Summary of Iterative Reasoning Preference Optimization, by Richard Yuanzhe Pang et al.
Iterative Reasoning Preference Optimization
by Richard Yuanzhe Pang, Weizhe Yuan, Kyunghyun Cho, He He, Sainbayar Sukhbaatar, Jason Weston
First submitted to arxiv on: 30 Apr 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 proposes an iterative approach to optimize the preference between competing Chain-of-Thought (CoT) candidates for reasoning tasks, building upon previous work in general instruction tuning. The method optimizes winning and losing steps that lead to correct answers using a modified DPO loss with an additional negative log-likelihood term. Experimental results show improved reasoning performance across iterations, outperforming other Llama-2-based models on GSM8K, MATH, and ARC-Challenge datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps make computers better at understanding and solving problems by optimizing how they choose between different ideas or solutions. The approach uses a special way of combining two types of loss functions to make the computer learn from its mistakes. The results show that this method can improve the computer’s ability to reason and solve problems, especially when working with large datasets. |
Keywords
» Artificial intelligence » Instruction tuning » Llama » Log likelihood
