Authors: Yichen Zhu, Minjie Zhu, Ning Liu, Zhicai Ou, Xiaofeng Mou, Jian Tang

Published on: January 04, 2024

Impact Score: 8.22

Arxiv code: Arxiv:2401.02330

Summary

• What is new: LLaVA-Phi uses a smaller language model to effectively engage in multi-modal dialogues.
• Why this is important: Previous models were too large and inefficient for real-time multi-modal (text and visual) dialogue tasks.
• What the research proposes: Utilizing the compact Phi-2 language model with 2.7B parameters trained on high-quality corpora for improved dialogue performance.
• Results: Achieved commendable performance on benchmarks for visual comprehension, reasoning, and knowledge-based perception.

Technical Details

Technological frameworks used: LLaVA-Phi

Models used: Phi-2

Data used: High-quality corpora

Potential Impact

Embodied agent applications, time-sensitive interactive systems, visual comprehension tools, and customer service technologies.

Want to implement this idea in a business?

We have generated a startup concept here: VisuoChat.