@yinzeyuan1037 Profile picture

Zeyuan Yin

@yinzeyuan1037

CS PhD student @michiganstateu. Prev ML @mbzuai, CS @2024_HUST, Intern @AlibabaGroup DAMO Academy. Looking for a AI research position @ USA

Similar User
点头象联系🌈网赚🌈偏门🌈灰产🌈赚钱🌈项目🌈副业🌈快钱🌈挣钱🌈区块链🌈灰产项目 photo

@Eric_Yao_

Hasan Hammoud photo

@hammh0a

Yueming Jin photo

@JinYueming

阳光 photo

@doublegirl2021

fiona.smol photo

@CryptoFi_18

Jonathan LEI photo

@xJonathanLEI

Zhen Zhang photo

@ZhenZHANG1120

Neil Han photo

@NeilHANYD

Zeyuan Yin Reposted

Teaching a big class this semester: efficientml.ai

Tweet Image 1

Zeyuan Yin Reposted

arxiv.org/abs/2409.02426 This is our latest work that tries to reveal what diffusion/denoising model is really doing in terms of learning a data distribution: Subspace Clustering and Subspace Denoising. What surprises me the most is the number of samples required to learn the…


Zeyuan Yin Reposted

1/7 As a CS PhD student, I've often reflected on how challenging it can be for newcomers to enter research, especially with limited mentorship. The questions, uncertainties, and lessons I've encountered inspired me to create a resource I wish I had when starting out.


Zeyuan Yin Reposted

I'm working on the next generation AI systems myself, not on LLMs. So technically, I'm telling you "compete with me", or rather, "work on the same thing as me, because that's the way to go, and theore the merrier!"


Zeyuan Yin Reposted

If you are a student interested in building the next generation of AI systems, don't work on LLMs

The Godfather of AI is at #VivaTech! Yann LeCun (@ylecun) advises students coming into the industry: "Don't work on LLM. This is in the hands of large companies, there's nothing you can bring to the table. You should work on next-gen AI systems that lift the limitations of LLMs.

Tweet Image 1


Zeyuan Yin Reposted

Looking to land a top job in GenAI? Focus is key. Instead of spreading yourself thin with 20 unrelated papers during your PhD, hone in on 2-3 representative publications in areas like SFT, RLHF, alignment, safety, or data selection. Make a name for yourself w. quality work.


Zeyuan Yin Reposted

Yes! We are looking for contributors for OpenDevin! Here are some ways to get started: 1. Join discussions on github, slack, or discord: github.com/openDevin/Open… 2. Take a look at the "good first issues" and try to work on them: github.com/OpenDevin/Open…

That's amazing . Are you guys looking for contributors , not sure how to start ?



Zeyuan Yin Reposted

I’m excited to share that I’m working on a new book about building applications with foundation models! AI Engineering builds upon Machine Learning Systems Design, but with a focus on large scale, ready made models. The book covers: - The new AI stack (e.g. how it differs from…

Tweet Image 1

To the moon and back

Doing an AI infra startup is like the moon mission: you calculate the burn carefully to get into orbit, you maneuver well to land a good exploration spot, and you make sure you have enough supplies to carry you back to earth to call it a success.



Zeyuan Yin Reposted

New YouTube video! Diffusion models are awesome! But how do they actually work? 🤔 Check out the video and learn with me! youtube.com/watch?v=i2qSxM…

Tweet Image 1

Zeyuan Yin Reposted

I've been enjoying Penzai the new Jax lib. It's very opinionated, but close to my ideal NN library. To test it out I ported the Tensor Puzzles to use NamedArrays. Feels so clean without the [:, None]'s srush.github.io/Tensor-Puzzles…

Tweet Image 1
Tweet Image 2

Zeyuan Yin Reposted
Tweet Image 1

Zeyuan Yin Reposted

What design choices matter when developing a visually-conditioned language model (VLM)? Check out our paper – Prismatic VLMs – and open-source training code, evaluation suite, and 42 pretrained VLMs at the 7B-13B scale! 📜 arxiv.org/abs/2402.07865 ⚙️ + 🤗 github.com/TRI-ML/prismat…

Tweet Image 1

Zeyuan Yin Reposted

🚀 🌐 Build your own video generation model like #Sora! Experience the power of replication without the price tag! Open-Sora delivers a low-cost implementation of Sora, cutting costs by a staggering 46%. Expand your sequences to nearly a million with this innovative open-source…


Zeyuan Yin Reposted

🔥Meet #EMMA the Embodied Multi-Modal Agent🤖 trained by imitation learning from a reflexion #LLM Agent🤖 in a parallel textworld📜using #DPO-#DAgger🚀Boosting the success rate of #ALFWorld visual-only tasks from ~20% ➡️ >80%! Accepted by #CVPR2024 Paper: arxiv.org/abs/2311.16714

Tweet Image 1

Zeyuan Yin Reposted

Our 2020 paper "Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention" with @angeloskath @apoorv2904 and @nik0spapp reached 1000 citations! proceedings.mlr.press/v119/katharopo…


Zeyuan Yin Reposted

Let's implement Mamba in Triton. (srush.github.io/annotated-mamb…) A gentle, (but mildly obsessive) tutorial notebook about GPU programming in Triton. We're getting close to mere mortals being able to do this 😂

Tweet Image 1
Tweet Image 2
Tweet Image 3
Tweet Image 4

Zeyuan Yin Reposted

Quadratic attention has been indispensable for information-dense modalities such as language... until now. Announcing Mamba: a new SSM arch. that has linear-time scaling, ultra long context, and most importantly--outperforms Transformers everywhere we've tried. With @tri_dao 1/

Tweet Image 1

Zeyuan Yin Reposted

How many people have actually read the Gemini 1.5's technical report? IMHO, the Kalamang translation result is a highlight.

Kalamang Translation One of the most exciting examples in the report involves translation of Kalamang. Kalamang is a language spoken by fewer than 200 speakers in western New Guinea in the east of Indonesian Papua (endangeredlanguages.com/lang/1891). Kalamang has almost no online…

Tweet Image 1
Tweet Image 2


Zeyuan Yin Reposted

🚨 Breaking Research Discovery! 🚨 Large Vision Language Models (#VLMs) amaze with coherence but hide risks. 🤯🗡️ 🔍 Meet "Shadowcast"🥷: A stealthy, mind-bending AI data poisoning method. 🕵️‍♂️💻 Project page 🔗: vlm-poison.github.io #LLMs #DataSecurity A 🧵 👇

Tweet Image 1
Tweet Image 2
Tweet Image 3
Tweet Image 4

Loading...

Something went wrong.


Something went wrong.