Summary of Recursive Chain-of-feedback Prevents Performance Degradation From Redundant Prompting, by Jinwoo Ahn et al.
Recursive Chain-of-Feedback Prevents Performance Degradation from Redundant Prompting
by Jinwoo Ahn, Kyuseung Shin
First submitted to arxiv on: 5 Feb 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 research explores how Large Language Models (LLMs) handle complex reasoning tasks, where they often struggle to provide logically sound steps towards a solution. To study this behavior, the authors introduce the Chain-of-Feedback (CoF) setting, which involves repeatedly prompting LLMs with meaningless feedback to assess their responses. Surprisingly, the results show that repeated feedback leads to decreased response quality and increased deviation from the intended outcome. To address these issues, the authors propose Recursive Chain-of-Feedback (R-CoF), a novel method that recursively revises incorrect responses by breaking down each reasoning step into smaller problems. Preliminary findings suggest that R-CoF can successfully answer most questions that LLMs initially fail to respond correctly, without requiring additional training data. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This research looks at how big language models do complex thinking tasks. These models often get stuck and don’t give good steps to solve the problem. To study this, the authors created a special setup called Chain-of-Feedback (CoF). They gave meaningless feedback to the model over and over to see what happens. The results showed that giving bad feedback makes the model’s answers worse. To fix this, the authors came up with a new way called Recursive Chain-of-Feedback (R-CoF). R-CoF breaks down each wrong step into smaller problems and fixes them one by one. It seems that R-CoF can help models answer tricky questions without needing more training. |
Keywords
» Artificial intelligence » Prompting
