Summary of Abstraction-of-thought Makes Language Models Better Reasoners, by Ruixin Hong et al.
Abstraction-of-Thought Makes Language Models Better Reasoners
by Ruixin Hong, Hongming Zhang, Xiaoman Pan, Dong Yu, Changshui Zhang
First submitted to arxiv on: 18 Jun 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 This paper bridges the gap between human abstract reasoning and language model capabilities by introducing Abstraction-of-Thought (AoT), a novel structured reasoning format that requires varying levels of abstraction within the reasoning process. AoT differs from the prevailing Chain-of-Thought (CoT) method, which only incorporates concrete details. The authors present AoT Collection, a finetuning dataset with 348k high-quality samples, to align language models with the AoT format. Experimental results show that AoT-aligned models outperform CoT-aligned models on many reasoning tasks from the Big-Bench Hard benchmark. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps computers understand abstract thinking, which is important for solving complex problems. The researchers create a new way of teaching language models to think abstractly called Abstraction-of-Thought (AoT). AoT is different from how we usually teach language models, which focuses on specific details. The team also created a large dataset with examples of abstract thinking, called AoT Collection. They tested different language models using this dataset and found that those taught to think abstractly did better on many tasks. |
Keywords
» Artificial intelligence » Language model
