{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:TFW2CZVWIT22CPHGV37PMWWS7A","short_pith_number":"pith:TFW2CZVW","schema_version":"1.0","canonical_sha256":"996da166b644f5a13ce6aefef65ad2f820d83584089eeebf9edb2ac20fb69133","source":{"kind":"arxiv","id":"2606.21077","version":1},"attestation_state":"computed","paper":{"title":"OTTER: A Red-Teaming System for Toxicity-Evading Jailbreak Prompt Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Fang Yu, Hsin-Ling Hsu, Jerry Wang, Nai-Chia Chen, Yi-Cheng Lai","submitted_at":"2026-06-19T03:55:08Z","abstract_excerpt":"Production LLMs increasingly rely on toxicity-based moderation filters as a primary defense, assuming that harmful intent correlates with toxic surface wording. We show this assumption is fundamentally brittle: surface toxicity and adversarial intent can be decoupled by replacing as few as five tokens. We present OTTER (Obfuscated Toxicity-Evading Token Evolution for Rewriting), a black-box red-teaming framework requiring only standard API access, directly targeting the practical constraints of industry security audits. Evaluated on 457 AdvBench prompts across four GPT models, OTTER raises ave"},"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":"2606.21077","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-06-19T03:55:08Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"a3c2c2b891d1ff38058e4c79e5dfb4b7cdb2dfc617af89f60844bc0c1def6dbd","abstract_canon_sha256":"525feceee5617cf75f050d2e04c137b9df000f4c023058602c9c7faa77924799"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-23T01:12:29.082068Z","signature_b64":"mnnC0Z0cTLz6CDJba8PyjqsjeFclF4xwy1vqkdFjnczQhnOrgMrh7HVxC07AZ4gYqaZCBC7n6lrOneIJy8whDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"996da166b644f5a13ce6aefef65ad2f820d83584089eeebf9edb2ac20fb69133","last_reissued_at":"2026-06-23T01:12:29.081565Z","signature_status":"signed_v1","first_computed_at":"2026-06-23T01:12:29.081565Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"OTTER: A Red-Teaming System for Toxicity-Evading Jailbreak Prompt Optimization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Fang Yu, Hsin-Ling Hsu, Jerry Wang, Nai-Chia Chen, Yi-Cheng Lai","submitted_at":"2026-06-19T03:55:08Z","abstract_excerpt":"Production LLMs increasingly rely on toxicity-based moderation filters as a primary defense, assuming that harmful intent correlates with toxic surface wording. We show this assumption is fundamentally brittle: surface toxicity and adversarial intent can be decoupled by replacing as few as five tokens. We present OTTER (Obfuscated Toxicity-Evading Token Evolution for Rewriting), a black-box red-teaming framework requiring only standard API access, directly targeting the practical constraints of industry security audits. Evaluated on 457 AdvBench prompts across four GPT models, OTTER raises ave"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.21077","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/2606.21077/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":"2606.21077","created_at":"2026-06-23T01:12:29.081651+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.21077v1","created_at":"2026-06-23T01:12:29.081651+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.21077","created_at":"2026-06-23T01:12:29.081651+00:00"},{"alias_kind":"pith_short_12","alias_value":"TFW2CZVWIT22","created_at":"2026-06-23T01:12:29.081651+00:00"},{"alias_kind":"pith_short_16","alias_value":"TFW2CZVWIT22CPHG","created_at":"2026-06-23T01:12:29.081651+00:00"},{"alias_kind":"pith_short_8","alias_value":"TFW2CZVW","created_at":"2026-06-23T01:12:29.081651+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/TFW2CZVWIT22CPHGV37PMWWS7A","json":"https://pith.science/pith/TFW2CZVWIT22CPHGV37PMWWS7A.json","graph_json":"https://pith.science/api/pith-number/TFW2CZVWIT22CPHGV37PMWWS7A/graph.json","events_json":"https://pith.science/api/pith-number/TFW2CZVWIT22CPHGV37PMWWS7A/events.json","paper":"https://pith.science/paper/TFW2CZVW"},"agent_actions":{"view_html":"https://pith.science/pith/TFW2CZVWIT22CPHGV37PMWWS7A","download_json":"https://pith.science/pith/TFW2CZVWIT22CPHGV37PMWWS7A.json","view_paper":"https://pith.science/paper/TFW2CZVW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.21077&json=true","fetch_graph":"https://pith.science/api/pith-number/TFW2CZVWIT22CPHGV37PMWWS7A/graph.json","fetch_events":"https://pith.science/api/pith-number/TFW2CZVWIT22CPHGV37PMWWS7A/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TFW2CZVWIT22CPHGV37PMWWS7A/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TFW2CZVWIT22CPHGV37PMWWS7A/action/storage_attestation","attest_author":"https://pith.science/pith/TFW2CZVWIT22CPHGV37PMWWS7A/action/author_attestation","sign_citation":"https://pith.science/pith/TFW2CZVWIT22CPHGV37PMWWS7A/action/citation_signature","submit_replication":"https://pith.science/pith/TFW2CZVWIT22CPHGV37PMWWS7A/action/replication_record"}},"created_at":"2026-06-23T01:12:29.081651+00:00","updated_at":"2026-06-23T01:12:29.081651+00:00"}