IAAIPoLar
Skip a Layer or Loop It? Learning Program-of-Layers in LLMs
Finds that pretrained LLM layers can be skipped or repeated to form a per-input program-of-layers without training, and learns a lightweight network that predicts such execution programs for each input.
Inside the method
Input → program predictor → sequence of skipped / repeated pretrained layers
Simplified conceptual schematic. Consult the paper for the complete architecture.
- Recurrence family
- Adaptive depth
- Depth control
- Per-input layer program (skip or loop)
- KV / state strategy
- See paper
Reading note
On mathematical reasoning benchmarks it improves accuracy over standard inference and prior dynamic-depth methods, often with fewer executed layers, and gains persist out of distribution.
Sources checked 2026-10-28. This catalog does not imply independent reproduction.
Related works
Cite this work
@misc{polar2026,
title = {Skip a Layer or Loop It? Learning Program-of-Layers in LLMs},
author = {Ziyue Li and Yang Li and Tianyi Zhou},
year = {2026},
eprint = {2606.06574},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2606.06574}
}