RAGOCR: Optical Compression of Retrieval-Augmented Text via Visual Representation

2026-08-01

AI Summary

This work addresses the high computational overhead in retrieval-augmented generation (RAG) caused by long input contexts by proposing a query-aware visual compression method that requires no fine-tuning of the generator. The approach converts retrieved text into compact visual representations and employs a dynamic resolution mechanism that adaptively adjusts visual granularity based on document relevance and complexity, thereby balancing efficient compression with preservation of critical information. Integrated with query-conditioned rendering and an offline-online collaborative compression architecture, the method achieves over a 15% absolute accuracy improvement on five question-answering benchmarks using only one-eighth of the input tokens required by standard RAG on the MedOmniKB corpus, consistently outperforming existing hard and soft compression baselines across varying retrieval depths.

Abstract

Retrieval-Augmented Generation (RAG) has become essential for knowledge-intensive question answering, yet scaling RAG pipelines remains challenging due to the prohibitive computational cost of processing lengthy retrieved contexts. Existing compression approaches face a fundamental trade-off: hard compression methods operate online in a query-aware fashion but achieve only modest compression rates and typically require fine-tuning the generative model, while soft compression methods attain higher ratios but rely on costly offline encoding that is entirely agnostic to the input query. To bridge this gap, we introduce RAGOCR, a novel framework that compresses retrieved documents into compact visual representations conditioned on the input query. To further balance compression rate and information fidelity, we introduce a query-aware dynamic resolution mechanism that adaptively allocates visual granularity based on each document's estimated relevance and complexity: highly relevant passages are rendered at higher resolution to preserve fine-grained details, while peripheral documents are aggressively compressed at lower resolution. Experiments on five QA benchmarks using the MedOmniKB retrieval corpus demonstrate that RAGOCR surpasses naive RAG by over 15\% in accuracy while requiring only one-eighth the number of input tokens, and consistently outperforms both hard and soft compression baselines across varying retrieval depths.

Problem

• Retrieval-Augmented Generation
• text compression
• computational cost
• query-awareness
• information fidelity

Innovation

• Retrieval-Augmented Generation
• visual compression
• query-aware resolution
• dynamic rendering
• token efficiency