New benchmark targets scientific figure plagiarism detection

Researchers have introduced SciFigPlag-Bench, a new benchmark designed to detect plagiarism in scientific figures. This benchmark goes beyond simple image similarity by evaluating the provenance of reused content, including how it has been transformed and where it appears in scholarly documents. SciFigPlag-Bench supports tasks such as pairwise detection, source attribution, and classification of reuse types, using a dataset of over 5,000 pairs of figures. Initial experiments with vision-language models highlight ongoing challenges in fine-grained provenance reasoning and spatial evidence grounding. AI

IMPACT This benchmark could improve the detection of academic misconduct by enabling more sophisticated analysis of visual content reuse in research papers.

RANK_REASON The cluster describes a new academic benchmark and associated paper published on arXiv. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• Litmaps
• ScienceCast
• SciFigPlag-Bench
• scite Smart Citations

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →