>
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X630-->
<meta property="og:image" content="static/image/your_banner_image.png" />
<meta property="og:image:width" content="1200"/>
<meta property="og:image:height" content="630"/>
<meta name="twitter:title" content="CLEA: Contrastive Learning from Exploratory Actions">
<meta name="twitter:description" content="Project website for the exploratory search for learning representations project">
<!-- Path to banner image, should be in the path listed below. Optimal dimenssions are 1200X600-->
<meta name="twitter:image" content="static/images/front-figure.png">
<meta name="twitter:card" content="summary_large_image">
<!-- Keywords for your paper to be indexed by-->
<meta name="keywords" content="Contrastive Learning, Exploratory Search, Human Preferences">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>CLEA: Contrastive Learning from Exploratory Actions</title>
<link rel="icon" type="image/x-icon" href="static/images/favicon.ico">
<link href="https://fonts.googleapis.com/css?family=Google+Sans|Noto+Sans|Castoro"
rel="stylesheet">
<link rel="stylesheet" href="static/css/bulma.min.css">
<link rel="stylesheet" href="static/css/bulma-carousel.min.css">
<link rel="stylesheet" href="static/css/bulma-slider.min.css">
<link rel="stylesheet" href="static/css/fontawesome.all.min.css">
<link rel="stylesheet"
href="https://cdn.jsdelivr.net/gh/jpswalsh/academicons@1/css/academicons.min.css">
<link rel="stylesheet" href="static/css/index.css">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.js"></script>
<script src="https://documentcloud.adobe.com/view-sdk/main.js"></script>
<script defer src="static/js/fontawesome.all.min.js"></script>
<script src="static/js/bulma-carousel.min.js"></script>
<script src="static/js/bulma-slider.min.js"></script>
<script type="text/javascript" async src="https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.7/MathJax.js?config=TeX-MML-AM_CHTML"></script>
<script type="text/x-mathjax-config">
MathJax.Hub.Config({
tex2jax: {
inlineMath: [['$','$'], ['\\(','\\)']],
processEscapes: true
}
});
</script>
<script src="static/js/index.js"></script>
</head>
<body>
<section class="hero">
<div class="hero-body">
<div class="container is-max-desktop">
<div class="columns is-centered">
<div class="column has-text-centered">
<h1 class="xtitle is-size-2 publication-title">Contrastive Learning from Exploratory Actions:</h1>
<h2 class="xtitle is-size-3 publication-title">Leveraging Natural Interactions for Preference Elicitation</h2>
<div class="is-size-5 publication-authors">
<!-- Paper authors -->
<span class="author-block">
<a href="https://ndennler.github.io/" target="_blank">Nathaniel Dennler</a></span>
<span class="author-block">
<a href="https://stefanosnikolaidis.net/" target="_blank">Stefanos Nikolaidis</a></span>
<span class="author-block">
<a href="https://maja-mataric.web.app/" target="_blank">Maja Matarić</a>
</span>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block">University of Southern California<br>2025 <i>IEEE/ACM Conference on Human-Robot Interaction (HRI)</i></span>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- Arxiv PDF link -->
<span class="link-block">
<a href="static/pdfs/HRI_CLEA.pdf" target="_blank"
class="external-link ">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Paper</span>
</a>
</span>
<!-- Supplementary PDF link -->
<span class="link-block">
<a href="static/pdfs/HRI_CLEA_appendix.pdf" target="_blank"
class="external-link ">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Supplementary</span>
</a>
</span>
<!-- Github link -->
<span class="link-block">
<a href="index.html" target="_blank"
class="external-link ">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code (Coming Soon!)</span>
</a>
</span>
<!-- ArXiv abstract Link -->
<span class="link-block">
<a href="https://github.com/interaction-lab/CLEA/tree/main/static/dataset" target="_blank"
class="external-link ">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Signal Dataset</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<!-- <section class="hero teaser">
<div class="container is-max-desktop">
<div class="marquee-wrapper">
<div class="container">
<div class="marquee-block">
<div class="marquee-inner to-left">
<span>
<div class="marquee-item">
<img src="static/images/carousel1.jpg" alt="MY ALT TEXT"/>
<p class="text-white">slay</p>
</div>
<div class="marquee-item">
<img src="static/images/carousel2.jpg" alt="MY ALT TEXT"/>
<p class="text-white">yas</p>
</div>
<div class="marquee-item">
<img src="static/images/carousel3.jpg" alt="MY ALT TEXT"/>
<p class="text-white">3</p>
</div>
<div class="marquee-item">
<img src="static/images/carousel4.jpg" alt="MY ALT TEXT"/>
<p class="text-white">4</p>
</div>
<div class="marquee-item">
<img src="static/images/carousel5.jpg" alt="MY ALT TEXT"/>
<p class="text-white">5</p>
</div>
</span>
<span>
<div class="marquee-item">
<img src="static/images/carousel1.jpg" alt="MY ALT TEXT"/>
<p class="text-white">1</p>
</div>
<div class="marquee-item">
<img src="static/images/carousel2.jpg" alt="MY ALT TEXT"/>
<p class="text-white">2</p>
</div>
<div class="marquee-item">
<img src="static/images/carousel3.jpg" alt="MY ALT TEXT"/>
<p class="text-white">3</p>
</div>
<div class="marquee-item">
<img src="static/images/carousel4.jpg" alt="MY ALT TEXT"/>
<p class="text-white">4</p>
</div>
<div class="marquee-item">
<img src="static/images/carousel5.jpg" alt="MY ALT TEXT"/>
<p class="text-white">5</p>
</div>
</span>
</div>
</div>
</div>
</section> -->
<!-- Paper abstract -->
<section class="section hero is-light">
<div class="container is-max-desktop">
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Abstract</h2>
<div class="content has-text-justified">
<p>
Robots that interact with humans must adapt to diverse user preferences.
Learning representations of robot behaviors can facilitate user-driven customization of the robot, but machine learning techniques require large amounts of manually labelled data.
Manually labelled data can be difficult to obtain because users are often unmotivated to engage in monotonous labeling processes.
In this work, we identify that <b>users learning to use a new robot automatically engage in <i>exploratory search processes</i> that generate data which can be used in place of manually-labelled data</b>.
We propose a method to learn representations called <i>Contrastive Learning from Exploratory Actions</i> (CLEA) that leverages this exploratory search data to learn representations of robot behaviors that facilitate user-driven customization.
We show that CLEA can learn representations that satisfy the criteria of effective robot representations: completeness, simplicity, minimality, and interpretability.
CLEA representations outperform self-supervised representations in their completeness, simplicity, minimality, and interpretability.
</p>
</div>
</div>
</div>
</div>
</section>
<!-- End paper abstract -->
<!-- Teaser video-->
<section class="hero teaser">
<div class="container is-max-desktop">
<div class="hero-body">
<br>
<video poster="" id="video1" autoplay controls loop height="100%">
<source src="static/videos/UserStudyInfo.mp4"
type="video/mp4">
</video>
</div>
</div>
</section>
<!-- End teaser video -->
<section class="hero teaser">
<div class="container is-max-desktop">
<div class="hero-body">
<h2 class="subtitle has-text-centered">
The above video shows the interface users used to design signals for a robot.
The robot helps users with an item finding task, and users showed diverse preferences and expectations for how the robot
should behave. While using this interface, users subtly expressed their preferences for the robot's behavior by exploring
different signal options before finally settling on their preferred signal.
In this work, <b>we harness this subtle source of information to learn user-aligned representations of robot behaviors</b>.
By using features that are more aligned with users, these users can more easily customize the robot to their preferences. Because it is easier to design signals,
users are then able to provide even more information on their preferences. The cycle results in continously improving both the robot customization process and the features that facilitate robot learning.
</h2>
<img src="static/images/framework.png" alt="an overview of the proposed framework. "/>
<h2 class="subtitle has-text-centered">
<a href="https://dl.acm.org/doi/10.1145/3610977.3634987">A good feature representation has four criteria</a>: completeness, simplicity, minimality, and explainability.
<b>Completeness</b> refers to the ability of the representation to capture all the information needed to predict the user's preferences.
<b>Simplicity</b> refers to the ability of the representation to reflect user's preferences with a simple linear model.
<b>Minimality</b> refers to the ability of the representation to be concise and not contain redundant information.
<b>Explainability</b> refers to the ability of the representation to be readily used with existing explainability techniques, for example "explanation by example".
</h2>
<h2 class="subtitle has-text-centered">
A simple alternative to CLEA is to learn feature representations directly from data using self-supervised approaches like autoencoders or variational autoencoders.
Other options are to use pre-trained models such as X-CLIP or Audio Spectrogram Transformers. We found that <b>using the data directly from users
allows us to learn representations that are better than these baselines </b> on all four criteria, highlighting the importance of collecting diverse user data.
</h2>
<h2 class="subtitle has-text-centered">
To evaluate CLEA, we collected a dataset of user rankings for robot signals. We then
trained a model to predict these rankings using the learned representations. We show the results for
the four criteria below.
</h2>
<img src="static/images/completeness.png" alt="a graph of test preference accuracy for each method"/>
<h2 class="subtitle has-text-centered">
<b>Completeness.</b> We show that CLEA representations are more complete than self-supervised representations in the graph above.
We evaluated this using the <i>Test Preference Accuracy</i> (TPA), the ability of a model to predict user choices on a held-out test set.
We show that CLEA representations are more complete than self-supervised representations because they achieve a higher TPA.
</h2>
<img src="static/images/minimality+simplicity.png" alt="A table and graph of AUC alignemnt for each method. The table contains dimensions 8,16,32,64,128. The graph shows the results for dimension 8 in detail."/>
<h2 class="subtitle has-text-centered">
<b>Simplicity and Minimality.</b> We show that CLEA representations are simpler and more minimal than self-supervised representations in the graph above.
We evaluated this using the <i>Area Under the Alignment Curve</i> (AUC Alignment). AUC Alignment measures the ability of a model to predict user choices on a held-out test set using a simple linear model.
We learn the linear model by using a bayesian update rule to update the model's weights. We show that CLEA representations are <b>simpler</b> than self-supervised representations because
they achieve a higher AUC Alignment across all dimesions and modalities. They are more <b>minimal</b> than self-supervised approaches because they achieve a higher AUC Alignment with fewer dimensions.
</h2>
<img src="static/images/explainability.png" alt="A graph of cosine similarities between top-ranked signals and exemplar signals"/>
<h2 class="subtitle has-text-centered">
<b>Explainability.</b> We adopt the framing of explanation by example to evaluate explainability.
A feature space is explainable if test examples are nearby training examples. We use the cosine similarity between
the top-ranked signals and exemplar signals to evaluate explainability. We found that the CLEA representations
are more explainable than self-supervised representations because they achieve a higher cosine similarity.
</h2>
<h2 class="subtitle has-text-centered">
Future work may explore how feature spaces may transfer across different robot tasks,
across different robot embodiments, or across users. We hope that the future of robotics
will be more user-friendly, and that robots will be able to adapt to diverse user preferences.
</h2>
</div>
</section>
<!-- Image carousel -->
<!-- <section class="hero is-small">
<div class="hero-body">
<div class="container">
<div id="results-carousel" class="carousel results-carousel">
<div class="item">
<img src="static/images/carousel1.jpg" alt="MY ALT TEXT"/>
<h2 class="subtitle has-text-centered">
First image description.
</h2>
</div>
<div class="item">
<img src="static/images/carousel2.jpg" alt="MY ALT TEXT"/>
<h2 class="subtitle has-text-centered">
Second image description.
</h2>
</div>
<div class="item">
<img src="static/images/carousel3.jpg" alt="MY ALT TEXT"/>
<h2 class="subtitle has-text-centered">
Third image description.
</h2>
</div>
<div class="item">
<img src="static/images/carousel4.jpg" alt="MY ALT TEXT"/>
<h2 class="subtitle has-text-centered">
Fourth image description.
</h2>
</div>
</div>
</div>
</div>
</section> -->
<!-- End image carousel -->
<!-- Youtube video -->
<!-- <section class="hero is-small is-light">
<div class="hero-body">
<div class="container">
<!-- Paper video. -->
<!-- <h2 class="title is-3">Video Presentation</h2>
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<div class="publication-video">
<iframe src="https://www.youtube.com/embed/IOJ1xDy0DEE?si=0sbrTXgzy0CU5u8O" frameborder="0" allow="autoplay; encrypted-media" allowfullscreen></iframe>
</div>
</div>
</div>
</div>
</div>
</section> -->
<!-- End youtube video -->
<!-- Video carousel -->
<!-- <section class="hero is-small">
<div class="hero-body">
<div class="container">
<h2 class="title is-3">Another Carousel</h2>
<div id="results-carousel" class="carousel results-carousel">
<div class="item item-video1">
<video poster="" id="video1" autoplay controls muted loop height="100%">
<source src="static/videos/carousel1.mp4"
type="video/mp4">
</video>
</div>
<div class="item item-video2">
<video poster="" id="video2" autoplay controls muted loop height="100%">
<source src="static/videos/carousel2.mp4"
type="video/mp4">
</video>
</div>
<div class="item item-video3">
<video poster="" id="video3" autoplay controls muted loop height="100%">\
<source src="static/videos/carousel3.mp4"
type="video/mp4">
</video>
</div>
</div>
</div>
</div>
</section> -->
<!-- End video carousel -->
<!-- Paper poster
<section class="hero is-small is-light">
<div class="hero-body">
<div class="container">
<h2 class="title">Poster</h2>
<iframe src="static/pdfs/sample.pdf" width="100%" height="550">
</iframe>
</div>
</div>
</section>
End paper poster -->
<!--BibTex citation -->
<!-- <section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>
@inproceedings{dennler2024exploratory,
title={Using Exploratory Search to Learn Represerations for Human Preferences},
author={Dennler, Nathaniel and Nikolaidis, Stefanos and Matari{\'c}, Maja},
booktitle={Companion of the 2024 ACM/IEEE International Conference on Human-Robot Interaction},
year={2024}
}</code></pre>
</div>
</section> -->
<!--End BibTex citation -->
<footer class="footer">
<div class="container">
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>
This page was built using the <a href="https://github.com/vinthony/project-page-template">modification version</a> of <a href="https://github.com/eliahuhorwitz/Academic-project-page-template" target="_blank">Academic Project Page Template</a> from <a href="https://github.com/vinthony">vinthony</a>.
You are free to borrow the code for this website, we just ask that you link back to this page in the footer. <br> This website is licensed under a <a rel="license" href="http://creativecommons.org/licenses/by-sa/4.0/" target="_blank">Creative
Commons Attribution-ShareAlike 4.0 International License</a>.
</p>
</div>
</div>
</div>
</div>
</footer>
<!-- Statcounter tracking code -->
<!-- You can add a tracker to track page visits by creating an account at statcounter.com -->
<!-- End of Statcounter Code -->
</body>
</html>