How Large Language Models Work
Edward Raff, Drew Farris +1
Description An examination of the internal mechanics of Large Language Models (LLMs) such as GPT and Gemini. It covers the technical transition from a...
Build a Reasoning Model (From Scratch)
Sebastian Raschka
Book Metadata
Publisher
Simon and Schuster
Published
2026
Duration
6 hr 15 min
ISBN
9781638358206
Genres
Computers
About This Book
Description A deep dive into the architecture and implementation of AI models capable of logical deduction and multi-step reasoning. It explains how to move beyond basic pattern matching to systems that can "think" through complex queries. Key Topics: Chain-of-thought prompting, reinforcement learning from human feedback (RLHF), and the integration of symbolic logic with neural networks. About the Technology: Reasoning models represent the next stage of Large Language Models (LLMs), focusing on accuracy and verifiable logic in specialized domains. About the Book: Follows a code-centric approach to building a model that performs structured reasoning tasks without relying on massive, opaque datasets. About the Reader: For machine learning engineers and researchers interested in the mechanics of AI cognition and logic. About the Author: Sebastian Raschka is a machine learning researcher and the author of several widely-used Python machine learning textbooks.
AI Summary
A hands‑on manual for building AI models that can perform logical deduction, chain‑of‑thought reasoning, and symbolic‑neural integration, showing how to create verifiable reasoning systems without massive opaque datasets.
Technical Reference
Summary
A hands‑on manual for building AI models that can perform logical deduction, chain‑of‑thought reasoning, and symbolic‑neural integration, showing how to create verifiable reasoning systems without massive opaque datasets.
Key Themes
- Chain‑of‑thought prompting for stepwise logic
- Reinforcement Learning from Human Feedback (RLHF)
- Symbolic‑neural hybrid architectures
- Interpretability and logical soundness evaluation
Ideal Reader
Machine learning engineers, AI researchers, and graduate students who want to design reasoning‑focused models rather than generic LLMs.
Details
stylecode‑first with explanatory sidebars and mathematical grounding
levelintermediate to expert, assuming solid ML fundamentals and Python proficiency
commitmentdeep dive; serves as both tutorial and reference manual
Why It Matters
The book provides a reproducible, code‑centric pathway to building AI systems that can reason transparently, addressing the growing demand for trustworthy models in high‑stakes domains such as healthcare, law, and scientific research.
Related Reading
Deep Learning by Ian Goodfellow, Yoshua Bengio, and Aaron CourvilleHuman Compatible by Stuart RussellReinforcement Learning: An Introduction by Sutton & BartoLogical Neural Networks literature by Artur d'Avila Garcez et al.
auto-generated