arXiv:2605.27195v1 Announce Type: new 
Abstract: Chart-to-data extraction with vision-language models (VLMs) is increasingly evaluated on benchmarks that show diminishing headroom (frontier VLMs exceed 89% on ChartQA) and with metrics that treat extracted points as unordered key-value pairs, ignoring the temporal structure of time series and penalizing small alignment shifts as catastrophic failures. We address both gaps with EpiCurveBench, a benchmark of 1,000 real-world epidemic curve images curated from diverse public-health sources, and EpiCurveSimilarity (ECS), an evaluation metric that aligns predicted and ground-truth series via dynamic programming, tolerating local temporal shifts and gaps while penalizing them proportionally. Evaluating six methods--three frontier closed VLMs, one open VLM, and two specialized chart-extraction systems--we find the strongest model reaches only 52.3% ECS, and that ECS spreads the four general-purpose VLMs over a 25-point range where key-value metrics (RMS, SCRM) compress them into a 5-point band. We further validate ECS against four downstream epidemiological summary statistics, finding that higher ECS predicts smaller errors in total counts, peak timing, and peak magnitude, and higher growth-rate fidelity; across all four, ECS correlates 1.5--3.6 times more strongly than Dynamic Time Warping, which lacks a gap penalty and therefore cannot distinguish a truncated prediction from a temporally faithful one. EpiCurveBench targets a high-impact public-health application--unlocking decades of outbreak data trapped in published figures--but the benchmark and metric apply directly to any structured time-series chart-extraction setting.

يمثل إدخال EpiCurveBench تقدمًا كبيرًا في تقييم نماذج الرؤية-اللغة (VLM) بشكل خاص لرقمنة منحنيات الأوبئة. يتكون هذا المعيار من 1,000 صورة حقيقية مأخوذة من بيانات الصحة العامة، مما يعالج القيود الموجودة في مقاييس التقييم الحالية التي تتجاهل الهيكل الزمني لبيانات السلاسل الزمنية.

La introducción de EpiCurveBench marca un avance significativo en la evaluación de modelos de visión-lenguaje (VLM) específicamente para la digitalización de curvas epidémicas. Este benchmark consta de 1,000 imágenes del mundo real obtenidas de datos de salud pública, abordando las limitaciones existentes en las métricas de evaluación actuales que pasan por alto la estructura temporal de los datos de series temporales.

L'introduction d'EpiCurveBench représente une avancée significative dans l'évaluation des modèles de vision-langage (VLM) spécifiquement pour la numérisation des courbes épidémiques. Ce benchmark se compose de 1 000 images réelles provenant de données de santé publique, abordant les limitations existantes des métriques d'évaluation actuelles qui négligent la structure temporelle des séries chronologiques.

The introduction of EpiCurveBench marks a significant advancement in the evaluation of vision-language models (VLMs) specifically for epidemic curve digitization. This benchmark consists of 1,000 real-world images sourced from public health data, addressing existing limitations in current evaluation metrics that overlook the temporal structure of time series data.