SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2601.07892 · 2026

Sherry: Hardware-Efficient 1.25-Bit Ternary Quantization via Fine-grained Sparsification

Xue Liu, Dapeng Wu, Hong Huang, Jianchen Zhu, Jinhai Yang, Guanghua Yu, Decheng Wu, Qiangqiang Hu

arXiv · PDF · Open in the Atlas

Code that ran

We have not lifted any functions out of this paper's repositories yet, so there is nothing we have run. The repositories linked to it are listed below.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The deployment of Large Language Models (LLMs) on resource-constrained edge devices is increasingly hindered by prohibitive memory and computational requirements. While ternary quantization offers a compelling solution by reducing weights to {-1, 0, +1}, current implementations suffer from a fundamental misalignment with commodity hardware. Most existing methods must choose between 2-bit aligned packing, which incurs significant bit wastage, or 1.67-bit irregular packing, which degrades inference speed. To resolve this tension, we propose Sherry, a hardware-efficient ternary quantization framework. Sherry introduces a 3:4 fine-grained sparsity that achieves a regularized 1.25-bit width by packing blocks of four weights into five bits, restoring powerof-two alignment. Furthermore, we identify weight trapping issue in sparse ternary training, which leads to representational collapse. To address this, Sherry introduces Arenas, an annealing residual synapse mechanism that maintains representational diversity during training. Empirical evaluations on LLaMA-3.2 across five benchmarks demonstrate that Sherry matches state-of-the-art ternary performance while significantly reducing model size. Notably, on an Intel i7-14700HX CPU, our 1B model achieves zero accuracy loss compared to SOTA baselines while providing 25% bit savings and 10% speed up. The code is available at https://github.com/Tencent/AngelSlim.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2601.07892")
get_code_for_paper("2601.07892")
have("2601.07892")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2601.07892.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2601.07892.svg)](https://syntology.ai/paper/2601.07892)

Connect an agent — have() is free.