CURV: Enhancing Chart Understanding Through Curriculum Visual Grounded Reasoning
CURV trains chart QA models to reason step by step while grounding each step in the chart image.
The paper frames chart question answering as a visual-grounded reasoning problem, not just a prompting problem. It introduces CCQA, a three-level curriculum dataset that moves from single-operation chart reasoning to complex multi-chart tasks. In experiments, CURV reports gains of up to 20.50% over baselines, with smaller gains on real-world and out-of-domain benchmarks. Code is listed as available in the paper. HF Daily Papers' note
The paper frames chart question answering as a visual-grounded reasoning problem, not just a prompting problem. It introduces CCQA, a three-level curriculum dataset that moves from single-operation chart reasoning to complex multi-chart tasks. In experiments, CURV reports gains of up to 20.50% over baselines, with smaller gains on real-world and out-of-domain benchmarks. Code is listed as available in the paper. HF Daily Papers' note
score 4