PlotGround: Grounding Plot Digitization in Real Scientific Figures and Their Source Data
PlotGround-1k tests plot digitization on 1,119 bioRxiv questions; the best model hits 87.5%.
PlotGround automatically builds plot-digitization benchmarks from real scientific figures and author-released source data. PlotGround-1k contains 1,119 human-verified questions from 1,066 bioRxiv preprints. Across 16 multimodal models, the best reaches 87.5% accuracy at a ±5% relative-error tolerance, while ±2% drops every model by 11–24 points. Supplying source tables instead of figures raises a coding agent's accuracy from 90.0% to 97.4% and cuts cost by 72%.
- Pipeline links real figures to author source tables and grounded questions
- PlotGround-1k has 1,119 questions from 1,066 bioRxiv preprints
- Best of 16 models reaches 87.5% at ±5% relative error
- Source tables raise accuracy from 90.0% to 97.4% and cut cost 72%
- Organizations
- bioRxiv
From the original article
1 min read at arxiv.orgScientific figures often encode quantitative results that are not readily available in machine-readable form, making accurate plot digitization important for verifying and reusing published findings. Yet it remains unclear how accurately current models recover plotted values from real scientific figures, as existing benchmarks rely largely on synthetic charts or cover only a limited range of chart types.
By Yaohui Zhang · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.