{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.11105", "kind": "paper", "url": "https://arxiv.org/abs/2606.11105"}], "authors": ["Haeji Jung", "Hila Gonen"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.11105", "source_url": "https://arxiv.org/abs/2606.11105"}, "categories": ["General AI", "Language & Knowledge", "cs.CL"], "description": {"en": "PhantomBench evaluates language models' ability to abstain from answering about non-existent entities. It comprises over 60,000 non-existent terms derived from real concepts across domains, and provides a pipeline for generating further instances."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.11105", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.11105", "title": "PhantomBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["University of British Columbia", "Canadian Institute for Advanced Research"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.11105", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.11105", "source_url": "https://openalex.org/W7164127218"}, "key": "claire-radar:2606.11105", "modality": null, "name": "PhantomBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "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_phantombench_848c0022", "origin_source": "arxiv", "recognition_confidence": "0.85", "record_sha256": "64f01d06dcc1a5c73545e70749cc1745235bc77ddc42fecc240024d5f513dab4", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.11105"}, "publisher": null, "released": "2026-06-09", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.11105", "source_url": "https://arxiv.org/abs/2606.11105"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-11105", "source": "claire_radar", "source_benchmark_id": "2606.11105", "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": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2606.11105", "observedAt": "2026-10-04T03:02:42.754940Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-24", "state": "stale"}}, "status": "stale"}, "capabilities": [], "confidence": "Medium", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "2a4e7d3c08cd8f27206db425569fa7bfe03ef54c20ad42a71dfe683da21bace1", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "PhantomBench evaluates language models' ability to abstain from answering about non-existent entities. It comprises over 60,000 non-existent terms derived from real concepts across domains, and provides a pipeline for generating further instances.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence"], "snippet": "We introduce PhantomBench, the first large-scale benchmark of its kind, comprising more than 60K non-existent terms and entities derived from real concepts across diverse domains."}, "familyId": "bmf_b3df0dc70b12", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_phantombench_848c0022", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2606.11105", "paper": "https://arxiv.org/abs/2606.11105", "pdf": "https://arxiv.org/pdf/2606.11105", "project": null, "report": "https://arxiv.org/abs/2606.11105"}, "metrics": [], "motivation": "Hallucinations, where language models (LMs) generate factually ungrounded responses, pose serious risks, as users tend to blindly rely on them.", "name": "PhantomBench", "oneLine": "PhantomBench evaluates language models' ability to abstain from answering about non-existent entities. It comprises over 60,000 non-existent terms derived from real concepts across domains, and provides a pipeline for generating further instances.", "paperTitle": "PhantomBench: Benchmarking the Non-existential Threat of Language Models", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.11105", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 0.85, "relation": "introduces", "releasedAt": "2026-06-09", "source": {"authors": ["Haeji Jung", "Hila Gonen"], "categories": ["cs.CL", "cs.AI"], "id": "2606.11105", "title": "PhantomBench: Benchmarking the Non-existential Threat of Language Models", "type": "arxiv", "url": "https://arxiv.org/abs/2606.11105"}, "sourceUpdatedAt": "2026-06-09T17:03:19Z", "topics": ["cs.CL"], "venueAttempts": [], "whyItMatters": "Addresses the evaluation gap in assessing models' calibration of knowledge boundaries, offering a practical tool for detecting hallucination tendencies in high-stakes applications and studying behavior on rare concepts."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
