Summary of Hdflow: Enhancing Llm Complex Problem-solving with Hybrid Thinking and Dynamic Workflows, by Wenlin Yao et al.
HDFlow: Enhancing LLM Complex Problem-Solving with Hybrid Thinking and Dynamic Workflows
by Wenlin Yao, Haitao Mi, Dong Yu
First submitted to arxiv on: 25 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 proposed HDFlow framework addresses the limitations of large language models (LLMs) in complex reasoning problems by combining fast and slow thinking modes adaptively. The approach consists of two key components: Dynamic Workflow, which decomposes complex problems into sub-tasks and assembles specialized LLMs or symbolic reasoning tools to solve them; and Hybrid Thinking, a framework that dynamically combines fast and slow thinking based on problem complexity. To train smaller LLMs to internalize hybrid reasoning strategies, an easy-to-scale dataset of 27K challenging reasoning problems is proposed, along with a hybrid thinking tuning method. Experimental results demonstrate the effectiveness of HDFlow in complex problem-solving, outperforming Chain-of-Thought on four reasoning benchmark datasets and providing a balance between computational efficiency and performance. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary HDFlow is a new way to help big language models think more deeply about tricky problems. It combines two ways of thinking: fast, instinctive thinking and slow, deliberate thinking. The framework has two parts: one that breaks down complex problems into smaller steps and assembles the right tools to solve them, and another that mixes fast and slow thinking based on how hard the problem is. To train these models, a big dataset of challenging reasoning problems was created, along with a way to fine-tune the models for better performance. |
