{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:G5ROO3OA7ID2Z3NPZLCKIO7KWE","short_pith_number":"pith:G5ROO3OA","schema_version":"1.0","canonical_sha256":"3762e76dc0fa07acedafcac4a43beab11a305df85795bb2e07dbe75c2233f3c0","source":{"kind":"arxiv","id":"2605.27025","version":1},"attestation_state":"computed","paper":{"title":"Attribute-Based Diagnosis of LLM Alignment with Hate Speech Annotations","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.MM"],"primary_cat":"cs.CL","authors_text":"Alexander Fraser, Faeze Ghorbanpour, Mohammad Amine Jradi","submitted_at":"2026-05-26T13:44:48Z","abstract_excerpt":"Hate speech annotation is costly, subjective, and prone to annotator disagreement, making large-scale dataset construction challenging. We systematically analyze how well large language models (LLMs) align with human judgments across ten theoretically grounded subjective attributes, such as dehumanization, violence, and sentiment, evaluating both small and large variants of Llama 3.1 and Qwen 2.5. Our analysis reveals a consistent split across all models: behaviorally explicit dimensions (insult, humiliate, attack-defend) correlate strongly with human annotations, while evaluative dimensions ("},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2605.27025","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-26T13:44:48Z","cross_cats_sorted":["cs.MM"],"title_canon_sha256":"7e2777e4e446bbeca691d087285ecd35ef70b27d4e929b98328b2089857eedaa","abstract_canon_sha256":"d8b5be17760d98842e429aa4b60cd81d305e20dcad07de9e85d9c74ab18712cb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:06:24.728817Z","signature_b64":"zHi23bvZ9pNgNunCIquMqSZcBj1ieDnIkcmF5YxKIqKvdBZAHHCvEzqq+ZNcR0NTpUd5AGR9BaGmA8F/ODe0Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3762e76dc0fa07acedafcac4a43beab11a305df85795bb2e07dbe75c2233f3c0","last_reissued_at":"2026-05-27T01:06:24.728227Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:06:24.728227Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Attribute-Based Diagnosis of LLM Alignment with Hate Speech Annotations","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.MM"],"primary_cat":"cs.CL","authors_text":"Alexander Fraser, Faeze Ghorbanpour, Mohammad Amine Jradi","submitted_at":"2026-05-26T13:44:48Z","abstract_excerpt":"Hate speech annotation is costly, subjective, and prone to annotator disagreement, making large-scale dataset construction challenging. We systematically analyze how well large language models (LLMs) align with human judgments across ten theoretically grounded subjective attributes, such as dehumanization, violence, and sentiment, evaluating both small and large variants of Llama 3.1 and Qwen 2.5. Our analysis reveals a consistent split across all models: behaviorally explicit dimensions (insult, humiliate, attack-defend) correlate strongly with human annotations, while evaluative dimensions ("},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.27025","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.27025/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2605.27025","created_at":"2026-05-27T01:06:24.728339+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.27025v1","created_at":"2026-05-27T01:06:24.728339+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.27025","created_at":"2026-05-27T01:06:24.728339+00:00"},{"alias_kind":"pith_short_12","alias_value":"G5ROO3OA7ID2","created_at":"2026-05-27T01:06:24.728339+00:00"},{"alias_kind":"pith_short_16","alias_value":"G5ROO3OA7ID2Z3NP","created_at":"2026-05-27T01:06:24.728339+00:00"},{"alias_kind":"pith_short_8","alias_value":"G5ROO3OA","created_at":"2026-05-27T01:06:24.728339+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/G5ROO3OA7ID2Z3NPZLCKIO7KWE","json":"https://pith.science/pith/G5ROO3OA7ID2Z3NPZLCKIO7KWE.json","graph_json":"https://pith.science/api/pith-number/G5ROO3OA7ID2Z3NPZLCKIO7KWE/graph.json","events_json":"https://pith.science/api/pith-number/G5ROO3OA7ID2Z3NPZLCKIO7KWE/events.json","paper":"https://pith.science/paper/G5ROO3OA"},"agent_actions":{"view_html":"https://pith.science/pith/G5ROO3OA7ID2Z3NPZLCKIO7KWE","download_json":"https://pith.science/pith/G5ROO3OA7ID2Z3NPZLCKIO7KWE.json","view_paper":"https://pith.science/paper/G5ROO3OA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.27025&json=true","fetch_graph":"https://pith.science/api/pith-number/G5ROO3OA7ID2Z3NPZLCKIO7KWE/graph.json","fetch_events":"https://pith.science/api/pith-number/G5ROO3OA7ID2Z3NPZLCKIO7KWE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/G5ROO3OA7ID2Z3NPZLCKIO7KWE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/G5ROO3OA7ID2Z3NPZLCKIO7KWE/action/storage_attestation","attest_author":"https://pith.science/pith/G5ROO3OA7ID2Z3NPZLCKIO7KWE/action/author_attestation","sign_citation":"https://pith.science/pith/G5ROO3OA7ID2Z3NPZLCKIO7KWE/action/citation_signature","submit_replication":"https://pith.science/pith/G5ROO3OA7ID2Z3NPZLCKIO7KWE/action/replication_record"}},"created_at":"2026-05-27T01:06:24.728339+00:00","updated_at":"2026-05-27T01:06:24.728339+00:00"}