{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:5OTLDPVKXDH26F66GTFJXN5XVI","short_pith_number":"pith:5OTLDPVK","schema_version":"1.0","canonical_sha256":"eba6b1beaab8cfaf17de34ca9bb7b7aa3cdc3191fe1b87c3c6d023a8e89a23f1","source":{"kind":"arxiv","id":"2607.17575","version":1},"attestation_state":"computed","paper":{"title":"A Dual-Hypothesis Reasoning Framework for LLM Guardrails","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Md Asiful Islam, Mihai Surdeanu","submitted_at":"2026-07-20T05:42:25Z","abstract_excerpt":"We propose ARBITER, a novel LLM guardrail framework that introduces two key ideas: (i) dual-hypothesis reasoning, a reasoning method for LLM guardrails that explicitly considers both safe and unsafe interpretations of a prompt before making a safety decision, and (ii) multi-component supervised fine-tuning (MC-SFT), a structured training loss for reasoning-based guardrails that decomposes LLM outputs into logical components and weights them according to their importance. Existing reasoning-based guardrails often rely on expensive procedures, such as generating reasoning traces using larger or "},"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":"2607.17575","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-20T05:42:25Z","cross_cats_sorted":[],"title_canon_sha256":"f0ff47c9046e373a9fae405704003135a4c3cafb9e57fbe5141b19fa4831a9c6","abstract_canon_sha256":"ed2b073526f99911f6c27e0eadd9a3d4223aa9fcc1f22f3c39448ed65a551141"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-21T01:21:53.091484Z","signature_b64":"ED70kauUz3g04iaR4bl4R39jp60bY3GpxcDRouYRol8BnGGxuWbEtathX5pw0PDgXMI2spmvKTB0QAjeksq1DA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eba6b1beaab8cfaf17de34ca9bb7b7aa3cdc3191fe1b87c3c6d023a8e89a23f1","last_reissued_at":"2026-07-21T01:21:53.090666Z","signature_status":"signed_v1","first_computed_at":"2026-07-21T01:21:53.090666Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Dual-Hypothesis Reasoning Framework for LLM Guardrails","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Md Asiful Islam, Mihai Surdeanu","submitted_at":"2026-07-20T05:42:25Z","abstract_excerpt":"We propose ARBITER, a novel LLM guardrail framework that introduces two key ideas: (i) dual-hypothesis reasoning, a reasoning method for LLM guardrails that explicitly considers both safe and unsafe interpretations of a prompt before making a safety decision, and (ii) multi-component supervised fine-tuning (MC-SFT), a structured training loss for reasoning-based guardrails that decomposes LLM outputs into logical components and weights them according to their importance. Existing reasoning-based guardrails often rely on expensive procedures, such as generating reasoning traces using larger or "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.17575","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/2607.17575/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":"2607.17575","created_at":"2026-07-21T01:21:53.091078+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.17575v1","created_at":"2026-07-21T01:21:53.091078+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.17575","created_at":"2026-07-21T01:21:53.091078+00:00"},{"alias_kind":"pith_short_12","alias_value":"5OTLDPVKXDH2","created_at":"2026-07-21T01:21:53.091078+00:00"},{"alias_kind":"pith_short_16","alias_value":"5OTLDPVKXDH26F66","created_at":"2026-07-21T01:21:53.091078+00:00"},{"alias_kind":"pith_short_8","alias_value":"5OTLDPVK","created_at":"2026-07-21T01:21:53.091078+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/5OTLDPVKXDH26F66GTFJXN5XVI","json":"https://pith.science/pith/5OTLDPVKXDH26F66GTFJXN5XVI.json","graph_json":"https://pith.science/api/pith-number/5OTLDPVKXDH26F66GTFJXN5XVI/graph.json","events_json":"https://pith.science/api/pith-number/5OTLDPVKXDH26F66GTFJXN5XVI/events.json","paper":"https://pith.science/paper/5OTLDPVK"},"agent_actions":{"view_html":"https://pith.science/pith/5OTLDPVKXDH26F66GTFJXN5XVI","download_json":"https://pith.science/pith/5OTLDPVKXDH26F66GTFJXN5XVI.json","view_paper":"https://pith.science/paper/5OTLDPVK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.17575&json=true","fetch_graph":"https://pith.science/api/pith-number/5OTLDPVKXDH26F66GTFJXN5XVI/graph.json","fetch_events":"https://pith.science/api/pith-number/5OTLDPVKXDH26F66GTFJXN5XVI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5OTLDPVKXDH26F66GTFJXN5XVI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5OTLDPVKXDH26F66GTFJXN5XVI/action/storage_attestation","attest_author":"https://pith.science/pith/5OTLDPVKXDH26F66GTFJXN5XVI/action/author_attestation","sign_citation":"https://pith.science/pith/5OTLDPVKXDH26F66GTFJXN5XVI/action/citation_signature","submit_replication":"https://pith.science/pith/5OTLDPVKXDH26F66GTFJXN5XVI/action/replication_record"}},"created_at":"2026-07-21T01:21:53.091078+00:00","updated_at":"2026-07-21T01:21:53.091078+00:00"}