/exp-design

Plan sensitivity analysis by executing an iterative ablation loop that triggers simplification and re-planning, quantifying success criteria across hyperparameter sweeps and multi-seed main runs to measure how variations affect your research outcomes.

Mechanism-first experiment planning is an approach that validates hypotheses through deep analysis of intermediate quantities, needed when you must select standard benchmarks and quantify success criteria across ablation and generalization runs.

Plan a complete research experiment suite by generating method candidates, selecting benchmarks and metrics, and running an ablation loop that covers sensitivity analysis, main validation, optional generalization, and multi-seed main runs.

Yes, experiment design requires reading and writing to a structured wiki project, generating per-block experiment pages, rebuilding context, adding tested_by graph edges, and logging a master design report for downstream execution.

Structure reproducible artifacts by generating per-block experiment pages within a wiki project, updating graph edges and idea linking, and producing a master design report that prepares the full experiment suite for downstream execution.

Yes, the ablation loop can trigger simplification and iterative re-planning of your experiment suite, allowing you to adjust method candidates and benchmark specifications based on sensitivity analysis and intermediate quantity results.