Summary of Improving Attributed Text Generation Of Large Language Models Via Preference Learning, by Dongfang Li et al.
Improving Attributed Text Generation of Large Language Models via Preference Learning
by Dongfang Li, Zetian Sun, Baotian Hu, Zhenyu Liu, Xinshuo Hu, Xuebo Liu, Min Zhang
First submitted to arxiv on: 27 Mar 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 paper addresses the challenge of generating unreliable content in large language models by introducing an Automatic Preference Optimization (APO) framework. Current attribution methods focus on retrieval stages and neglect human-like citation mechanisms, which can bolster credibility. The authors model the attribution task as preference learning and propose APO to reduce misinformation and hallucinations. They create a curated dataset with 6,330 examples and synthesize 95,263 pairs of attribution preferences. Inspired by human citations, they propose a progressive preference optimization method that leverages fine-grained information. Experiments on three datasets (ASQA, StrategyQA, and ELI5) demonstrate APO achieves state-of-the-art citation F1 with higher answer quality. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large language models are very smart computers that can understand and generate human-like text. However, they sometimes make mistakes and create false or misleading information. Researchers want to fix this by making the computer explain why it thinks something is true. The new approach uses a technique called “preference learning” to help the computer decide what information to include in its responses. The team created a large dataset of examples and then used an automatic method to generate even more examples. They also developed a way for the computer to optimize its performance by using fine-grained details. Tests showed that this new approach performs better than previous methods. |
Keywords
» Artificial intelligence » Optimization
