{"record": {"aliases": ["When Similar Means Different: Evaluating LLMs on Arabic--Hebrew Cognates"], "artifacts": [{"id": "arxiv:2606.13218", "kind": "paper", "url": "https://arxiv.org/abs/2606.13218"}], "authors": ["Junhong Liang", "Noor Abo Mokh", "Bashar Alhafni"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.13218", "source_url": "https://arxiv.org/abs/2606.13218"}, "categories": ["General AI", "Language & Knowledge", "cs.CL"], "description": {"en": "A benchmark for evaluating Arabic–Hebrew cognate identification and semantic disambiguation across multiple input representations."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.13218", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.13218", "title": "SemCog Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2606.13218", "modality": null, "name": "SemCog Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_when-similar-means-different-evaluating-ll_f658dbc8", "origin_source": "arxiv", "recognition_confidence": "0.6", "record_sha256": "a957cbcb8a6ad751576a4ff7ea810321d99a18abc964acff77a438170843bd42", "relation": "introduces", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T18:27:54.325266Z", "source_url": "https://arxiv.org/abs/2606.13218"}, "publisher": null, "released": "2026-06-11", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.13218", "source_url": "https://arxiv.org/abs/2606.13218"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-13218", "source": "claire_radar", "source_benchmark_id": "2606.13218", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["When Similar Means Different: Evaluating LLMs on Arabic--Hebrew Cognates"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The public release and focused multilingual reasoning challenge support near-term reuse.", "score": 3}, "capabilities": [], "confidence": "Medium", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T18:27:54.325266Z", "inputHash": "c5862c2b3643c6763ea15698083e710d64abc33758ecf38a94f91389fe750934", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "we introduce SemCog Bench", "canonicalNameSource": "abstract", "decisionReason": "The supplied abstract provides a formal name, defined tasks, accuracy-based comparison, and an explicit public code-and-data reuse path.", "evaluationEvidence": {"reuse": "Our code and data are publicly available.", "scoring": "models achieve high accuracy on true cognates, performance drops sharply on false friends and loanwords", "task": "a curated benchmark of 1,858 Arabic--Hebrew word pairs with sentence-level annotations for cognate identification and semantic disambiguation"}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T18:27:54.325266Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "A benchmark for evaluating Arabic–Hebrew cognate identification and semantic disambiguation across multiple input representations.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "To evaluate this capability, we introduce SemCog Bench, a curated benchmark of 1,858 Arabic--Hebrew word pairs with sentence-level annotations for cognate identification and semantic disambiguation."}, "familyId": "bmf_a19a5ee53493", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_when-similar-means-different-evaluating-ll_f658dbc8", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2606.13218", "pdf": "https://arxiv.org/pdf/2606.13218", "project": null, "report": "https://arxiv.org/abs/2606.13218"}, "metrics": [], "motivation": "Arabic and Hebrew, as closely related Semitic languages, share a substantial lexicon of true cognates, misleading false friends, and modern loanwords.", "name": "SemCog Bench", "oneLine": "A benchmark for evaluating Arabic–Hebrew cognate identification and semantic disambiguation across multiple input representations.", "paperTitle": "When Similar Means Different: Evaluating LLMs on Arabic--Hebrew Cognates", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.13218", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "readiness": "Paper only", "recognitionConfidence": 0.6, "relation": "introduces", "releasedAt": "2026-06-11", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Junhong Liang", "Noor Abo Mokh", "Bashar Alhafni"], "categories": ["cs.CL"], "id": "2606.13218", "title": "When Similar Means Different: Evaluating LLMs on Arabic--Hebrew Cognates", "type": "arxiv", "url": "https://arxiv.org/abs/2606.13218"}, "sourceUpdatedAt": "2026-06-11T11:33:09Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "It measures whether language models can distinguish genuine cross-lingual semantic relationships from misleading surface-form similarity."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
