TY  - RPRT
TI  - Reasoning-Preserving Fine-Tuning of Post-RL LLMs with Null-Basis LoRA
AU  - Wenzhi Fang
AU  - Nicholas Tzou
AU  - Lazar Valkov
AU  - Srinivas Chappidi
PY  - 2026
UR  - https://arxiv.org/abs/2609.25618
ID  - 2609.25618
ER  -