TARA: Task-Adaptive Rank Allocation for Efficient Large Language Model Fine-Tuning in Geo-Information Text Classification

Wang, Canhui; Shen, Juntao; Feng, Yicong; Huang, Jin; Jing, Yanwu; Chen, Weiwei; Zhang, Wanqiang; Wang, Min

doi:10.3390/ijgi15080372

This is an early access version, the complete PDF, HTML, and XML versions will be available soon.

Open AccessArticle

TARA: Task-Adaptive Rank Allocation for Efficient Large Language Model Fine-Tuning in Geo-Information Text Classification

by

Canhui Wang

¹,

Juntao Shen

²,

Yicong Feng

³,

Jin Huang

²,

Yanwu Jing

¹,

Weiwei Chen

¹,

Wanqiang Zhang

⁴ and

Min Wang

^5,*

¹

Surveying and Mapping Geographic Information Center, Sichuan Institute of Geological Survey, Chengdu 610072, China

²

School of Electrical Engineering, Southwest Jiaotong University, Chengdu 611756, China

³

Information Centre of Natural Resources Department of Sichuan Province, Chengdu 610072, China

⁴

Sichuan DiTing AI Technology Co., Ltd., Chengdu 610041, China

⁵

College of Artificial Intelligence and Electronic Engineering, Sichuan Technology and Business University, Chengdu 611745, China

^*

Author to whom correspondence should be addressed.

ISPRS Int. J. Geo-Inf. 2026, 15(8), 372; https://doi.org/10.3390/ijgi15080372

Submission received: 14 May 2026 / Revised: 24 July 2026 / Accepted: 7 August 2026 / Published: 18 August 2026

Download Versions Notes

Abstract

Geo-information texts, including geospatial data-use regulations and Earth observation metadata, are central to data governance and compliance auditing in remote sensing ecosystems. Full fine-tuning of large pre-trained language models is often computationally impractical, while standard LoRA reduces cost but assigns a fixed rank to all adapted modules, ignoring differences across layers and projection types. This paper proposes TARA, a task-adaptive rank allocation method for LoRA-based fine-tuning. TARA assigns learnable importance scores to rank dimensions and uses Gumbel–Sigmoid sampling with the Straight-Through Estimator to learn discrete rank masks under a global sparsity constraint. We further construct RSRegulation, a geospatial regulatory compliance benchmark containing 4032 English-language samples derived from 168 clauses across seven regulatory and policy sources with clause-level data isolation. Across five random seeds, TARA achieves 95.30 ± 0.10% accuracy and 95.44 ± 0.10% F1 with a maximum trainable adapter budget of 1.57 M parameters. The learned soft allocation corresponds to approximately 0.38 M effective adapter parameters and 75.8% soft rank compression. Physical hard pruning reduces the deployed adapter to 0.086 M parameters while retaining 95.12 ± 0.11% accuracy and 95.21 ± 0.10% F1. Layer-wise analysis shows that value projections retain higher ranks than query projections under the current task and backbone, revealing a task-dependent non-uniform allocation pattern.

Keywords: geo-information text classification; pre-trained language models; parameter-efficient fine-tuning; adaptive rank allocation; geospatial regulatory compliance; LoRA

Share and Cite

MDPI and ACS Style

Wang, C.; Shen, J.; Feng, Y.; Huang, J.; Jing, Y.; Chen, W.; Zhang, W.; Wang, M. TARA: Task-Adaptive Rank Allocation for Efficient Large Language Model Fine-Tuning in Geo-Information Text Classification. ISPRS Int. J. Geo-Inf. 2026, 15, 372. https://doi.org/10.3390/ijgi15080372

AMA Style

Wang C, Shen J, Feng Y, Huang J, Jing Y, Chen W, Zhang W, Wang M. TARA: Task-Adaptive Rank Allocation for Efficient Large Language Model Fine-Tuning in Geo-Information Text Classification. ISPRS International Journal of Geo-Information. 2026; 15(8):372. https://doi.org/10.3390/ijgi15080372

Chicago/Turabian Style

Wang, Canhui, Juntao Shen, Yicong Feng, Jin Huang, Yanwu Jing, Weiwei Chen, Wanqiang Zhang, and Min Wang. 2026. "TARA: Task-Adaptive Rank Allocation for Efficient Large Language Model Fine-Tuning in Geo-Information Text Classification" ISPRS International Journal of Geo-Information 15, no. 8: 372. https://doi.org/10.3390/ijgi15080372

APA Style

Wang, C., Shen, J., Feng, Y., Huang, J., Jing, Y., Chen, W., Zhang, W., & Wang, M. (2026). TARA: Task-Adaptive Rank Allocation for Efficient Large Language Model Fine-Tuning in Geo-Information Text Classification. ISPRS International Journal of Geo-Information, 15(8), 372. https://doi.org/10.3390/ijgi15080372

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Article metric data becomes available approximately 24 hours after publication online.

Article Menu

TARA: Task-Adaptive Rank Allocation for Efficient Large Language Model Fine-Tuning in Geo-Information Text Classification

Abstract

Share and Cite

Article Metrics

Further Information

Guidelines

MDPI Initiatives

Follow MDPI