FastRM: An efficient and automatic explainability framework for multimodal generative models
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866913821866590208 |
|---|---|
| author | Stan, Gabriela Ben-Melech Aflalo, Estelle Luo, Man Rosenman, Shachar Le, Tiep Paul, Sayak Tseng, Shao-Yen Lal, Vasudev |
| author_facet | Stan, Gabriela Ben-Melech Aflalo, Estelle Luo, Man Rosenman, Shachar Le, Tiep Paul, Sayak Tseng, Shao-Yen Lal, Vasudev |
| contents | Large Vision Language Models (LVLMs) have demonstrated remarkable reasoning capabilities over textual and visual inputs. However, these models remain prone to generating misinformation. Identifying and mitigating ungrounded responses is crucial for developing trustworthy AI. Traditional explainability methods such as gradient-based relevancy maps, offer insight into the decision process of models, but are often computationally expensive and unsuitable for real-time output validation. In this work, we introduce FastRM, an efficient method for predicting explainable Relevancy Maps of LVLMs. Furthermore, FastRM provides both quantitative and qualitative assessment of model confidence. Experimental results demonstrate that FastRM achieves a 99.8% reduction in computation time and a 44.4% reduction in memory footprint compared to traditional relevancy map generation. FastRM allows explainable AI to be more practical and scalable, thereby promoting its deployment in real-world applications and enabling users to more effectively evaluate the reliability of model outputs. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2412_01487 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | FastRM: An efficient and automatic explainability framework for multimodal generative models Stan, Gabriela Ben-Melech Aflalo, Estelle Luo, Man Rosenman, Shachar Le, Tiep Paul, Sayak Tseng, Shao-Yen Lal, Vasudev Artificial Intelligence Large Vision Language Models (LVLMs) have demonstrated remarkable reasoning capabilities over textual and visual inputs. However, these models remain prone to generating misinformation. Identifying and mitigating ungrounded responses is crucial for developing trustworthy AI. Traditional explainability methods such as gradient-based relevancy maps, offer insight into the decision process of models, but are often computationally expensive and unsuitable for real-time output validation. In this work, we introduce FastRM, an efficient method for predicting explainable Relevancy Maps of LVLMs. Furthermore, FastRM provides both quantitative and qualitative assessment of model confidence. Experimental results demonstrate that FastRM achieves a 99.8% reduction in computation time and a 44.4% reduction in memory footprint compared to traditional relevancy map generation. FastRM allows explainable AI to be more practical and scalable, thereby promoting its deployment in real-world applications and enabling users to more effectively evaluate the reliability of model outputs. |
| title | FastRM: An efficient and automatic explainability framework for multimodal generative models |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2412.01487 |