Unified Models Generate Intermediate Images

{
  "author": null,
  "context": "Paradigm where VLMs explicitly generate pixel-level intermediate images for visual reasoning, incurring high computational overhead.",
  "category": "ai",
  "enrichment": {
    "is_stale": true
  },
  "norm_label": "Unified Models Generate Intermediate Images",
  "source_url": null,
  "graphify_id": "unified_models_approach",
  "source_file": "guo-atlas-2026-fulltext.md",
  "imported_from": "/Users/antonborzov/Documents/Research.nosync/papers/extract_typed_out/guo-atlas-2026-fulltext/graph.json",
  "extracted_type": "concept",
  "source_location": "Fig. 1.I, Introduction",
  "graphify_file_type": "concept"
}