{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:DV4KRPXPYNTYI3MEQANMISBF52","short_pith_number":"pith:DV4KRPXP","schema_version":"1.0","canonical_sha256":"1d78a8beefc367846d84801ac44825eebd6cdd26d0802251831c6c417d8f1a3c","source":{"kind":"arxiv","id":"2605.19966","version":1},"attestation_state":"computed","paper":{"title":"Detecting Fluent Optimization-Based Adversarial Prompts via Sequential Entropy Changes","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Miguel R. D. Rodrigues, Mohammed Alshaalan","submitted_at":"2026-05-19T15:15:51Z","abstract_excerpt":"Optimization-based adversarial suffixes can jailbreak aligned large language models (LLMs) while remaining fluent, weakening static and windowed perplexity-based detectors. We cast adversarial suffix detection as an online change-point detection problem over the token-level next-token entropy stream. Using the LLM system prompt to estimate a robust baseline, we standardize user-token entropies and apply a one-sided CUSUM statistic. The resulting detector, CPD Online (CPD), is model-agnostic, training-free, runs online, and localizes the adversarial suffix onset. On a benchmark of 1,012 optimiz"},"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":"2605.19966","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-19T15:15:51Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"4cb8f42048c82d0363e499226af1679d3a09865f1f480bd0e7966ea51810bd50","abstract_canon_sha256":"01faa17f9f97c2293014b299fd0d4593e37c13ac72a13ce3b8e0604d4834b002"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T02:05:57.411636Z","signature_b64":"eWmN7DdUTZ0Lou+Us+DQPqeHE7cyIqdGmcguqLH3woy9zeAtP/peKgSYlGXiI90QoxT2s9ZBr958MJXV77sRCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1d78a8beefc367846d84801ac44825eebd6cdd26d0802251831c6c417d8f1a3c","last_reissued_at":"2026-05-20T02:05:57.411042Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T02:05:57.411042Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Detecting Fluent Optimization-Based Adversarial Prompts via Sequential Entropy Changes","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Miguel R. D. Rodrigues, Mohammed Alshaalan","submitted_at":"2026-05-19T15:15:51Z","abstract_excerpt":"Optimization-based adversarial suffixes can jailbreak aligned large language models (LLMs) while remaining fluent, weakening static and windowed perplexity-based detectors. We cast adversarial suffix detection as an online change-point detection problem over the token-level next-token entropy stream. Using the LLM system prompt to estimate a robust baseline, we standardize user-token entropies and apply a one-sided CUSUM statistic. The resulting detector, CPD Online (CPD), is model-agnostic, training-free, runs online, and localizes the adversarial suffix onset. On a benchmark of 1,012 optimiz"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.19966","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/2605.19966/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":"2605.19966","created_at":"2026-05-20T02:05:57.411134+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.19966v1","created_at":"2026-05-20T02:05:57.411134+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.19966","created_at":"2026-05-20T02:05:57.411134+00:00"},{"alias_kind":"pith_short_12","alias_value":"DV4KRPXPYNTY","created_at":"2026-05-20T02:05:57.411134+00:00"},{"alias_kind":"pith_short_16","alias_value":"DV4KRPXPYNTYI3ME","created_at":"2026-05-20T02:05:57.411134+00:00"},{"alias_kind":"pith_short_8","alias_value":"DV4KRPXP","created_at":"2026-05-20T02:05:57.411134+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/DV4KRPXPYNTYI3MEQANMISBF52","json":"https://pith.science/pith/DV4KRPXPYNTYI3MEQANMISBF52.json","graph_json":"https://pith.science/api/pith-number/DV4KRPXPYNTYI3MEQANMISBF52/graph.json","events_json":"https://pith.science/api/pith-number/DV4KRPXPYNTYI3MEQANMISBF52/events.json","paper":"https://pith.science/paper/DV4KRPXP"},"agent_actions":{"view_html":"https://pith.science/pith/DV4KRPXPYNTYI3MEQANMISBF52","download_json":"https://pith.science/pith/DV4KRPXPYNTYI3MEQANMISBF52.json","view_paper":"https://pith.science/paper/DV4KRPXP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.19966&json=true","fetch_graph":"https://pith.science/api/pith-number/DV4KRPXPYNTYI3MEQANMISBF52/graph.json","fetch_events":"https://pith.science/api/pith-number/DV4KRPXPYNTYI3MEQANMISBF52/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DV4KRPXPYNTYI3MEQANMISBF52/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DV4KRPXPYNTYI3MEQANMISBF52/action/storage_attestation","attest_author":"https://pith.science/pith/DV4KRPXPYNTYI3MEQANMISBF52/action/author_attestation","sign_citation":"https://pith.science/pith/DV4KRPXPYNTYI3MEQANMISBF52/action/citation_signature","submit_replication":"https://pith.science/pith/DV4KRPXPYNTYI3MEQANMISBF52/action/replication_record"}},"created_at":"2026-05-20T02:05:57.411134+00:00","updated_at":"2026-05-20T02:05:57.411134+00:00"}