-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
526 lines (470 loc) · 19.8 KB
/
Copy pathindex.html
File metadata and controls
526 lines (470 loc) · 19.8 KB
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
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML//EN">
<!-- placeholder: June 2023 checkpoint -->
<html>
<head>
<!-- Google tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-KQ4C08LNW4"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-KQ4C08LNW4');
</script>
<!-- <link href='https://fonts.googleapis.com/css?family=News+Cycle:400,700' rel='stylesheet' type='text/css'> -->
<!-- <link rel="stylesheet" href="css/bootstrap.min.css"> -->
<meta name="description" content="Cognitive Bias Ranking evaluates and compares Large Language Models across key cognitive bias categories to assess robustness in reasoning.">
<link rel="stylesheet" href="css/llmrankings.css">
<!-- <link rel="stylesheet" href="css/csrankings-extra.css"> -->
<link rel="stylesheet" href="css/modern.css">
<title>LLM Rankings</title>
<style type="text/css">
/* .row.switch-xs {
transform: rotate(180deg);
direction: rtl;
} */
@media (max-width: 768px) {
.col-centered {
display:inline-block;
float: none;
text-align: left;
margin-right:-4px;
}
.row-centered {
text-align: center;
}
}
/* .row.switch-xs > [class^="col-"] {
transform: rotate(-180deg);
direction: ltr;
}
} */
body {
font-family: Helvetica, Arial;
}
table td {
vertical-align: top;
}
.table-fit {
white-space: nowrap;
width: 5%;
}
td {
padding-top: 2px;
padding-bottom: 2px;
}
a {
color : blue ;
}
.csr-barchart {
display: none;
}
.area_group_selector { cursor : pointer; }
/* .hovertip { cursor : pointer;
color : blue } */
/* .hovertip {
cursor: pointer;
color: #2a7a4e; /* Deep indigo blue */
/* font-size: 14px;
font-weight: bold;
transition: transform 0.2s ease, color 0.2s ease;
display: inline-block;
margin-right: 6px;
} */
/* .llm-row:hover .hovertip {
color: #2a7a4e; /* Muted green on hover */
/* transform: scale(1.2); /* Slight grow effect on hover */
/*} */
.bluefont {
color : blue;
}
input { cursor : pointer; }
/* latin-ext */
@font-face {
font-family: 'News Cycle';
font-style: normal;
font-weight: 400;
src: local('News Cycle'), local('NewsCycle'), url(https://fonts.gstatic.com/s/newscycle/v13/d03oiboZGiaNuMDvH253CgsYbbCjybiHxArTLjt7FRU.woff2) format('woff2');
unicode-range: U+0100-024F, U+1E00-1EFF, U+20A0-20AB, U+20AD-20CF, U+2C60-2C7F, U+A720-A7FF;
}
/* latin */
@font-face {
font-family: 'News Cycle';
font-style: normal;
font-weight: 400;
src: local('News Cycle'), local('NewsCycle'), url(https://fonts.gstatic.com/s/newscycle/v13/9Xe8dq6pQDsPyVH2D3tMQgzyDMXhdD8sAj6OAJTFsBI.woff2) format('woff2');
unicode-range: U+0000-00FF, U+0131, U+0152-0153, U+02C6, U+02DA, U+02DC, U+2000-206F, U+2074, U+20AC, U+2212, U+2215;
}
/* latin-ext */
@font-face {
font-family: 'News Cycle';
font-style: normal;
font-weight: 700;
src: local('News Cycle Bold'), local('NewsCycle-Bold'), url(https://fonts.gstatic.com/s/newscycle/v13/G28Ny31cr5orMqEQy6ljtzrEaqfC9P2pvLXik1Kbr9s.woff2) format('woff2');
unicode-range: U+0100-024F, U+1E00-1EFF, U+20A0-20AB, U+20AD-20CF, U+2C60-2C7F, U+A720-A7FF;
}
/* latin */
@font-face {
font-family: 'News Cycle';
font-style: normal;
font-weight: 700;
src: local('News Cycle Bold'), local('NewsCycle-Bold'), url(https://fonts.gstatic.com/s/newscycle/v13/G28Ny31cr5orMqEQy6ljt2aVI6zN22yiurzcBKxPjFE.woff2) format('woff2');
unicode-range: U+0000-00FF, U+0131, U+0152-0153, U+02C6, U+02DA, U+02DC, U+2000-206F, U+2074, U+20AC, U+2212, U+2215;
}
.note {
font-weight: bold;
}
.areaname {
font-variant:small-caps;
font-size: -1;
}
.barchart {
font-size: +1;
}
</style>
<script defer>
var trackOutboundLink = function(url, new_window) {
/* Setting a default value for backward compatibility */
if (new_window === undefined) {
new_window = false;
}
ga('send', 'event', 'outbound', 'click', url, {'transport': 'beacon',
'hitCallback':
function () {
if (!new_window) {
document.location = url;
}
}
});
if (new_window){
window.open(url);
}
}
</script>
<noscript><img height="1" width="1" style="display:none"
src="https://www.facebook.com/tr?id=1437880419634625&ev=PageView&noscript=1"
/></noscript>
<script src="js/jquery-3.3.1.min.js" type="text/javascript"></script>
<script src="js/he-1.2.0.js" type="text/javascript"></script>
<script defer src="js/vega-5.20.2.min.js" type="text/javascript"></script>
<script defer src="js/vega-lite-5.1.1.min.js" type="text/javascript"></script>
<script defer src="js/vega-embed-6.18.2.min.js" type="text/javascript"></script>
<script defer src="js/papaparse.min.js" type="text/javascript"></script>
<script defer src="continents.js" type="text/javascript"></script>
<!-- <script defer src="js/bootstrap.min.js" type="text/javascript"></script>
<script defer src="js/navigo.min.js"></script>
<script src="csrankings.js" type="text/javascript" defer></script> -->
<!-- Google tag (gtag.js) -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-YCPQLE0NGZ"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-YCPQLE0NGZ');
</script>
</head>
<body>
<div style="display:none;">
<script type="text/javascript" id="mapmyvisitors"
src="//mapmyvisitors.com/map.js?d=9qw4uhD3sxNMwI2hK_oJ98FigDOu2dlnZBUzKnEly0I&cl=ffffff&w=a">
</script>
</div>
<div class="container">
<form id="rankform" name="rankform" action="" method="get" target="_blank">
<div class="page-header">
<h1>Ranking LLMs on Cognitive Biases</h1>
<div> <!-- <div class="panel panel-info"> -->
<div> <!-- <div class="panel-heading"> -->
<p class="welcome-text">
As Large Language Models (LLMs) are increasingly embedded in real-world decision-making processes, it becomes crucial to examine the extent to which they exhibit cognitive biases in their responses which are systematic distortions commonly observed in human judgment. This platform presents a large-scale evaluation of eight well-established cognitive biases across a diverse set of LLMs, analyzing over 2.8 million of responses generated through controlled prompt variations. </p>
<p class="welcome-text">Our evaluation framework is designed to measure model susceptibility to Anchoring, Availability, Confirmation, Framing, Interpretation, Overattribution, Prospect Theory, and Representativeness biases. The analysis investigates how both model size and prompt specificity play a significant role in bias expression. Each model's resistance score is computed from its performance across a curated dataset of psychologist-authored decision scenarios. Higher scores indicate stronger resistance to producing biased output. </p>
<p class="welcome-text">By clicking the triangles (<span class="hovertip">►</span>), the table reveals how bias susceptibility changes across prompts with varying levels of detail, as structured by the <a href="https://aclanthology.org/2023.findings-emnlp.946.pdf" class="orange-link">TELeR Taxonomy</a> . These results provide transparent, empirical insights into model reliability and trustworthiness, highlighting how model choice and thoughtful prompt design can mitigate common reasoning pitfalls. Our <a href="https://arxiv.org/abs/2509.22856" class="orange-link">paper </a>details the experiments and results.</p>
<div class="row">
<div class="col-xs-12 col-sm-12 col-md-12 col-lg-12 text-center">
Select Model Temperature for ranking*.
<label for="temp">Temperature: </label>
<select id="temp" name="temp">
<optgroup label="temperature">
<option value="0.2" selected="selected">0.2</option>
<option value="0.4">0.4</option>
<option value="0.6">0.6</option>
<option value="0.8">0.8</option>
</optgroup>
</select>
</div>
</div>
</div>
</div>
<hr style="border: none; height: 1px; background-color: #ccc; ">
<div style="display: flex; flex-wrap: wrap; gap: 20px; align-items: flex-start;">
<!-- Left Table (fixed width, does not stretch) -->
<div style="flex: 0 0 450px; max-width: 100%;" class="text-left">
<div class="table-responsive">
<table class="table table-sm table-striped" style="width: 100%;">
<thead>
<tr>
<th colspan="2">
<!-- <p style="font-size: 125%">All Capabilities
<span class="toggle-buttons" data-group="all">
<button type="button" class="toggle-button area_group_selector" id="all_areas_off" data-group="all" data-toggle="off">Off</button>
<button type="button" class="toggle-button area_group_selector" id="all_areas_on" data-group="all" data-toggle="on">On</button>
</span>
</p> -->
</th>
</tr>
</thead>
<thead>
<tr>
<th colspan="2">
<p class="ai-area">Cognitive Bias Resistance
<!-- [<a class="area_group_selector" id="ai_areas_off">off</a> | <a class="area_group_selector" id="ai_areas_on">on</a>]-->
<span class="toggle-buttons" data-group="all">
<button type="button" class="toggle-button area_group_selector" id="ai_areas_off" data-group="all" data-toggle="off">Off</button>
<button type="button" class="toggle-button area_group_selector" id="ai_areas_on" data-group="all" data-toggle="on">On</button>
</span>
</p>
</th>
</tr>
</thead>
<tbody class="ai-area">
<tr>
<td>
<!-- <span class="hovertip" id="factual-widget">►</span> -->
<label for="ai">
Representativeness Bias
</label>
<div class="table-responsive" style="display:none;" id="rep">
</div>
</td>
<td width="40px">
<input type="checkbox" class="parent" name="rep" id="rep" value="1.0"/>
</td>
</tr>
<tr>
<td>
<!-- <span class="hovertip" id="harmless-widget">►
</span> -->
<label for="vision">
Interpretation Bias
</label>
<div class="table-responsive" style="display:none;" id="inter">
</div>
</td>
<td>
<input type="checkbox" class="parent" name="inter" id="inter" value="1.0"/>
</td>
</tr>
<tr>
<td>
<!-- <span class="hovertip" id="anchoring-widget">
►
</span> -->
<label for="anchoring">
Anchoring Bias Mitigation
</label>
<div class="table-responsive" style="display:none;" id="anchoringb">
</div>
</td>
<td>
<input type="checkbox" name="anchor" id="anchor" value="1.0"/>
</td>
</tr>
<tr>
<td>
<!-- <span class="hovertip" id="availibility-widget">
► -->
</span>
<label for="availability">
Availability Bias
</label>
<div class="table-responsive" style="display:none;" id="avail">
</div>
</td>
<td>
<input type="checkbox" name="av" id="av" value="1.0"/>
</td>
</tr>
<tr>
<td>
<!-- <span class="hovertip" id="nlp-widget">
►
</span> -->
<label for="confirmation">
Confirmation Bias
</label>
<div class="table-responsive" style="display:none;" id="confirmationb">
</div>
</td>
<td>
<input type="checkbox" name="confirm" id="confirm" value="1.0"/>
</td>
</tr>
<tr>
<td>
<!-- <span class="hovertip" id="framing-widget">
►
</span> -->
<label for="framing">
Framing Bias
</label>
<div class="table-responsive" style="display:none;" id="fram">
</div>
</td>
<td>
<input type="checkbox" name="fra" id="fra" value="1.0"/>
</td>
</tr>
<tr>
<td>
<!-- <span class="hovertip" id="nlp-widget">
►
</span> -->
<label for="nlp">
Prospect Theory Bias
</label>
<div class="table-responsive" style="display:none;" id="nlp-conferences">
</div>
</td>
<td>
<input type="checkbox" name="nlp" id="nlp" value="1.0"/>
</td>
<tr>
<td>
<!-- <span class="hovertip" id="nlp-widget">
►
</span> -->
<label for="nlp">
Fundamental Attribution Error Bias
</label>
<div class="table-responsive" style="display:none;" id="nlp-conferences">
</div>
</td>
<td>
<input type="checkbox" name="attr" id="attr" value="1.0"/>
</td>
</tr>
</tbody>
</tbody>
</table>
</div>
</div>
<!-- <div class="col-xs-6 col-sm-push-6 col-sm-6 col-md-6 col-lg-6 text-center">
<div id="success">
</div>
</div> -->
<div style="flex: 1; min-width: 300px;" class="text-center">
<div style="max-height: 80vh; overflow-y: auto; padding-right: 10px;">
<table class="table table-sm table-striped" id="llm-ranking-table" style="border-collapse: separate; border-spacing: 0 4px; width: 100%;">
<thead>
<tr>
<th style="text-align: left; font-weight: bold;">Rank</th>
<th style="text-align: left; font-weight: bold;">LLM</th>
<th style="text-align: right; font-weight: bold;">Score</th>
</tr>
</thead>
<tbody id="llm-table-body">
<!-- JS will populate LLM rows here -->
</tbody>
</table>
</div>
</div>
</div>
<hr style="border: none; height: 1px; background-color: #ccc; margin: 10px 0;">
<div class="page-footer">
<p>
[ * GPT o4 mini was evaluated at a temperature of "1.0", which was the lowest setting available for the model]
</p>
<!-- <div style="background-color: #f9f9f9; border: 1px solid #ccc; padding: 8px; border-radius: 6px; margin-top: 40px; font-family: sans-serif; position: relative;">
<p style="margin-bottom: 8px; font-size: 1em;">
If you use these rankings/results in your manuscript, please cite the following papers:
</p>
<button id="copy-button" type="button" onclick="copyBibtex(event)" style="position: absolute; top: 8px; right: 8px; color: black; border: none; padding: 5px 5px; font-size: 0.6em; border-radius: 4px; cursor: pointer;">
Copy
</button>
<pre id="bibtex-block" style="font-size: 0.8em; background-color: #fff; border: 1px solid #e0e0e0; padding: 5px; border-radius: 4px; overflow-x: auto; white-space: pre-wrap;">
@inproceedings{DBLP:conf/emnlp/SantuF23,
author = {Shubhra Kanti Karmaker Santu and
Dongji Feng},
editor = {Houda Bouamor and
Juan Pino and
Kalika Bali},
title = {TELeR: {A} General Taxonomy of {LLM} Prompts for Benchmarking Complex
Tasks},
booktitle = {Findings of the Association for Computational Linguistics: {EMNLP}
2023, Singapore, December 6-10, 2023},
pages = {14197--14203},
publisher = {Association for Computational Linguistics},
year = {2023},
url = {https://doi.org/10.18653/v1/2023.findings-emnlp.946},
doi = {10.18653/V1/2023.FINDINGS-EMNLP.946},
timestamp = {Sun, 06 Oct 2024 21:00:53 +0200},
biburl = {https://dblp.org/rec/conf/emnlp/SantuF23.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
</pre>
<pre id="bibtex-block" style="font-size: 0.8em; background-color: #fff; border: 1px solid #e0e0e0; padding: 5px; border-radius: 4px; overflow-x: auto; white-space: pre-wrap;">
@misc{knipper2025biasdetailsassessmentcognitive,
title={The Bias is in the Details: An Assessment of Cognitive Bias in LLMs},
author={R. Alexander Knipper and Charles S. Knipper and Kaiqi Zhang and Valerie Sims and Clint Bowers and Santu Karmaker},
year={2025},
eprint={2509.22856},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2509.22856},
}
</pre>
</div> -->
<div style="display: flex; flex-wrap: wrap; gap: 20px; margin-top: 40px;">
<!-- First citation box -->
<div style="flex: 1 1 400px; background-color: #f9f9f9; border: 1px solid #ccc; padding: 8px; border-radius: 6px; font-family: sans-serif; position: relative; min-width: 300px;">
<p style="margin-bottom: 8px; font-size: 1em;">If you use these rankings/results in your manuscript, please cite:</p>
<button id="copy-button-1" type="button" onclick="copyBibtex(event, 'bibtex-block-1')" style="position: absolute; top: 8px; right: 8px; color: black; border: none; padding: 5px 5px; font-size: 0.6em; border-radius: 4px; cursor: pointer;">
Copy
</button>
<pre id="bibtex-block-1"
style="font-size: 0.8em; background-color: #fff; border: 1px solid #e0e0e0;
padding: 5px; border-radius: 4px; overflow-x: auto;
white-space: pre; text-align: left;">
@inproceedings{DBLP:conf/emnlp/SantuF23,
author = {Shubhra Kanti Karmaker Santu and Dongji Feng},
editor = {Houda Bouamor and Juan Pino and Kalika Bali},
title = {TELeR: {A} General Taxonomy of {LLM} Prompts for Benchmarking Complex Tasks},
booktitle = {Findings of the Association for Computational Linguistics: {EMNLP} 2023, Singapore, December 6-10, 2023},
pages = {14197--14203},
publisher = {Association for Computational Linguistics},
year = {2023},
url = {https://doi.org/10.18653/v1/2023.findings-emnlp.946},
doi = {10.18653/V1/2023.FINDINGS-EMNLP.946},
timestamp = {Sun, 06 Oct 2024 21:00:53 +0200},
biburl = {https://dblp.org/rec/conf/emnlp/SantuF23.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}
</pre>
</div>
<!-- Second citation box -->
<div style="flex: 1 1 400px; background-color: #f9f9f9; border: 1px solid #ccc; padding: 8px; border-radius: 6px; font-family: sans-serif; position: relative; min-width: 300px;">
<p style="margin-bottom: 8px; font-size: 1em;">and cite:</p>
<button id="copy-button-2" type="button" onclick="copyBibtex(event, 'bibtex-block-2')" style="position: absolute; top: 8px; right: 8px; color: black; border: none; padding: 5px 5px; font-size: 0.6em; border-radius: 4px; cursor: pointer;">
Copy
</button>
<pre id="bibtex-block-2" style="font-size: 0.8em; background-color: #fff; border: 1px solid #e0e0e0; padding: 5px; border-radius: 4px; overflow-x: auto; white-space: pre-wrap;">
@misc{knipper2025biasdetailsassessmentcognitive,
title={The Bias is in the Details: An Assessment of Cognitive Bias in LLMs},
author={R. Alexander Knipper and Charles S. Knipper and Kaiqi Zhang and Valerie Sims and Clint Bowers and Santu Karmaker},
year={2025},
eprint={2509.22856},
archivePrefix={arXiv},
primaryClass={cs.CL},
url={https://arxiv.org/abs/2509.22856},
}
</pre>
</div>
</div>
<p>
<br />
This ranking platform is developed by the <a href="https://karmake2.github.io/bridgeAI/" class="orange-link">Bridge-AI</a> Lab, Department of Computer Science, University of Central Florida, as part of ongoing research into fairness, bias mitigation, and trustworthiness in large language models. The evaluations aim to inform the broader AI community by highlighting model behavior under various cognitive bias scenarios. All results are based on systematically designed test sets and transparent scoring criteria.
<!--(see the <b><a href="faq.html">FAQ</a></b> for more details).-->
</p>
</div>
<script src="js/llm-rankings.js"></script>
</body>
</html>