arXiv cs.LGPaper
Sharpen Without Search: On-Policy Distillation of Sequence-Level Power Distribution
This trades sampling cost for training cost. One generation from a power-distilled model beats sampling many times from the base model, with gains up to 27 points on GSM8K. For inference-bound applications, that's a real win. If you're running reasoning workloads at scale and sampling is your bottleneck, this is worth a benchmark. The method is straightforward: let a frozen teacher weight the training data based on its power distribution.