A confidence-convergence decision layer for reliability-aware sequential recognition under limited observations

This dataset supports the manuscript “A confidence-convergence decision layer for reliability-aware sequential recognition under limited observations”. It contains the figure source data tables, sensitivity-test data, publication-ready figure files, Python scripts for figure generation and data auditing, and the Supplementary Information files. The source data are provided as CSV files and are mapped to the corresponding main figures, including overall performance, coverage-risk behaviour, observation-window effects, confidence evolution, ablation analysis, and robustness analyses. The included scripts allow regeneration of the main figures and verification of the figure source data, supporting transparency and reproducibility of the reported results.本数据集为论文《面向有限观测场景下可靠性感知序列识别的置信度-收敛决策层》提供支撑。数据集包含图表源数据表、敏感性测试数据、可直接用于出版的图表文件、用于图表生成与数据审核的Python脚本，以及补充信息文件。源数据以CSV（逗号分隔值）文件格式提供，并与对应主图表形成映射关系，涵盖整体性能、覆盖度-风险行为、观测窗口效应、置信度演化、消融实验分析与鲁棒性分析等类别。所附带的脚本可实现主图表的重新生成与图表源数据的核验，为研究结果的透明度与可复现性提供保障。