-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathindex.html
398 lines (332 loc) · 16.4 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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="description" content="Exploring Spatial Schema Intuitions in Large Language and Vision Models">
<meta name="keywords" content="Spatial Schemas, LLM, ACL">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Exploring Spatial Schemas in Large Language Models</title>
<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">
<link rel="icon" href="./static/images/favicon.svg">
<script src="https://ajax.googleapis.com/ajax/libs/jquery/3.5.1/jquery.min.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 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="title is-1 publication-title">Exploring Spatial Schema Intuitions in Large Language and Vision
Models</h1>
<div class="is-size-5 publication-authors">
<span class="author-block">
<a href="https://scholar.google.de/citations?user=qEXBDy4AAAAJ&hl=en&oi=ao">Philipp Wicke</a><sup>*,1,2</sup>,</span>
<span class="author-block">
<a href="https://scholar.google.de/citations?user=CqvC1uoAAAAJ&hl=en&oi=ao">Lennart Wachowiak</a><sup>*,3</sup>,</span>
<span class="author-block">
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup>*</sup>Both authors contributed equally</span>
<span class="author-block"><sup>1</sup>CIS, Ludwig-Maximilian-University, Munich</span>
<span class="author-block"><sup>2</sup>Munich Center for Machine Learning (MCML)</span>
<span class="author-block"><sup>3</sup>King’s College London</span>
</div>
<div class="column has-text-centered">
<div class="publication-links">
<!-- PDF Link. -->
<span class="link-block">
<a href="https://arxiv.org/pdf/2402.00956.pdf" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fas fa-file-pdf"></i>
</span>
<span>Paper</span>
</a>
</span>
<span class="link-block">
<a href="https://arxiv.org/abs/2402.00956" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="ai ai-arxiv"></i>
</span>
<span>arXiv</span>
</a>
</span>
<!-- Code Link. -->
<span class="link-block">
<a href="https://github.com/PhilWicke/Image-Schemas-in-LLMs" class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="fab fa-github"></i>
</span>
<span>Code</span>
</a>
</span>
</div>
</div>
</div>
</div>
</div>
</div>
</section>
<section class="section">
<div class="container is-max-desktop">
<!-- Abstract. -->
<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>
Demonstrating exceptional versatility in diverse NLP tasks,
large language models exhibit notable proficiency without
task-specific fine-tuning. Despite this adaptability, the
question of embodiment in LLMs remains underexplored,
distinguishing them from embodied systems in robotics where
sensory perception directly informs physical action.
Our investigation navigates the intriguing terrain of whether
LLMs, despite their non-embodied nature, effectively capture
implicit human intuitions about fundamental, spatial building
blocks of language. We employ insights from spatial cognitive
foundations developed through early sensorimotor experiences,
guiding our exploration through the reproduction of three
psycholinguistic experiments. Surprisingly, correlations
between model outputs and human responses emerge, revealing
adaptability without a tangible connection to embodied experiences.
Notable distinctions include polarized language model responses
and reduced correlations in vision language models. This research
contributes to a nuanced understanding of the interplay between
language, sensory experiences, and cognition in cutting-edge
language models.
</p>
</div>
</div>
</div>
<!--/ Abstract. -->
<!-- Selected Experiments. -->
<div class="columns is-centered has-text-centered">
<div class="column">
<h2 class="title is-3">Reproduced Experiments</h2>
</div>
</div>
<div class="columns is-centered has-text-centered">
<div class="column">
<h3 class="title is-5">Experiment 01</h3>
<p style="text-align: left;">
In the first experiment, we replicate
studies by <b>Gibbs et al. (<a
href="https://academic.oup.com/jos/article-abstract/11/4/231/1640108?login=false"
target="_blank">1994</a>)</b>,
who explore intuitions about image schemas with the verb <i>to stand</i>.
In their study, participants rated relatedness between 32 phrases and image schemas
(balance, verticality, center-periphery, resistance, linkage) on
a Likert scale (1-7). Each participant rated all phrases for one schema before moving
to the next, with schema orders counterbalanced. The experiment was
repeated with a synonym replacing <i>stand</i> in each phrase. Phrases are represented
by image schema profiles, showing average participant-rated relatedness that can be correlated with LLM performance. <br><br>
<b>Example: </b><br>
<tt>Consider the notion of Verticality. Verticality refers to the
sense of an extension along an up-down orientation. How strongly
is the phrase "stand at attention" related to this notion on a
scale from 1 (not at all related) to 7 (very strongly related)?</tt>
</p>
</div>
<div class="column">
<h3 class="title is-5">Experiment 02</h3>
<p style="text-align: left;">
The second experiment replicates a study by <b>Beitel et al. (<a
href="https://www.jbe-platform.com/content/books/9789027283696-cilt.177.11bei"
target="_blank">2001</a>)</b>,
The authors adopt the experimental approach of Experiment 01 but focus
on phrases containing the preposition <i>on</i>. They introduce a new
set of image schemas (support, pressure, constraint, covering, visibility)
relevant to this context. Unlike the general definitions, participants
now refer to example sentences, each accompanied by statements explaining
how the image schema relates to it. For instance, support is introduced
with an example: "In the case of the use of 'on' in 'the book is on the
desk': the support relation refers to the desk supporting the book." <br><br>
<b>Example: </b><br>
<tt>In the case of the use of "on" in "the book is on the desk": the VISIBILITY relation refers to the book being visible on the desk.
On a scale from 1 (not at all appropriate) to 7 (very appropriate), how appropriate is the concept VISIBILITY in regards to the phrase: "There is a physician on call"?</tt>
</p>
<!-- Content for Exp 02 -->
</div>
<div class="column">
<h3 class="title is-5">Experiment 03</h3>
<!-- Content for Exp 03 -->
<p style="text-align: left;">
In their study, <b>Richardson et al. (<a href="https://escholarship.org/uc/item/9vs820bx"
target="_blank">2001</a>)</b>
offer experimental support for image schemas through a task where
participants select images corresponding to 30 verbs categorized by
concreteness and primary directionality (horizontal, vertical, neutral).
In their study, participants choose from four images
marked with arrows to indicate directionality on the horizontal
and vertical axes. The results are analyzed in relation to
the primary direction (horizontal/vertical). The study demonstrates
the connection between linguistic concreteness and spatial
representation in a visual task, requiring an evaluation using state-of-the-art vision language models. <br><br>
<b>Example (Pseudo-Visual Condition): </b><br>
<tt>Given the event "lifted", which of the following arrows best represents this event: ↑, ↓, ←, →. A research participant would choose the arrow: </tt>
</p>
</div>
</div>
<!--/ Selected Experiments. -->
<!-- Paper video. -->
<div class="columns is-centered has-text-centered">
<div class="column is-four-fifths">
<h2 class="title is-3">Selected Models</h2>
<div class="publication-image">
<img src="static/images/models.png" alt="Selected Models Image" width="80%">
</div>
</div>
</div>
<!--/ Paper video. -->
</div>
</section>
<!-- Exp 1 and 2 -->
<section class="section has-text-centered" id="scientific-data-viewer">
<div class="container">
<h2 class="title">Results: Experiment 01 and 02</h2>
<div id="conditionButtons">
<button class="conditionButton text" onclick="selectCondition2('Exp1')">Experiment 01</button>
<button class="conditionButton pseudo-visual" onclick="selectCondition2('Exp2')">Experiment 02</button>
</div>
<div id="modelButtons2" class="hidden">
<!-- Models will be dynamically added here based on the selected condition -->
</div>
<div id="imageContainer">
<img id="selectedImage2" src="images/default2.png" alt="Click on a model to select results">
</div>
</div>
<script>
const models2 = {
Exp1: ['Llama-2-7b-chat', 'Llama-2-13b-chat', 'Llama-2-70b-chat', 'GPT-3-base', 'GPT-3-instruct', 'GPT-3-instruct(logprobs-averaged)', "GPT-4"],
Exp2: ['Llama-2-7b-chat', 'Llama-2-13b-chat', 'Llama-2-70b-chat', 'GPT-3-base', 'GPT-3-instruct', 'GPT-3-instruct(logprobs-averaged)', "GPT-4"]
};
function selectCondition2(condition) {
var modelButtons = document.getElementById("modelButtons2");
modelButtons.innerHTML = "";
// Add each model as a button
models2[condition].forEach(model => {
var button = document.createElement("button");
button.className = "modelButton";
button.textContent = model;
button.onclick = function () {
selectModel2(condition, model);
};
modelButtons.appendChild(button);
});
// Show the model buttons
modelButtons.classList.remove("hidden");
updateImage2(condition, models2[condition][0]);
}
function selectModel2(condition, model) {
var modelButtons = document.getElementById("modelButtons2");
// Toggle the selected state for the model buttons
modelButtons.childNodes.forEach(button => {
button.classList.remove("selectedModel");
if (button.textContent === model) {
button.classList.add("selectedModel");
}
});
updateImage2(condition, model);
}
function updateImage2(condition, model) {
var imagePath = `images/${condition}/${condition}_${model}.png`;
document.getElementById("selectedImage2").src = imagePath;
}
</script>
</section>
<!-- Add this section to the existing HTML code -->
<section class="section has-text-centered" id="scientific-data-viewer">
<div class="container">
<h2 class="title">Results: Experiment 03</h2>
<div id="conditionButtons">
<button class="conditionButton text" onclick="selectCondition('TEXT')">TEXT</button>
<button class="conditionButton pseudo-visual" onclick="selectCondition('PSEUDO-VISUAL')">PSEUDO-VISUAL</button>
<button class="conditionButton visual" onclick="selectCondition('VISUAL')">VISUAL</button>
</div>
<div id="modelButtons" class="hidden">
<!-- Models will be dynamically added here based on the selected condition -->
</div>
<div id="imageContainer">
<img id="selectedImage" src="images/default.png" alt="Click on a model to select results">
</div>
</div>
<script>
const models = {
TEXT: ['Llama-2-13b', 'Llama-2-70b', 'Llama-2-13b-chat', 'Llama-2-70b-chat', 'GPT-3-base', 'GPT-3-instruct', "GPT-4"],
"PSEUDO-VISUAL": ['Llama-2-13b', 'Llama-2-70b', 'Llama-2-13b-chat', 'Llama-2-70b-chat', 'GPT-3-base', 'GPT-3-instruct', "GPT-4"],
VISUAL: ['idefics-80b', 'idefics-80b-instruct', 'GPT-4-vision']
};
function selectCondition(condition) {
var modelButtons = document.getElementById("modelButtons");
modelButtons.innerHTML = "";
// Add each model as a button
models[condition].forEach(model => {
var button = document.createElement("button");
button.className = "modelButton";
button.textContent = model;
button.onclick = function () {
selectModel(condition, model);
};
modelButtons.appendChild(button);
});
// Show the model buttons
modelButtons.classList.remove("hidden");
updateImage(condition, models[condition][0]);
}
function selectModel(condition, model) {
var modelButtons = document.getElementById("modelButtons");
// Toggle the selected state for the model buttons
modelButtons.childNodes.forEach(button => {
button.classList.remove("selectedModel");
if (button.textContent === model) {
button.classList.add("selectedModel");
}
});
updateImage(condition, model);
}
function updateImage(condition, model) {
var imagePath = `images/${condition}/${condition}_${model}_Choices.png`;
document.getElementById("selectedImage").src = imagePath;
}
</script>
</div>
</section>
<section class="section" id="BibTeX">
<div class="container is-max-desktop content">
<h2 class="title">BibTeX</h2>
<pre><code>@article{wickewacho2024exp,
author = {Wicke, Philipp and Wachowiak, Lennart},
title = {Exploring Spatial Schemas in Large Language Models},
booktitle = {Findings of the Association for Computational Linguistics: ACL 2024},
address = {Bangkok, Thailand},
publisher = {Association for Computational Linguistics},
journal = {arXiv preprint arXiv:2402.00956},
year = {2024}
}</code></pre>
</div>
</section>
<footer class="footer">
<div class="container">
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>
This website was built using <a href="https://github.com/nerfies/nerfies.github.io">nerfies code</a>.
</p>
</div>
</div>
</div>
</div>
</footer>
</body>
</html>