Setareh Cohan

Ph.D. Student @UBC
ML/RL Research @ubc_cs

[Guy Tevet @GuyTvt]

CLoSD meets @twominutepapers—what an honor! 🤩 Thank you, Dr. Károly Zsolnai-Fehér, for making dreams come true! 🙏
https://t.co/1NvFsJdCEc

[AI at Meta @AIatMeta]

New release from Meta FAIR — Meta Motivo is a first-of-its-kind behavioral foundation model for controlling virtual physics-based humanoid agents for a wide range of complex whole-body tasks.

The model is capable of expressing human-like behaviors and achieves performance competitive with task-specific methods and outperforms state-of-the-art unsupervised RL and model-based baselines.

Try the demo ➡️ https://t.co/AbbFGJ6rqr
Get the model and code ➡️ https://t.co/TFVLxYrhIX

We’re excited about how this research could pave the way for fully embodied agents, leading to more lifelike NPCs, democratization of character animation and new types of immersive experiences.

[Ricky T. Q. Chen @RickyTQChen]

Want to learn continuous & discrete Flow Matching? We've just released:

📙 A guide covering Flow Matching basics & advanced methods https://t.co/uf3zDfhdKe.

💻 An open source codebase with image & text examples https://t.co/Nfsoss8ZP2.

🗣️ A Flow Matching tutorial #NeurIPS2024.

[Michael Black @Michael_J_Black]

My talk on “A history of human motion in Computer Vision: From puppets to LLMs” is available here with an extensive bibliography: https://t.co/J8nzXtI9Fa

[Jaakko Lehtinen @jaakkolehtinen]

In this new work, we seek to understand what guidance really means in the context of diffusion models. No, it’s not about the prompts!

https://t.co/3v7BNMjOfu

@NVIDIAAI @CSAalto @FCAI_fi

[Setareh Cohan @setarehcohan]

Flexible Motion In-betweening with Diffusion Models🧎🚶🧍Done together with @GuyTvt, @rdednl, @xbpeng4, and @Mvandepanne

📄Paper: https://t.co/4QqS7l3lsV
🌐Website: https://t.co/GfkFaTvmM2

Grateful for the insights and feedback provided by everyone!

[Waymo @Waymo]

Today, we’re introducing Waymax, a first-of-its-kind simulator developed specifically for solving autonomous driving research problems around planning and sim agents. Discover how researchers can access Waymax: https://t.co/AVQA1eaahN

[RL Theory Virtual Seminars @RLtheory]

We are back!

Upcoming:
05/30: Ying Jin (Stanford)
06/06: Yujia Jin (Stanford)
06/13: Zeyu Jia (MIT)
06/20: Han Zhong (Peking)
06/27: Kefan Dong (Stanford)
07/04: Qinghua Liu (Princeton)
07/11: Ahmed Touati & Yann Ollivier (Meta AI)
07/18: Ayush Sekhari (MIT)

[Mahesh Sathiamoorthy @madiator]

Someone wrote article on Transformers from scratch: https://t.co/Zyq4nuHzsN

I mean, really from scratch..

[Philipp Wu @philippswu]

Introducing Masked Trajectory Modeling (MTM), a new general-purpose framework for sequential decision making. A single transformer trained with MTM can exhibit multiple capabilities by simply choosing different masking patterns at inference time. Accepted at ICML 2023. 🧵👇

[Google DeepMind @GoogleDeepMind]

Football players can tackle, get up, kick and chase a ball in one seamless motion. How could robots master these motor skills? ⚽

We trained AI agents to demonstrate these agile behaviours using end-to-end reinforcement learning.

Find out more: https://t.co/LkYtaMeUEd

[Sergey Levine @svlevine]

We released a new version of implicit Q-learning (IQL) with diffusion-based policies to get even better results with less hyperparameter tuning. For paper&code:
paper: https://t.co/CIdmJXwC5j
code: https://t.co/cd3vp2557Q
by Philippe Hansen-Estruch, @ikostrikov, et al.
A thread:

[NVIDIA AI Developer @NVIDIAAIDev]

Check out NVIDIA’s newest generative model for complex open-world 3D scenes.

It will generate novel scenes, which you can then edit and style for a wide range of uses, including potential use cases in VR & robotics.
https://t.co/gt8Vnp1KYx

[Ben Parr @benparr]

🚨 HUGE news in AI: Google just launched Generative AI across ALL of Google Workspace -- Gmail, Docs, Sheets, Slides, Images -- EVERYTHING.

They made a video showing off the new AI's capabilities. It's AWESOME.

[Volodymyr Kuleshov 🇺🇦 @volokuleshov]

Question to the panel—what are the open problems in diffusion generative models? Top answer—generalizing to discrete sequences and coming up with good corruption processes for that domain.

[Setareh Cohan @setarehcohan]

I couldn’t make it to #NeurIPS2022 because my visa wasn't approved. Glad my poster made it!😅