1 paper · 1 filter
Nathan Hu, Jake Ward, Thomas Icard +1
While reasoning models are increasingly ubiquitous, the effects of reasoning training on a model's internal mechanisms remain poorly understood. In this work, we introduce transcod…