Summary of Let Go Of Your Labels with Unsupervised Transfer, by Artyom Gadetsky et al.
Let Go of Your Labels with Unsupervised Transfer
by Artyom Gadetsky, Yulun Jiang, Maria Brbic
First submitted to arxiv on: 11 Jun 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: None
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 paper proposes a novel method called TURTLE that enables fully unsupervised transfer learning from pre-trained vision-language models. This breakthrough allows for solving new tasks without human guidance, as long as the data has underlying labels. The authors demonstrate the effectiveness of TURTLE by achieving state-of-the-art performance on 26 benchmark datasets, surpassing zero-shot transfer and prompt tuning baselines. The method employs representation spaces from two foundation models to guide the search for the underlying labeling, showcasing the power and potential of unsupervised transfer learning. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary TURTLE is a new way to use pre-trained models without needing human help. It works by finding the best labels for a dataset that makes sense based on how the model represents things. This helps solve tasks without any training or special setup. The paper tests TURTLE on many different datasets and shows it does really well, beating other methods that require more information. |
Keywords
» Artificial intelligence » Prompt » Transfer learning » Unsupervised » Zero shot
