Summary of Adacoder: Adaptive Prompt Compression For Programmatic Visual Question Answering, by Mahiro Ukai et al.
AdaCoder: Adaptive Prompt Compression for Programmatic Visual Question Answering
by Mahiro Ukai, Shuhei Kurita, Atsushi Hashimoto, Yoshitaka Ushiku, Nakamasa Inoue
First submitted to arxiv on: 28 Jul 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: Computer Vision and Pattern Recognition (cs.CV); Multimedia (cs.MM)
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 In this paper, researchers propose a novel approach to address the limitation of requiring long input prompts for visual programmatic models (VPMs) in natural language processing tasks like visual question answering. They introduce AdaCoder, an adaptive prompt compression framework that generates compressed preprompts for VPMs using large language models (LLMs). The framework operates in two phases: compression and inference. In the compression phase, a set of compressed preprompts is generated based on specific question types, while in the inference phase, the framework predicts the question type and selects the corresponding compressed preprompt to generate code for answering the question. AdaCoder utilizes a single frozen LLM and pre-defined prompts, eliminating the need for additional training and maintaining adaptability across different powerful black-box LLMs such as GPT and Claude. Experimental results demonstrate that AdaCoder reduces token length by 71.1% while maintaining or even improving performance in visual question answering tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper is about making it easier to answer questions using pictures. Right now, computers need a lot of information to understand what you’re asking. The authors propose a new way to give them this information that’s more efficient and effective. They created something called AdaCoder that helps computers generate code to answer questions by reducing the amount of information they need. This means computers can answer questions faster and better using pictures. |
Keywords
» Artificial intelligence » Claude » Gpt » Inference » Natural language processing » Prompt » Question answering » Token
