Summary of Behavioral Bias Of Vision-language Models: a Behavioral Finance View, by Yuhang Xiao et al.
Behavioral Bias of Vision-Language Models: A Behavioral Finance View
by Yuhang Xiao, Yudi Lin, Ming-Chang Chiu
First submitted to arxiv on: 23 Sep 2024
Categories
- Main: Computation and Language (cs.CL)
- 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 paper explores the potential biases of Large Vision-Language Models (LVLMs) in various domains, particularly in the context of behavioral finance. It proposes an end-to-end framework to assess LVLMs’ reasoning capabilities and identify biases such as recency bias and authority bias. The study finds that open-source models like LLaVA-NeXT, MobileVLM-V2, Mini-Gemini, MiniCPM-Llama3-V 2.5, and Phi-3-vision-128k exhibit significant biases, whereas the proprietary model GPT-4o is less affected. This highlights areas where open-source models can be improved. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary LVLMs are AI models that combine language and vision capabilities to create more human-like intelligence. However, these models may have unintended biases when used in different domains. The paper investigates two common financial biases – recency bias (focusing on recent events) and authority bias (following expert opinions). It develops a new framework for evaluating LVLMs’ reasoning abilities and identifies biases present in several open-source models. This research can help improve AI models to better serve society. |
Keywords
» Artificial intelligence » Gemini » Gpt
