Summary of Lookback Lens: Detecting and Mitigating Contextual Hallucinations in Large Language Models Using Only Attention Maps, by Yung-sung Chuang et al.
Lookback Lens: Detecting and Mitigating Contextual Hallucinations in Large Language Models Using Only Attention Maps
by Yung-Sung Chuang, Linlu Qiu, Cheng-Yu Hsieh, Ranjay Krishna, Yoon Kim, James Glass
First submitted to arxiv on: 9 Jul 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI); 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 Medium Difficulty Summary: This paper presents a novel approach to detect contextual hallucinations in large language models (LLMs) when answering questions or summarizing articles. The proposed model, called Lookback Lens, utilizes attention weights ratio features as input and is found to be effective in detecting hallucinations across different tasks and models. The detection process is based on the intuition that LLMs attend more to context information when generating accurate responses and less when engaging in contextual hallucinations. This paper demonstrates the effectiveness of the Lookback Lens detector in mitigating hallucinations, achieving a reduction of 9.6% in the XSum summarization task. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Low Difficulty Summary: Sometimes, artificial intelligence models can make up answers that aren’t true based on what they’re given to read or write about. This paper shows how to detect when these models are making things up. The approach uses a special way of looking at how the model pays attention to the information it’s given versus what it generates itself. This method works well across different tasks and even different-sized models. By using this method, we can reduce the number of made-up answers by 9.6% in one particular task. |
Keywords
» Artificial intelligence » Attention » Summarization
