1. CharXiv
现有数据集通常关注过于简化和同质化的图表,并且问题往往基于模板生成,这导致了对MLLMs图表理解能力的过度乐观评估。为了解决这个问题,作者提出了一个新的评估套件CharXiv,它包含了从arXiv论文中精选的2323个自然、具有挑战性和多样性的图表,并设计了两种类型的问题:描述性问题和推理问题,以全面评估MLLMs在图表理解方面的能力。
- paper:Charting Gaps in Realistic Chart Understanding in Multimodal LLMs
- link:https://arxiv.org/abs/2406.18521
- dataset:https://huggingface.co/datasets/princeton-nlp/CharXiv
2. OneChart
数据集类型:图表结构提取、图表推理
- paper:OneChart: Purify the Chart Structural Extraction via One Auxiliary Token
- link:https://arxiv.org/abs/2404.09987
- dataset:https://github.com/LingyvKong/OneChart
3. ChartLlama
- paper:ChartLlama: A Multimodal LLM for Chart Understanding and Generation
- link:https://arxiv.org/pdf/2311.16483
- dataset:https://huggingface.co/datasets/listen2you002/ChartLlama-Dataset
4. ChartX
- ChartX & ChartVLM: A Versatile Benchmark and Foundation Model for Complicated Chart Reasoning
- link:https://arxiv.org/pdf/2402.12185
- dataset:https://github.com/UniModal4Reasoning/ChartVLM