The story’s score and claims

Comments on No. 80 of 140 · Ian’s Blog · edition of Friday 20 November 2026

This weight does not spark joy

Summary

Deletion tuning fine-tunes a language model by zeroing a very sparse set of weights in the feed-forward gate matrices, rather than updating values as full fine-tuning or LoRA do. Lu et al. report it forgets less and needs a much smaller mask.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules