{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.10645", "kind": "paper", "url": "https://arxiv.org/abs/2607.10645"}, {"id": "gh:karpovilia/mafiascope", "kind": "repo", "url": "https://github.com/karpovilia/mafiascope"}, {"kind": "website", "url": "https://karpovilia.github.io/mafiascope/"}], "authors": ["Ilia Karpov", "Maksim Grigorenko", "Vadim Ayuyev"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.10645", "source_url": "https://arxiv.org/abs/2607.10645"}, "categories": ["General AI", "Language & Knowledge", "cs.CL"], "description": {"en": "MafiaScope evaluates LLM agents in the social deduction game Mafia. It probes each agent's private beliefs after every public utterance, scoring them against ground truth without influencing the game. The testbed provides an open-source engine, interactive visualizer, recorded games, and counterfactual replay corpus."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.10645", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.10645", "title": "MafiaScope"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["National Research University Higher School of Economics"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.10645", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.10645", "source_url": "https://openalex.org/W7168270908"}, "key": "claire-radar:2607.10645", "modality": null, "name": "MafiaScope", "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_mafiascope_6f84f9bf", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "c22398ae5c9f522aecbc22ae05d8780e0d89b9e644f4b2af431972dbcf6cbc39", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.10645"}, "publisher": null, "released": "2026-07-12", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.10645", "source_url": "https://arxiv.org/abs/2607.10645"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-10645", "source": "claire_radar", "source_benchmark_id": "2607.10645", "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/karpovilia/mafiascope", "githubScope": "benchmark_repo", "githubStars": 0, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2607.10645", "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": "e824aa18d8728d8ce133e2b350308995e6708abafda4ae2fe48e5e64a53b88ea", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "MafiaScope evaluates LLM agents in the social deduction game Mafia. It probes each agent's private beliefs after every public utterance, scoring them against ground truth without influencing the game. The testbed provides an open-source engine, interactive visualizer, recorded games, and counterfactual replay corpus.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "public artifact URL"], "snippet": "We present MafiaScope, an open testbed that turns the social deduction game Mafia into a measurement instrument for machine Theory of Mind."}, "familyId": "bmf_a84e60ac04f2", "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/karpovilia/mafiascope", "url": "https://github.com/karpovilia/mafiascope"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/karpovilia/mafiascope", "url": "https://huggingface.co/papers/2607.10645"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/karpovilia/mafiascope.", "url": "https://arxiv.org/abs/2607.10645"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/karpovilia/mafiascope", "url": "https://arxiv.org/abs/2607.10645"}], "scope": "benchmark_repo", "url": "https://github.com/karpovilia/mafiascope", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_mafiascope_6f84f9bf", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/karpovilia/mafiascope", "data": null, "hfPaper": "https://huggingface.co/papers/2607.10645", "paper": "https://arxiv.org/abs/2607.10645", "pdf": "https://arxiv.org/pdf/2607.10645", "project": "https://karpovilia.github.io/mafiascope/", "report": "https://arxiv.org/abs/2607.10645"}, "metrics": [], "motivation": "An LLM agent's public behaviour reveals little about its social reasoning: an agent that votes correctly may be guessing, and an agent that lies well leaves no trace of what it actually believes.", "name": "MafiaScope", "oneLine": "MafiaScope evaluates LLM agents in the social deduction game Mafia. It probes each agent's private beliefs after every public utterance, scoring them against ground truth without influencing the game. The testbed provides an open-source engine, interactive visualizer, recorded games, and counterfactual replay corpus.", "paperTitle": "MafiaScope: Non-Invasive, Time-Resolved Belief Probing for LLM Agents in Social Deduction Games", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.10645", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "MafiaScope Team", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/karpovilia/mafiascope"}], "ranking": {"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": 539, "score": 23}, "method": "attention-ranking-v13", "rank": 539, "score": 23}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-07-12", "source": {"authors": ["Ilia Karpov"], "categories": ["cs.CL", "cs.MA", "cs.SI"], "id": "2607.10645", "title": "MafiaScope: Non-Invasive, Time-Resolved Belief Probing for LLM Agents in Social Deduction Games", "type": "arxiv", "url": "https://arxiv.org/abs/2607.10645"}, "sourceUpdatedAt": "2026-07-18T17:54:54Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "Machine Theory of Mind is difficult to measure from observable behavior alone. MafiaScope separates incorrect belief formation from incorrect action under correct beliefs, a distinction invisible in dialogue or outcome data, enabling more precise evaluation of social reasoning in LLM agents."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
