Tag
研究团队提出 MisVisAgentBench，首个配对可执行基准，用于评估误导性图表对 Web Agent 执行的影响。在 13 个视觉-语言 Web Agent 上，替换为误导图表后平均任务成功率下降超过 34 个百分点，且失败大多沿着被误导的操作分支发生。
DocHop introduces a benchmark for evaluating multi-hop reasoning in information-dense documents, focusing on chart-context integration in multimodal large language models, with experiments showing a significant gap between model and human performance.
This paper introduces CoCoEvolve, a self-supervised method that improves cross-representation understanding across charts, tables, and code by enforcing one-to-one consistency between representations, with training-time and test-time co-evolution objectives.
CURV proposes a curriculum learning framework that enhances chart question answering by developing intrinsic visual grounded reasoning in multimodal LLMs, achieving significant improvements on real-world benchmarks.
A 4B open-source model beats Mythos 5 on the CharXiv chart understanding benchmark, showing strong performance from a freely available small model.
MIT researchers developed ChartNet, a dataset of over a million charts, to train vision-language models to interpret charts more accurately. Their open-source models outperform much larger commercial models on chart understanding tasks.
HakushoBench is a Japanese chart and table VQA benchmark built from governmental white papers to evaluate vision-language models' understanding of complex visual data, challenging open-weight models with a 58.6% accuracy and a 34.9-point gap to proprietary models.