{"slug": "lensvlm-selective-context-expansion-for-compressed-visual-representation-oftext", "title": "LensVLM: Selective Context Expansion for Compressed Visual Representation OfText", "summary": "Researchers submitted LensVLM, an inference framework and post-training recipe built on Qwen3.5-9B-Base, to arXiv on 7 May 2026, enabling vision language models to scan compressed rendered text and selectively expand only relevant images to uncompressed form via learned tools. LensVLM maintains accuracy comparable to the full-text upper bound at 4.3x effective compression and outperforms retrieval-based, text- and visual-compression baselines up to 10.1x effective compression across seven text QA benchmarks, with gains growing as compression increases on multimodal document and code tasks. The analysis found training makes visual compression robust to rendering choices and that text expansion suits rendered text while high-resolution image expansion suits native documents whose layout cues carry task-relevant information.", "body_md": "# Computer Science > Computer Vision and Pattern Recognition\n\n  [Submitted on 7 May 2026]\n\n# Title:LensVLM: Selective Context Expansion for Compressed Visual Representation of Text\n\n[View PDF](https://arxiv.org/pdf/2605.07019)\n\n[HTML (experimental)](https://arxiv.org/html/2605.07019v1)\n\nAbstract:Vision Language Models (VLMs) offer the exciting possibility of processing text as rendered images, bypassing the need for tokenizing the text into long token sequences. Since VLM image encoders map fixed-size images to a fixed number of visual tokens, varying rendering resolution provides a fine-grained compression knob. However, accuracy deteriorates quickly as compression increases: characters shrink below the vision encoder's effective resolution, making them indistinguishable. To address this, we propose LensVLM, an inference framework and post-training recipe that enables VLMs to scan compressed images, then selectively expand only the relevant images to their uncompressed form via learned tools. Building on Qwen3.5-9B-Base, LensVLM maintains accuracy comparable to the full-text upper bound at 4.3x effective compression and outperforms retrieval-based, text- and visual-compression baselines up to 10.1x effective compression across seven text QA benchmarks. LensVLM also generalizes to multimodal document and code understanding tasks, with the accuracy gain over baselines growing as compression increases. Our analysis validates this approach: training makes visual compression robust to rendering choices, and as compression grows the model increasingly relies on expanded content rather than unreliable visual reading. The analysis also yields practical tool-choice guidance: text expansion is preferable for rendered text, while high-resolution image expansion suits native documents whose layout cues carry task-relevant information.\n    \n\n### References & Citations\n\nLoading...\n\n# Bibliographic and Citation Tools\n\nBibliographic Explorer \n\n*(*[What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))\nConnected Papers \n\n*(*[What is Connected Papers?](https://www.connectedpapers.com/about))\nLitmaps \n\n*(*[What is Litmaps?](https://www.litmaps.co/))\nscite Smart Citations \n\n*(*[What are Smart Citations?](https://www.scite.ai/))\n# Code, Data and Media Associated with this Article\n\nalphaXiv \n\n*(*[What is alphaXiv?](https://alphaxiv.org/))\nCatalyzeX Code Finder for Papers \n\n*(*[What is CatalyzeX?](https://www.catalyzex.com))\nDagsHub \n\n*(*[What is DagsHub?](https://dagshub.com/))\nGotit.pub \n\n*(*[What is GotitPub?](http://gotit.pub/faq))\nHugging Face \n\n*(*[What is Huggingface?](https://huggingface.co/huggingface))\nScienceCast \n\n*(*[What is ScienceCast?](https://sciencecast.org/welcome))\n# Demos\n\n# Recommenders and Search Tools\n\nInfluence Flower \n\n*(*[What are Influence Flowers?](https://influencemap.cmlab.dev/))\nCORE Recommender \n\n*(*[What is CORE?](https://core.ac.uk/services/recommender))\n# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.\n\nHave an idea for a project that will add value for arXiv's community? [**Learn more about arXivLabs**](https://info.arxiv.org/labs/index.html).", "url": "https://wpnews.pro/news/lensvlm-selective-context-expansion-for-compressed-visual-representation-oftext", "canonical_source": "https://arxiv.org/abs/2605.07019", "published_at": "2026-09-24 05:18:05+00:00", "updated_at": "2026-09-24 05:31:23.758867+00:00", "lang": "en", "topics": ["computer-vision", "natural-language-processing", "large-language-models", "ai-research", "machine-learning"], "entities": ["LensVLM", "Qwen3.5-9B-Base", "arXiv"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/lensvlm-selective-context-expansion-for-compressed-visual-representation-oftext", "markdown": "https://wpnews.pro/news/lensvlm-selective-context-expansion-for-compressed-visual-representation-oftext.md", "text": "https://wpnews.pro/news/lensvlm-selective-context-expansion-for-compressed-visual-representation-oftext.txt", "jsonld": "https://wpnews.pro/news/lensvlm-selective-context-expansion-for-compressed-visual-representation-oftext.jsonld"}}