Summary of Characterizing Context Influence and Hallucination in Summarization, by James Flemings et al.
Characterizing Context Influence and Hallucination in Summarization
by James Flemings, Wanrong Zhang, Bo Jiang, Zafar Takhirov, Murali Annavaram
First submitted to arxiv on: 3 Oct 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Machine Learning (cs.LG)
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 investigates two significant concerns surrounding Large Language Models (LLMs): hallucination and private information leakage. Despite LLMs achieving impressive performance in various tasks, they can generate content that contradicts context or inadvertently leak private information due to input regurgitation. To address these concerns simultaneously, the authors introduce a definition for context influence and Context-Influence Decoding (CID). They show that amplifying context and prior knowledge increases context influence on an LLM, providing a lower bound of private information leakage. Experimental evaluations demonstrate that improving F1 ROGUE-L scores on CNN-DM for LLaMA 3 also leads to increased context influence. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Large Language Models (LLMs) have made huge progress in many areas, but they can also make mistakes. They might create text that doesn’t match what we’re talking about, or reveal private information. Until now, researchers have mostly looked at these problems separately, not together. This paper explores both issues and shows how the quality of context affects LLMs’ performance. By looking at this relationship, the authors can provide a better understanding of why LLMs make mistakes and what we can do to improve them. |
Keywords
» Artificial intelligence » Cnn » Hallucination » Llama
