{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:2XBZGLH2C4H2CFQXL4SZI4Q3XH","short_pith_number":"pith:2XBZGLH2","schema_version":"1.0","canonical_sha256":"d5c3932cfa170fa116175f2594721bb9fa370ccbd9095157baaaab6f3e6c8ba5","source":{"kind":"arxiv","id":"2511.12487","version":2},"attestation_state":"computed","paper":{"title":"ToxSearch: Evolving Prompts for Toxicity Search in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.NE","authors_text":"Onkar Shelar, Travis Desell","submitted_at":"2025-11-16T07:47:31Z","abstract_excerpt":"Large Language Models remain vulnerable to adversarial prompts that elicit toxic content even after safety alignment. We present ToxSearch, a black-box evolutionary framework that tests model safety by evolving prompts in a synchronous steady-state loop. The system employs a diverse set of operators, including lexical substitutions, negation, back-translation, paraphrasing, and two semantic crossover operators, while a moderation oracle provides fitness guidance. Operator-level analysis shows heterogeneous behavior: lexical substitutions offer the best yield-variance trade-off, semantic-simila"},"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":"2511.12487","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.NE","submitted_at":"2025-11-16T07:47:31Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"b30ed043a64b28d74daaea932d6b7cf13d22fa916f8fca6309b7faa89bcd5889","abstract_canon_sha256":"1c147ba145dacfcc93972e94f52c0ec8861635d9904c6d22d562535e83b52b2e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T03:05:04.329249Z","signature_b64":"Djv8vW3iY1xEbtgOoN09RbY4YDY3ZIsid722cpcajesmWI413tD3sBWks4Lourh3MtrbopC0Tdgm4MrUhNnLBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d5c3932cfa170fa116175f2594721bb9fa370ccbd9095157baaaab6f3e6c8ba5","last_reissued_at":"2026-06-02T03:05:04.328762Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T03:05:04.328762Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"ToxSearch: Evolving Prompts for Toxicity Search in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.NE","authors_text":"Onkar Shelar, Travis Desell","submitted_at":"2025-11-16T07:47:31Z","abstract_excerpt":"Large Language Models remain vulnerable to adversarial prompts that elicit toxic content even after safety alignment. We present ToxSearch, a black-box evolutionary framework that tests model safety by evolving prompts in a synchronous steady-state loop. The system employs a diverse set of operators, including lexical substitutions, negation, back-translation, paraphrasing, and two semantic crossover operators, while a moderation oracle provides fitness guidance. Operator-level analysis shows heterogeneous behavior: lexical substitutions offer the best yield-variance trade-off, semantic-simila"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2511.12487","kind":"arxiv","version":2},"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/2511.12487/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":"2511.12487","created_at":"2026-06-02T03:05:04.328818+00:00"},{"alias_kind":"arxiv_version","alias_value":"2511.12487v2","created_at":"2026-06-02T03:05:04.328818+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2511.12487","created_at":"2026-06-02T03:05:04.328818+00:00"},{"alias_kind":"pith_short_12","alias_value":"2XBZGLH2C4H2","created_at":"2026-06-02T03:05:04.328818+00:00"},{"alias_kind":"pith_short_16","alias_value":"2XBZGLH2C4H2CFQX","created_at":"2026-06-02T03:05:04.328818+00:00"},{"alias_kind":"pith_short_8","alias_value":"2XBZGLH2","created_at":"2026-06-02T03:05:04.328818+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2601.20981","citing_title":"Diversifying Toxicity Search in Large Language Models Through Speciation","ref_index":25,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/2XBZGLH2C4H2CFQXL4SZI4Q3XH","json":"https://pith.science/pith/2XBZGLH2C4H2CFQXL4SZI4Q3XH.json","graph_json":"https://pith.science/api/pith-number/2XBZGLH2C4H2CFQXL4SZI4Q3XH/graph.json","events_json":"https://pith.science/api/pith-number/2XBZGLH2C4H2CFQXL4SZI4Q3XH/events.json","paper":"https://pith.science/paper/2XBZGLH2"},"agent_actions":{"view_html":"https://pith.science/pith/2XBZGLH2C4H2CFQXL4SZI4Q3XH","download_json":"https://pith.science/pith/2XBZGLH2C4H2CFQXL4SZI4Q3XH.json","view_paper":"https://pith.science/paper/2XBZGLH2","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2511.12487&json=true","fetch_graph":"https://pith.science/api/pith-number/2XBZGLH2C4H2CFQXL4SZI4Q3XH/graph.json","fetch_events":"https://pith.science/api/pith-number/2XBZGLH2C4H2CFQXL4SZI4Q3XH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/2XBZGLH2C4H2CFQXL4SZI4Q3XH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/2XBZGLH2C4H2CFQXL4SZI4Q3XH/action/storage_attestation","attest_author":"https://pith.science/pith/2XBZGLH2C4H2CFQXL4SZI4Q3XH/action/author_attestation","sign_citation":"https://pith.science/pith/2XBZGLH2C4H2CFQXL4SZI4Q3XH/action/citation_signature","submit_replication":"https://pith.science/pith/2XBZGLH2C4H2CFQXL4SZI4Q3XH/action/replication_record"}},"created_at":"2026-06-02T03:05:04.328818+00:00","updated_at":"2026-06-02T03:05:04.328818+00:00"}