Summary of Chain Of Thoughtlessness? An Analysis Of Cot in Planning, by Kaya Stechly et al.
Chain of Thoughtlessness? An Analysis of CoT in Planning
by Kaya Stechly, Karthik Valmeekam, Subbarao Kambhampati
First submitted to arxiv on: 8 May 2024
Categories
- Main: Artificial Intelligence (cs.AI)
- Secondary: None
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 A novel case study examines the efficacy of chain of thought prompting (CoT) on reasoning problems from the classical planning domain of Blocksworld. Two state-of-the-art large language models (LLMs) are evaluated across two axes: generality of prompts and complexity of queried problems. The results indicate that meaningful performance improvements only arise when prompts are highly specific to their problem class, and these gains quickly deteriorate as query-specified stack size exceeds example sizes. Additionally, scalable variants of three domains commonly studied in previous CoT papers are created, revealing similar failure modes. These findings suggest that CoT’s performance boosts do not stem from the model learning general algorithmic procedures via demonstrations, but rather depend on carefully crafting highly problem-specific prompts. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary A new study looks at how well large language models (LLMs) can solve problems with a method called chain of thought prompting. This method shows the LLM an example solution to help it learn. The study uses simple problems from Blocksworld, a type of planning puzzle. It finds that this method only works well when the prompts are very specific to the problem being solved. If the prompts are too general or try to solve more complex problems, the LLM doesn’t improve much. This means that while chain of thought prompting can be helpful, it’s not a magic solution that makes LLMs better at solving all types of problems. |
Keywords
» Artificial intelligence » Prompting
