Ghosted Layers: Unconstrained Activation Alignment for Recovering Layer-pruned LLMs

Vincent-Daniel Yun, Junhyuk Jo, Sai Praneeth Karimireddy, Sunwoo Lee

ICML Workshop, 2026

arXiv

This work proposes Ghosted Layers, a training-free recovery module for layer-pruned Transformer LLMs that derives a closed-form optimal linear operator from a small calibration set to correct the boundary activation mismatch introduced by pruning, achieving the unconstrained optimum of the alignment objective and consistently improving accuracy and perplexity over prior training-free baselines while preserving pruning’s efficiency gains.

BibTeX

@article{yun2026ghosted,
  title={Ghosted Layers: Unconstrained Activation Alignment for Recovering Layer-Pruned LLMs},
  author={Yun, Vincent-Daniel and Jo, Junhyuk and Karimireddy, Sai Praneeth and Lee, Sunwoo},
  journal={arXiv preprint arXiv:2605.15491},
  year={2026}
}