Summary of Efficient Generative Modeling with Residual Vector Quantization-based Tokens, by Jaehyeon Kim et al.
Efficient Generative Modeling with Residual Vector Quantization-Based Tokens
by Jaehyeon Kim, Taehong Moon, Keon Lee, Jaewoong Cho
First submitted to arxiv on: 13 Dec 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 This research paper explores the use of Residual Vector Quantization (RVQ) for high-fidelity generation in vector-quantized generative models. The authors introduce ResGen, an efficient RVQ-based discrete diffusion model that generates high-fidelity samples without compromising sampling speed. The key idea is to predict the vector embedding of collective tokens rather than individual ones. The proposed method uses token masking and multi-token prediction within a principled probabilistic framework using discrete diffusion process and variational inference. The authors validate the efficacy and generalizability on two challenging tasks: conditional image generation on ImageNet 256×256 and zero-shot text-to-speech synthesis. Experimental results show that ResGen outperforms autoregressive counterparts in both tasks, delivering superior performance without compromising sampling speed. Furthermore, as RVQ depth increases, generative models exhibit enhanced generation fidelity or faster sampling speeds compared to similarly sized baseline models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary ResGen is a new way to make computers generate images and speech that look and sound more realistic. It’s like taking a bunch of small pieces of information and putting them together to create something new. The researchers made a special kind of computer model called Residual Vector Quantization (RVQ) that can do this. They tested it on two big tasks: making pictures from words, and turning text into spoken language. Their results show that RVQ works better than other methods at doing these things. It also lets computers make more realistic images and speech as they get more powerful. |
Keywords
» Artificial intelligence » Autoregressive » Diffusion » Diffusion model » Embedding » Image generation » Inference » Quantization » Token » Zero shot
