-
Notifications
You must be signed in to change notification settings - Fork 0
Expand file tree
/
Copy pathindex.html
More file actions
377 lines (344 loc) · 14.9 KB
/
Copy pathindex.html
File metadata and controls
377 lines (344 loc) · 14.9 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
<!DOCTYPE html>
<html>
<head>
<meta charset="utf-8">
<meta name="description"
content="Benchmarking Agent Memory in Interdependent Multi-Session Agentic Tasks">
<meta name="keywords" content="MemoryArena, AgentMemory, LLM">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>MemoryArena: Benchmarking Agent Memory in Interdependent Multi-Session Agentic Tasks</title>
<!-- Global site tag (gtag.js) - Google Analytics -->
<script async src="https://www.googletagmanager.com/gtag/js?id=G-PYVRSFMDRL"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag() {
dataLayer.push(arguments);
}
gtag('js', new Date());
gtag('config', 'G-PYVRSFMDRL');
</script>
<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>
<nav class="navbar" role="navigation" aria-label="main navigation">
<div class="navbar-brand">
<a role="button" class="navbar-burger" aria-label="menu" aria-expanded="false">
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
<span aria-hidden="true"></span>
</a>
</div>
<div class="navbar-menu">
<div class="navbar-start" style="flex-grow: 1; justify-content: center;">
<a class="navbar-item" href="#home">
<span class="icon">
<i class="fas fa-home"></i>
</span>
</a>
</div>
</div>
</nav>
<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" style="display: flex; align-items: center; justify-content: center; gap: 0.5rem; flex-wrap: wrap;">
<span style="display: flex; align-items: center; gap: 0.5rem;">
<img src="./static/images/memoryarena_logo.png" alt="MemoryArena Logo" style="max-width: 50px; height: auto;">
MemoryArena:
</span>
<span style="flex-basis: 100%; text-align: center;">
Benchmarking Agent Memory in Interdependent Multi-Session Agentic Tasks
</span>
</h1>
<div class="is-size-5 publication-authors">
<div class="author-row">
<span class="author-block">
<a href="#">Zexue He</a><sup>1*</sup>,</span>
<span class="author-block">
<a href="#">Yu Wang</a><sup>2*</sup>,</span>
<span class="author-block">
<a href="#">Churan Zhi</a><sup>2*</sup>,</span>
<span class="author-block">
<a href="#">Yuanzhe Hu</a><sup>2*</sup>,</span>
<span class="author-block">
<a href="#">Tzu-Ping Chen</a><sup>2*</sup>,</span>
<span class="author-block">
<a href="#">Lang Yin</a><sup>3*</sup>,</span>
</div>
<div class="author-row">
<span class="author-block">
<a href="#">Ze Chen</a><sup>4</sup>,</span>
<span class="author-block">
<a href="#">Tong Arthur Wu</a><sup>5</sup>,</span>
<span class="author-block">
<a href="#">Siru Ouyang</a><sup>3</sup>,</span>
<span class="author-block">
<a href="#">Zihan Wang</a><sup>6</sup>,</span>
</div>
<div class="author-row">
<span class="author-block">
<a href="#">Jiaxin Pei</a><sup>1</sup>,</span>
<span class="author-block">
<a href="#">Julian McAuley</a><sup>2</sup>,</span>
<span class="author-block">
<a href="#">Yejin Choi</a><sup>1</sup>,</span>
<span class="author-block">
<a href="#">Alex Pentland</a><sup>1</sup></span>
</div>
</div>
<div class="is-size-5 publication-authors">
<span class="author-block"><sup>1</sup>Stanford University</span>
<span class="author-block"><sup>2</sup>UCSD</span>
<span class="author-block"><sup>3</sup>UIUC</span>
<span class="author-block"><sup>4</sup>Princeton University</span>
<span class="author-block"><sup>5</sup>University of Pittsburgh</span>
<span class="author-block"><sup>6</sup>2077AI</span>
</div>
<div class="is-size-6 publication-authors" style="margin-top: 1rem; color: #666;">
<p><sup>*</sup>Equal contribution Corresponding author: zexueh@stanford.edu</p>
</div>
<div class="column has-text-centered" style="margin-top: 2rem;">
<div class="publication-links">
<!-- Paper Link. -->
<span class="link-block">
<a href="https://arxiv.org/abs/2602.16313"
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>
<!-- Data Link. -->
<span class="link-block">
<a href="https://huggingface.co/datasets/ZexueHe/memoryarena"
class="external-link button is-normal is-rounded is-dark">
<span class="icon">
<i class="data ai-arxiv"></i>
</span>
<span>Data</span>
</a>
</span>
<!-- GitHub Link. -->
<span class="link-block">
<a href=" https://github.com/ZexueHe/MemoryArena.git"
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="hero teaser">
<div class="container is-max-desktop">
<div class="hero-body">
<div class="columns is-centered">
<div class="column is-full-width">
<img src="./static/images/memoryarena.png" alt="MemoryArena Teaser Figure" style="max-width: 100%; height: auto;">
<h2 class="subtitle has-text-centered" style="margin-top: 1.5rem;">
MemoryArena: A Benchmark for Agent Memory in Multi-Session Agentic Tasks
</h2>
</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>
Existing evaluations of agents with memory typically assess <strong>memorization</strong> and <strong>action</strong> in isolation. One class of benchmarks evaluates memorization by testing recall of past conversations or text but fails to capture how memory is used to guide future decisions. Another class focuses on agents acting in single-session tasks without the need for long-term memory. However, in realistic settings, memorization and action are tightly coupled: agents acquire memory while interacting with the environment, and subsequently rely on that memory to solve future tasks.
</p>
<p>
To capture this setting, We introduce <span class="memoryarena">MemoryArena</span>, a unified evaluation gym for benchmarking agent memory in multi-session Memory-Agent-Environment loops. The benchmark consists of human-crafted agentic tasks with explicitly interdependent subtasks, where agents must learn from earlier actions and feedback by distilling experiences into memory, and subsequently use that memory to guide later actions to solve the overall task. MemoryArena supports evaluation across web navigation, preference-constrained planning, progressive information search, and sequential formal reasoning, and reveals that agents with near-saturated performance on existing long-context memory benchmarks like LoCoMo perform poorly in our agentic setting, exposing a gap in current evaluations for agents with memory.
</p>
</div>
</div>
</div>
<!--/ Abstract. -->
<!-- Dataset. -->
<div class="columns is-centered" style="margin-top: 3rem;">
<div class="column is-four-fifths">
<h2 class="title is-3 has-text-centered">Dataset</h2>
<div class="content has-text-justified">
<h3 class="title is-5">Overview</h3>
<p>
This dataset contains structured multi-session agentic tasks with questions (list), answers (list), and necessary background context or misc. Each row in the test file represents an agentic task (dict) with multiple subtasks, their corresponding answers, and background information.
</p>
<h3 class="title is-5">Dataset Structure</h3>
<p>Each line in the HuggingFace test file is a dictionary with the following fields:</p>
<ul>
<li><strong>id</strong> (int): Unique identifier for each agentic task entry</li>
<li><strong>questions</strong> (list of str): List of sub-task queries</li>
<li><strong>answers</strong> (list of str): List of corresponding answers for each subtask</li>
<li><strong>backgrounds</strong> (str or list of str): Background/context information for each task
<ul>
<li>Bundled Shopping and Progressive Search: no necessary backgrounds.</li>
<li>Travel Planner: the travel details of the base person in each task serve as the background information for all subtasks.</li>
<li>Formal Reasoning (Math and Phys): each subtask may have its background information.</li>
</ul>
</li>
</ul>
<h3 class="title is-5">Usage</h3>
<p>Load with Hugging Face Datasets:</p>
<pre><code class="language-python">from datasets import load_dataset
ds = load_dataset("ZexueHe/memoryarena", "bundled_shopping")
ds = load_dataset("ZexueHe/memoryarena", "progressive_search")
ds = load_dataset("ZexueHe/memoryarena", "group_travel_planner")
ds = load_dataset("ZexueHe/memoryarena", "formal_reasoning_math")
ds = load_dataset("ZexueHe/memoryarena", "formal_reasoning_phys")
</code></pre>
<h3 class="title is-5">Example Tasks</h3>
<p>Bundled Webshop:</p>
<pre><code class="language-json">{
"id": 0,
"questions": [
"search subtask 1",
"search subtask 1",
...
],
"answers": [
"search subtask result 1",
"search subtask result 2",
...
]
}
</code></pre>
<p>Progressive Search:</p>
<pre><code class="language-json">{
"id": 0,
"questions": [
"buy subtask item1",
"buy subtask item 2",
...
],
"answers": [
{"target_asin": "B00TUDFEW2", "attributes": ["Almond Flour", ...]},
{"target_asin": "B08957C9ZH", "attributes": [...]},
...
]
}
</code></pre>
<p>Group Travel Planner:</p>
<pre><code class="language-json">{
"id": 0,
"base_person": {
"name": "Jennifer",
"query": "I am Jennifer. Please help me plan a trip from St. Petersburg to Rockford spanning 3 days...",
"daily_plans": [
{"days": 1, "current_city": "from St. Petersburg to Rockford", "transportation": "..."},
{"days": 2, "current_city": "Rockford", "transportation": "..."},
...
]
},
"questions": [
"I am Eric.\n I'm joining Jennifer for this trip...",
"I am Emma.\n I'm traveling with Jennifer and Eric...",
...
],
"answers": [
[
{"days": 1, "current_city": "from St. Petersburg to Rockford", "transportation": "..."},
{"days": 2, "current_city": "Rockford", "transportation": "..."},
...
],
[
{"days": 1, "current_city": "from St. Petersburg to Rockford", "transportation": "..."},
{"days": 2, "current_city": "Rockford", "transportation": "..."},
...
],
...
]
}
</code></pre>
<p>Formal Reasoning (Math and Phys):</p>
<pre><code class="language-json">{
"id": 0,
"paper_name": "paper_id",
"backgrounds": [
"necessary definitions, formulations, and relevant context of subtask 1",
"necessary definitions, formulations, and relevant context of subtask 2",
...
],
"questions": [
"Math subtask question 1",
"Math subtask question 2",
...
],
"answers": [
"Math result for subtask 1",
"Math result for subtask 2",
...
]
}
</code></pre>
<h3 class="title is-5">License</h3>
<p>
This dataset is licensed under the
<a href="https://creativecommons.org/licenses/by/4.0/">Creative Commons Attribution 4.0 International (CC-BY-4.0)</a>
license.
</p>
<h3 class="title is-5">Citation</h3>
<p>If you use this dataset, PLEASE CITE THE NEW BIBTEX:</p>
<pre><code class="language-bibtex">@article{he2026memoryarena,
title={MemoryArena: Benchmarking Agent Memory in Interdependent Multi-Session Agentic Tasks},
author={He, Zexue and Wang, Yu and Zhi, Churan and Hu, Yuanzhe and Chen, Tzu-Ping and Yin, Lang and Chen, Ze and Wu, Tong Arthur and Ouyang, Siru and Wang, Zihan and others},
journal={arXiv preprint arXiv:2602.16313},
year={2026}
}
</code></pre>
</div>
</div>
</div>
<!--/ Dataset. -->
</div>
</section>
<footer class="footer">
<div class="container">
<div class="content has-text-centered">
<a class="icon-link" href="https://github.com" disabled>
<i class="fab fa-github"></i>
</a>
</div>
<div class="columns is-centered">
<div class="column is-8">
<div class="content">
<p>
This website is licensed under a <a rel="license"
href="http://creativecommons.org/licenses/by-sa/4.0/">Creative
Commons Attribution-ShareAlike 4.0 International License</a>.
</p>
</div>
</div>
</div>
</div>
</footer>
</body>
</html>