arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2502.12929cs.LGcs.AIcs.CL
收藏

Flow-of-Options:通过思考选项实现多样化与改进的LLM推理

Flow-of-Options: Diversified and Improved LLM Reasoning by Thinking Through Options

Flagship Pioneering(旗舰先锋)

Lakshmi Nair, Ian Trase, Mark Kim

AI总结:

提出名为Flow-of-Options (FoO) 的推理方法,通过压缩可解释的任务表示强制实现LLM推理的多样性,在数据科学和治疗化学等任务上显著超越基线且成本极低,并具备广泛适用性。

摘要:

我们提出了一种名为Flow-of-Options (FoO) 的新型推理方法,旨在解决大型语言模型(LLMs)中的内在偏差。Flow-of-Options 使 LLMs 能够在其推理中系统地探索多种可能性,这通过一个为自主解决机器学习(ML)任务而开发的基于 FoO 的智能体框架得到了验证。FoO 通过压缩且可解释的任务表示,在 LLM 解决方案中强制实现多样性,与最先进的基线相比,在标准数据科学任务上提升了 38.2% - 69.2%,在治疗化学任务上提升了 37.4% - 47.9%。由于每个任务的总操作成本低于 1 美元,我们的框架非常适合对成本敏感的应用。除了表格分类和回归之外,我们还展示了基于 FoO 的智能体系统在强化学习和图像生成等任务上的更广泛适用性。我们的代码已在以下网址开源:https://github.com/flagshippioneering/Flow-of-Options。

英文摘要:

We present a novel reasoning approach called Flow-of-Options (FoO), designed to address intrinsic biases in Large Language Models (LLMs). Flow-of-Options enables LLMs to systematically explore a diverse range of possibilities in their reasoning, as demonstrated by an FoO-based agentic framework developed for autonomously solving Machine Learning (ML) tasks. FoO enforces diversity in LLM solutions through compressed and interpretable task representations, resulting in improvements of 38.2% - 69.2% on standard data science tasks, and 37.4% - 47.9% on therapeutic chemistry tasks, as compared to state-of-the-art baselines. With an overall operation cost under $1 per task, our framework is well-suited for cost-sensitive applications. Going beyond tabular classification and regression, we show the broader applicability of our FoO-based agentic system to tasks such as reinforcement learning and image generation. Our code is open-sourced at: https://github.com/flagshippioneering/Flow-of-Options.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑