Finetuning

Priors, not rules. Only firm guardrails: held-out eval you never train on, no leakage, trust evo's recorded numbers over the run's self-report. Override anything else against the gate.

Pick the technique by reward shape

Decide on the reward first, technique second. Choosing the comfortable technique over the matching one is the most common failure.

Reward shape	Technique
Verifiable (exact match, unit tests, parser-decidable)	RL (GRPO / RLOO / PPO), reward includes format, so the model learns to emit verifier-acceptable shape
Preference pairs (chosen vs rejected)	DPO / KTO / ORPO, cheaper than full RL, no rollouts
Demonstrations only (curated traces, chat data)	SFT, install format/tone/capability the base lacks
Have a scorer + want SFT stability	RFT, sample, filter by reward, SFT on survivors

"SFT-then-RL" is not a law. For a competent base model on a verifiable benchmark, RL-from-base often beats SFT-then-RL end-to-end.

Research the literature before the first commit

The decision tree above is the structural prior. The empirical answer for this model on this benchmark usually has a recent paper, blog, or HF Space recipe behind it -- and what beats baseline on a 4B base model in 2026 is not what the agent's pre-training data captures. Before picking the technique for exp_0001 (the first experiment after baseline), invoke evo:ideator with a literature brief:

Task(
    subagent_type="evo:ideator",
    prompt="brief=literature\n"
           "model_family=<e.g. Qwen3-4B-Base, Llama-3.1-8B-Base>\n"
           "benchmark=<name + URL/paper if known>\n"
           "objective=<one line: what beats baseline looks like>\n"
           "constraints=<budget, data sources allowed, gated models forbidden, etc>"
)