mlsanityData to 7 Oct 2026

Text or Pixels? It Takes Half: On the Token Efficiency of Visual Text Inputs in Multimodal LLMs

arXiv 2510.18279 ↗First posted 21 Oct 2025cs.CL cs.AIcited by 8

On the map: Multimodal LLMs / vision-language models → Visual token pruning and efficiency

Abstract (arXiv)

Large language models (LLMs) and their multimodal variants can now process visual inputs, including images of text. This raises an intriguing question: can we compress textual inputs by feeding them as images to reduce token usage while preserving performance? In this paper, we show that visual text representations are a practical and surprisingly effective form of input compression for decoder LLMs. We exploit the idea of rendering long text inputs as a single image and provide it directly to the model. This leads to dramatically reduced number of decoder tokens required, offering a new form of input compression. Through experiments on two distinct benchmarks RULER (long-context retrieval) and CNN/DailyMail (document summarization) we demonstrate that this text-as-image method yields substantial token savings (often nearly half) without degrading task performance.

Key facts. Cited by 8 papers in the map (data through 7 Oct 2026). Placed in Visual token pruning and efficiency, one of 16 directions in Multimodal LLMs / vision-language models. Builds on 20 of its 35 references that are in the map. First posted on arXiv 21 Oct 2025.

Builds on

20 of 35 references.

References · 20

  1. 2017Attention Is All You Needcited by 59,932
  2. 2019BERTScore: Evaluating Text Generation with BERTcited by 5,462
  3. 2022PaLM: Scaling Language Modeling with Pathwayscited by 5,304
  4. 2024Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolutioncited by 4,151
  5. 2025Gemini 2.5: Pushing the Frontier with Advanced Reasoning, Multimodality, Long Context, and Next Generation Agentic Capabilitiescited by 3,655
  6. 2023Lost in the Middle: How Language Models Use Long Contextscited by 3,383
  7. 2024RULER: What's the Real Context Size of Your Long-Context Language Models?cited by 1,270
  8. 2023LongLLMLingua: Accelerating and Enhancing LLMs in Long Context Scenarios via Prompt Compressioncited by 443
  9. 2024LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compressioncited by 343
  10. 2024DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Headscited by 283
  11. 2024BABILong: Testing the Limits of LLMs with Long Context Reasoning-in-a-Haystackcited by 211
  12. 2024xRAG: Extreme Context Compression for Retrieval-augmented Generation with One Tokencited by 123
8 more
  1. 2024Retrieval Augmented Generation or Long-Context LLMs? A Comprehensive Study and Hybrid Approachcited by 117
  2. 2023SLM: Bridge the thin gap between speech and text foundation modelscited by 67
  3. 2024Characterizing Prompt Compression Methods for Long Context Inferencecited by 18
  4. 2024Enhancing Vision-Language Model Reliability with Uncertainty-Guided Dropout Decodingcited by 14
  5. 2024RefreshKV: Updating Small KV Cache During Long-form Generationcited by 11
  6. 2024From Text to Pixel: Advancing Long-Context Understanding in MLLMscited by 9
  7. 2024Enhancing and Accelerating Large Language Models via Instruction-Aware Contextual Compressioncited by 4
  8. 2025Context-Efficient Retrieval with Factual Decompositioncited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 7.

8 papers.

  1. 2025VTCBench: Can Vision-Language Models Understand Long Context with Vision-Text Compression?cited by 11
  2. 2026VISTA-Bench: Do Vision-Language Models Really Understand Visualized Text as Well as Pure Text?cited by 8
  3. 2026When Text-as-Vision Meets Semantic IDs in Generative Recommendation: An Empirical Studycited by 3
  4. 2026Same Semantics, Different Paths: Self-Improving Alignment for Vision-Text Compressioncited by 2
  5. 2026Global Context Compression with Interleaved Vision-Text Transformationcited by 1
  6. 2026Token-Efficient Multimodal Reasoning via Image Prompt Packagingcited by 0
  7. 2026LoMo: Local Modality Substitution for Deeper Vision-Language Fusioncited by 0
  8. 2026Dual-Latent Memory Routing for Vision-Language Reasoningcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024FocusLLaVA: A Coarse-to-Fine Approach for Efficient and Effective Visual Token Compression
  2. 2025A Survey of Token Compression for Efficient Multimodal Large Language Models
  3. 2025VisionSelector: End-to-End Learnable Visual Token Compression for Efficient Multimodal LLMs
  4. 2025Can Visual Input Be Compressed? A Visual Token Compression Benchmark for Large Multimodal Models
  5. 2026Vision Token Reduction via Attention-Driven Self-Compression for Efficient Multimodal Large Language Models
  6. 2025A Comprehensive Study of Decoder-Only LLMs for Text-to-Image Generation
  7. 2025Token Sequence Compression for Efficient Multimodal Computing
  8. 2025TokenCarve: Information-Preserving Visual Token Compression in Multimodal Large Language Models