论文进阶
benchmark
基准测试，用于衡量与比较系统在特定任务上表现的标准数据集及配套评估协议。
English Definition
"A standardized dataset and evaluation protocol used to measure and compare systems on a specific task."
用法说明 · 针对中文母语者
在 HCI/AI 论文中通常以专有名词形式出现（如 ExChart-Bench），兼具数据集与评估流程双重含义。中文常译作'基准'或'基准测试'。
真实用例 · 3 条
"ExChart-Bench, a benchmark spanning real-world and synthetic charts across diverse styles and encodings"
论文Pipeline extracted — review needed·介绍新构建的 benchmark 覆盖范围
"A systematic evaluation of widely used MLLMs on ExChart-Bench reveals a consistent bottleneck"
论文Pipeline extracted — review needed·用 benchmark 做系统性评估
"ExChart-Bench helps surface these gaps and provides a principled way to evaluate both general-purpose MLLMs and specialized models trained with our proposed framework."
论文Pipeline extracted — review needed·benchmark 的双重作用（暴露差距 + 提供评估手段）
由 pipeline 自动采集，待人工 review