IAAILoopQ
LoopQ: Quantization for Looped Language Models
A loop-aware post-training quantization framework that keeps one shared quantized backbone while adding loop-aware activation scaling, selective transformations, cross-loop state alignment and trajectory-aware calibration to limit recursive error.
Inside the method
Shared W4A4 backbone × R + loop-aware scaling and cross-loop alignment
Simplified conceptual schematic. Consult the paper for the complete architecture.
- Recurrence family
- Full-stack recurrence
- Depth control
- Fixed loops of the base model
- KV / state strategy
- See paper
Reading note
Under W4A4, reports a 67.8% average relative improvement in five-task accuracy across four backbones and about 75% relative perplexity reduction versus the strongest static PTQ baseline.
Sources checked 2026-10-28. This catalog does not imply independent reproduction.
Related works
Cite this work
@misc{loopq2026,
title = {LoopQ: Quantization for Looped Language Models},
author = {Rui Fang and Hsi-Wen Chen and Chia-Ming Cheng and Ming-Syan Chen},
year = {2026},
eprint = {2605.16343},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.16343}
}