{
  "$schema": "https://themachinepress.com/schemas/story-v1.schema.json",
  "schema_version": "1.0.0",
  "document_type": "machine_press_story",
  "story": {
    "story_id": "mp-2026-10-20-005",
    "source_story_id": "tmp-story-openvlharness-multimodal-memory",
    "edition_id": "mp-2026-10-20-morning-0102",
    "edition_url": "https://themachinepress.com/edition/2026-10-20",
    "position": 5,
    "story_type": "dispatch",
    "section": "developer-tools",
    "editorial_classification": "editorial",
    "headline": "The Visual Agent Keeps the Evidence It Found",
    "slug": "the-visual-agent-keeps-the-evidence-it-found",
    "dek": "OpenVLHarness turns tool output into addressable images, data and web references that persist across a multimodal reasoning trace.",
    "summary": "OpenVLHarness turns tool output into addressable images, data and web references that persist across a multimodal reasoning trace.",
    "body_text": "More inference did little on 22 visual benchmarks when the missing ingredient was external evidence. OpenVLHarness adds a capability layer that renders heterogeneous tool results into model-readable observations and persistent memory that keeps images, data and references addressable. Across eight evaluated backbones, the authors report average gains of 9.7 to 12.6 points over matched no-tool baselines. In one comparison, a medium-effort smaller model beat maximum-effort systems while using no more than 37 percent of their model-token cost. These are harness-specific results, but they show why tool access and evidence retention must be evaluated together.",
    "why_it_matters": "OpenVLHarness turns tool output into addressable images, data and web references that persist across a multimodal reasoning trace.",
    "limitations": [],
    "importance": 8,
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-20-005/the-visual-agent-keeps-the-evidence-it-found",
    "json_url": "https://themachinepress.com/story/mp-2026-10-20-005.json",
    "first_published_at": "2026-10-20T09:00:00.000-04:00",
    "modified_at": "2026-10-20T09:00:00.000-04:00",
    "content_status": "new",
    "is_carryover": false,
    "carryover_reason": null,
    "key_claims": [
      {
        "claim_id": "claim-mp-2026-10-20-005-001",
        "text": "OpenVLHarness turns tool output into addressable images, data and web references that persist across a multimodal reasoning trace.",
        "source_ids": [
          "source-2026-10-20-005"
        ],
        "qualification": null
      }
    ],
    "source_ids": [
      "source-2026-10-20-005"
    ],
    "tags": [
      "visual agents",
      "tool use",
      "multimodal memory"
    ],
    "image_url": null,
    "corrections": []
  },
  "sources": [
    {
      "source_id": "source-2026-10-20-005",
      "title": "arXiv preprint 2610.26860, original v1",
      "publisher": "arXiv",
      "url": "https://arxiv.org/abs/2610.26860v1",
      "canonical_url": "https://arxiv.org/abs/2610.26860v1",
      "source_type": "primary_research",
      "is_primary_source": true,
      "published_at": "2026-10-19T13:59:58.000-04:00",
      "accessed_at": "2026-10-20T08:15:06.176-04:00",
      "supports_claim_ids": [
        "claim-mp-2026-10-20-005-001"
      ]
    }
  ],
  "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 Visual Agent Keeps the Evidence It Found",
    "publisher": "The Machine Press",
    "published_at": "2026-10-20T09:00:00.000-04:00",
    "canonical_url": "https://themachinepress.com/story/mp-2026-10-20-005/the-visual-agent-keeps-the-evidence-it-found"
  }
}