EpiCurveBench - Manually Annotated Set: leaderboard
EpiCurveBench (2026) chart digitization on its 100 stylistically diverse epidemic curves that were published without source tables and annotated by hand, scored by EpiCurveSimilarity, an alignment metric that tolerates small time shifts and normalizes value errors by the axis range. General-purpose models run at high reasoning effort; scale 0 to 100, higher is better.
Metric: EpiCurveSimilarity, ECS (%): edit distance with real penalty between each extracted and ground-truth series (gap penalty 1, value differences normalized by the y-axis range and counted fully above 1%), scaled to 0-100 and averaged over matched series per chart (0 for a missed series), then over Set 2 only: the 100 stylistically diverse epidemic curves without source tables, each digitized independently by two annotators with WebPlotDigitizer; the model returns the chart data as a TSV from one generic extraction prompt; the general-purpose VLMs at high reasoning effort, the specialized chart extractors in their single configuration; higher is better. Source: arxiv.org. Saturation forecast: Around August 2027. 6 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 2.5 Pro | High | 56.9 | #280 |
| 2 | GPT-5.2 | High | 50 | #190 |
| 3 | Claude Opus 4.5 | High | 41.4 | #165 |
| 4 | Qwen 3 VL 235B A22B Instruct | 34.7 | #541 | |
| 5 | OneChart | 18.5 | ||
| 6 | TinyChart | 18.4 | #9894 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=epicurvebench-manually-annotated-set · How It Works · Data refreshed daily, snapshot 2026-11-22.