COCOcap

COCOcap is a benchmark for handling several input types together (images, documents, text). The official description says: "COCO Captioning - image captioning on MS-COCO dataset". Below we describe in simple words what benchmarks of this type generally examine.

COCOcap measures handling several input types together (images, documents, text). In detail, it checks:

• questions that mix text with images or document pages
• reading and extracting information from screenshots and scans
• understanding layout, titles, logos and figures in real documents
• staying accurate when the text content alone is not enough

The score is a composite index published by the benchmark organiser rather than a plain percentage. Interpret it only against the other models in the table — higher is better.

Official methodology note: COCO Captioning - image captioning on MS-COCO dataset.

What is the COCOcap benchmark?

COCOcap is a benchmark for handling several input types together (images, documents, text). The official description says: 'COCO Captioning - image captioning on MS-COCO dataset'. Below we describe in simple words what benchmarks of this type generally examine.

What does COCOcap test?

COCOcap mainly measures handling several input types together (images, documents, text). Concretely, it checks: questions that mix text with images or document pages; reading and extracting information from screenshots and scans; understanding layout, titles, logos and figures in real documents.

How is the COCOcap score calculated?

The score is a composite index published by the benchmark organiser rather than a plain percentage. Interpret it only against the other models in the table — higher is better. The declared maximum score of this benchmark is 150.0.

Where do I get COCOcap data and scores?

The official source is https://cocodataset.org/#home. The official benchmark page/reference is linked on this page. Every score on Open LLM Radar is also available as open data: download the scores JSON (http://open-llm-radar-api:8000/api/v1/benchmarks/cococap/scores) or pull it directly from the Open LLM Radar API with any HTTP client.