80.
We are taking a big step towards scaling LLMs that can unlearn on demand.
We are taking a big step towards scaling LLMs that can unlearn on demand. Cleanly deleting data from LLMs has proven impossible: training entangles every source in shared weights. NULLs (Natively Unlearnable LLMs) escapes this, keeping mill