ZeroSense:How Vision matters in Long Context Compression

TL;DR

This paper introduces a new evaluation framework and benchmark for visual-text compression methods, revealing that high token compression ratios do not necessarily preserve text quality, and emphasizing the importance of vision in long-context modeling.

Contribution

The work proposes a decoupled evaluation framework and ZeroSense Benchmark to accurately assess visual-text compression quality independently of downstream task performance.

Findings

VTC quality and downstream task accuracy often diverge.

Existing metrics may overestimate VTC effectiveness.

Vision plays a crucial role in long-context compression.

Abstract

Recent visual-text compression (VTC) methods, typified by DeepSeek-OCR, report impressive high token compression ratios for long-context modeling tasks by leveraging text-to-image rendering. However, existing evaluation protocols heavily rely on downstream task performance. Such evaluation metrics fail to accurately measure text preservation due to the strong inherent linguistic priors of Multimodal Large Language Models (MLLMs). In this work, we introduce a new evaluation framework that decouples MLLMs' capabilities to faithfully assess VTC quality. Within this framework, we further introduce the ZeroSense Benchmark to ensure low semantic correlation of testing samples. By eliminating contextual dependencies, our benchmark guarantees that the evaluation results are purely reflective of VTC quality, unaffected by the semantic inference capabilities of downstream models. Extensive experiments across multiple datasets demonstrate that VTC quality and downstream task accuracy diverge significantly, highlighting the necessity of our decoupled evaluation framework.