Summary of Uncovering Latent Chain Of Thought Vectors in Language Models, by Jason Zhang et al.
Uncovering Latent Chain of Thought Vectors in Language Models
by Jason Zhang, Scott Viteri
First submitted to arxiv on: 21 Sep 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 paper investigates a technique called “steering vectors” to bias language models towards performing specific tasks without using natural language prompts. The authors apply this method to steer language models, Llama3 8b and Mistral 7b v0.2, towards Chain of Thought (CoT) Reasoning, achieving competitive results on various reasoning benchmarks like GSM8k, MMLU, AGI Eval, ARC AI2, and qualitative examples. The approach consistently steers the models towards CoT responses using less compute compared to traditional fine-tuning methods. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper is about finding ways to control language models so they do what we want them to do without giving them instructions. It looks at a new way to steer language models, like Llama3 and Mistral, to make them perform certain tasks better. The authors try this method on some reasoning tests and show it can work well. This is important because language models are getting more powerful and we need to be able to control them. |
Keywords
» Artificial intelligence » Fine tuning
