Summary of Longsafety: Enhance Safety For Long-context Llms, by Mianqiu Huang et al.
LongSafety: Enhance Safety for Long-Context LLMs
by Mianqiu Huang, Xiaoran Liu, Shaojun Zhou, Mozhi Zhang, Qipeng Guo, Linyang Li, Chenkun Tan, Yang Gao, Pengyu Wang, Linlin Li, Qun Liu, Yaqian Zhou, Xipeng Qiu, Xuanjing Huang
First submitted to arxiv on: 11 Nov 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 Recent advancements in model architectures and length extrapolation techniques have extended the context length of large language models (LLMs), enabling their application in complex tasks. However, despite these advancements, the safety issues in long-context scenarios remain underexplored. This paper introduces LongSafety, a comprehensive dataset for long-context LLMs, containing 10 tasks and 17k samples with an average length of 40.9k tokens. Training with LongSafety enhances long-context safety performance while preserving general capabilities. The study highlights the importance of addressing safety concerns in long-context scenarios, demonstrating that long-context safety does not equal alignment with short-context safety data. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Long language models can understand very long pieces of text, which is useful for tasks like writing stories or summarizing books. But using these models also raises some safety concerns. Imagine if a model was trained to be good at being mean and it became really good at it – that could be a problem! To address this issue, researchers created a new dataset called LongSafety, which has 10 tasks and thousands of samples of text for the models to practice with. The study shows that using LongSafety makes the models safer and better at understanding very long pieces of text. |
Keywords
» Artificial intelligence » Alignment » Context length
