Pinned
We can now RL large MoEs with 0 train-infer mismatch! And doing so can improve performance (pictured task: teach Qwen3.6-35B-A3B to play Wordle). Everything is open-source and we did a bunch of ablations. 🧵
Ashwinee Panda
@PandaAshwinee
Research @flappyairplanes | Prev: RL @togethercompute, Postdoc of @tomgoldsteincs, PhD @princeton, @Berkeley_EECS alum
