WORLDWIDE_SanGuai / Research bulletin

[ worldwide.arakne.de ]

世界很大,讀一小點。

SanGuai 的三日研究佈告板。每天更新,讀完就走。

EST. 202672H上架 72 小時
人工智慧 / 2611.13004

G-STAR:以群組縮放詞元容量實現深度壓縮的自迴歸生成

G-STAR: Group-Scaled Token Capacity for Deeply Compressed Autoregressive Generation

作者: Xu Huang, Ye Huang, Xiaotong Li, Zhengrong Yue, Jialv Zou, Minhao Jing, Haodong Li, Wenhao Yan et al.論文發布: 2026-11-08離架倒數: 14 小時

研究速讀

此論文提出G-STAR,基於乘積量化推導出在固定名義資訊預算下,群組大小應隨單軸空間壓縮倍數平方成長,並據此設計空間與群組共同縮放的框架,聲稱可達32倍甚至有效64倍壓縮。作者另提出半自迴歸解碼器,結合平行特徵計算與輕量因果建模,以處理大群組內的相依性。摘要稱其在五項文字生圖基準上具競爭力,並在1024×1024解析度下較先前離散自迴歸模型快20倍。

值得留意

  1. 摘要提出群組大小隨單軸空間壓縮倍數平方縮放的設計原則,並稱可支援32倍與有效64倍壓縮。
  2. 為處理大群組內部相依性,提出結合平行特徵計算與輕量因果建模的半自迴歸解碼器。
  3. 作者稱在GenEval、DPG-Bench、PRISM、LongText-Bench與Qwen-Image-Bench上表現具競爭力,並稱單張H200上每張1024×1024影像約2.5秒、較先前方法快20倍。

閱讀限制

僅閱讀標題與摘要,未檢視方法細節、實驗設定、基準比較方式或消融結果,因此無法驗證壓縮倍率、速度與基準表現等宣稱。

原始摘要 [EN]

Discrete autoregressive models offer a scalable framework for image generation. However, they face a fundamental trade-off: long token sequences preserve representation quality but slow generation, while aggressive spatial compression reduces representation capacity. Based on product quantization, we show that, under a fixed nominal information budget, group size should scale quadratically with the per-axis spatial compression factor. Building on this principle, we propose GSTAR, a spatial--group co-scaling framework that enables \(32\times\) and even effective \(64\times\) compression while preserving strong reconstruction, understanding, and generation performance. Since larger groups introduce intra-group dependencies that must be modeled efficiently, we analyze their impact on inference latency and generation quality and introduce a semi-autoregressive decoder that combines parallel feature computation with lightweight causal modeling. With these designs, our highly compressed models perform competitively with leading models across five representative text-to-image benchmarks (GenEval, DPG-Bench, PRISM, LongText-Bench, and Qwen-Image-Bench). At \(1024\times1024\) resolution, our model delivers a \(20\times\) speedup over prior discrete autoregressive models, generating each image in 2.5 seconds on a single NVIDIA H200. Together, these results show that highly compressed discrete representations can enable high-quality autoregressive generation at low latency.

AI 摘要僅根據公開標題與摘要整理;研究結論來自作者,請以原文為準。