An agentic multi-stage pipeline produces Colon-Bench—528 densely annotated colonoscopy windows spanning 14 lesions, 300k boxes, 213k masks and clinical text—then benchmarks MLLMs and a colon-skill prompt that gains up to 9.7% VQA.
In: MultiMedia Modeling
1 Pith paper cite this work, alongside 1 external citations. Polarity classification is still indexing.
1
Pith paper citing it
1
external citations · OpenAlex
fields
eess.IV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Colon-Bench: An Agentic Workflow for Scalable Dense Lesion Annotation in Full-Procedure Colonoscopy Videos
An agentic multi-stage pipeline produces Colon-Bench—528 densely annotated colonoscopy windows spanning 14 lesions, 300k boxes, 213k masks and clinical text—then benchmarks MLLMs and a colon-skill prompt that gains up to 9.7% VQA.