Summary of Revisiting the Power Of Prompt For Visual Tuning, by Yuzhu Wang et al.
Revisiting the Power of Prompt for Visual Tuning
by Yuzhu Wang, Lechao Cheng, Chaowei Fang, Dingwen Zhang, Manni Duan, Meng Wang
First submitted to arxiv on: 4 Feb 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 This study introduces Self-Prompt Tuning (SPT), a novel approach that leverages learnable prompt tokens to customize pre-trained models for downstream tasks. By initializing prompts with downstream token prototypes, SPT substantially improves performance in fine-tuning and outperforms existing methods by a remarkable margin. In fact, it surpasses full fine-tuning in 19 out of 24 tasks using less than 0.4% of learnable parameters on the FGVC and VTAB-1K benchmarks. Moreover, SPT is robust to prompt lengths and scales well with model capacity and training data size. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This study shows how we can make computers better at doing specific jobs by giving them special instructions called prompts. These prompts help the computer learn new things without needing a lot of extra information. The researchers found that if they start with some special words, the computer will get even better at its job. They tested this idea and it worked really well! In fact, it was much better than just letting the computer figure everything out on its own. This could be very useful for things like recognizing pictures or understanding speech. |
Keywords
* Artificial intelligence * Fine tuning * Prompt * Token
