@misc{indiciae6a4928e6bddb,
  title = {Align, Then Correct: Training-Free Two-Stage Low-Rank Compensation for Extremely Quantized Large Language Models},
  author = {Seobin Song and Geonho Lee and Janghwan Lee and Jungwook Choi},
  year = {2026},
  url = {https://arxiv.org/abs/2610.08164},
  note = {Source identifier: 2610.08164}
}