Commit
•
a675894
1
Parent(s):
905f8ef
Create 2306.05425.atom
Browse files- arxiv/2306.05425.atom +71 -0
arxiv/2306.05425.atom
ADDED
@@ -0,0 +1,71 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
<?xml version="1.0" encoding="UTF-8"?>
|
2 |
+
<feed xmlns="http://www.w3.org/2005/Atom">
|
3 |
+
<link href="http://arxiv.org/api/query?search_query%3D%26id_list%3D2306.05425%26start%3D0%26max_results%3D1" rel="self" type="application/atom+xml"/>
|
4 |
+
<title type="html">ArXiv Query: search_query=&id_list=2306.05425&start=0&max_results=1</title>
|
5 |
+
<id>http://arxiv.org/api/UKr4sG8yTUNI8GqLUahlRDrZ5vk</id>
|
6 |
+
<updated>2023-06-09T00:00:00-04:00</updated>
|
7 |
+
<opensearch:totalResults xmlns:opensearch="http://a9.com/-/spec/opensearch/1.1/">1</opensearch:totalResults>
|
8 |
+
<opensearch:startIndex xmlns:opensearch="http://a9.com/-/spec/opensearch/1.1/">0</opensearch:startIndex>
|
9 |
+
<opensearch:itemsPerPage xmlns:opensearch="http://a9.com/-/spec/opensearch/1.1/">1</opensearch:itemsPerPage>
|
10 |
+
<entry>
|
11 |
+
<id>http://arxiv.org/abs/2306.05425v1</id>
|
12 |
+
<updated>2023-06-08T17:59:56Z</updated>
|
13 |
+
<published>2023-06-08T17:59:56Z</published>
|
14 |
+
<title>MIMIC-IT: Multi-Modal In-Context Instruction Tuning</title>
|
15 |
+
<summary> High-quality instructions and responses are essential for the zero-shot
|
16 |
+
performance of large language models on interactive natural language tasks. For
|
17 |
+
interactive vision-language tasks involving intricate visual scenes, a large
|
18 |
+
quantity of diverse and creative instruction-response pairs should be
|
19 |
+
imperative to tune vision-language models (VLMs). Nevertheless, the current
|
20 |
+
availability of vision-language instruction-response pairs in terms of
|
21 |
+
quantity, diversity, and creativity remains limited, posing challenges to the
|
22 |
+
generalization of interactive VLMs. Here we present MultI-Modal In-Context
|
23 |
+
Instruction Tuning (MIMIC-IT), a dataset comprising 2.8 million multimodal
|
24 |
+
instruction-response pairs, with 2.2 million unique instructions derived from
|
25 |
+
images and videos. Each pair is accompanied by multi-modal in-context
|
26 |
+
information, forming conversational contexts aimed at empowering VLMs in
|
27 |
+
perception, reasoning, and planning. The instruction-response collection
|
28 |
+
process, dubbed as Syphus, is scaled using an automatic annotation pipeline
|
29 |
+
that combines human expertise with GPT's capabilities. Using the MIMIC-IT
|
30 |
+
dataset, we train a large VLM named Otter. Based on extensive evaluations
|
31 |
+
conducted on vision-language benchmarks, it has been observed that Otter
|
32 |
+
demonstrates remarkable proficiency in multi-modal perception, reasoning, and
|
33 |
+
in-context learning. Human evaluation reveals it effectively aligns with the
|
34 |
+
user's intentions. We release the MIMIC-IT dataset, instruction-response
|
35 |
+
collection pipeline, benchmarks, and the Otter model.
|
36 |
+
</summary>
|
37 |
+
<author>
|
38 |
+
<name>Bo Li</name>
|
39 |
+
</author>
|
40 |
+
<author>
|
41 |
+
<name>Yuanhan Zhang</name>
|
42 |
+
</author>
|
43 |
+
<author>
|
44 |
+
<name>Liangyu Chen</name>
|
45 |
+
</author>
|
46 |
+
<author>
|
47 |
+
<name>Jinghao Wang</name>
|
48 |
+
</author>
|
49 |
+
<author>
|
50 |
+
<name>Fanyi Pu</name>
|
51 |
+
</author>
|
52 |
+
<author>
|
53 |
+
<name>Jingkang Yang</name>
|
54 |
+
</author>
|
55 |
+
<author>
|
56 |
+
<name>Chunyuan Li</name>
|
57 |
+
</author>
|
58 |
+
<author>
|
59 |
+
<name>Ziwei Liu</name>
|
60 |
+
</author>
|
61 |
+
<arxiv:comment xmlns:arxiv="http://arxiv.org/schemas/atom">Project page: https://otter-ntu.github.io/ Dataset & code:
|
62 |
+
https://github.com/Luodian/otter Initial release, work in progress</arxiv:comment>
|
63 |
+
<link href="http://arxiv.org/abs/2306.05425v1" rel="alternate" type="text/html"/>
|
64 |
+
<link title="pdf" href="http://arxiv.org/pdf/2306.05425v1" rel="related" type="application/pdf"/>
|
65 |
+
<arxiv:primary_category xmlns:arxiv="http://arxiv.org/schemas/atom" term="cs.CV" scheme="http://arxiv.org/schemas/atom"/>
|
66 |
+
<category term="cs.CV" scheme="http://arxiv.org/schemas/atom"/>
|
67 |
+
<category term="cs.AI" scheme="http://arxiv.org/schemas/atom"/>
|
68 |
+
<category term="cs.CL" scheme="http://arxiv.org/schemas/atom"/>
|
69 |
+
<category term="cs.HC" scheme="http://arxiv.org/schemas/atom"/>
|
70 |
+
</entry>
|
71 |
+
</feed>
|