Ziyi Yang
@yzy_ai
Followers
425
Following
321
Media
11
Statuses
85
Research Scientist @DbrxMosaicAI | @Stanford Ph.D.
Joined November 2016
🚀 #ICML2025 is here! Join the @databricks social event to chat LLMs, reasoning, multimodal AI & beyond.📅 Jul 15 · 6–8:30 PM .Secure your spot now 👉 #AgentBricks #LLM #Reasoning #Agent.
0
0
4
Thrilled to cross 3K citations just as I start my next chapter at @databricks! 🚀 Huge thanks to collaborators & mentors who’ve fueled this journey. Can’t wait to turn research into impact at DBRX🔥
1
0
9
Another great work by @ZinengTang , generative contrastive learning for vision+language.
We are thrilled to announce TULIP!. 🌷 A state of the vision language encoders coupled with generative model for stronger representation learning.
1
0
4
RT @NeurIPSConf: NeurIPS acknowledges that the cultural generalization made by the keynote speaker today reinforces implicit biases by maki….
0
389
0
We announced Phi 3.5 series today! .1️⃣ Multilingual Mini 3.8B: 2️⃣ MoE 16x3.8B (active 6.6B): 3️⃣ multi-frame vision LLM:
We released phi 3.5: mini+MoE+vision. A better mini model with multilingual support: A new MoE model: A new vision model supporting multiple images:
0
6
42
The poster is up! See you there🎉. @ZinengTang @mohitban47
CoDi-2 is selected as #CVPR2024 Highlight. Come joint us in today’s poster session Arch 4A-E #314 in 5pm to 6:30 pm!. @yzy_ai @nlpyang @ChenguangZhu2 @mohitban47.
1
1
6
Happy to meet&chat on LLM+multimodal at CVPR poster session today! #314 for CoDi-2.
CoDi-2 is selected as #CVPR2024 Highlight. Come joint us in today’s poster session Arch 4A-E #314 in 5pm to 6:30 pm!. @yzy_ai @nlpyang @ChenguangZhu2 @mohitban47.
0
0
2
Check out our latest work on LLM online preference alignment🔥🔥🔥.
🚀Excited to share the work during my internship at Microsoft!. We proposed Self-Exploring Language Models (SELM) that actively explore potentially high-reward responses in online RLHF. A thread🧵: (1/8). Paper: Code:
0
1
3
Proud to lead the post-training of Phi-3 models. Shout out to our amazing Microsoft GenAI team! We are so back🔥🔥🔥.
Microsoft announces Phi-3. A Highly Capable Language Model Locally on Your Phone. We introduce phi-3-mini, a 3.8 billion parameter language model trained on 3.3 trillion tokens, whose overall performance, as measured by both academic benchmarks and internal testing,
0
0
8
RT @Diyi_Yang: Learning social skills is out of reach for most people🙁 . How can we make social skill training more accessible?. We introdu….
0
61
0
My friend @xycking at @character_ai is hiring researcher/engineer/designer/PM etc! Open roles:
0
0
1
…and this time how much longer its gonna take humans to make the “atomic bomb”🤔.
0
0
1
UDOP (@CVPR’23 spotlight), Universal Document Processing, is on @huggingface🥳.Great thanks for the support @NielsRogge. HF model repo: Paper: Project: @ZinengTang @mohitban47 @studyfang_ @nlpyang.
A powerful new document AI model finally landed in @huggingface! LayoutLMv3 and Donut decided to have a baby and call it UDOP. It's a vision encoder, text decoder Transformer similar to Donut 🍩 combined with an OCR engine as in LayoutLM to get the best of both worlds. 1/2
1
4
26
Super excited to share that CoDi-2 is accepted to @CVPR, see you in Seattle this summer! #CVPR2024 #CVPR #LLMs. Project Page: @ZinengTang @mohitban47 @nlpyang.
Codi-2: an interleaved, multimodal LLM that supports arbitrary audio, image, and video mixture in I/O. You can give instructions like:. - "Given video frames [video], what happens next? Generate sound" ==> next frame, audio.- "An [image A] on [image B]" ==> composite image of A &
1
1
12
Arrived at #NeurIPS2023! Will present CoDi on Tue 9am at @Microsoft booth & 5:15pm poster session with my intern @ZinengTang. I’m also hiring interns for spring/summer, DM if you wanna chat at the venue!. CoDi: CoDi-2:
3
0
6