Summary of Non-autoregressive Sequence-to-sequence Vision-language Models, by Kunyu Shi et al.
Non-autoregressive Sequence-to-Sequence Vision-Language Models
by Kunyu Shi, Qi Dong, Luis Goncalves, Zhuowen Tu, Stefano Soatto
First submitted to arxiv on: 4 Mar 2024
Categories
- Main: Computer Vision and Pattern Recognition (cs.CV)
- Secondary: Artificial Intelligence (cs.AI)
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 sequence-to-sequence vision-language model that addresses the limitations of current models due to their autoregressive nature. The proposed model, NARVL, uses parallel decoding and Query-CTC loss to marginalize over multiple inference paths in the decoder, allowing it to model the joint distribution of tokens rather than conditional distributions. This approach enables faster inference times, reducing complexity from linear to constant time. NARVL achieves performance on-par with state-of-the-art autoregressive models while improving inference speed. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making computers better at understanding and generating text that describes pictures. Right now, the best way to do this requires a lot of calculations, which makes it slow. The researchers created a new model called NARVL that can do this task faster by using a different method. It’s like going from a step-by-step recipe to making a whole meal at once! This new approach is just as good as the old one, but it gets the job done much quicker. |
Keywords
* Artificial intelligence * Autoregressive * Decoder * Inference * Language model
