Summary of Benchmarking the Attribution Quality Of Vision Models, by Robin Hesse et al.
Benchmarking the Attribution Quality of Vision Models
by Robin Hesse, Simone Schaub-Meyer, Stefan Roth
First submitted to arxiv on: 16 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Machine Learning (cs.LG)
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 The proposed novel evaluation protocol overcomes limitations of the incremental-deletion protocol, enabling evaluation of 23 attribution methods and their quality with regards to different vision backbone designs. The findings show that intrinsically explainable models outperform standard models, while raw attribution values exhibit higher quality than previous work suggests. Additionally, consistent changes in attribution quality are observed when varying network design, highlighting the importance of certain standard design choices for promoting attribution quality. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper proposes a new way to test how well computer vision models understand what they’re doing. It’s like trying to figure out why a dog is barking by looking at all the things that might make it bark. The researchers found that some models are better than others at explaining themselves, and that changing certain parts of the model makes its explanations better or worse. This can help people build better AI models in the future. |
