Yao Mark Mu

Ph.D. candidate at The University of Hong Kong. Obtained M.Phil Degree at Tsinghua University. Interest: Reinforcement Learning and Representation Learning.

[Mengkang Hu @aaron_mkhu]

1/ 📢Save Time and Money for LLM-Planning! 💰💻

🙌Introducing 🌳TREE-PLANNER🌳: A novel framework for efficient and effective close-loop task planning with LLMs

📟Project Page: https://t.co/tWgZfeEf44
📜Paper: https://t.co/Yqu0A41cVZ

More details 👇

[Yao Mark Mu @YaoMarkMu1]

Really happy to announce that EmbodiedGPT has been accepted as a Spotlight at NeurIPS2023! 🎉 #NeurIPS2023

[Yao Mark Mu @YaoMarkMu1]

🤝 Seeking #Postdoc opportunities! Please DM or email [[email protected]] 🙏
🧠 Research interest: Embodied AI, Reinforcement Learning, Robot Control, Autonomous Driving 🤖

[Edward Hu @edward_s_hu]

Dropped into a new environment, how can an agent discover and accomplish diverse tasks within it?

In our ICLR'23 spotlight paper, we propose Planning Exploratory Goals (PEG), a RL agent that solves hard tasks like block stacking with no supervision!

https://t.co/JybMdxRGwA

[Pan Lu @COLM2026 @lupantech]

🚀Meet Chameleon! An innovative plug-and-play framework enhancing #GPT4 and #ChatGPT like #AutoGPT for compositional reasoning, blending off-the-shelf tools with tailored LLM models 🔧✨🧠. New SOTA on #ScienceQA and TabMWP! 📈

🔗https://t.co/IBCvtHfz7Q
📜https://t.co/b0fETXZHfq

[Yao Mark Mu @YaoMarkMu1]

@ShoufaChen Cool!

[ShoufaChen @ShoufaChen]

Segment Anything and Edit at 🚀Part Level

👉 Edit anything at the part level.
👉Visual ChatGPT controller.
github: https://t.co/iVDzoYM81C
blog: https://t.co/kYd6dNImEe

[Yao Mark Mu @YaoMarkMu1]

@GoogleAI @du_yilun really great work! will Unipi open source the codebase?

[Yao Mark Mu @YaoMarkMu1]

Project page: https://t.co/yUgIKw3Y61

[Yao Mark Mu @YaoMarkMu1]

Metadiffuser has been accepted by #ICML2023

[Yao Mark Mu @YaoMarkMu1]

Adaptdiffuser has been select Oral presentation in ICML2023！
arxiv link: https://t.co/Es7A2CRVbf

[Cameron R. Wolfe, Ph.D. @cwolferesearch]

Can large language models (LLMs) train themselves? Recent research indicates that the answer might be yes… 🧵 [1/7]

[Shunyu Yao @ShunyuYao12]

Excited about this work on emergent communication (EC)! EC's been a tricky subject (i.e. lots of toy papers), but IMO the true potential is unleashing soon.

Simplest reason: we're running out of language by humans on Internet. Have to use machine's self-generated language soon!

[LangChain @LangChain]

🤖LangChain x AutoGPT🤖

First BabyAGI, then CAMEL, then generative agents, now finally AutoGPT...

Implementing the core AutoGPT prompt + agent executor in LangChain primitives (to allow for swapping between models, vectorstores, tools, etc)

Notebook: https://t.co/2hag1gISd8

[Jianwei Yang @jw2yang4ai]

Thanks to @_akhaliq , @huggingface , @Gradio for the support! We had moved our SEEM demo to huggingface! It supports panoptic segmentation, various types of referring segmentation with text, audio, stroke, example image, and the combs! Have a try here: https://t.co/IJb9CYFKUo

[Andreea Bobu @andreea7b]

We’re organizing a very exciting workshop at ICML this year! If your work touches upon interactive learning or learning from human feedback, this workshop is for you!

[Yao Mark Mu @YaoMarkMu1]

Our new paper, EC^2, has been published in CVPR2023. It presents a novel video-language per-training scheme via emergent communication for few-shot embodied control.
Project page: https://t.co/CpuPCdfQeO Paper:https://t.co/kam0XChQJk

[Danijar Hafner @danijarh]

Excited to announce DreamerV3 🌍, a scalable and general RL algorithm that masters a wide range of applications with fixed hyperparameters!

Applied out of the box, it solves the Minecraft Diamond challenge without human data. 💎

👇 Thread
https://t.co/zTAy1HZvIW

[Ethan Caballero @ethanCaballero]

U-Net is dead. Transformers are SotA Diffusion Models now.

"Scalable Diffusion Models with Transformers"

paper: https://t.co/O11p80rmlb
website: https://t.co/hJZeDVlPDp

[AK @_akhaliq]

GeoCode: Interpretable Shape Programs
abs: https://t.co/hyMBxFsW5n
project page: https://t.co/PCPguHSQyj