DRPixelBench
A deep research benchmark for visual reasoning over scientific papers. It requires models to understand text, charts, tables, and formulas simultaneously, testing cross-modal evidence integration and multi-step reasoning.
Timeline (last 90 days)
DRPixelBench is a benchmark specifically targeting scientific papers that requires models to simultaneously understand text, figures, tables, and formulas.
After training, DR Pixel improves retrieval recall on DRPixelBench by about 15 percentage points and final accuracy by more than 10 percentage points.
All Facts (2)
DRPixelBench is a benchmark specifically targeting scientific papers that requires models to simultaneously understand text, figures, tables, and formulas.
50%UnverifiedAfter training, DR Pixel improves retrieval recall on DRPixelBench by about 15 percentage points and final accuracy by more than 10 percentage points.
50%