Arctic-TILT. Business Document Understanding at Sub-Billion Scale

TL;DR

Arctic-TILT is a compact, cost-efficient model that achieves state-of-the-art accuracy in business document understanding tasks involving large, visually rich PDFs, suitable for large-scale enterprise deployment.

Contribution

The paper introduces Arctic-TILT, a small yet powerful model capable of matching larger models' accuracy on document understanding tasks, with efficient fine-tuning and deployment.

Findings

Achieves accuracy comparable to models 1000x larger.

Establishes state-of-the-art results on seven benchmarks.

Supports quick inference and reliable confidence scoring.

Abstract

The vast portion of workloads employing LLMs involves answering questions grounded on PDF or scan content. We introduce the Arctic-TILT achieving accuracy on par with models 1000$\times$ its size on these use cases. It can be fine-tuned and deployed on a single 24GB GPU, lowering operational costs while processing Visually Rich Documents with up to 400k tokens. The model establishes state-of-the-art results on seven diverse Document Understanding benchmarks, as well as provides reliable confidence scores and quick inference, which are essential for processing files in large-scale or time-sensitive enterprise environments.