{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-11-013",
    "source_story_id": "tmp-story-risebench-visual-editing",
    "edition_id": "mp-2026-10-11-morning-0093",
    "edition_url": "https://themachinepress.com/edition/2026-10-11",
    "position": 13,
    "story_type": "dispatch",
    "section": "benchmarks-evals",
    "editorial_classification": "editorial",
    "headline": "The Best Image Editor Still Missed Nearly Half the Reasoning Tests",
    "slug": "the-best-image-editor-still-missed-nearly-half-the-reasoning-tests",
    "dek": "RISEBench++ tests temporal, causal, spatial, logical and counterfactual editing across 1,000 bilingual cases.",
    "summary": "RISEBench++ tests temporal, causal, spatial, logical and counterfactual editing across 1,000 bilingual cases.",
    "body_text": "RISEBench++ organizes reasoning-informed visual editing into six dimensions, 12 subcategories and 65 task types, including multi-image and multi-turn inputs. Its 1,000 human-annotated cases are released in English and Chinese and scored for instruction reasoning, appearance consistency and visual plausibility using human review and calibrated model judging. Across 58 systems—34 open models, 19 closed models and five agentic methods—the strongest reported result was 56.6% accuracy. A training-free planning and verification agent improved on many baselines. The benchmark exposes current limitations; its model-judge component and task taxonomy still shape what counts as success.",
    "why_it_matters": "RISEBench++ tests temporal, causal, spatial, logical and counterfactual editing across 1,000 bilingual cases.",
    "limitations": [],
    "importance": 9,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-11-013/the-best-image-editor-still-missed-nearly-half-the-reasoning-tests",
    "json_url": "https://themachinepress.com/story/mp-2026-10-11-013.json",
    "first_published_at": "2026-10-11T09:00:00.000-04:00",
    "modified_at": "2026-10-11T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-11-013-001",
        "text": "RISEBench++ tests temporal, causal, spatial, logical and counterfactual editing across 1,000 bilingual cases.",
        "source_ids": [
          "source-2026-10-11-013"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-10-11-013"
    ],
    "tags": [
      "image editing",
      "reasoning benchmark",
      "multimodal models"
    ],
    "image_url": "https://themachinepress.com/issues/2026-10-11/risebench-wireframe.webp",
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-11-013",
      "title": "arXiv preprint 2610.12343, original v1",
      "publisher": "arXiv / submitting research authors",
      "url": "https://arxiv.org/abs/2610.12343v1",
      "canonical_url": "https://arxiv.org/abs/2610.12343v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-08T13:15:58.000-04:00",
      "accessed_at": "2026-10-11T08:21:27.764-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-11-013-001"
      ]
    }
  ],
  "corrections": [],
  "publisher": {
    "name": "The Machine Press",
    "url": "https://themachinepress.com",
    "description": "A daily newspaper for the age of artificial intelligence."
  },
  "cite_this_report": {
    "title": "The Best Image Editor Still Missed Nearly Half the Reasoning Tests",
    "publisher": "The Machine Press",
    "published_at": "2026-10-11T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-11-013/the-best-image-editor-still-missed-nearly-half-the-reasoning-tests"
  }
}