Ge Yang (@episodeyang) 's Twitter Profile
Ge Yang

@episodeyang

I am planting acorns one at a time with policy gradient.

ID: 424092529

linkhttp://www.episodeyang.com calendar_today29-11-2011 09:35:25

2,2K Tweet

2,2K Followers

2,2K Following

Ge Yang (@episodeyang) 's Twitter Profile Photo

Look at what Yue Wang got to work with his students! And the best part is you can also train your robot on these data : ) and there is no VXF involved in this video :-P

Ge Yang (@episodeyang) 's Twitter Profile Photo

Check this out — they use a sparse MoE with k=2, to allow post-training pruning that reduce inference cost. It is quite clever 👏 Moritz Reuss and Jyo Pari !

Sholto Douglas (@_sholtodouglas) 's Twitter Profile Photo

A distillation of our mental models that we use to think about the systems perspective on training and inference at scale. The most important takeaway - you should be able to describe everything about your model with simple equations, and deeply understand how long it should

Kimi.ai (@kimi_moonshot) 's Twitter Profile Photo

🚀 Introducing our new tech report: Muon is Scalable for LLM Training We found that Muon optimizer can be scaled up using the follow techniques: • Adding weight decay • Carefully adjusting the per-parameter update scale ✨ Highlights: • ~2x computational efficiency vs AdamW

🚀 Introducing our new tech report: Muon is Scalable for LLM Training

We found that Muon optimizer can be scaled up using the follow techniques: 
• Adding weight decay
• Carefully adjusting the per-parameter update scale

✨ Highlights:
• ~2x computational efficiency vs AdamW
Ge Yang (@episodeyang) 's Twitter Profile Photo

What really excites me about this is that Atlas vector search will become even better, making it easier for a lot of smaller teams.

Xuxin Cheng (@xuxin_cheng) 's Twitter Profile Photo

Meet 𝐀𝐌𝐎 — our universal whole‑body controller that unleashes the 𝐟𝐮𝐥𝐥  kinematic workspace of humanoid robots to the physical world. AMO is a single policy trained with RL + Hybrid Mocap & Trajectory‑Opt. Accepted to #RSS2025. Try our open models & more 👉