2026-10-13T12:21:01Z https://www.scidb.cn/oai

10.57760/sciencedb.j00133.00697 2026-06-29T08:58:18Z

2026-06-29 Multi-Modal Fusion Prompter Model doi:10.57760/sciencedb.j00133.00697 en This dataset contains the PyTorch source code of the MFPM model proposed in the paper "Research on Missing Modal Completion and Dual Branch Dynamic Fusion Sentiment Analysis Model Based on Hint Learning", as well as the raw output results and chart drawing data of all experiments. The experiment covers four benchmark datasets: CMU-MOSI, CMU-MOSEI, IEMOCAP, and CH-SIMS, including comparative experiments, ablation experiments, robustness testing (with a loss rate of 5%~95%), generalization experiments, and detailed indicators of sensitivity analysis (ACC, F1, MAE, Corr, etc.) in 64 scenarios of complete modality, single modality loss, and bimodal loss. The data is stored in Excel format for easy reproduction of paper charts and conclusions.  Multimodal sentiment analysis; Modality missing; Prompt learning; Dynamic fusion; Robustness zheng bing PUBLIC https://creativecommons.org/licenses/by-nc-sa/4.0/ dataset Science Data Bank