GRRR: The Geometry of Reshaping, Rotation, and Routing in Decoder LLM post-training
arXiv:2609.22146v1 Announce Type: new Abstract: We study how post-training changes the weights of Large Language Models (LLMs) relative to their pretrained weights. Across 12 post-training chains with supervised fine-tuning (SFT) and reinforcement learning (RL), we express each weight update in the…
Read the full story at arXiv cs.LG ↗
Timeline · 1 report
- 2026-09-22 04:00 · arXiv cs.LG
GRRR: The Geometry of Reshaping, Rotation, and Routing in Decoder LLM post-training