Month in 4 Papers (April 2023–4)

This series of posts is designed to bring you the newest findings and developments in the NLP field. I’ll delve into four significant research papers each month, offering a comprehensive summary. Be sure to visit my blog regularly or subscribe to my newsletter for monthly updates. Let’s dive in!

Visual Prompt Planning

📝 Low-code LLM: Visual Programming over LLMs

It could be hard to write a perfect prompt in models like ChatGPT. The main reason could be that we don’t know what the model is thinking. Will I get the desired output by changing one word? Or, I am asking the wrong question! Visual Programming can mitigate this problem. It uses a Planning LLM to lay out a step-by-step plan based on the user’s prompt. Then, the user can edit the instructions to get the best result. Users get control over the generation process using a friendly UI on multiple tasks. It can write essays, code, be an assistant, and embed knowledge. It is a great step to make the LLMs more user-friendly and control their Chain-of-Thoughts. [examples] (There are more examples in the paper)

Attention-free LM

📝 Hyena Hierarchy: Towards Larger Convolutional Language Models