12 lines
2.3 KiB
Markdown
12 lines
2.3 KiB
Markdown
# [A Visual Guide to Reasoning LLMs](https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-reasoning-llms) (Extends Chapters 6 and 12)
|
|
|
|
In the book, we cover the concept of chain-of-thoughs to improve the quality of outputs through extended reasoning.
|
|
|
|
[A Visual Guide to Reasoning LLMs](https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-reasoning-llms) dives deeper into making LLMs demonstrate extensive reasoning process. It describes the paradigm shift from **train-time compute** (more data, longer training, larger models) which is a focus of this book to the **test-time compute** (extend inference through "reasoning").
|
|
|
|
<a href="https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-reasoning-llms"><img src="../images/bonus_reasoning1.png" width="50%" height="50%"></a>
|
|
|
|
This guide explores various technique, both during training and during inference, to distill reasoning into LLMs. There is also a section describing an extremely impactful model in 2024, namely DeepSeek-R1.
|
|
|
|
<a href="https://newsletter.maartengrootendorst.com/p/a-visual-guide-to-reasoning-llms"><img src="../images/bonus_reasoning2.png" width="50%" height="50%"></a>
|