{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:AQWWNNOBBF4BCJRABSNOAGHT4X","short_pith_number":"pith:AQWWNNOB","schema_version":"1.0","canonical_sha256":"042d66b5c109781126200c9ae018f3e5dd21dcd515ca9c4ac05b05e7e3dbc368","source":{"kind":"arxiv","id":"2608.01388","version":1},"attestation_state":"computed","paper":{"title":"Why Formal Monitors Fail: Attack Distribution Entropy as a Coverage Bound for LTL-Based LLM Agent Safety","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Ruiyang Zhang","submitted_at":"2026-08-02T17:08:21Z","abstract_excerpt":"Runtime safety monitors based on Linear Temporal Logic (LTL) and finite automata (FSA) are increasingly deployed to intercept unsafe tool-call sequences in LLM agents. Yet the same monitor achieves 68-75% attack coverage on some model architectures and near-zero on others, with no explanation from capability scores, training data, or prompt design. We provide the missing theory. We prove that the recall of any fixed-invariant FSA monitor is bounded above by the concentration of the attack distribution: the fraction of attacks covered by the k most frequent trigger-completion patterns. When att"},"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":"2608.01388","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2026-08-02T17:08:21Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"53fda36e3238900eb2421b8acd3c703cdbbe2604eff5fdfed596b5d576e9d180","abstract_canon_sha256":"ae9b0559e587bc27387d72c192e24661a5cb89453a537c710cdfee99f82f7166"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-04T02:04:55.793876Z","signature_b64":"XK8JOT88XiHteGG1KiiNYvqCzKuU/ahCw8xkVevsxowRrSCixtwFnVXz99VzcizqhiXOJjyoZPwmDwniqj/uCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"042d66b5c109781126200c9ae018f3e5dd21dcd515ca9c4ac05b05e7e3dbc368","last_reissued_at":"2026-08-04T02:04:55.792437Z","signature_status":"signed_v1","first_computed_at":"2026-08-04T02:04:55.792437Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Why Formal Monitors Fail: Attack Distribution Entropy as a Coverage Bound for LTL-Based LLM Agent Safety","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Ruiyang Zhang","submitted_at":"2026-08-02T17:08:21Z","abstract_excerpt":"Runtime safety monitors based on Linear Temporal Logic (LTL) and finite automata (FSA) are increasingly deployed to intercept unsafe tool-call sequences in LLM agents. Yet the same monitor achieves 68-75% attack coverage on some model architectures and near-zero on others, with no explanation from capability scores, training data, or prompt design. We provide the missing theory. We prove that the recall of any fixed-invariant FSA monitor is bounded above by the concentration of the attack distribution: the fraction of attacks covered by the k most frequent trigger-completion patterns. When att"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.01388","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/2608.01388/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":"2608.01388","created_at":"2026-08-04T02:04:55.793828+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.01388v1","created_at":"2026-08-04T02:04:55.793828+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.01388","created_at":"2026-08-04T02:04:55.793828+00:00"},{"alias_kind":"pith_short_12","alias_value":"AQWWNNOBBF4B","created_at":"2026-08-04T02:04:55.793828+00:00"},{"alias_kind":"pith_short_16","alias_value":"AQWWNNOBBF4BCJRA","created_at":"2026-08-04T02:04:55.793828+00:00"},{"alias_kind":"pith_short_8","alias_value":"AQWWNNOB","created_at":"2026-08-04T02:04:55.793828+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/AQWWNNOBBF4BCJRABSNOAGHT4X","json":"https://pith.science/pith/AQWWNNOBBF4BCJRABSNOAGHT4X.json","graph_json":"https://pith.science/api/pith-number/AQWWNNOBBF4BCJRABSNOAGHT4X/graph.json","events_json":"https://pith.science/api/pith-number/AQWWNNOBBF4BCJRABSNOAGHT4X/events.json","paper":"https://pith.science/paper/AQWWNNOB"},"agent_actions":{"view_html":"https://pith.science/pith/AQWWNNOBBF4BCJRABSNOAGHT4X","download_json":"https://pith.science/pith/AQWWNNOBBF4BCJRABSNOAGHT4X.json","view_paper":"https://pith.science/paper/AQWWNNOB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.01388&json=true","fetch_graph":"https://pith.science/api/pith-number/AQWWNNOBBF4BCJRABSNOAGHT4X/graph.json","fetch_events":"https://pith.science/api/pith-number/AQWWNNOBBF4BCJRABSNOAGHT4X/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AQWWNNOBBF4BCJRABSNOAGHT4X/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AQWWNNOBBF4BCJRABSNOAGHT4X/action/storage_attestation","attest_author":"https://pith.science/pith/AQWWNNOBBF4BCJRABSNOAGHT4X/action/author_attestation","sign_citation":"https://pith.science/pith/AQWWNNOBBF4BCJRABSNOAGHT4X/action/citation_signature","submit_replication":"https://pith.science/pith/AQWWNNOBBF4BCJRABSNOAGHT4X/action/replication_record"}},"created_at":"2026-08-04T02:04:55.793828+00:00","updated_at":"2026-08-04T02:04:55.793828+00:00"}