MLPapersOfTheWeek/README.md

ML-Papers-of-the-Week

Highlighting top ML papers of the week. Top ML Papers of the Week (Jan 1-8):

1. GoogleAI introduces Muse, a new text-to-image generation model based on masked generative transformers; significantly more efficient than other diffusion models like Imagen and DALLE-2.
   • Paper: https://arxiv.org/abs/2301.00704
2. Microsoft introduces VALL-E, a text-to-audio model that performs state-of-the-art zero-shot performance; the text-to-speech synthesis task is treated as a conditional language modeling task:
   • https://valle-demo.github.io/
3. A new paper shows the potential of enhancing LLMs by retrieving relevant external knowledge based on decomposed reasoning steps obtained through chain-of-thought prompting.
   • https://arxiv.org/abs/2301.00303
4. Presents a technique for compressing large language models while not sacrificing performance; "pruned to at least 50% sparsity in one-shot, without any retraining."
   • https://arxiv.org/pdf/2301.00774.pdf
5. ConvNeXt V2 is a performant model based on a fully convolutional masked autoencoder framework and other architectural improvements. CNNs are sticking back!
   • https://arxiv.org/abs/2301.00808
6. With more capabilities, we are starting to see a wider range of applications with LLMs. This paper utilized large language models for conducting corporate lobbying activities.
   • https://arxiv.org/abs/2301.01181
7. This work aims to better understand how deep learning models overfit or memorize examples; interesting phenomena observed; important work toward a mechanistic theory of memorization.
   • https://transformer-circuits.pub/2023/toy-double-descent/index.html
8. StitchNet is a novel paradigm to create new coherent neural networks by reusing pretrained fragments of existing NNs.
   • https://arxiv.org/abs/2301.01947
9. Proposes integrated decomposition, an approach to improve Science Q&A through a human-in-the-loop workflow for refining compositional LM programs.
   • https://arxiv.org/abs/2301.01751
10. A Succinct Summary of Reinforcement Learning. A nice little overview of some important ideas in RL.
    • https://arxiv.org/abs/2301.01379