Summary of Self-rewarding Language Models, by Weizhe Yuan et al.
Self-Rewarding Language Models
by Weizhe Yuan, Richard Yuanzhe Pang, Kyunghyun Cho, Xian Li, Sainbayar Sukhbaatar, Jing Xu, Jason Weston
First submitted to arxiv on: 18 Jan 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 proposed approach to achieving superhuman language agents involves training reward models using a novel self-rewarding mechanism, where the language model itself provides rewards during training. By leveraging large language models (LLMs) as judges and prompting them with iterative DPO training, the authors demonstrate improved instruction following ability and high-quality reward provision. The resulting fine-tuned LLaMA 2 70B model outperforms several existing systems on the AlpacaEval 2.0 leaderboard. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary We’re working on making super smart computers that can understand and respond to human language. To do this, we need a way for these computers to learn from their own mistakes and improve over time. One idea is to have the computer give itself rewards when it does something correct, like answering a question correctly. This approach is called self-rewarding, and it’s an important step towards making our computers super smart. |
Keywords
» Artificial intelligence » Language model » Llama » Prompting
