PhD Student at the Learning and Intelligent Systems Lab @TUBerlin. Intern @allen_ai. Working on discovering new things with machine learning and robots.
🚨The submission deadline of our workshop has been extended until 18th of June.
Find out more at: https://t.co/aaMTcHZ0vi
Looking forward to your contributions!
#RSS2026 Call for participants 📢
Super excited to announce our RSS 2026 Workshop: Sampling-Based Optimization for Robotics!
The workshop will be held on July 17 in Sydney.
More info: https://t.co/Vh44l4RUoL。
#RSS2026 Call for participants 📢
Super excited to announce our RSS 2026 Workshop: Sampling-Based Optimization for Robotics!
The workshop will be held on July 17 in Sydney.
More info: https://t.co/Vh44l4RUoL。
@chris_j_paxton@DJiafei Totally agree, felt like the first paper that really evaluated all these models side-by-side. I wonder, what happened to GR00T though?
Just came back home from @RL_Conference and I really enjoyed it.
Summary: Too many great speakers to name them all, too many great people to name them all, and we won the Best Paper award at the @ibrlworkshop! 🏆
Thanks to all the great people :-)
Really enjoyed #UAI2025 in Rio last week!
Met a lot of great researchers, presented "Stein Variational Evolution Strategies" as an oral, and appreciated Copacabana.
Big thanks again to the organizers! Already looking forward to the next conferences in Brazil.🇧🇷
Introducing Stein Variational Evolution Strategies
Exciting new work on leveraging Stein Variational methods for discovering diverse solutions with ES 🦎
Work led by the amazing @corbraun 🤗 Check out his thread for many more details 🎨
🧑💻Code: https://t.co/B5RSE3ieUQ
📝 NB: https://t.co/bMwNXhPWN7
Credits @maxencefaldor 🙏
I am beyond excited to share our new work: Stein Variational Evolution Strategies
I'm really excited about this for many reasons:
1. I got to collaborate with the amazing @RobertTLange and @Marc__Toussaint.
2. It will be an oral presentation at #UAI2025 next month!
A thread. 🧵
You may wonder: "But why not just run multiple CMA-ES in parallel?"
Well, the SVGD-inspired coordination is key!
It pushes the populations apart to find diverse, high-quality solutions, preventing them from all collapsing to a single mode. Our method finds the whole landscape.