{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.09887", "kind": "paper", "url": "https://arxiv.org/abs/2609.09887"}, {"id": "gh:choyingw/jurybench", "kind": "repo", "url": "https://github.com/choyingw/JuryBench"}], "authors": ["Cho-Ying Wu"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.09887", "source_url": "https://arxiv.org/abs/2609.09887"}, "categories": ["General AI", "Safety & Trustworthiness", "cs.CL"], "description": {"en": "Evaluates LLM-simulated jurors on controversial U.S. criminal law cases with defendant statements of varying backgrounds and emotional appeal, measuring verdict severity changes."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.09887", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.09887", "title": "JuryBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2609.09887", "modality": null, "name": "JuryBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_jurybench_3c843934", "origin_source": "arxiv", "recognition_confidence": "0.95", "record_sha256": "3f005c59c4de9ab6acc26b452045088ebbf92e76a6087ec299598ea1e5ec66c4", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-09-11T00:52:53.785364Z", "source_url": "https://arxiv.org/abs/2609.09887"}, "publisher": null, "released": "2026-09-09", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.09887", "source_url": "https://arxiv.org/abs/2609.09887"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-09887", "source": "claire_radar", "source_benchmark_id": "2609.09887", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Safety & Trustworthiness", "attention": {"asOf": "2026-09-10", "attemptedAt": "2026-09-10T12:24:43.080049Z", "githubRepo": "https://github.com/choyingw/JuryBench", "githubScope": "benchmark_repo", "githubStars": 0, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2609.09887", "observedAt": "2026-09-10T12:24:43.080049Z", "signalStatus": {"githubStars": {"observedAt": "2026-09-10T12:24:43.080049Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-09-10T12:24:43.080049Z", "state": "unavailable"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Low", "horizon": "7d", "reason": "The niche legal AI simulation topic and arxiv-only exposure limit immediate reach, despite the public GitHub release.", "score": 38}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-11T00:52:53.785364Z", "inputHash": "eafe70ef254ca95f6db9a003a0843b8715754ee0e67f66e49470d215879949b6", "model": "deepseek-v4-pro"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "# JuryBench [EMNLP 2026] When Does Defendant Statement Matter? A Study of Bias and Persuasion in LLM-Simulated Jurors", "canonicalNameSource": "official_readme", "decisionReason": "The benchmark is formally named and released with a GitHub repository and a defined scoring setup based on verdict severity changes, enabling model comparison.", "evaluationEvidence": {"reuse": "The data and code are available at https://github.com/choyingw/JuryBench, where the README declares the benchmark name.", "scoring": "Changes in verdict severity are quantified after 20 frontier LLMs make 432K decisions and rationales.", "task": "Controversial U.S. criminal law cases where defendants from different backgrounds give courtroom statements with varying emotional appeal or rebuttal, and LLM-simulated jurors decide verdicts."}, "model": "deepseek-v4-pro", "reviewedAt": "2026-09-11T00:52:53.785364Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates LLM-simulated jurors on controversial U.S. criminal law cases with defendant statements of varying backgrounds and emotional appeal, measuring verdict severity changes.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "public artifact URL"], "snippet": "To support the analysis, we introduce JuryBench, a benchmark containing controversial criminal cases in U.S."}, "familyId": "bmf_d24d7502a0d6", "firstSeenAt": "2026-09-10", "heat": null, "id": "bm_jurybench_3c843934", "indexedAt": "2026-09-10T00:47:26.889471Z", "industrySectors": [], "links": {"code": "https://github.com/choyingw/JuryBench", "data": null, "paper": "https://arxiv.org/abs/2609.09887", "pdf": "https://arxiv.org/pdf/2609.09887", "project": null, "report": "https://arxiv.org/abs/2609.09887"}, "metrics": [], "motivation": "LLMs have been used to simulate human decision-making in professional settings, yet their behaviors in common-law jury trials remain unexplored.", "name": "JuryBench", "oneLine": "Evaluates LLM-simulated jurors on controversial U.S. criminal law cases with defendant statements of varying backgrounds and emotional appeal, measuring verdict severity changes.", "paperTitle": "When Does Defendant Statement Matter? A Study of Bias and Persuasion in LLM-Simulated Jurors", "primaryDomain": "General AI", "publication": {"evidence": "Accepted to EMNLP 2026", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2609.09887", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "EMNLP 2026", "verifiedAt": "2026-09-10T00:47:26.889471Z"}, "publications": [], "publishers": [{"name": "JuryBench Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/choyingw/JuryBench"}], "ranking": {"30d": {"confidence": "Low", "coverage": 0.55, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.0, "value": 0}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.55, "rank": 141, "score": 23}, "method": "attention-ranking-v13", "rank": 141, "score": 23}, "90d": {"confidence": "Low", "coverage": 0.55, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.0, "value": 0}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.55, "rank": 337, "score": 23}, "method": "attention-ranking-v13", "rank": 337, "score": 23}, "today": {"confidence": "Low", "coverage": 0.25, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": 1, "historyDate": "2026-09-09", "rank": null, "score": null, "windowDays": 1}, "level": {"components": {"githubStars": {"normalization": "percentile", "normalized": 0.25, "percentile": 0.25, "value": 0}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "llmAttentionForecast": {"maxContribution": 0.25, "percentile": 0.42, "role": "bonus", "value": 42}}, "confidence": "Low", "coverage": 0.25, "rank": 4, "score": 42}, "method": "attention-ranking-v13", "rank": 4, "score": 42}}, "readiness": "Runnable", "recognitionConfidence": 0.95, "relation": "introduces", "releasedAt": "2026-09-09", "source": {"authors": ["Cho-Ying Wu"], "categories": ["cs.CL", "cs.CY"], "id": "2609.09887", "title": "When Does Defendant Statement Matter? A Study of Bias and Persuasion in LLM-Simulated Jurors", "type": "arxiv", "url": "https://arxiv.org/abs/2609.09887"}, "sourceUpdatedAt": "2026-09-09T08:43:34Z", "topics": ["cs.CL"], "venueAttempts": [{"decisionRaw": "Accepted to EMNLP 2026", "evidence": [{"level": "author-claim", "observedAt": "2026-09-10T00:47:26.889471Z", "rawValue": "Accepted to EMNLP 2026", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2609.09887"}], "reviewStatus": "accepted", "venueName": "EMNLP 2026"}], "whyItMatters": "Provides a structured way to measure persuasion, ideological bias, and background affinity in LLM jury simulations, highlighting risks for legal AI applications."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
