Enregistré dans:
Détails bibliographiques
Auteurs principaux: Liang, Yunhao, Ying, Ruixuan, Li, Bo, Li, Hong, Yan, Kai, Li, Qingwen, Yang, Min, Satoshi, Okamoto, Cui, Zhe, Ni, Shiwen
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:https://arxiv.org/abs/2601.03714
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866908752422109184
author Liang, Yunhao
Ying, Ruixuan
Li, Bo
Li, Hong
Yan, Kai
Li, Qingwen
Yang, Min
Satoshi, Okamoto
Cui, Zhe
Ni, Shiwen
author_facet Liang, Yunhao
Ying, Ruixuan
Li, Bo
Li, Hong
Yan, Kai
Li, Qingwen
Yang, Min
Satoshi, Okamoto
Cui, Zhe
Ni, Shiwen
contents DeepSeek-OCR utilizes an optical 2D mapping approach to achieve high-ratio vision-text compression, claiming to decode text tokens exceeding ten times the input visual tokens. While this suggests a promising solution for the LLM long-context bottleneck, we investigate a critical question: "Visual merit or linguistic crutch - which drives DeepSeek-OCR's performance?" By employing sentence-level and word-level semantic corruption, we isolate the model's intrinsic OCR capabilities from its language priors. Results demonstrate that without linguistic support, DeepSeek-OCR's performance plummets from approximately 90% to 20%. Comparative benchmarking against 13 baseline models reveals that traditional pipeline OCR methods exhibit significantly higher robustness to such semantic perturbations than end-to-end methods. Furthermore, we find that lower visual token counts correlate with increased reliance on priors, exacerbating hallucination risks. Context stress testing also reveals a total model collapse around 10,000 text tokens, suggesting that current optical compression techniques may paradoxically aggravate the long-context bottleneck. This study empirically defines DeepSeek-OCR's capability boundaries and offers essential insights for future optimizations of the vision-text compression paradigm. We release all data, results and scripts used in this study at https://github.com/dududuck00/DeepSeekOCR.
format Preprint
id arxiv_https___arxiv_org_abs_2601_03714
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Visual Merit or Linguistic Crutch? A Close Look at DeepSeek-OCR
Liang, Yunhao
Ying, Ruixuan
Li, Bo
Li, Hong
Yan, Kai
Li, Qingwen
Yang, Min
Satoshi, Okamoto
Cui, Zhe
Ni, Shiwen
Computation and Language
Computer Vision and Pattern Recognition
DeepSeek-OCR utilizes an optical 2D mapping approach to achieve high-ratio vision-text compression, claiming to decode text tokens exceeding ten times the input visual tokens. While this suggests a promising solution for the LLM long-context bottleneck, we investigate a critical question: "Visual merit or linguistic crutch - which drives DeepSeek-OCR's performance?" By employing sentence-level and word-level semantic corruption, we isolate the model's intrinsic OCR capabilities from its language priors. Results demonstrate that without linguistic support, DeepSeek-OCR's performance plummets from approximately 90% to 20%. Comparative benchmarking against 13 baseline models reveals that traditional pipeline OCR methods exhibit significantly higher robustness to such semantic perturbations than end-to-end methods. Furthermore, we find that lower visual token counts correlate with increased reliance on priors, exacerbating hallucination risks. Context stress testing also reveals a total model collapse around 10,000 text tokens, suggesting that current optical compression techniques may paradoxically aggravate the long-context bottleneck. This study empirically defines DeepSeek-OCR's capability boundaries and offers essential insights for future optimizations of the vision-text compression paradigm. We release all data, results and scripts used in this study at https://github.com/dududuck00/DeepSeekOCR.
title Visual Merit or Linguistic Crutch? A Close Look at DeepSeek-OCR
topic Computation and Language
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2601.03714