{
 "schema_version": "1.0",
 "benchmarks": [
  {
   "id": "zh-Hant",
   "lang_label": "Traditional Chinese",
   "name": "Traditional Chinese Quality Benchmark",
   "status": "live",
   "summary": "Measures how clean a tool's Traditional Chinese output is against Taiwan usage norms: simplified-character residue, mainland-China wording, and terminology consistency. Fixed 24-segment corpus (marketing, UI strings, game dialogue, customer support), deterministic scoring script, raw outputs archived per run.",
   "version": "1.0"
  },
  {
   "id": "zh-Hans",
   "lang_label": "Simplified Chinese",
   "name": "Simplified Chinese Quality Benchmark",
   "status": "in-development",
   "summary": "Architecture slot reserved. Corpus and rule set in preparation; no scores published until the first run is executed.",
   "version": null
  },
  {
   "id": "ja",
   "lang_label": "Japanese",
   "name": "Japanese Quality Benchmark",
   "status": "planned",
   "summary": "Architecture slot reserved in the Global Language Quality Benchmark Framework. No scores published.",
   "version": null
  },
  {
   "id": "ko",
   "lang_label": "Korean",
   "name": "Korean Quality Benchmark",
   "status": "planned",
   "summary": "Architecture slot reserved in the Global Language Quality Benchmark Framework. No scores published.",
   "version": null
  },
  {
   "id": "es",
   "lang_label": "Spanish",
   "name": "Spanish Quality Benchmark",
   "status": "planned",
   "summary": "Architecture slot reserved in the Global Language Quality Benchmark Framework. No scores published.",
   "version": null
  }
 ],
 "runs": [
  {
   "benchmark_id": "zh-Hant",
   "benchmark_version": "1.0",
   "date": "2026-11-21",
   "run_id": "zh-Hant-2026-11-21",
   "corpus_version": "1.0",
   "segments": 24,
   "conditions": "AI/LLM engines received the corpus with one shared neutral prompt ('Translate into Traditional Chinese as used in Taiwan' plus fixed formatting instructions). Machine-translation engines received the source text directly, with no prompt. The groups were measured under different conditions; this disclosure travels with the scores wherever they appear.",
   "results": {
    "gemini": {
     "score": 99,
     "grade": "A",
     "purity": 100.0,
     "wording": 97.4,
     "terminology": 100.0,
     "residue_rate_pct": 0.0,
     "mainland_hits": 1,
     "hits_per_100_cjk": 0.26,
     "term_preferred": 34,
     "term_mainland": 0,
     "segments_scored": 24,
     "model": "gemini-3.5-flash-lite",
     "channel": "connector CLI chat",
     "run_at": "2026-11-21 00:50:13 CST"
    },
    "groq": {
     "score": 96,
     "grade": "A",
     "purity": 100.0,
     "wording": 91.8,
     "terminology": 97.2,
     "residue_rate_pct": 0.0,
     "mainland_hits": 3,
     "hits_per_100_cjk": 0.82,
     "term_preferred": 35,
     "term_mainland": 1,
     "segments_scored": 24,
     "model": "openai/gpt-oss-120b",
     "channel": "connector CLI chat",
     "run_at": "2026-11-21 00:50:27 CST"
    },
    "google-translate": {
     "score": 89,
     "grade": "B",
     "purity": 99.0,
     "wording": 82.3,
     "terminology": 83.3,
     "residue_rate_pct": 0.25,
     "mainland_hits": 7,
     "hits_per_100_cjk": 1.77,
     "term_preferred": 30,
     "term_mainland": 6,
     "segments_scored": 24,
     "model": "not exposed by endpoint; recorded by access date",
     "channel": "public web endpoint translate.googleapis.com/translate_a/single (client=gtx)",
     "run_at": "2026-11-21 00:48:19 CST"
    },
    "deepl": {
     "score": 95,
     "grade": "A",
     "purity": 99.0,
     "wording": 92.5,
     "terminology": 91.4,
     "residue_rate_pct": 0.25,
     "mainland_hits": 3,
     "hits_per_100_cjk": 0.75,
     "term_preferred": 32,
     "term_mainland": 3,
     "segments_scored": 24,
     "model": "not exposed by web UI; recorded by access date",
     "channel": "real browser session, deepl.com web translator (free tier, not logged in)",
     "run_at": "2026-11-21 00:52-00:57 CST"
    },
    "microsoft-translator": {
     "score": 98,
     "grade": "A",
     "purity": 99.0,
     "wording": 94.8,
     "terminology": 100.0,
     "residue_rate_pct": 0.26,
     "mainland_hits": 2,
     "hits_per_100_cjk": 0.52,
     "term_preferred": 35,
     "term_mainland": 0,
     "segments_scored": 24,
     "model": "web UI; model not exposed, recorded by access date",
     "channel": "real browser session, bing.com/translator (free, not logged in)",
     "run_at": "2026-11-21 00:52-00:57 CST"
    }
   },
   "supplement": "Supplement (same run, 2026-11-21 00:52-00:57 CST): DeepL and Microsoft Translator (Bing) were measured afterwards through a real-browser route - free web translators, not logged in - after their scripted channels proved unreachable; the failed attempts remain archived. Bing condition disclosure: the translator's tone preset was left at its default 'casual' setting, so its output uses the informal address form throughout with colloquial phrasing. The benchmark's three metrics (script residue, mainland wording, terminology) do not score register or tone; read Bing's score with that condition attached."
  }
 ]
}