Test your interpretability techniques by de-censoring Chinese models

2 points| allenleee | 1 month ago |lesswrong.com