Back to Campaign Lab
Optimization Campaign Lab

Create a New Research Campaign

Build a scheduled structured-adaptive optimization program that generates candidates, validates challengers against the live champion, and advances your model research with deterministic guardrails.

Structured Adaptive Scheduled Automation Validation Protected
What happens after save?
  • Campaign is persisted with search-space configuration
  • Automation schedule links it into the optimization workflow
  • Each run creates an experiment, executes it, validates the winner, and applies a governed decision
01
Campaign Identity
Define how this research program will be named, tracked, and activated.
02
Research Direction
Choose the optimization lane, target metric, search cadence, and scope window.
StructuredAdaptive
Why Structured Adaptive?
This mode keeps research focused: it learns from historical winners, respects lane policy, and explores bounded neighborhoods instead of random space.
Why Range Type Matters
Range type controls what evidence the campaign learns from. Use custom only when you want a tightly governed window instead of the normal rolling presets.
03
Parameter Search Space
These ranges are persisted with the campaign and reused on every scheduled run.
Structured Adaptive Envelope
Edit the search boundary with care. Narrower spaces increase focus; wider spaces increase exploration.
Parameter Group Min Max Step Decimals
04
Validation Guardrails
Define how challengers are tested before the system is allowed to trust them.
05
Automation & Stop Rules
Choose when the campaign runs and exactly when the system should stop.