{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.28315", "kind": "paper", "url": "https://arxiv.org/abs/2605.28315"}, {"id": "gh:jasonnlp/hardmtbench", "kind": "repo", "url": "https://github.com/jasonNLP/HardMTBench"}], "authors": ["Zheng Li", "Mao Zheng", "Mingyang Song", "Tianxiang Fei"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.28315", "source_url": "https://arxiv.org/abs/2605.28315"}, "categories": ["Factuality", "General AI", "Language & Knowledge", "cs.CL"], "description": {"en": "HardMTBench is a difficulty-aware diagnostic benchmark for Chinese-English domain translation, covering 12 domains with 20,000 directional test items and annotated hardness scores based on domain knowledge, translation difficulty, and terminology load."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.28315", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.28315", "title": "HardMTBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.28315", "modality": null, "name": "HardMTBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_hardmtbench_3eaf8a99", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "f5483b27e2b2387b1b269e149df567efdc4d200cee43f304144225d96269a93f", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.28315"}, "publisher": null, "released": "2026-05-27", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.28315", "source_url": "https://arxiv.org/abs/2605.28315"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-28315", "source": "claire_radar", "source_benchmark_id": "2605.28315", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/jasonNLP/HardMTBench", "githubScope": "benchmark_repo", "githubStars": 2, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2605.28315", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "fresh"}, "capabilities": ["Factuality"], "confidence": "High", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "b9f7f5ff378782dbc14a56d78cc76f8200856ce363e51fbafcf0000647111757", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "HardMTBench is a difficulty-aware diagnostic benchmark for Chinese-English domain translation, covering 12 domains with 20,000 directional test items and annotated hardness scores based on domain knowledge, translation difficulty, and terminology load.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce HardMTBench, a difficulty-aware diagnostic benchmark for bidirectional Chinese-English domain translation."}, "familyId": "bmf_91528a6e04ed", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/jasonNLP/HardMTBench", "url": "https://github.com/jasonNLP/HardMTBench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/jasonNLP/HardMTBench", "url": "https://huggingface.co/papers/2605.28315"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/jasonNLP/HardMTBench.", "url": "https://arxiv.org/abs/2605.28315"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/jasonNLP/HardMTBench", "url": "https://arxiv.org/abs/2605.28315"}], "scope": "benchmark_repo", "url": "https://github.com/jasonNLP/HardMTBench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_hardmtbench_3eaf8a99", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/jasonNLP/HardMTBench", "data": null, "hfPaper": "https://huggingface.co/papers/2605.28315", "paper": "https://arxiv.org/abs/2605.28315", "pdf": "https://arxiv.org/pdf/2605.28315", "project": null, "report": "https://arxiv.org/abs/2605.28315"}, "metrics": [], "motivation": "General-purpose machine translation benchmarks such as FLORES-200 have reached a saturation regime on Chinese-English pairs, where modern large language models cluster within a narrow band of high scores.", "name": "HardMTBench", "oneLine": "HardMTBench is a difficulty-aware diagnostic benchmark for Chinese-English domain translation, covering 12 domains with 20,000 directional test items and annotated hardness scores based on domain knowledge, translation difficulty, and terminology load.", "paperTitle": "HardMTBench: Stress-Testing Chinese-English Translation on Knowledge-Intensive Domains", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.28315", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-05-27", "source": {"authors": ["Zheng Li", "Mao Zheng", "Mingyang Song", "Tianxiang Fei"], "categories": ["cs.CL"], "id": "2605.28315", "title": "HardMTBench: Stress-Testing Chinese-English Translation on Knowledge-Intensive Domains", "type": "arxiv", "url": "https://arxiv.org/abs/2605.28315"}, "sourceUpdatedAt": "2026-05-27T11:16:19Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "Addresses the saturation of general MT benchmarks on Chinese-English by widening score separation, exposing domain-specific weaknesses in knowledge-intensive areas that quality-only metrics miss."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
