Series

The LLM Training Lifecycle

From raw text to instruction-following assistant — a stage-by-stage breakdown of how large language models are pre-trained, mid-trained, fine-tuned, and aligned, and what changes at each stage.

Fundamentals

llm-trainingpretrainingsftrlhfdpoalignment

1
LLM Training Stages: Pre-training, Mid-training, SFT, RL, and DPO
What actually happens at each stage of training a large language model — what data, what objective, what the model learns, and why the stages are ordered the way they are.
2025-06-20
1
LLM Training Stages: Pre-training, Mid-training, SFT, RL, and DPO
What actually happens at each stage of training a large language model — what data, what objective, what the model learns, and why the stages are ordered the way they are.
2025-06-20
2
Why Long-Context Mid-Training Is Its Own Stage: RoPE Scaling, Attention Entropy, and Lost-in-the-Middle
Training a model to handle 128K context isn't just running inference on longer sequences — it requires a dedicated mid-training phase because positional encoding, attention entropy, and information retrieval all break in distinct ways beyond the training window.
2025-06-20
2
Why Long-Context Mid-Training Is Its Own Stage: RoPE Scaling, Attention Entropy, and Lost-in-the-Middle
Training a model to handle 128K context isn't just running inference on longer sequences — it requires a dedicated mid-training phase because positional encoding, attention entropy, and information retrieval all break in distinct ways beyond the training window.
2025-06-20
3
The LLM Alignment Pipeline: SFT, Reward Models, and RL End to End
Training a helpful, harmless, honest LLM requires three sequential stages that each build on the previous one. Here's how SFT, reward modeling, and RL fit together as a system — and where each stage can fail.
2024-06-19
3
The LLM Alignment Pipeline: SFT, Reward Models, and RL End to End
Training a helpful, harmless, honest LLM requires three sequential stages that each build on the previous one. Here's how SFT, reward modeling, and RL fit together as a system — and where each stage can fail.
2024-06-19