-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathzh.html
336 lines (295 loc) · 14.2 KB
/
zh.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
<!DOCTYPE html>
<html lang="en">
<head>
<script async src="https://www.googletagmanager.com/gtag/js?id=G-C1CRWDNJ1J"></script>
<script>
window.dataLayer = window.dataLayer || [];
function gtag(){dataLayer.push(arguments);}
gtag('js', new Date());
gtag('config', 'G-C1CRWDNJ1J');
</script>
<link rel="preconnect" href="https://fonts.gstatic.com" crossorigin>
<link href="https://fonts.googleapis.com/css2?family=Noto+Sans+SC:[email protected]&display=swap" rel="stylesheet">
<meta charset="UTF-8">
<meta name="viewport" content="width=device-width, initial-scale=1.0">
<title>Chinese reading task about ML</title>
<style>
body {
font-family: Arial, sans-serif;
background-color: #f4f4f9;
color: #333;
margin: 0;
padding: 20px;
}
.container {
max-width: 800px;
margin: 0 auto;
background-color: #fff;
padding: 20px;
border-radius: 8px;
box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);
}
h1 {
color: #0056b3;
text-align: center;
}
p {
line-height: 1.6;
}
.zh-text {
font-size: 1.3em;
font-family: 'Noto Sans SC';
font-weight: 300;
margin: 0 0 5px 0;
}
.pinyin {
padding-top: 5px;
padding-bottom: 5px;
font-style: italic;
color: #888;
}
table {
width: 100%;
border-collapse: collapse;
margin-top: 20px;
}
th, td {
padding: 12px;
border: 1px solid #ddd;
text-align: left;
}
th {
background-color: #0056b3;
color: #fff;
}
td {
background-color: #f9f9f9;
}
td.zh {
font-family: 'Noto Sans SC';
font-size: 1.2em;
font-weight: 400;
}
</style>
</head>
<body>
<div class="container">
<h1>MPO: Boosting LLM Agents with Meta Plan Optimization</h1>
<div><p class='zh-text'>1. 最近的大语言模型进展使得基于LLM的代理能够成功处理互动式规划任务。</p>
<p class='zh-text'>2. 然而，现有方法常常受到规划幻觉的困扰，并且每个新代理都需要重新训练。</p>
<p class='zh-text'>3. 为解决这些挑战，我们提出了元规划优化（MPO）框架，通过直接引入显式指导来增强代理的规划能力。</p>
<p class='zh-text'>4. MPO利用元规划提供高层次的通用指导，帮助代理规划，并根据任务执行反馈持续优化元规划。</p>
<p class='zh-text'>5. 实验表明，MPO在两项代表性任务中显著优于现有基准，并提高了任务完成效率和泛化能力。</p></div>
<div class="pinyin">
<p>1. 最近的大语言模型进展使得基于LLM的代理能够成功处理互动式规划任务。然而，现有方法常常受到规划幻觉的困扰，并且每个新代理都需要重新训练。为解决这些挑战，我们提出了元规划优化（MPO）框架，通过直接引入显式指导来增强代理的规划能力。MPO利用元规划提供高层次的通用指导，帮助代理规划，并根据任务执行反馈持续优化元规划。实验表明，MPO在两项代表性任务中显著优于现有基准，并提高了任务完成效率和泛化能力。
Zuìjìn de dà yǔyán móxíng jìnzhǎn shǐdé jīyú LLM de dàilǐ nénggòu chénggōng chǔlǐ hùdòngshì guīhuà rènwù</p>
<p>2. Rán'ér, xiànyǒu fāngfǎ chángcháng shòudào guīhuà huànjué de kùnhuò, bìngqiě měi gè xīn dàilǐ dōu xūyào chóngxīn xùnliàn</p>
<p>3. Wèi jiějué zhèxiē tiǎozhàn, wǒmen tíchūle yuán guīhuà yōuhuà (MPO) kuàngjià, tōngguò zhíjiē yǐnrù xiǎnshì zhǐdǎo lái zēngqiáng dàilǐ de guīhuà nénglì</p>
<p>4. MPO lìyòng yuán guīhuà tígōng gāo céngcì de tōngyòng zhǐdǎo, bāngzhù dàilǐ guīhuà, bìnggēnjù rènwù zhíxíng fǎnkuì chíxù yōuhuà yuán guīhuà</p>
<p>5. Shíyàn biǎomíng, MPO zài liǎng xiàng dàibiǎoxìng rènwù zhōng xiǎnzhù yōuhuàn xiànzhùn bǐzhǔn, bìng tígāole rènwù wánchéng xiàolǜ hé fànhuà nénglì</p>
</div>
<div><p>1. Recent advancements in large language models have enabled LLM-based agents to successfully handle interactive planning tasks.</p>
<p>2. However, existing methods often suffer from planning hallucinations, and each new agent requires retraining.</p>
<p>3. To address these challenges, we propose the Meta-Planning Optimization (MPO) framework, which enhances the agent's planning capability by directly introducing explicit guidance.</p>
<p>4. MPO leverages meta-planning to provide high-level, general guidance to assist the agent in planning and continuously optimizes the meta-planning based on task execution feedback.</p>
<p>5. Experiments demonstrate that MPO significantly outperforms existing benchmarks in two representative tasks and improves task completion efficiency and generalization ability.</p></div>
<h2>Vocabulary</h2>
<table>
<thead>
<tr>
<th>Word</th>
<th>Pinyin</th>
<th>Translation</th>
</tr>
</thead>
<tbody>
<tr>
<td class="zh">大语言模型</td>
<td>dà yǔyán móxíng</td>
<td>large language model</td>
</tr>
<tr>
<td class="zh">基于</td>
<td>jīyú</td>
<td>based on</td>
</tr>
<tr>
<td class="zh">代理</td>
<td>dàilǐ</td>
<td>agent</td>
</tr>
<tr>
<td class="zh">互动式</td>
<td>hùdòngshì</td>
<td>interactive</td>
</tr>
<tr>
<td class="zh">规划</td>
<td>guīhuà</td>
<td>planning</td>
</tr>
<tr>
<td class="zh">任务</td>
<td>rènwù</td>
<td>task</td>
</tr>
<tr>
<td class="zh">幻觉</td>
<td>huànjué</td>
<td>hallucination</td>
</tr>
<tr>
<td class="zh">困扰</td>
<td>kùnrǎo</td>
<td>trouble</td>
</tr>
<tr>
<td class="zh">重新</td>
<td>chóngxīn</td>
<td>renew</td>
</tr>
<tr>
<td class="zh">训练</td>
<td>xùnliàn</td>
<td>training</td>
</tr>
<tr>
<td class="zh">挑战</td>
<td>tiǎozhàn</td>
<td>challenge</td>
</tr>
<tr>
<td class="zh">提出</td>
<td>tíchū</td>
<td>propose</td>
</tr>
<tr>
<td class="zh">框架</td>
<td>kuàngjià</td>
<td>framework</td>
</tr>
<tr>
<td class="zh">显式</td>
<td>xiǎnshì</td>
<td>explicit</td>
</tr>
<tr>
<td class="zh">指导</td>
<td>zhǐdǎo</td>
<td>guidance</td>
</tr>
<tr>
<td class="zh">增强</td>
<td>zēngqiáng</td>
<td>enhance</td>
</tr>
<tr>
<td class="zh">能力</td>
<td>nénglì</td>
<td>ability</td>
</tr>
<tr>
<td class="zh">利用</td>
<td>lìyòng</td>
<td>utilize</td>
</tr>
<tr>
<td class="zh">提供</td>
<td>tígōng</td>
<td>provide</td>
</tr>
<tr>
<td class="zh">高层次</td>
<td>gāo céngcì</td>
<td>high-level</td>
</tr>
<tr>
<td class="zh">通用</td>
<td>tōngyòng</td>
<td>general</td>
</tr>
<tr>
<td class="zh">帮助</td>
<td>bāngzhù</td>
<td>help</td>
</tr>
<tr>
<td class="zh">执行</td>
<td>zhíxíng</td>
<td>execute</td>
</tr>
<tr>
<td class="zh">反馈</td>
<td>fǎnkuì</td>
<td>feedback</td>
</tr>
<tr>
<td class="zh">持续</td>
<td>chíxù</td>
<td>continuous</td>
</tr>
<tr>
<td class="zh">优化</td>
<td>yōuhuà</td>
<td>optimize</td>
</tr>
<tr>
<td class="zh">实验</td>
<td>shíyàn</td>
<td>experiment</td>
</tr>
<tr>
<td class="zh">表明</td>
<td>biǎomíng</td>
<td>indicate</td>
</tr>
<tr>
<td class="zh">代表性</td>
<td>dàibiǎoxìng</td>
<td>representative</td>
</tr>
<tr>
<td class="zh">显著</td>
<td>xiǎnzhù</td>
<td>significant</td>
</tr>
<tr>
<td class="zh">优于</td>
<td>yōuyú</td>
<td>superior to</td>
</tr>
<tr>
<td class="zh">现有</td>
<td>xiànyǒu</td>
<td>existing</td>
</tr>
<tr>
<td class="zh">基准</td>
<td>jīzhǔn</td>
<td>benchmark</td>
</tr>
<tr>
<td class="zh">提高</td>
<td>tígāo</td>
<td>improve</td>
</tr>
<tr>
<td class="zh">完成</td>
<td>wánchéng</td>
<td>complete</td>
</tr>
<tr>
<td class="zh">效率</td>
<td>xiàolǜ</td>
<td>efficiency</td>
</tr>
<tr>
<td class="zh">泛化</td>
<td>fànhuà</td>
<td>generalize</td>
</tr>
</tbody>
</table>
</div>
</body>
</html>