{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.01552", "kind": "paper", "url": "https://arxiv.org/abs/2606.01552"}, {"id": "gh:rabbitrose/rolecde", "kind": "repo", "url": "https://github.com/rabbitrose/RoleCDE"}], "authors": ["Huayi Lai", "Shichao Song", "Simin Niu", "Hanyu Wang", "Jiawei Yang", "Zhouxing Wang", "Zhiqiang Yin", "Xun Liang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.01552", "source_url": "https://arxiv.org/abs/2606.01552"}, "categories": ["General AI", "Language & Knowledge", "Safety"], "description": {"en": "RoleCDE evaluates role-playing agents under structured conflicts between role-specific values and alignment-oriented constraints. It comprises approximately 8,000 role profiles and 240,000 dilemma instances across three difficulty levels and eight role categories, with scoring via LLM-as-a-judge."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.01552", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.01552", "title": "RoleCDE"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Renmin University of China"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.01552", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.01552", "source_url": "https://openalex.org/W7163140343"}, "key": "claire-radar:2606.01552", "modality": null, "name": "RoleCDE", "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_rolecde_7c0f211a", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "2082540e13468dfb9a5ab12e919b07e5cc06e0d1f0b0080d97d2d94fe44124e1", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.01552"}, "publisher": null, "released": "2026-06-01", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.01552", "source_url": "https://arxiv.org/abs/2606.01552"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-01552", "source": "claire_radar", "source_benchmark_id": "2606.01552", "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/rabbitrose/RoleCDE", "githubScope": "benchmark_repo", "githubStars": 1, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2606.01552", "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": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "c20da1b38ee79974202a55eca82e8b9607810ea76939e9cab77f375de94bd6df", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "RoleCDE evaluates role-playing agents under structured conflicts between role-specific values and alignment-oriented constraints. It comprises approximately 8,000 role profiles and 240,000 dilemma instances across three difficulty levels and eight role categories, with scoring via LLM-as-a-judge.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "To address this gap, we introduce RoleCDE, the first benchmark designed to evaluate RPAs under structured conflicts between role-specific values and alignment-oriented constraints."}, "familyId": "bmf_ce6bd9dfe8c8", "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/rabbitrose/RoleCDE", "url": "https://github.com/rabbitrose/RoleCDE"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/rabbitrose/RoleCDE", "url": "https://huggingface.co/papers/2606.01552"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/rabbitrose/RoleCDE", "url": "https://arxiv.org/abs/2606.01552"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/rabbitrose/RoleCDE.", "url": "https://arxiv.org/abs/2606.01552"}], "scope": "benchmark_repo", "url": "https://github.com/rabbitrose/RoleCDE", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_rolecde_7c0f211a", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/rabbitrose/RoleCDE", "data": null, "hfPaper": "https://huggingface.co/papers/2606.01552", "paper": "https://arxiv.org/abs/2606.01552", "pdf": "https://arxiv.org/pdf/2606.01552", "project": null, "report": "https://arxiv.org/abs/2606.01552"}, "metrics": [], "motivation": "Role-playing agents(RPAs) are widely used to steer large language models(LLMs) toward role-consistent behavior, yet existing benchmarks mainly evaluate surface-level fidelity and offer limited insight into decision making under role-alignment value conflicts.", "name": "RoleCDE", "oneLine": "RoleCDE evaluates role-playing agents under structured conflicts between role-specific values and alignment-oriented constraints. It comprises approximately 8,000 role profiles and 240,000 dilemma instances across three difficulty levels and eight role categories, with scoring via LLM-as-a-judge.", "paperTitle": "RoleCDE:Benchmarking and Mitigating Role-Alignment Trade-offs in Role-Playing Agents", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.01552", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "RoleCDE Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/rabbitrose/RoleCDE"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-01", "source": {"authors": ["Huayi Lai", "Shichao Song", "Simin Niu", "Hanyu Wang", "Jiawei Yang", "Zhouxing Wang", "Zhiqiang Yin", "Xun Liang"], "categories": ["cs.AI"], "id": "2606.01552", "title": "RoleCDE:Benchmarking and Mitigating Role-Alignment Trade-offs in Role-Playing Agents", "type": "arxiv", "url": "https://arxiv.org/abs/2606.01552"}, "sourceUpdatedAt": "2026-06-01T01:56:27Z", "topics": ["Safety"], "venueAttempts": [], "whyItMatters": "Existing benchmarks focus on surface fidelity and lack coverage of decision-making under role-alignment value conflicts. RoleCDE provides a systematic evaluation of how agents resolve such conflicts, revealing systematic biases and offering a tool for improving alignment and role consistency."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
