Flow-of-Options: Diversified and Improved LLM Reasoning by Thinking Through Options
Lakshmi Nair, Ian Trase, Mark Kim
2025-02-19
Summary
This paper talks about a new way to make AI language models think more creatively and less biased called Flow-of-Options (FoO). It's like teaching a super-smart computer to brainstorm different ideas before making a decision, just like humans do.
What's the problem?
Big AI language models often get stuck thinking in one way because of built-in biases. This can make them less effective at solving complex problems, especially in areas like data science and chemistry where there might be many possible solutions.
What's the solution?
The researchers created FoO, which makes AI explore lots of different options when solving a problem. They tested it on various tasks, including some tricky science and chemistry problems. FoO helped the AI come up with more diverse and better solutions than other top-notch AI systems. It even works for things like teaching AI to play games or create images.
Why it matters?
This matters because it makes AI smarter and more flexible in solving real-world problems. FoO could help scientists discover new medicines, make data analysis more accurate, and even create more creative AI art. Plus, it's cheap to use, costing less than a dollar per task, which means more people and companies could benefit from this advanced AI technology. It's a big step towards making AI think more like humans, considering many options before making decisions.
Abstract
We present a novel reasoning approach called Flow-of-Options (FoO), designed to address intrinsic biases in Large Language Models (LLMs). FoO enables LLMs to systematically explore a diverse range of possibilities in their reasoning, as demonstrated by an FoO-based agentic system for autonomously solving Machine Learning tasks (AutoML). Our framework outperforms state-of-the-art baselines, achieving improvements of 38.2% - 69.2% on standard data science tasks, and 37.4% - 47.9% on therapeutic chemistry tasks. With an overall operation cost under $1 per task, our framework is well-suited for cost-sensitive applications. Beyond classification and regression, we illustrate the broader applicability of our FoO-based agentic system to tasks such as reinforcement learning and image generation. Our framework presents significant advancements compared to current state-of-the-art agentic systems for AutoML, due to the benefits of FoO in enforcing diversity in LLM solutions through compressed, explainable representations that also support long-term memory when combined with case-based reasoning.