{"record": {"aliases": ["Pancasila-Dilemmas: Evaluating Large Language Models on Indonesian Human Value Dilemmas Grounded in Pancasila"], "artifacts": [{"id": "arxiv:2607.18066", "kind": "paper", "url": "https://arxiv.org/abs/2607.18066"}, {"id": "gh:tjunlp-lab/pancasila-dilemmas", "kind": "repo", "url": "https://github.com/tjunlp-lab/Pancasila-Dilemmas"}], "categories": ["General AI", "Language & Knowledge", "cs.CL"], "description": {"en": "An evaluation dataset of 1,834 dilemma questions from Indonesian news that measures LLM alignment with the five Pancasila values, scored by PMS and MVAS against answers from Indonesian annotators."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.18066", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.18066", "title": "Pancasila-Dilemmas"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2607.18066", "modality": null, "name": "Pancasila-Dilemmas", "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_pancasila-dilemmas-evaluating-large-langua_ef785f1c", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "5ac727bd9c640b85351252db094d6471846da108ada74c3faa39e226e67b54c2", "relation": "evaluates_only", "review_model": "claude-code-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T20:45:26.375696Z", "source_url": "https://arxiv.org/abs/2607.18066"}, "publisher": null, "released": "2026-07-20", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.18066", "source_url": "https://arxiv.org/abs/2607.18066"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-18066", "source": "claire_radar", "source_benchmark_id": "2607.18066", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["Pancasila-Dilemmas: Evaluating Large Language Models on Indonesian Human Value Dilemmas Grounded in Pancasila"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "A culturally specific alignment dataset with a public repo and 50 evaluated models. Its audience is regional and niche.", "score": 35}, "capabilities": [], "confidence": "Medium", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T20:45:26.375696Z", "inputHash": "193af94c6543286602e8d80795f7f60b57d6a37efa1fcaf9c192a524ba8d0e6a", "model": "claude-code-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "Pancasila-Dilemmas: Evaluating Large Language Models on Indonesian Human Value Dilemmas Grounded in Pancasila", "canonicalNameSource": "paper_title", "decisionReason": "It has a formal name, a defined task, concrete metrics (PMS and MVAS), and a public GitHub dataset link.", "evaluationEvidence": {"reuse": "The dataset is publicly available at https://github.com/tjunlp-lab/Pancasila-Dilemmas", "scoring": "less than 0.5 Probability Match Score (PMS) and 0.72 Max-Vote Agreement Score (MVAS)", "task": "an evaluation dataset of 1,834 questions derived from Indonesian news, classified by 5 values of Pancasila"}, "model": "claude-code-default", "reviewedAt": "2026-09-24T20:45:26.375696Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "An evaluation dataset of 1,834 dilemma questions from Indonesian news that measures LLM alignment with the five Pancasila values, scored by PMS and MVAS against answers from Indonesian annotators.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "public artifact URL", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "In this paper, we introduce Pancasila-Dilemmas, an evaluation dataset of 1,834 questions derived from Indonesian news, classified by 5 values of Pancasila: Religion, Humanity, Unity, Democracy, and Social Justice."}, "familyId": "bmf_98ade5379b53", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_pancasila-dilemmas-evaluating-large-langua_ef785f1c", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/tjunlp-lab/Pancasila-Dilemmas", "data": null, "paper": "https://arxiv.org/abs/2607.18066", "pdf": "https://arxiv.org/pdf/2607.18066", "project": null, "report": "https://arxiv.org/abs/2607.18066"}, "metrics": [], "motivation": "The value alignment of large language models (LLMs) is crucial for ensuring responses align with human intention and value preferences.", "name": "Pancasila-Dilemmas", "oneLine": "An evaluation dataset of 1,834 dilemma questions from Indonesian news that measures LLM alignment with the five Pancasila values, scored by PMS and MVAS against answers from Indonesian annotators.", "paperTitle": "Pancasila-Dilemmas: Evaluating Large Language Models on Indonesian Human Value Dilemmas Grounded in Pancasila", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.18066", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "publishers": [{"name": "tjunlp-lab", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/tjunlp-lab/Pancasila-Dilemmas"}], "readiness": "Runnable", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-07-20", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Supryadi", "Irfan", "Julianti", "Darren Keanly Martin", "Jayvin Fernando", "Yuqi Ren", "Deyi Xiong"], "categories": ["cs.CL"], "id": "2607.18066", "title": "Pancasila-Dilemmas: Evaluating Large Language Models on Indonesian Human Value Dilemmas Grounded in Pancasila", "type": "arxiv", "url": "https://arxiv.org/abs/2607.18066"}, "sourceUpdatedAt": "2026-07-20T15:33:24Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "Most value-alignment evaluations focus on Western or universal values. This dataset offers a public, country-specific test for Indonesian values, and results for 50 LLMs show large gaps, especially on Religion and Unity."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
