{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RFUG5JBNRA72UOLZ45EOZLODJ4","short_pith_number":"pith:RFUG5JBN","schema_version":"1.0","canonical_sha256":"89686ea42d883faa3979e748ecadc34f1009e48dea79b97cad36298a0b21d8a9","source":{"kind":"arxiv","id":"2605.25420","version":1},"attestation_state":"computed","paper":{"title":"SomaliBench Eval: Measuring English-to-Somali Refusal Gaps in Open-Weight Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY"],"primary_cat":"cs.CL","authors_text":"Khalid Yusuf Dahir","submitted_at":"2026-05-25T04:45:44Z","abstract_excerpt":"Large language model safety evaluation remains heavily English-centered, leaving low-resource languages under-measured even when models are deployed globally. We evaluate four open-weight instruction-tuned models on SomaliBench v0, a native-author-verified benchmark of 100 harmful-intent prompts paired across English and Somali. Each of Llama-3.1-8B-Instruct, Gemma-2-9B-Instruct, Qwen-2.5-7B-Instruct, and Aya-23-8B is run locally with temperature 0 and the same English \"helpful, harmless, and honest\" (HHH) system prompt. A pinned Claude Sonnet snapshot (claude-sonnet-4-5-20250929) classifies e"},"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":"2605.25420","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-25T04:45:44Z","cross_cats_sorted":["cs.AI","cs.CY"],"title_canon_sha256":"4404a1f05b8cdd4a70401f944b48c1c43f394076b25409cb141dd578658a54e7","abstract_canon_sha256":"2aa6e47fb1d1c661757f3e7b583512c07b176b139ed37a11e40c055bef7e64eb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:04:34.080543Z","signature_b64":"E7nvlaDqw45fbySZWKzSZqiqUTMehVeXw7lT5nvAQVFfL/zPUNwZ4sFEhK2B8GCmq3vdpJGNJ13UCOuK/ysGDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"89686ea42d883faa3979e748ecadc34f1009e48dea79b97cad36298a0b21d8a9","last_reissued_at":"2026-05-26T02:04:34.079569Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:04:34.079569Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SomaliBench Eval: Measuring English-to-Somali Refusal Gaps in Open-Weight Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CY"],"primary_cat":"cs.CL","authors_text":"Khalid Yusuf Dahir","submitted_at":"2026-05-25T04:45:44Z","abstract_excerpt":"Large language model safety evaluation remains heavily English-centered, leaving low-resource languages under-measured even when models are deployed globally. We evaluate four open-weight instruction-tuned models on SomaliBench v0, a native-author-verified benchmark of 100 harmful-intent prompts paired across English and Somali. Each of Llama-3.1-8B-Instruct, Gemma-2-9B-Instruct, Qwen-2.5-7B-Instruct, and Aya-23-8B is run locally with temperature 0 and the same English \"helpful, harmless, and honest\" (HHH) system prompt. A pinned Claude Sonnet snapshot (claude-sonnet-4-5-20250929) classifies e"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.25420","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/2605.25420/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":"2605.25420","created_at":"2026-05-26T02:04:34.079745+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.25420v1","created_at":"2026-05-26T02:04:34.079745+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.25420","created_at":"2026-05-26T02:04:34.079745+00:00"},{"alias_kind":"pith_short_12","alias_value":"RFUG5JBNRA72","created_at":"2026-05-26T02:04:34.079745+00:00"},{"alias_kind":"pith_short_16","alias_value":"RFUG5JBNRA72UOLZ","created_at":"2026-05-26T02:04:34.079745+00:00"},{"alias_kind":"pith_short_8","alias_value":"RFUG5JBN","created_at":"2026-05-26T02:04:34.079745+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/RFUG5JBNRA72UOLZ45EOZLODJ4","json":"https://pith.science/pith/RFUG5JBNRA72UOLZ45EOZLODJ4.json","graph_json":"https://pith.science/api/pith-number/RFUG5JBNRA72UOLZ45EOZLODJ4/graph.json","events_json":"https://pith.science/api/pith-number/RFUG5JBNRA72UOLZ45EOZLODJ4/events.json","paper":"https://pith.science/paper/RFUG5JBN"},"agent_actions":{"view_html":"https://pith.science/pith/RFUG5JBNRA72UOLZ45EOZLODJ4","download_json":"https://pith.science/pith/RFUG5JBNRA72UOLZ45EOZLODJ4.json","view_paper":"https://pith.science/paper/RFUG5JBN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.25420&json=true","fetch_graph":"https://pith.science/api/pith-number/RFUG5JBNRA72UOLZ45EOZLODJ4/graph.json","fetch_events":"https://pith.science/api/pith-number/RFUG5JBNRA72UOLZ45EOZLODJ4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RFUG5JBNRA72UOLZ45EOZLODJ4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RFUG5JBNRA72UOLZ45EOZLODJ4/action/storage_attestation","attest_author":"https://pith.science/pith/RFUG5JBNRA72UOLZ45EOZLODJ4/action/author_attestation","sign_citation":"https://pith.science/pith/RFUG5JBNRA72UOLZ45EOZLODJ4/action/citation_signature","submit_replication":"https://pith.science/pith/RFUG5JBNRA72UOLZ45EOZLODJ4/action/replication_record"}},"created_at":"2026-05-26T02:04:34.079745+00:00","updated_at":"2026-05-26T02:04:34.079745+00:00"}