Rui Qian

Senior Research Scientist @Apple | CS PhD @Cornell

[Rui Qian @RuiQian3]

@PKUCXK 我现在还在那个31楼鹅腿群🤣

[Rui Qian @RuiQian3]

@xxunhuang will there be a streaming link? so pity can not be there!

[Aran Komatsuzaki @arankomatsuzaki]

Apple presents Manzano: Simple & scalable unified multimodal LLM

• Hybrid vision tokenizer (continuous ↔ discrete) cuts task conflict
• SOTA on text-rich benchmarks, competitive in gen vs GPT-4o/Nano Banana
• One model for both understanding & generation
• Joint recipe: pretrain + refine + SFT
• Scales cleanly (300M → 30B) with consistent gains

[Rui Qian @RuiQian3]

At @Apple AIML, we always care about model cost and efficiency, including generation models. Some attempts we made recently to simplify MMDiT, moving towards faster and stronger generation models!

[𝚐𝔪𝟾𝚡𝚡𝟾 @gm8xx8]

DiT-Air: Revisiting the Efficiency of Diffusion Model Architecture Design in Text to Image Generation



[Rui Qian @RuiQian3]

Some cool stuff we were experimenting with last year! #Imagen #Imagen2

[AK @_akhaliq]

Greedy Growing Enables High-Resolution Pixel-Based Diffusion Models

We address the long-standing problem of how to learn effective pixel-based image diffusion models at scale, introducing a remarkably simple greedy growing method for stable training of large-scale,

[Rui Qian @RuiQian3]

Happy to be part of the team and contribute to Imagen 3 and Veo! Amazing experience working with amazing team! #GoogleIO #Google #DeepMindAI

[Rui Qian @RuiQian3]

This is amazing...

[AK @_akhaliq]

Generative Agents: Interactive Simulacra of Human Behavior

abs: https://t.co/h9vI44UEke
project page: https://t.co/khS5i3jsHN

[Rui Qian @RuiQian3]

Of course this tweet is also generated by gpt-4. 😆😅@BelongieLab

[Rui Qian @RuiQian3]

Leveraging #ChatGPT (GPT-4 model) to tackle complex #LaTeX package conflicts and resolve formatting issues in subfigure/table while drafting my thesis. It's been a game-changer! 📚🤖 Huge shoutout to #OpenAI for creating such a helpful tool!👏 #AcademicTwitter #ThesisWriting

[Rui Qian @RuiQian3]

@NYCTSubway agree with the long F-training gap every weekend. People living on Roosevelt island need more help to travel during weekends!

[Roosevelt Islander @Rooseveltisland]

Every weekend

[Rui Qian @RuiQian3]

@ghostynewt Wow your love Xi'an food?!

[Rui Qian @RuiQian3]

@Rooseveltisland @iDig2Learn @BigReuse @JulieMenin @SeawrightForNY Can adults join as well?

[Rui Qian @RuiQian3]

Please check out our latest work on expert-curated fine-grained audio-visual categorization!

[Grant Van Horn @Grant_Van_Horn]

Today (10/26) we're presenting our work on Fine-Grained Audiovisual Categorization with the SSW60 Dataset at #ECCV2022 poster #115 from 3:30-5:30pm.
What are the benefits of image, audio, and video modalities, and how should you train your model? Drop by and find out! 🐦🤖🐤

[Grant Van Horn @Grant_Van_Horn]

Today (10/26) we're presenting our work on Fine-Grained Audiovisual Categorization with the SSW60 Dataset at #ECCV2022 poster #115 from 3:30-5:30pm.
What are the benefits of image, audio, and video modalities, and how should you train your model? Drop by and find out! 🐦🤖🐤

[Rui Qian @RuiQian3]

@ruomingpang The link seems down right now?

[Rui Qian @RuiQian3]

@YifanJiang17 Thanks!

[Rui Qian @RuiQian3]

Check out our latest research on multimodal open-vocabulary video understanding research!

[Yin Cui @YinCuiCV]

Can we use audio and motion modality to improve open-vocabulary video classification?

We equip CLIP with cross-modal fusion to leverage multimodal information. Our method MOV archives SOTA results on UCF and HMDB zero-shot action recognition.

https://t.co/rEp4YRXE6n

[AK @_akhaliq]

Multimodal Open-Vocabulary Video Classification via Pre-Trained Vision and Language Models
abs: https://t.co/nw8EPuxWtp

achieves sota results on UCF and HMDB zero-shot video classification benchmarks, outperforming traditional zero-shot methods and recent methods based on VLMs

[Rui Qian @RuiQian3]

@BelongieLab

[Rui Qian @RuiQian3]

@Rooseveltisland @JulieMenin @SeawrightForNY @MTA @RIOCny @RiocCeo Very annoying. It was soooo crowded to get on tram to go back to roosevelt island today.