Summary of Rho-1: Not All Tokens Are What You Need, by Zhenghao Lin et al.
Rho-1: Not All Tokens Are What You Need
by Zhenghao Lin, Zhibin Gou, Yeyun Gong, Xiao Liu, Yelong Shen, Ruochen Xu, Chen Lin, Yujiu Yang, Jian Jiao, Nan Duan, Weizhu Chen
First submitted to arxiv on: 11 Apr 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 abstract introduces a new language model called Rho-1 that employs Selective Language Modeling (SLM) to selectively train on useful tokens aligned with the desired distribution. Unlike traditional LMs, Rho-1 uses a reference model to score pretraining tokens and then trains the language model with a focused loss on tokens with higher scores. The approach yields an absolute improvement in few-shot accuracy of up to 30% in nine math tasks after continual pretraining on the OpenWebMath corpus. When fine-tuned, Rho-1 achieves state-of-the-art results on the MATH dataset, matching DeepSeekMath with only 3% of the pretraining tokens. The SLM approach also increases efficiency and performance when continual pretraining is done on general tokens, resulting in a 6.8% average enhancement across 15 diverse tasks. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper introduces a new way to train language models called Selective Language Modeling (SLM). Instead of training on every word, Rho-1 only trains on words that are important for understanding what we want the model to do. This helps the model learn faster and more accurately. The results show that Rho-1 is much better at doing math problems than other models that were trained in a different way. It’s also more efficient and can be used for many tasks. |
Keywords
» Artificial intelligence » Few shot » Language model » Pretraining
