Chapter 1: LLM Behaviour

การทำความเข้าใจพฤติกรรมเชิงลึกของโมเดล (Model Dynamics)

Key Topics

In-context Learning

• How LLMs learn from examples in the prompt
• Few-shot vs Zero-shot patterns
• Context window utilization

Long-context Window Management

• Handling long documents and conversations
• Attention mechanisms for long contexts
• Memory efficiency techniques

Hallucination Reduction

• Understanding why hallucinations occur
• Techniques to minimize false outputs
• Verification and validation strategies

Determinism & Consistency

• Controlling output variability
• Temperature and top-p settings
• Reproducible results in production

Practical Applications

• Binding LLM outputs to logic in production systems
• Creating reliable AI-driven workflows
• Managing model expectations