FTibSuite: A Comprehensive Resource Suite for Tibetan Vision–Language Modeling

Guixian Xu 1,2 ∗ \ast Yide Liang 2 ∗ \ast Zeli Su 2 ∗ ⁣ † \ast\dagger Xuexian Song 4 ∗ \ast Ziyin Zhang 3 ∗ ⁣ † \ast\dagger Yushuang Dong 2 Ting Zhang 2 Xu Han 2 1 Hainan International College, Minzu University of China 2 School of Information Engineering, Minzu University of China 3 Shanghai Jiao Tong University 4 Institute of Automation, Chinese Academy of Sciences {guixian_xu,24302278,zeli_su,yushuangdong,hanxu}@muc.edu.cn [email protected] [email protected] [email protected] · 2026

Read FTibSuite: A Comprehensive Resource Suite for Tibetan Vision–Language Modeling by Guixian Xu 1,2 ∗ \ast Yide Liang 2 ∗ \ast Zeli Su 2 ∗ ⁣ † \ast\dagger Xuexian Song 4 ∗ \ast Ziyin Zhang 3 ∗ ⁣ † \ast\dagger Yushuang Dong 2 Ting Zhang 2 Xu Han 2 1 Hainan International College, Minzu University of China 2 School of Information Engineering, Minzu University of China 3 Shanghai Jiao Tong University 4 Institute of Automation, Chinese Academy of Sciences {guixian_xu,24302278,zeli_su,yushuangdong,hanxu}@muc.edu.cn [email protected] [email protected] [email protected] (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-26601.txt.

Description

Guixian Xu and seven colleagues build FTibSuite, the first standard suite for Tibetan vision-language work. Tibetan stays underserved for lack of training and evaluation setup. FTibSuite bundles human-verified multimodal training corpora as FTibData, covering instruction tuning, image-text alignment, and continual pretraining. FTibBench adapts five mainstream multimodal benchmarks to Tibetan with layered quality control cutting translation noise. FTibVLM, a baseline using Qwen3-VL-8B-Instruct with a three-stage adaptation pipeline, lifts MMBench accuracy to 67.78 from 42.97, and POPE-random accuracy to 80.56 from 47.53, while keeping Chinese abilities nearly intact.

Catalog record

CREATORGuixian Xu 1,2 ∗ \ast Yide Liang 2 ∗ \ast Zeli Su 2 ∗ ⁣ † \ast\dagger Xuexian Song 4 ∗ \ast Ziyin Zhang 3 ∗ ⁣ † \ast\dagger Yushuang Dong 2 Ting Zhang 2 Xu Han 2 1 Hainan International College, Minzu University of China 2 School of Information Engineering, Minzu University of China 3 Shanghai Jiao Tong University 4 Institute of Automation, Chinese Academy of Sciences {guixian_xu,24302278,zeli_su,yushuangdong,hanxu}@muc.edu.cn [email protected] [email protected] [email protected]
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE54253
LICENSEhttp://creativecommons.org/licenses/by-sa/4.0/
RIGHTSCC BY SA
SOURCEarXiv

Source: https://arxiv.org/abs/2605.26601

Browse the full library · About the Tower of Babel