Summary of Memory-inspired Temporal Prompt Interaction For Text-image Classification, by Xinyao Yu et al.
Memory-Inspired Temporal Prompt Interaction for Text-Image Classification
by Xinyao Yu, Hao Sun, Ziwei Niu, Rui Qin, Zhenjia Bai, Yen-Wei Chen, Lanfen Lin
First submitted to arxiv on: 26 Jan 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 proposed Memory-Inspired Temporal Prompt Interaction (MITP) method addresses the computational cost issue associated with fine-tuning large-scale multimodal models for downstream tasks. By mimicking human memory strategy, MITP consists of two stages: acquiring and consolidation/activation. The acquiring stage uses temporal prompts on intermediate layers to facilitate information exchange between modalities, while the consolidation/activation stage leverages similarity-based prompt interaction and prompt generation strategies. This approach achieves competitive results on various datasets with reduced memory usage (2.0M trainable parameters) compared to pre-trained foundation models. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large-scale pre-trained multimodal models are great at many tasks, but they’re big and need lots of computing power to work well. Researchers want to make them smaller and faster without sacrificing performance. To do this, they created a new way to “talk” to these models called Memory-Inspired Temporal Prompt Interaction (MITP). MITP is like how humans remember things – it has two parts: getting information and storing and using that information. The method uses special prompts at different levels to help the model learn and remember better, without needing as much computing power. This makes it a more efficient way to use these models for many tasks. |
Keywords
» Artificial intelligence » Fine tuning » Prompt
