appendix-a

appendix A  References and further reading

 

A.1 Chapter 1: Understanding reasoning models

A.1.1 References

This is the announcement article for OpenAI's o1 model, which is regarded as the first LLM-based reasoning model:

DeepSeek-R1 is the first open source reasoning model to be accompanied by a comprehensive technical report showing that reasoning emerges from reinforcement learning with verifiable rewards (a topic covered in more detail in chapter 5):

  • DeepSeek-AI, Daya Guo, Dejian Yang, et al., “DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning” (Jan. 22, 2025), https://arxiv.org/abs/2501.12948

This is the OpenAI CEO’s comment on the reasoning (“chain-of-thought”) capabilities of future models: “We will next ship GPT-4.5, the model we called Orion internally, as our last non-chain-of-thought model.”

A research paper by AI researchers at Apple found that reasoning models are sophisticated (but very capable) pattern matchers:

A.1.2 Further reading

A.2 Chapter 2: Generating text with a pretrained LLM

A.2.1 References

A.2.2 Further reading

A.3 Chapter 3: Evaluating reasoning models

A.3.1 References

A.3.2 Further reading

A.4 Chapter 4: Improving reasoning with inference-time scaling

A.4.1 References

A.4.2 Further reading

A.5 Chapter 5: Inference-time scaling via self-refinement

A.5.1 References

A.5.2 Further reading

A.6 Chapter 6: Training reasoning models with reinforcement learning

A.6.1 References