{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QPCPLFWW6ZZWSBGJJ5UA7SNCE3","short_pith_number":"pith:QPCPLFWW","schema_version":"1.0","canonical_sha256":"83c4f596d6f6736904c94f680fc9a226f1431fd883133d3118f2010bb2bcfc2d","source":{"kind":"arxiv","id":"2605.21602","version":1},"attestation_state":"computed","paper":{"title":"Benchmarking and Improving Monitors for Out-Of-Distribution Alignment Failure in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.AI","authors_text":"Cassidy Laidlaw, Dylan Feng, Pragya Srivastava","submitted_at":"2026-05-20T18:08:21Z","abstract_excerpt":"Many safety and alignment failures of large language models (LLMs) occur due to out-of-distribution (OOD) situations: unusual prompt or response patterns that are unforeseen by model developers. We systematically study whether LLM monitoring pipelines can detect these OOD alignment failures by introducing a benchmark called Misalignment Out Of Distribution (MOOD). It is difficult to find failures that are truly OOD for off-the-shelf models trained on vast safety datasets. We sidestep this by including a restricted training set in MOOD that we use to train our own monitors, as well as seven tes"},"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":"2605.21602","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-20T18:08:21Z","cross_cats_sorted":["cs.SE"],"title_canon_sha256":"cc97ddd8d2f188556db2e6e5098adf58220df7cc5228dbbb3929d920c5019aa0","abstract_canon_sha256":"ad7de2a6a426d93bbc32c830e6e17e7574b995113123ed0b2d5408e2de46bde6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:03:24.866311Z","signature_b64":"uQN0dWEt2dmznhF05vjzHKDFfcV1CpkOuKAFBTN8e1e0RkiaLiY3OuBvILIG+DwM3WrZaFHE6LLAbDPJ9nRyAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"83c4f596d6f6736904c94f680fc9a226f1431fd883133d3118f2010bb2bcfc2d","last_reissued_at":"2026-05-22T01:03:24.865833Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:03:24.865833Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Benchmarking and Improving Monitors for Out-Of-Distribution Alignment Failure in LLMs","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.SE"],"primary_cat":"cs.AI","authors_text":"Cassidy Laidlaw, Dylan Feng, Pragya Srivastava","submitted_at":"2026-05-20T18:08:21Z","abstract_excerpt":"Many safety and alignment failures of large language models (LLMs) occur due to out-of-distribution (OOD) situations: unusual prompt or response patterns that are unforeseen by model developers. We systematically study whether LLM monitoring pipelines can detect these OOD alignment failures by introducing a benchmark called Misalignment Out Of Distribution (MOOD). It is difficult to find failures that are truly OOD for off-the-shelf models trained on vast safety datasets. We sidestep this by including a restricted training set in MOOD that we use to train our own monitors, as well as seven tes"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.21602","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/2605.21602/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":"2605.21602","created_at":"2026-05-22T01:03:24.865897+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.21602v1","created_at":"2026-05-22T01:03:24.865897+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.21602","created_at":"2026-05-22T01:03:24.865897+00:00"},{"alias_kind":"pith_short_12","alias_value":"QPCPLFWW6ZZW","created_at":"2026-05-22T01:03:24.865897+00:00"},{"alias_kind":"pith_short_16","alias_value":"QPCPLFWW6ZZWSBGJ","created_at":"2026-05-22T01:03:24.865897+00:00"},{"alias_kind":"pith_short_8","alias_value":"QPCPLFWW","created_at":"2026-05-22T01:03:24.865897+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/QPCPLFWW6ZZWSBGJJ5UA7SNCE3","json":"https://pith.science/pith/QPCPLFWW6ZZWSBGJJ5UA7SNCE3.json","graph_json":"https://pith.science/api/pith-number/QPCPLFWW6ZZWSBGJJ5UA7SNCE3/graph.json","events_json":"https://pith.science/api/pith-number/QPCPLFWW6ZZWSBGJJ5UA7SNCE3/events.json","paper":"https://pith.science/paper/QPCPLFWW"},"agent_actions":{"view_html":"https://pith.science/pith/QPCPLFWW6ZZWSBGJJ5UA7SNCE3","download_json":"https://pith.science/pith/QPCPLFWW6ZZWSBGJJ5UA7SNCE3.json","view_paper":"https://pith.science/paper/QPCPLFWW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.21602&json=true","fetch_graph":"https://pith.science/api/pith-number/QPCPLFWW6ZZWSBGJJ5UA7SNCE3/graph.json","fetch_events":"https://pith.science/api/pith-number/QPCPLFWW6ZZWSBGJJ5UA7SNCE3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QPCPLFWW6ZZWSBGJJ5UA7SNCE3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QPCPLFWW6ZZWSBGJJ5UA7SNCE3/action/storage_attestation","attest_author":"https://pith.science/pith/QPCPLFWW6ZZWSBGJJ5UA7SNCE3/action/author_attestation","sign_citation":"https://pith.science/pith/QPCPLFWW6ZZWSBGJJ5UA7SNCE3/action/citation_signature","submit_replication":"https://pith.science/pith/QPCPLFWW6ZZWSBGJJ5UA7SNCE3/action/replication_record"}},"created_at":"2026-05-22T01:03:24.865897+00:00","updated_at":"2026-05-22T01:03:24.865897+00:00"}