Summary of Lapt: Label-driven Automated Prompt Tuning For Ood Detection with Vision-language Models, by Yabin Zhang et al.
LAPT: Label-driven Automated Prompt Tuning for OOD Detection with Vision-Language Models
by Yabin Zhang, Wenjie Zhu, Chenhang He, Lei Zhang
First submitted to arxiv on: 12 Jul 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI); 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 This paper introduces Label-driven Automated Prompt Tuning (LAPT), a novel approach to out-of-distribution (OOD) detection using Vision-Language Models (VLMs) like CLIP. LAPT reduces the need for manual prompt engineering, which is time-consuming and requires domain expertise. The framework uses in-distribution class names and negative labels mined automatically through image synthesis and retrieval methods. A simple cross-entropy loss optimizes prompts, while cross-modal and cross-distribution mixing strategies reduce noise and explore intermediate spaces. LAPT operates autonomously, requiring only ID class names as input, eliminating manual intervention. The paper demonstrates consistent outperformance of manually crafted prompts, setting a new standard for OOD detection. Additionally, LAPT enhances ID classification accuracy, generalization robustness to covariate shifts, and performance in challenging full-spectrum OOD detection tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine having a special tool that helps machines understand when they’re shown something they’ve never seen before. This is called out-of-distribution (OOD) detection. The problem is that making these tools work well requires a lot of manual effort and expertise. A team of researchers has created a new way to make OOD detection easier and more accurate, using a combination of language and vision models like CLIP. Their approach, called LAPT, uses automated methods to create prompts (questions) that help the machine understand what it’s seeing. This makes it possible to detect unknown classes without needing human intervention. The team tested their approach and found it outperformed manual methods, making it a significant step forward in OOD detection. |
Keywords
* Artificial intelligence * Classification * Cross entropy * Generalization * Image synthesis * Prompt
