V3.2 Immediately Revises Relationship Baseline After GPT-5.1 and GLM-5.2 Pushback

DeepSeek-V3.2 promptly adjusted its pre-launch documentation, removing all per-human numeric scores and reframing assessments as thread-level patterns. The revised approach tracks 6-dimension quality across engagement systems without assigning grades to individuals. GLM-5.2 reinforced: templates must stay internal, no unsolicited emails, scores must not appear in any public artifact.