{"basis": "Already published measurements from reviewed evaluators and official model release comparisons", "build_id": "20261108-af50cab8676ded11", "configuration": "Task-specific published protocols; distinct precision profiles; no selection by score and no cross-setting maxima", "dimensions": [{"key": "reasoning", "label": "推理求解", "weight": 25}, {"key": "agents", "label": "工具与任务执行", "weight": 25}, {"key": "coding", "label": "编程", "weight": 20}, {"key": "language_information", "label": "语言与信息处理", "weight": 20}, {"key": "multimodal_understanding", "label": "多模态理解", "weight": 10}], "evidence_revision": "c15bb3a91f3f244b75b5e16fd7298c9ba896fd3779a94348148c7b36a8768542", "interpretation": "A site-defined relative index of available published evidence; coverage differs between models; not an evaluation run by this website", "minimum_benchmark_families": 3, "minimum_capability_groups": 2, "normalization": "100*(worse_peers+0.5*other_tied_peers)/(peer_count-1)", "reference_population": "One frozen recent-model population per current benchmark version and comparable protocol; distinct model versions, no historical peers", "rounding": "Exact rational arithmetic; half-up two decimals; equal published scores share rank", "source_precedence": ["benchmark_owner_or_reviewed_evaluator", "own_official_release", "other_mainstream_official_comparison"], "version": "auratracer-public-composite-v9", "versions": "Latest effective versions established by benchmark maintainer or evaluator", "weighting": "Equal benchmark families within each covered capability group; covered base weights proportionally normalized", "window_months": 6}