79.
Having worked on unlearning for multiple years, it was clear that post-training "fixes" alone were a dead-end. Moโฆ
Having worked on unlearning for multiple years, it was clear that post-training "fixes" alone were a dead-end. Model learning is way too entangled. With ๐ก๐จ๐๐๐ we decided to architect unlearnability into the model, and scaled it to 1B+