Build a Reasoning Model (from Scratch)

Sebastian Raschka is a machine learning and AI researcher. He is known for making complex topics easy to understand. With a focus on practical application, Sebastian brings a clear, step-by-step approach to building advanced models. He translates deep learning theory into hands-on projects that help readers build their own AI systems from scratch.

• 1 UNDERSTANDING REASONING MODELS 2 GENERATING TEXT WITH A PRE-TRAINED LLM 3 EVALUATING REASONING MODELS 4 IMPROVING REASONING WITH INFERENCE-TIME SCALING 5 INFERENCE-TIME SCALING VIA SELF-REFINEMENT 6 TRAINING REASONING MODELS WITH REINFORCEMENT LEARNING 7 IMPROVING GRPO FOR REINFORCEMENT LEARNING 8 DISTILLING REASONING MODELS FOR EFFICIENT REASONING APPENDICES APPENDIX A: REFERENCES AND FURTHER READING APPENDIX B: EXERCISE SOLUTIONS APPENDIX C: QWEN3 LLM SOURCE CODE APPENDIX D: USING LARGER LLMS APPENDIX E: BATCHED INFERENCE APPENDIX F: COMMON APPROACHES TO MODEL EVALUATION

“Truly inspirational! It motivates you to put your new skills into action.” Benjamin Muskalla, Senior Engineer, GitHub  “The most understandable and comprehensive explanation of language models yet! It's unique and practical teaching style achieves a level of understanding you can’t get any other way.” Cameron Wolfe, Senior Scientist, Netflix