Ashwinee Panda
@PandaAshwinee
๊ฐ์
February 2020
654 ํ๋ก์ ์ค 3.9K ํฌ
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. ๐งต
๋ ๋ณด๊ธฐ
