{
  "schema_version": "1.1",
  "id": "archive:https://arxiv.org/abs/2611.14469v1",
  "slug": "2611-14469v1-1um73ov",
  "url": "https://feed7.dev/p/2611-14469v1-1um73ov",
  "title": "C$^2$-Routing: Complementary Cropping and Compression for Efficient Vision-Language Inference",
  "why_included": "C²-Routing adaptively crops or compresses images before vision-language inference, with a fallback gate for risky reductions. Its reported savings are measured against already token-capped inputs across three models and 18 datasets.",
  "summary": "C²-Routing chooses between cropping for local detail and compression for broader context, with a safety gate that can fall back to the original input. Across **three VLMs** and **18 datasets**, it reduced FLOPs by 36.53% and visual tokens by 42.36% on average, with a 0.77% mean relative task-score decrease.",
  "practical_implication": "For agents that inspect screenshots, documents, or other images, adaptive preprocessing could reduce inference work while preserving relevant detail. The safety gate is part of the approach, so a simple fixed token cap may not capture the same tradeoff.",
  "agent_context": "C²-Routing chooses between cropping for local detail and compression for broader context, with a safety gate that can fall back to the original input. Across **three VLMs** and **18 datasets**, it reduced FLOPs by 36.53% and visual tokens by 42.36% on average, with a 0.77% mean relative task-score decrease.\n\nFor agents that inspect screenshots, documents, or other images, adaptive preprocessing could reduce inference work while preserving relevant detail. The safety gate is part of the approach, so a simple fixed token cap may not capture the same tradeoff.\n\nThe comparisons used a full-image baseline capped at 1,024 visual tokens, and experiments ran on RTX 3090 GPUs. The material does not establish how the method performs on other workloads or hardware.",
  "source": {
    "name": "arXiv",
    "url": "https://arxiv.org/abs/2611.14469v1",
    "published_at": "2026-11-10T18:32:28.000Z"
  },
  "source_class": "blog_post",
  "content_type": "Paper",
  "layer": "context",
  "domains": [
    "image"
  ],
  "topics": [
    "context-engineering",
    "model-selection"
  ],
  "verification": {
    "status": "needs_review",
    "label": "Needs Review",
    "method": "unverified",
    "verified_at": null
  },
  "uncertainty": [
    "The comparisons used a full-image baseline capped at 1,024 visual tokens, and experiments ran on RTX 3090 GPUs. The material does not establish how the method performs on other workloads or hardware."
  ],
  "connected_context": {
    "meaning": "The results make adaptive image preprocessing a measurable option for reducing visual inference work, with the reported average task-score decrease and a fallback gate defining the tradeoff. They support testing the approach on screenshot or document workflows, but do not establish transfer beyond the studied models, datasets, token cap, and hardware.",
    "corpus_size": 1059,
    "generated_at": "2026-11-11T11:09:47.282Z",
    "connections": [
      {
        "title": "ReToken: One Token to Improve Vision-Language Models for Visual Retrieval",
        "source_name": "arXiv",
        "source_url": "https://arxiv.org/abs/2607.28627v1",
        "feed7_url": "https://feed7.dev/p/2607-28627v1-0pc8021",
        "reason": "C²-Routing reduces input cost by choosing cropping or compression before inference; ReToken selects relevant visual tokens from cached context. Both limit visual evidence processed, but at different stages and with different mechanisms."
      }
    ]
  },
  "lifecycle": "Current",
  "published_at": "2026-11-10T18:32:28.000Z",
  "modified_at": "2026-11-10T18:32:28.000Z",
  "supersedes": [],
  "expires_at": null,
  "formats": {
    "html": "https://feed7.dev/p/2611-14469v1-1um73ov",
    "json": "https://feed7.dev/p/2611-14469v1-1um73ov.json",
    "markdown": "https://feed7.dev/p/2611-14469v1-1um73ov.md"
  }
}