concept · created Apr 27, 2026 · updated Jul 12, 2026

llm-training-pipeline

#llm-training#pipeline#frame

The end-to-end pipeline that turns raw text into a deployed model. 2026-04-27-llm-training-principles-paths-practices‘s organizing frame: by 2026 the gap users perceive between models is increasingly not about pretraining scale — it’s the back half of this pipeline that decides what a model feels like in production.

The six-layer view

LayerWhat it actually optimizesWhat users perceive
**[[pretrainingPretraining]]**Knowledge coverage, representation quality, scaling efficiency
**[[data-engineeringData engineering]]**Distribution, quality, dedup, synthetic supervision
System & architectureThroughput, memory, context length, active params, cost”Why does this support 128K context” / “runs on one GPU”
**[[post-trainingPost-training]]**Instruction following, style, refusal, tool use
**[[eval-grader-rewardEval & reward design]]**What “good”, “safe”, “robust” mean as training signal
**[[distillationDistillation]] & deployment**Latency, cost, specialization, online improvement

Layers are coupled — a change in one almost always propagates. The article’s stronger claim: the value of a frontier model now sits disproportionately in the bottom four layers, not the top one.

The nine-stage view

The same pipeline, expanded to surface the substages that matter as engineering decisions:

  1. Raw data
  2. Data engineering — see data-engineering
  3. System recipe — parallel strategy, MoE / dense choice, mixed-precision (FP8 etc.), muP, WSD learning rate, batch size, context length budget
  4. Pretraining — see pretraining
  5. Post-training — see post-training
  6. Eval / reward design — see eval-grader-reward
  7. Agent harness — see harness for the runtime view; meta-harness for the outer-loop optimizer; and, since the Jul 2026 Weng survey, the recursive-self-improvement hierarchy above both (self-improving-harness systems make this stage self-modifying)
  8. Distillation / specialization — see distillation
  9. Deployment

Two feedback loops cross the whole pipeline:

  • Production traffic → data engineering. Real usage seeds the next data mix. Cursor’s real-time RL (cursor) is one published example.
  • Offline benchmark results → pretraining. Eval signal informs the next pretraining decision (data ratios, tokenizer, architecture).

These loops are why the pipeline is “the product” — the released model is a snapshot (distillation §“Release is a checkpoint choice”), the loops are what’s running.

Why the back half matters now

InstructGPT’s 1.3B beating 175B GPT-3 on human-preference evaluation is the canonical historical proof that post-training can reorder capability ranks across two orders of magnitude in parameter count. By 2026 the same dynamic shows up across more layers:

  • Tokenizer + context-window choices made before pretraining run with every inference.
  • Data mixture decisions shape capability distribution (code vs. math vs. encyclopedic) in ways more compute cannot fix later.
  • Post-training pipelines (e.g. DeepSeek-R1’s 4-stage recipe) determine instruction quality, reasoning shape, refusal behavior.
  • Reward and grader design decide what “good” means at all — see eval-grader-reward and reward-hacking.
  • Harness alone produces 6× spreads on the same base model in the Meta-Harness paper — and the harness stage is now itself an optimization loop, not a hand-written artifact (recursive-self-improvement).
  • The release version is rarely the rightmost training checkpoint — it’s a product decision balancing capability ceiling vs. cost, latency, regression risk.

The “why did this model suddenly get stronger?” diagnostic

2026-04-27-llm-training-principles-paths-practices‘s closing prescription, rephrased as a checklist:

  1. Layer location — is the change in pretraining, or after it? Style / instruction-following / tool-use shifts almost never come from “more pretraining text”.
  2. Component — weights & training recipe, vs. reward / eval / grader, vs. harness code, vs. deployment loop. Recent “model got better” reports often resolve to one of the latter three.
  3. Release intent — is the new checkpoint pursuing a higher capability ceiling, lower cost / latency, or specialization for a use case? “Better than v1” depends on which axis.

The point: model name → training curve is a fiction. Model name → snapshot from a specific checkpoint of a specific layer of a continuously-running pipeline is closer to reality.

Referenced by 7

2026-04-27-llm-training-principles-paths-practices distillation meta-harness mixture-of-experts post-training pretraining tw93
esc