Arnav Chavan on X: "How can LLMs learn new skills from offline data without losing what they already know? Introducing Subspace Tuning (SST): a simple approach that improves the learning–forgetting trade-off by controlling where weight updates happen. Paper: https://t.co/JfelwtlobJ Blog: https://t.co/n9t1DliFBk" / X
x.com · saved by 1 readers
How can LLMs learn new skills from offline data without losing what they already know? Introducing Subspace Tuning (SST): a simple approach that improves the learning–forgetting trade-off by controlling where weight updates happen. Paper: https://t.co/66hCsCIgeE Blog: https://t.co/tzi6j86Yzg 🧵 1…