Ryousuke Yamada (@FragileGoodwill) - Twitter Profili

Ryousuke Yamada retweetledi

Yuki@y_m_asano·30 Nis

🎉[openings] I’m hiring postdoctoral researchers to join our @FunAILab at UTN through the Alexander von Humboldt Research Fellowship (@AvHStiftung), via the Henriette Herz Scouting Programme. As a Henriette Herz Scout, I can nominate outstanding international researchers for this fellowship route. I’m especially keen to hear from candidates working on multimodal learning, video and image pretraining, and post-training. Fellows would be hosted in our lab at UTN and work closely with us on these topics. Key requirements: * finished your doctoral studies less than 4 years ago or will finish in the next 6 months * did not live/work in Germany in the last 10 years * applications from female, trans* and/or non-binary candidates are highly encouraged! Interested? Please send a short note with your CV, PhD year, current affiliation, 2–3 key publications, and a few lines on how your work connects. Please share! 🔀

English

1

23

61

5.5K

Ryousuke Yamada retweetledi

Yuki@y_m_asano·30 Mar

[new CVPR'26 paper] 🔄 SSL works great when you have tons of data. But in 3D… we don’t. High-quality 3D scans are expensive, slow, and hard to scale. So what if we could pretrain 3D models without any real 3D scans? 1/

English

3

31

215

24.4K

Ryousuke Yamada retweetledi

Andrei Bursuc@abursuc·30 Mar

Can we get a SoTA point cloud encoder without actual 3D scans? Yes, we can, thanks to a clever mix of: - data generation: video point clouds from web videos - LAM3C: a new self-supervised learning strategy for such data Super fun #cvpr2026 project led by @FragileGoodwill 👇

Yuki@y_m_asano

[new CVPR'26 paper] 🔄 SSL works great when you have tons of data. But in 3D… we don’t. High-quality 3D scans are expensive, slow, and hard to scale. So what if we could pretrain 3D models without any real 3D scans? 1/

English

1

7

28

3.4K

Ryousuke Yamada retweetledi

Hirokatsu Kataoka | 片岡裕雄@HirokatuKataoka·30 Mar

This work lies at the intersection of reconstruction × 3D point clouds (PC) × visual pre-training under limited data. Video-generated PCs offer a scalable alternative for PC pre-training without (sans in French) relying on 3D scans. ryosuke-yamada.github.io/lam3c/

Yuki@y_m_asano

[new CVPR'26 paper] 🔄 SSL works great when you have tons of data. But in 3D… we don’t. High-quality 3D scans are expensive, slow, and hard to scale. So what if we could pretrain 3D models without any real 3D scans? 1/

English

1

7

28

7K

Ryousuke Yamada@FragileGoodwill·31 Mar

Super excited to share our project (LAM3C x RoomTours)!! Can we learn 3D point cloud representations without real 3D scans? Our work explores exactly this 👇 arxiv.org/abs/2512.23042 #CVPR2026 @CVPR

Yuki@y_m_asano

[new CVPR'26 paper] 🔄 SSL works great when you have tons of data. But in 3D… we don’t. High-quality 3D scans are expensive, slow, and hard to scale. So what if we could pretrain 3D models without any real 3D scans? 1/

English

0

4

10

1.6K

Ryousuke Yamada retweetledi

Seungwook Han@seungwookh·12 Mar

Can language models learn useful priors without ever seeing language? We pre-pre-train transformers on neural cellular automata — fully synthetic, zero language. This improves language modeling by up to 6%, speeds up convergence by 40%, and strengthens downstream reasoning. Surprisingly, it even beats pre-pre-training on natural text! Blog: hanseungwook.github.io/blog/nca-pre-p… (1/n)

English

47

261

1.7K

253.5K

Ryousuke Yamada@FragileGoodwill·4 Mar

@XiaoyangWu_ @_akhaliq Amazing work😀 Thanks for this great research

English

1

0

57

Xiaoyang Wu@XiaoyangWu_·4 Mar

Thanks AK!

AK@_akhaliq

Utonia Toward One Encoder for All Point Clouds paper: huggingface.co/papers/2603.03…

English

2

1

18

8.3K

Ryousuke Yamada retweetledi

cvpaper.challenge | AI/CV研究コミュニティ@CVpaperChalleng·3 Mar

【4/7開催】ASPIRE Computer Vision Workshop @ Tokyo 日時：4月7日（火）10:30–18:00 会場：東京科学大学大岡山キャンパス蔵前会館参加希望の方は以下より事前登録をお願いします。 research-p.com/event/2690 ※現地参加は定員制です。お早めにご登録ください。

日本語

0

6

15

2.9K

Ryousuke Yamada@FragileGoodwill·24 Şub

@FumiyaUchiyama Amazing!! Congrats 🥳

English

1

0

1

123

Fumiya UCHIYAMA@FumiyaUchiyama·24 Şub

Our paper "CLIP-like Model as a Foundational Density Ratio Estimator" has been accepted to #CVPR2026 (main)🎉 We reinterpreted CLIP/SigLIP as density ratio estimator and proposed applications for transfer learning / semantic diversity metric for an image! arxiv.org/abs/2506.22881

English

2

15

78

6.8K

Ryousuke Yamada retweetledi

Takuma Yagi (八木拓真) Ph.D.@takuma_yagi_·24 Şub

Our work on fine-grained Hand-Object Interaction understanding VideoQA benchmark (HanDyVQA) has been accepted to #CVPR2026 main track! We created 11K QA pairs and 4K segmentation masks that require holistic understanding of HOIs as dynamic process. Dataset already available!

English

1

8

47

3.5K

Ryousuke Yamada retweetledi

Dawid Kopiczko@dawkopi·16 Şub

Common knowledge in ML: more unique training data → better generalization. Turns out this doesn't hold for long-CoT SFT. Under a fixed update budget, repeating a small dataset multiple times beats training on more unique samples. And it's not even close.

English

3

26

312

20K

Ryousuke Yamada@FragileGoodwill·24 Şub

@IdeKohsuke Congrats 👏

English

1

0

1

223

kohsukeIde@IdeKohsuke·23 Şub

Two papers (1 main, 1 findings) have been accepted to CVPR 2026! 🎉 More details coming soon! #CVPR2026

English

1

2

33

6.3K

Ryousuke Yamada@FragileGoodwill·21 Şub

@Masakichi333210 @CVPR Congrats 👏

English

1

0

167

Masaki Kawamura@Masakichi333210·21 Şub

Our paper "PowerCLIP: Powerset Alignment for Contrastive Pre-Training" has been accepted to @CVPR 2026! 🎉 See you in Denver!

Masaki Kawamura@Masakichi333210

🚀 New arXiv preprint! PowerCLIP is the first method to align **powersets of image region subsets with textual phrase structures**, enabling fine-grained compositional and robust image-text understanding beyond simple global or token-to-patch alignment.

English

4

17

90

19.4K

Ryousuke Yamada@FragileGoodwill·21 Şub

@ulo_fd Wow, おめでとうございます👍

日本語

0

40

R.Yana@ulo_fd·21 Şub

CVPR2本採択めでたい🎉

日本語

1

0

28

1.1K

Ryousuke Yamada@FragileGoodwill·21 Şub

@shibatashi_ Congrats 🎉

English

0

1

216

shibata@shibatashi_·21 Şub

Our paper has been accepted to CVPR 2026 main (this work was done during my visit to CMU)

English

1

2

77

5.2K

Ryousuke Yamada@FragileGoodwill·21 Şub

@dahlian0 Congrats 🥳

English

1

0

1

264

Risa Shinoda@dahlian0·21 Şub

My first author paper has been accepted to CVPR!🦋✨

English

3

5

126

7.4K

Ryousuke Yamada retweetledi

Ko Watanabe 🇩🇪@ko_watanabe_jp·8 Oca

貴重なニュルンベルク工科大学の体験記！ドイツの生活がすごく伝わる記事でした、Airbnb含め、物件詐欺がよくあるので書いてあることは本当に参考になるし、気をつけるべき点など把握できそうです・・・！ x.com/FragileGoodwil…

Ryousuke Yamada@FragileGoodwill

今更ながら、海外研究留学 Advent Calendar 2025にお誘いいただき、寄稿しました!! adventar.org/calendars/12626 ドイツ（ニュルンベルク）での家探し＆一人暮らし立ち上げの話を書きました🇩🇪 現地での仮住まいなどリアル体験をまとめています note.com/ryousukeeee/n/…

日本語

0

2

6

982

Ryousuke Yamada@FragileGoodwill·4 Oca

今更ながら、海外研究留学 Advent Calendar 2025にお誘いいただき、寄稿しました!! adventar.org/calendars/12626 ドイツ（ニュルンベルク）での家探し＆一人暮らし立ち上げの話を書きました🇩🇪 現地での仮住まいなどリアル体験をまとめています note.com/ryousukeeee/n/…

日本語

0

1

3

1.9K

Ryousuke Yamada retweetledi

特徴量の気持ち@t_Signull·11 Ara

生データを見ろおじさん｜八木拓真 @t_Signull cvpaper.challenge アドベントカレンダー 11日目の記事になります。 #CVPAdv2025 note.com/takumayagi/n/n…

日本語

1

54

177

62.5K

Ryousuke Yamada retweetledi

Masaki Kawamura@Masakichi333210·1 Ara

🚀 New arXiv preprint! PowerCLIP is the first method to align **powersets of image region subsets with textual phrase structures**, enabling fine-grained compositional and robust image-text understanding beyond simple global or token-to-patch alignment.

English

1

20

66

29.4K

Ryousuke Yamada

Keşfet