{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:46ZH5J3DCSHDZ5IZJIQNFGPWKY","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"2764522f062cb65b72d1d80f8268bf772c217383bbfe748685d269510792a907","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-20T03:16:15Z","title_canon_sha256":"6a07405683e4a1960a6b9c251f529d38836165cf38021aeff1203c88d960b62b"},"schema_version":"1.0","source":{"id":"2605.20654","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.20654","created_at":"2026-05-21T01:04:47Z"},{"alias_kind":"arxiv_version","alias_value":"2605.20654v1","created_at":"2026-05-21T01:04:47Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.20654","created_at":"2026-05-21T01:04:47Z"},{"alias_kind":"pith_short_12","alias_value":"46ZH5J3DCSHD","created_at":"2026-05-21T01:04:47Z"},{"alias_kind":"pith_short_16","alias_value":"46ZH5J3DCSHDZ5IZ","created_at":"2026-05-21T01:04:47Z"},{"alias_kind":"pith_short_8","alias_value":"46ZH5J3D","created_at":"2026-05-21T01:04:47Z"}],"graph_snapshots":[{"event_id":"sha256:28a961c3b45ca9185b9d9a5d7e8244ad3cf0ba2806916dba2fa0623d73415a34","target":"graph","created_at":"2026-05-21T01:04:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2605.20654/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"While Large Language Models (LLMs) demonstrate remarkable capabilities, they remain susceptible to sophisticated, multi-step jailbreak attacks that circumvent conventional surface-level safety alignment by exploiting the internal generation process. To address these vulnerabilities, we propose Reflector, a principled two-stage framework that internalizes self-reflection within the generation trajectory. Reflector first leverages teacher-guided generation to produce high-quality reflection data for supervised fine-tuning (SFT), establishing structured reflection patterns. It subsequently uses R","authors_text":"Bo Zou, Chaochao Lu, Chao Yang, Jiachen Ma, Jiawen Zhang, Xiangtian Li","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-20T03:16:15Z","title":"REFLECTOR: Internalizing Step-wise Reflection against Indirect Jailbreak"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.20654","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:ac03ea71ccc33d7610923ca49574447d0add99bd43b4b6d580a1344699ffece3","target":"record","created_at":"2026-05-21T01:04:47Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"2764522f062cb65b72d1d80f8268bf772c217383bbfe748685d269510792a907","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-20T03:16:15Z","title_canon_sha256":"6a07405683e4a1960a6b9c251f529d38836165cf38021aeff1203c88d960b62b"},"schema_version":"1.0","source":{"id":"2605.20654","kind":"arxiv","version":1}},"canonical_sha256":"e7b27ea763148e3cf5194a20d299f65632e04095210591de3889ef9c922df6d4","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"e7b27ea763148e3cf5194a20d299f65632e04095210591de3889ef9c922df6d4","first_computed_at":"2026-05-21T01:04:47.238834Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-21T01:04:47.238834Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"UlwFxTP8qClEC44ofOpZdpnueVHlzxLGfziyBwz3mzf7WQEl21VMQxng/mxnTscjP9EeIKO05xcTGen5sgR2BQ==","signature_status":"signed_v1","signed_at":"2026-05-21T01:04:47.239587Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.20654","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ac03ea71ccc33d7610923ca49574447d0add99bd43b4b6d580a1344699ffece3","sha256:28a961c3b45ca9185b9d9a5d7e8244ad3cf0ba2806916dba2fa0623d73415a34"],"state_sha256":"4937e971959588192cad565054b76d753c3b1bc5e3bc1b68726cc3375e53b757"}