DocSLM: A Small Vision-Language Model for Long Multimodal Document Understanding

MCML Authors

Tanveer Hannan

→ Group Thomas Seidl
 Database Systems, Data Mining and AI

Thomas Seidl

Prof. Dr.

Core PI

Database Systems, Data Mining and AI

Abstract

Large Vision-Language Models (LVLMs) have demonstrated strong multimodal reasoning capabilities on long and complex documents. However, their high memory footprint makes them impractical for deployment on resourceconstrained edge devices. We present DocSLM, an efficient Small Vision-Language Model designed for long document understanding under constrained memory resources. DocSLM incorporates a Hierarchical Multimodal Compressor that jointly encodes visual, textual, and layout information from each page into a fixed-length sequence, greatly reducing memory consumption while preserving both local and global semantics. To enable scalable processing over arbitrarily long inputs, we further introduce a Streaming Abstention mechanism that operates on document segments sequentially and filters low-confidence responses through an entropy-based uncertainty calibrator. Across multiple long multimodal document benchmarks, DocSLM matches or surpasses state-of-the-art methods while using 82% fewer visual tokens, 75% fewer parameters, and 71% lower latency-delivering reliable multimodal document understanding on lightweight edge devices. Code is available in the supplementary.

inproceedings HMP+26

---

Findings @CVPR 2026

Findings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Denver, CO, USA, Jun 03-07, 2026.

---

Authors

T. Hannan • D. Mallios • P. Pathak • F. Sardari • T. Seidl • G. Bertasius • M. Fayyaz • S. Sengupta

---

Links

DOI

---

In Collaboration

 Meta
 Microsoft

---

Research Area

 A3 | Computational Models