A/B verdict tool for recorded optimization-run ledgers. Use when the user wants two (or more) runs of the GPU-optimization loop compared and judged: knowledge-base ON vs OFF ablations, "which config/side won?", "did KB actually help — real effect or noise?", round-by-round or convergence-speed analy
A/B verdict tool for recorded optimization-run ledgers. Use when the user wants two (or more) runs of the GPU-optimization loop compared and judged: knowledge-base ON vs OFF ablations, "which config/side won?", "did KB actually help — real effect or noise?", round-by-round or convergence-speed analysis, or a sweep summary across several run pairs. Trigger equally on Korean phrasing (온오프 비교, 어느 쪽이