Reasoning Models and DeepSeek R1 from scratch

Graphics in 5 Minutes · Advanced ·🧠 Large Language Models ·1y ago

Skills: LLM Foundations80%

Key Takeaways

Explains reasoning models like DeepSeek R1, including large language models, math problems, and Chain-of-Thought Prompting

Original Description

How do reasoning models like DeepSeek R1 work? A short cartoon that explains reasoning models. 0:05 - large language models 0:25 - math problems 0:55 - superhuman performance 1:10 - AlphaZero 2:52 - Math as a game 3:16 - DeepSeek R1-Zero 3:38 - GRPO 4:34 - Chain-of-Thought Prompting (CoT) 4:51 - think-answer template 7:21 - DeepSeek R1 7:54 - GPQA 8:23 - towards superhuman performance

Watch on YouTube ↗ (saves to browser)

Sign in to unlock AI tutor explanation · ⚡30

More on: LLM Foundations

View skill →

Getting Started with Vertex AI Gemini 1.5 Flash

I TRAINED AN AI TO SOLVE 2+2 (w/ Live Coding)

I TRAINED AN AI TO SOLVE 2+2 (w/ Live Coding)

Open Assistant Live Coding (Open-Source ChatGPT Replication)

Open Assistant Live Coding (Open-Source ChatGPT Replication)

How To Create A Chatbot Using Python In 5 Minutes | Build Chatbot With Python | Simplilearn

How To Create A Chatbot Using Python In 5 Minutes | Build Chatbot With Python | Simplilearn

How to use the ChatGPT API with Python!!

How to use the ChatGPT API with Python!!

Nicholas Renotte

Gemini 2.5: Create an interactive plot of economic data

Gemini 2.5: Create an interactive plot of economic data

Google DeepMind

Related Reads

ai books beginners: what careful buyers should check

Learn what to check when buying AI books as a beginner to make informed purchasing decisions

Knowledge Distillation — Deep Dive + Problem: Template Matching Score

Learn Knowledge Distillation to compress Large Language Models and improve deployment efficiency

Integrating Open-Weight LLMs via a Unified API: A Practical Guide

Learn to integrate open-weight LLMs using a unified API for indie devs and small teams

Testing the Reliability of ChatGPT for Text Annotation and Classification: ACautionary Remark

Learn to test the reliability of ChatGPT for text annotation and classification tasks, understanding its limitations and potential biases.

Chapters (12)

0:05 large language models

0:25 math problems

0:55 superhuman performance

1:10 AlphaZero

2:52 Math as a game

3:16 DeepSeek R1-Zero

3:38 GRPO

4:34 Chain-of-Thought Prompting (CoT)

4:51 think-answer template

7:21 DeepSeek R1

7:54 GPQA

8:23 towards superhuman performance

5 Levels of AI Agents - From Simple LLM Calls to Multi-Agent Systems

Dave Ebbelaar (LLM Eng)