Summary of 360zhinao Technical Report, by 360zhinao Team
360Zhinao Technical Report
by 360Zhinao Team
First submitted to arxiv on: 22 May 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 We introduce the 360Zhinao model with a parameter size of 7 billion, featuring context lengths spanning 4K, 32K, and 360K. The models are available at this GitHub URL. To facilitate rapid development in pretraining, we establish an ablation environment to evaluate experiment runs with minimal model size. We optimize data cleaning and composition strategies to pretrain the 360Zhinao-7B-Base model on 3.4 trillion tokens. Our approach emphasizes balancing quantity and quality through filtering and reformatting. With tailored data, the 360Zhinao-7B model’s context window can be extended to 32K and 360K. We apply Reinforcement Learning from Human Feedback (RLHF) and Self-Following Training (SFT) to specific tasks, achieving competitive performance among models of similar size. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary We’re introducing a new type of AI model called 360Zhinao. It’s really big and can process lots of information at once. We made it work better by cleaning up the data it uses and making sure it’s learning from good examples. This helps the model learn faster and do tasks like language translation and text summarization more accurately. The model is available online, and we’re excited to see what people will use it for! |
Keywords
» Artificial intelligence » Context window » Pretraining » Reinforcement learning from human feedback » Rlhf » Summarization » Translation
