Summary of A Concept-based Explainability Framework For Large Multimodal Models, by Jayneel Parekh et al.
A Concept-Based Explainability Framework for Large Multimodal Models
by Jayneel Parekh, Pegah Khayatan, Mustafa Shukor, Alasdair Newson, Matthieu Cord
First submitted to arxiv on: 12 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Computer Vision and Pattern Recognition (cs.CV)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Medium Difficulty summary: This paper presents a novel framework for interpreting the internal representations of large multimodal models (LMMs), which combine unimodal encoders and large language models. The proposed dictionary learning approach is applied to token representations, yielding “multi-modal concepts” that are well-grounded in both vision and text. These concepts are qualitatively and quantitatively evaluated, demonstrating their usefulness for interpreting test samples’ representations. Additionally, the framework’s ability to disentangle different concepts and ground them visually and textually is examined. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty summary: This research helps us better understand how big models that combine words and pictures work from the inside out. Right now, we don’t fully know what these models are thinking when they process information. The scientists in this study developed a new way to “listen” to these models’ internal thoughts, which they call “multi-modal concepts.” These ideas are connected to both pictures and words, making them helpful for understanding how the models make decisions. |
Keywords
» Artificial intelligence » Multi modal » Token
