{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:WALUUJPAWITKAAYHWTQYHUJWI4","short_pith_number":"pith:WALUUJPA","schema_version":"1.0","canonical_sha256":"b0174a25e0b226a00307b4e183d13647340829d8b0ca8752a2fc88c47f6c06cd","source":{"kind":"arxiv","id":"2605.25984","version":1},"attestation_state":"computed","paper":{"title":"SafeCtrl-RL: Inference-Time Adaptive Behaviour Control for LLM Dialogue via RL-Driven Prompt Optimisation","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Michael Orme, Yanchao Yu, Zhiyuan Tan","submitted_at":"2026-05-25T16:03:38Z","abstract_excerpt":"Ensuring safe and contextually appropriate behaviour in Large Language Models (LLMs) remains a critical challenge for real-world deployment. We present \\textbf{SafeCtrl-RL}, an inference-time behavioural control framework that enables adaptive safety regulation without model retraining or parameter modification. The method formulates dialogue generation as a sequential decision process, where a reinforcement learning agent dynamically selects prompt adjustment strategies based on contextual feedback. This allows unsafe behaviours to be suppressed through iterative refinement, which we conceptu"},"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":"2605.25984","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-25T16:03:38Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c721ec7835743df008dd439da75a7026d59ab3a57a015abf71850a99afa8c911","abstract_canon_sha256":"1ae199964bebe2546359d1ee8780070599f0a8eef6b5c103f6e81ec65af76409"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-26T02:05:21.853061Z","signature_b64":"/Nu1T0Djbl+L6S42jnz4U+7EvqyHZzftItZC08KBlwYFDTryxZlc/s+GpsqjTJK7QP+H67OmhnHX0ZLGJYzGDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b0174a25e0b226a00307b4e183d13647340829d8b0ca8752a2fc88c47f6c06cd","last_reissued_at":"2026-05-26T02:05:21.852433Z","signature_status":"signed_v1","first_computed_at":"2026-05-26T02:05:21.852433Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SafeCtrl-RL: Inference-Time Adaptive Behaviour Control for LLM Dialogue via RL-Driven Prompt Optimisation","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Michael Orme, Yanchao Yu, Zhiyuan Tan","submitted_at":"2026-05-25T16:03:38Z","abstract_excerpt":"Ensuring safe and contextually appropriate behaviour in Large Language Models (LLMs) remains a critical challenge for real-world deployment. We present \\textbf{SafeCtrl-RL}, an inference-time behavioural control framework that enables adaptive safety regulation without model retraining or parameter modification. The method formulates dialogue generation as a sequential decision process, where a reinforcement learning agent dynamically selects prompt adjustment strategies based on contextual feedback. This allows unsafe behaviours to be suppressed through iterative refinement, which we conceptu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.25984","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/2605.25984/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":"2605.25984","created_at":"2026-05-26T02:05:21.852555+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.25984v1","created_at":"2026-05-26T02:05:21.852555+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.25984","created_at":"2026-05-26T02:05:21.852555+00:00"},{"alias_kind":"pith_short_12","alias_value":"WALUUJPAWITK","created_at":"2026-05-26T02:05:21.852555+00:00"},{"alias_kind":"pith_short_16","alias_value":"WALUUJPAWITKAAYH","created_at":"2026-05-26T02:05:21.852555+00:00"},{"alias_kind":"pith_short_8","alias_value":"WALUUJPA","created_at":"2026-05-26T02:05:21.852555+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/WALUUJPAWITKAAYHWTQYHUJWI4","json":"https://pith.science/pith/WALUUJPAWITKAAYHWTQYHUJWI4.json","graph_json":"https://pith.science/api/pith-number/WALUUJPAWITKAAYHWTQYHUJWI4/graph.json","events_json":"https://pith.science/api/pith-number/WALUUJPAWITKAAYHWTQYHUJWI4/events.json","paper":"https://pith.science/paper/WALUUJPA"},"agent_actions":{"view_html":"https://pith.science/pith/WALUUJPAWITKAAYHWTQYHUJWI4","download_json":"https://pith.science/pith/WALUUJPAWITKAAYHWTQYHUJWI4.json","view_paper":"https://pith.science/paper/WALUUJPA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.25984&json=true","fetch_graph":"https://pith.science/api/pith-number/WALUUJPAWITKAAYHWTQYHUJWI4/graph.json","fetch_events":"https://pith.science/api/pith-number/WALUUJPAWITKAAYHWTQYHUJWI4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WALUUJPAWITKAAYHWTQYHUJWI4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WALUUJPAWITKAAYHWTQYHUJWI4/action/storage_attestation","attest_author":"https://pith.science/pith/WALUUJPAWITKAAYHWTQYHUJWI4/action/author_attestation","sign_citation":"https://pith.science/pith/WALUUJPAWITKAAYHWTQYHUJWI4/action/citation_signature","submit_replication":"https://pith.science/pith/WALUUJPAWITKAAYHWTQYHUJWI4/action/replication_record"}},"created_at":"2026-05-26T02:05:21.852555+00:00","updated_at":"2026-05-26T02:05:21.852555+00:00"}