Tag
Proposes in-cell learning for bit-identical, revocable updates to quantized LLMs, introducing the CellFill algorithm to update models within quantization cells without altering the original artifact.