{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:JYM7UZ4ACXTSYU4YMU6QUC3L2R","short_pith_number":"pith:JYM7UZ4A","canonical_record":{"source":{"id":"2406.05644","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-09T05:04:37Z","cross_cats_sorted":["cs.AI","cs.CR","cs.CY"],"title_canon_sha256":"0c9c82d1d8aac8f68d7da6ecf925ca6eb533b08fee4b8979043f316ab867fab2","abstract_canon_sha256":"7920e54f5969de4dbbe044f81efe7bc1e21d97c91b79be996b8b98785eceb5a0"},"schema_version":"1.0"},"canonical_sha256":"4e19fa678015e72c5398653d0a0b6bd45e5881e6834685e44d309aa38cdfe77f","source":{"kind":"arxiv","id":"2406.05644","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.05644","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"arxiv_version","alias_value":"2406.05644v2","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.05644","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"pith_short_12","alias_value":"JYM7UZ4ACXTS","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"pith_short_16","alias_value":"JYM7UZ4ACXTSYU4Y","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"pith_short_8","alias_value":"JYM7UZ4A","created_at":"2026-07-05T08:31:16Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:JYM7UZ4ACXTSYU4YMU6QUC3L2R","target":"record","payload":{"canonical_record":{"source":{"id":"2406.05644","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-09T05:04:37Z","cross_cats_sorted":["cs.AI","cs.CR","cs.CY"],"title_canon_sha256":"0c9c82d1d8aac8f68d7da6ecf925ca6eb533b08fee4b8979043f316ab867fab2","abstract_canon_sha256":"7920e54f5969de4dbbe044f81efe7bc1e21d97c91b79be996b8b98785eceb5a0"},"schema_version":"1.0"},"canonical_sha256":"4e19fa678015e72c5398653d0a0b6bd45e5881e6834685e44d309aa38cdfe77f","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:31:16.378905Z","signature_b64":"VV2Z6c9gb+89nSPQMKXowbrFOgNuKrURMA6VFtBA+IF0w1WH93Dze6eVYTBCjC9CaviZ7vjcNKn8DlDA7DbrAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4e19fa678015e72c5398653d0a0b6bd45e5881e6834685e44d309aa38cdfe77f","last_reissued_at":"2026-07-05T08:31:16.378420Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:31:16.378420Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2406.05644","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:31:16Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"I9EDXx5xN2nKbNB5oKHygvrF+iJyUWfCqpnqd3QC3bFKVeYMkVoJ8hbU6aOSHvbJyWbasF2rhG6aVHM78uvMDw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T01:49:01.164750Z"},"content_sha256":"f024d2b5a73c8dad78f530aa5293f8f36213618aead6f3438efb030373401ab0","schema_version":"1.0","event_id":"sha256:f024d2b5a73c8dad78f530aa5293f8f36213618aead6f3438efb030373401ab0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:JYM7UZ4ACXTSYU4YMU6QUC3L2R","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"How Alignment and Jailbreak Work: Explain LLM Safety through Intermediate Hidden States","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR","cs.CY"],"primary_cat":"cs.CL","authors_text":"Fei Huang, Haiyang Yu, Rongwu Xu, Xinghua Zhang, Yongbin Li, Zhenhong Zhou","submitted_at":"2024-06-09T05:04:37Z","abstract_excerpt":"Large language models (LLMs) rely on safety alignment to avoid responding to malicious user inputs. Unfortunately, jailbreak can circumvent safety guardrails, resulting in LLMs generating harmful content and raising concerns about LLM safety. Due to language models with intensive parameters often regarded as black boxes, the mechanisms of alignment and jailbreak are challenging to elucidate. In this paper, we employ weak classifiers to explain LLM safety through the intermediate hidden states. We first confirm that LLMs learn ethical concepts during pre-training rather than alignment and can i"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.05644","kind":"arxiv","version":2},"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/2406.05644/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T08:31:16Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7u3n5DOtueToMtwX4RKTAaZEWM5fr6hTA8oc0olACJ37s8hM+y5mZ4UAPe2T3bQgNEU+vZh3Ig6juPtmhw96AA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T01:49:01.165254Z"},"content_sha256":"3decc2028569a1b7fe81b571e6c12de390649395c2308e1c328fc57f79848f98","schema_version":"1.0","event_id":"sha256:3decc2028569a1b7fe81b571e6c12de390649395c2308e1c328fc57f79848f98"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/JYM7UZ4ACXTSYU4YMU6QUC3L2R/bundle.json","state_url":"https://pith.science/pith/JYM7UZ4ACXTSYU4YMU6QUC3L2R/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/JYM7UZ4ACXTSYU4YMU6QUC3L2R/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-10T01:49:01Z","links":{"resolver":"https://pith.science/pith/JYM7UZ4ACXTSYU4YMU6QUC3L2R","bundle":"https://pith.science/pith/JYM7UZ4ACXTSYU4YMU6QUC3L2R/bundle.json","state":"https://pith.science/pith/JYM7UZ4ACXTSYU4YMU6QUC3L2R/state.json","well_known_bundle":"https://pith.science/.well-known/pith/JYM7UZ4ACXTSYU4YMU6QUC3L2R/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:JYM7UZ4ACXTSYU4YMU6QUC3L2R","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":"7920e54f5969de4dbbe044f81efe7bc1e21d97c91b79be996b8b98785eceb5a0","cross_cats_sorted":["cs.AI","cs.CR","cs.CY"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-09T05:04:37Z","title_canon_sha256":"0c9c82d1d8aac8f68d7da6ecf925ca6eb533b08fee4b8979043f316ab867fab2"},"schema_version":"1.0","source":{"id":"2406.05644","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.05644","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"arxiv_version","alias_value":"2406.05644v2","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.05644","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"pith_short_12","alias_value":"JYM7UZ4ACXTS","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"pith_short_16","alias_value":"JYM7UZ4ACXTSYU4Y","created_at":"2026-07-05T08:31:16Z"},{"alias_kind":"pith_short_8","alias_value":"JYM7UZ4A","created_at":"2026-07-05T08:31:16Z"}],"graph_snapshots":[{"event_id":"sha256:3decc2028569a1b7fe81b571e6c12de390649395c2308e1c328fc57f79848f98","target":"graph","created_at":"2026-07-05T08:31:16Z","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/2406.05644/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Large language models (LLMs) rely on safety alignment to avoid responding to malicious user inputs. Unfortunately, jailbreak can circumvent safety guardrails, resulting in LLMs generating harmful content and raising concerns about LLM safety. Due to language models with intensive parameters often regarded as black boxes, the mechanisms of alignment and jailbreak are challenging to elucidate. In this paper, we employ weak classifiers to explain LLM safety through the intermediate hidden states. We first confirm that LLMs learn ethical concepts during pre-training rather than alignment and can i","authors_text":"Fei Huang, Haiyang Yu, Rongwu Xu, Xinghua Zhang, Yongbin Li, Zhenhong Zhou","cross_cats":["cs.AI","cs.CR","cs.CY"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-09T05:04:37Z","title":"How Alignment and Jailbreak Work: Explain LLM Safety through Intermediate Hidden States"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.05644","kind":"arxiv","version":2},"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:f024d2b5a73c8dad78f530aa5293f8f36213618aead6f3438efb030373401ab0","target":"record","created_at":"2026-07-05T08:31:16Z","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":"7920e54f5969de4dbbe044f81efe7bc1e21d97c91b79be996b8b98785eceb5a0","cross_cats_sorted":["cs.AI","cs.CR","cs.CY"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-09T05:04:37Z","title_canon_sha256":"0c9c82d1d8aac8f68d7da6ecf925ca6eb533b08fee4b8979043f316ab867fab2"},"schema_version":"1.0","source":{"id":"2406.05644","kind":"arxiv","version":2}},"canonical_sha256":"4e19fa678015e72c5398653d0a0b6bd45e5881e6834685e44d309aa38cdfe77f","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"4e19fa678015e72c5398653d0a0b6bd45e5881e6834685e44d309aa38cdfe77f","first_computed_at":"2026-07-05T08:31:16.378420Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T08:31:16.378420Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"VV2Z6c9gb+89nSPQMKXowbrFOgNuKrURMA6VFtBA+IF0w1WH93Dze6eVYTBCjC9CaviZ7vjcNKn8DlDA7DbrAA==","signature_status":"signed_v1","signed_at":"2026-07-05T08:31:16.378905Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.05644","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f024d2b5a73c8dad78f530aa5293f8f36213618aead6f3438efb030373401ab0","sha256:3decc2028569a1b7fe81b571e6c12de390649395c2308e1c328fc57f79848f98"],"state_sha256":"865d99bdd0e54ef1f4f470e63bd5b0d799b092aa519042643b1640bf07c635dc"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"3ivuUe+R4b5tG059MoG4+w7bZ7MQWAw4NHhw/IlmKfvPyK+nq/F90pgVZHnmDjl49x6kIb9GE5X0rANc2MoPDg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T01:49:01.171838Z","bundle_sha256":"ddcfd93137632a448bac5f3ad443d5f1bdb015a42c93de40e4a93b71407097ef"}}