openai/gpt-chat-latest → B (9:1)

Side B fixes user-visible vote comparison highlighting by correcting ranking-group gradient calculation, replacing the history ratio bars with slider-based rendering that preserves left/right polarity, updating UI logic/CSS, and adding multiple tests covering orientation, slider mapping, and ranking invariants. Side A only removes a now-redundant zero-ratio early return in the reducer and updates the associated test expectations, a comparatively small cleanup with limited lasting impact.