ali (@waterloo_intern)
@waterloo_intern
model perf @baseten || eng @uwaterloo
GPU kernels, inference, and the occasional (peer-reviewed) shitpost
99 正在关注 5.1K 粉丝
we distilled 2.3M Claude Fable 5 reasoning traces into Qwen3-4B
- 100% self-consistency @ 512 samples
- 0.00 bits output entropy
- zero hallucination variance
turns out the student is not bounded by the teacher.
it also converged on one universal truth.
we open-sourced the model weights👇
显示更多
