Atsuyuki Miyai @UTokyo
@AtsuMiyaiAM
Followers
161
Following
180
Media
8
Statuses
67
1st-year PhD student at The University of Tokyo 🇯🇵, Yamasaki Lab and Aizawa Lab | Computer Vision | Trustworthy AI | Account for Research Purposes
The University of Tokyo, Japan
Joined July 2023
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
#MaybellineTeddyLand
• 435817 Tweets
TEDDYTINT X CHARLOTTE
• 251897 Tweets
#MAYBELLINExCHARLOTTE
• 244258 Tweets
#jjk268
• 225368 Tweets
台風の影響
• 181777 Tweets
Morara
• 155286 Tweets
Sukuna
• 106766 Tweets
Megumi
• 94402 Tweets
Karime
• 90585 Tweets
TEDDYTINT X FAYE
• 86352 Tweets
デコピン
• 77348 Tweets
Yuji
• 72613 Tweets
Yuta
• 62951 Tweets
焼肉の日
• 50087 Tweets
Fireboy
• 49650 Tweets
Gomita
• 41940 Tweets
Luciano
• 35160 Tweets
安全第一
• 31978 Tweets
मेजर ध्यानचंद
• 29765 Tweets
राष्ट्रीय खेल
• 29272 Tweets
#NationalSportsDay
• 28523 Tweets
芸能界復帰
• 21092 Tweets
Goma
• 20866 Tweets
臨時休業
• 20086 Tweets
伯仲燦然
• 19919 Tweets
台風のせい
• 19218 Tweets
Uraume
• 17422 Tweets
VisiPRABOWO BangunBANGSA
• 16260 Tweets
MENUJUbangsa MAJUsejahtera
• 15728 Tweets
PAGTATAGDocu TopMovieInPH
• 15574 Tweets
試合中止
• 11001 Tweets
ابو شجاع
• 10558 Tweets
Pinned Tweet
How are OOD Detection, Open-set Recognition, Anomaly Detection, etc. evolving in the CLIP & GPT-4V eras?🤔
📣 Check our preprint survey on Generalized OOD detection v2!
We encapsulate the evolution of these tasks in VLM era, identifying challenges!
5
17
60
#NeurIPS2023
, “LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning”の論文を更新しました。CoOpに背景などの情報を分布外データとして学習するロスを加えるだけで、ImageNetベンチマークにおいて1ショット (15分の訓練)でも既存のCLIPベースの手法を超える精度を達成しました！！
1
6
53
🎉April Fool is the birthday of Unsolvable Problem Detection! UPD examines the VLM’s ability to withhold answers when faced with unsolvable problems. Please enjoy VLMs with unsolvable problems today!
paper page:
code:
Unsolvable Problem Detection
Evaluating Trustworthiness of Vision Language Models
This paper introduces a novel and significant challenge for Vision Language Models (VLMs), termed Unsolvable Problem Detection (UPD). UPD examines the VLM's ability to withhold answers when
2
34
156
1
10
45
Happy to share that our work "LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning" has been accepted to
#NeurIPS2023
! 🎉
I would like to express my deepest appreciation to my co-authors and reviewers.
arXiv:
2
8
33
Today, I had a poster presentation on LoCoOp at
#NeurIPS2023
!
It was very exciting to see so many people come to see our poster! Thanks for coming!
0
3
21
🎉Happy to share our MM-UPD Leaderboard🏆
We are now accepting submissions from the community! We look forward to your submissions😀
Many thanks to
@mervenoyann
@clefourrier
for the invitation and for guiding us!
MMUPD is now hosted on
@huggingface
Hub as a leaderboard 🏆
amazing to see LLaVA-1.6 is outperforming proprietary models on many subtasks 🤩
link in the next one!
2
4
28
1
5
18
I will be attending
#ICLR2024
in person next week and will present UPD at the Workshop on Reliable and Responsible Foundation Models on May 11th.
Please drop by if you can! ☕️"
自分なりに精一杯頑張ります！！時間が合えば是非!
🎉April Fool is the birthday of Unsolvable Problem Detection! UPD examines the VLM’s ability to withhold answers when faced with unsolvable problems. Please enjoy VLMs with unsolvable problems today!
paper page:
code:
1
10
45
1
4
18
We update our
#NeurIPS2023
, “LoCoOp: Few-Shot Out-of-Distribution Detection via Prompt Learning”. LoCoOp performs OOD regularization with OOD regions (e.g., background) in ID training images and outperforms existing methods even in a 1-shot setting on ImageNet benchmarks.
1
1
13
#MIRU2024
の招待講演にLoCoOpが選ばれました。学部4年生の頃から参加させていただいて、今回が初招待公演なので大変光栄です。
また、ポスターでは我々のOODチームの最新作 (まだ未公開)の話も少し織り交ぜるつもりなので、是非お越しください!
🕐: 8/6, 10:20 -
code:
1
2
11
Our poster on UPD has been out!
Please enjoy the Workshop on Reliable and Responsible Foundation Models!
🕐 11th (Saturday), 10:15~11:15, 14:45~15:45
📍 A3
📕 full version:
🤗
0
5
6
Thanks,
@mervenoyann
, for sharing our MM-UPD Bench! Glad to the interest from the Hugging Face community🤗
@huggingface
!
Yes, MM-UPD reveals a new dimension of the VLMs' performances!
A great vision language benchmark: MM-UPD evaluates how model responds to unsolvable problems 🤓
As of now, most VLMs, including GPT-4V and LLaVA-Next-34B, struggle with refusing to answer
4
10
58
0
2
6
This work is joint work with very very amazing and cool members!!
@JingkangY
, Jingyang Zhang, Yifei Ming,
@YK_officia1
, Prof. Go Irie,
@SharonYixuanLi
, Prof. Hai Li,
@liuziwei7
, and my supervisor Prof. Kiyoharu Aizawa
0
1
6
Big thanks to my awesome co-authors including the first four OpenOOD team members!
@JingkangY
,
@BeFunky2345
,
@ming5_alvin
,
@YueqianL
,
@YK_officia1
, Prof. Go Irie,
@JotyShafiq
,
@SharonYixuanLi
, Prof. Hai Li,
@liuziwei7
, Prof. Toshihiko Yamasaki, and Prof. Kiyoharu Aizawa
0
1
4
Our UPD has been integrated into the LMMs-Eval codebase! Thanks, LMMs-Eval team!
lmms-eval:
paper:
Accelerating the Development of Large Multimoal Models with LMMs-Eval
Repo:
Blog:
We are offering a one command evaluation API for fast and thorough evaluation of LMMs over 39 datasets (increasingly).
1
24
112
1
2
4
This is joint work with
@YK_officia1
(UTokyo, LY Corp.), Go Irie (Tokyo Univ. of Science), and Kiyoharu Aizawa (UTokyo).
If you are interested in our work, please check the paper and code for details.
paper:
code:
0
1
1
Our comprehensive reviews found that OOD Detection and Anomaly Detection have become the demanding challenges in the VLM era🔥
We hope these findings will foster collaborative efforts among each community 🤝
We summarize the paper list at !
1
0
3
⚠️ Is Your LLM Quantization Safe? ⚠️
Don’t miss my junior Kazuki’s wonderful work at ICML workshop
@NG_AI_Safety
.
This is a collaboration work with
@ETH
!
paper:
code:
workshop:
#ICML2024
#NextGenAISafety
Our work appears at the NextGenAISafety Workshop @ ICML
Honored to be selected for an oral presentation!
#ICML2024
#NextGenAISafety
0
5
16
0
0
3
論文をプロモーションするのも大切かなと思ったのでアカウントを作成しました。主にワークの紹介での使用になるかと思います。 自分を知ってくれてたり、ワークに興味を持ってくれた方は気軽にFollowしていただけると幸いです。
1
1
2