{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:PV7FUORUXPF6UTL22FKJHFMOM4","short_pith_number":"pith:PV7FUORU","canonical_record":{"source":{"id":"2406.14393","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-20T15:12:27Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"6dec27b904ea45aa3a41146713abd7d708c48459cbfda4303f6241d7b744f661","abstract_canon_sha256":"46d2ccb7dec9b5aecb43af69910fbd1552a174443131f1809ded55e6f22f88d5"},"schema_version":"1.0"},"canonical_sha256":"7d7e5a3a34bbcbea4d7ad15493958e671fa87d98128063eb058b764cd99a9acd","source":{"kind":"arxiv","id":"2406.14393","version":5},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.14393","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"arxiv_version","alias_value":"2406.14393v5","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.14393","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"pith_short_12","alias_value":"PV7FUORUXPF6","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"pith_short_16","alias_value":"PV7FUORUXPF6UTL2","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"pith_short_8","alias_value":"PV7FUORU","created_at":"2026-07-05T10:51:23Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:PV7FUORUXPF6UTL22FKJHFMOM4","target":"record","payload":{"canonical_record":{"source":{"id":"2406.14393","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-20T15:12:27Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"6dec27b904ea45aa3a41146713abd7d708c48459cbfda4303f6241d7b744f661","abstract_canon_sha256":"46d2ccb7dec9b5aecb43af69910fbd1552a174443131f1809ded55e6f22f88d5"},"schema_version":"1.0"},"canonical_sha256":"7d7e5a3a34bbcbea4d7ad15493958e671fa87d98128063eb058b764cd99a9acd","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:51:23.406073Z","signature_b64":"89FQPJh5+kkPotux1Oi83OQhfSDgmJKti9uEXf1Gu0KJAStLGD2ahiFPh4k8zETfcbgDCMdZ180T+Hr/XHdWBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7d7e5a3a34bbcbea4d7ad15493958e671fa87d98128063eb058b764cd99a9acd","last_reissued_at":"2026-07-05T10:51:23.405581Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:51:23.405581Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2406.14393","source_version":5,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:51:23Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"E+ckhmTNq1TS2nKm1Y+41I59ifEbbLsMpnFqutbI0CPviYUPG8gYUDmBqvCjm8srUIl7fvCNsX3RfWq/8EJ4Cg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T07:13:34.368857Z"},"content_sha256":"8405d326a918ce6cabe264958c9b43d9fb7ba9442fca04ac89b7b3d57b184ef0","schema_version":"1.0","event_id":"sha256:8405d326a918ce6cabe264958c9b43d9fb7ba9442fca04ac89b7b3d57b184ef0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:PV7FUORUXPF6UTL22FKJHFMOM4","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Jailbreaking as a Reward Misspecification Problem","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Jiahui Gao, Lei Li, Lingpeng Kong, Qi Liu, Zhenguo Li, Zhihui Xie","submitted_at":"2024-06-20T15:12:27Z","abstract_excerpt":"The widespread adoption of large language models (LLMs) has raised concerns about their safety and reliability, particularly regarding their vulnerability to adversarial attacks. In this paper, we propose a novel perspective that attributes this vulnerability to reward misspecification during the alignment process. This misspecification occurs when the reward function fails to accurately capture the intended behavior, leading to misaligned model outputs. We introduce a metric ReGap to quantify the extent of reward misspecification and demonstrate its effectiveness and robustness in detecting h"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.14393","kind":"arxiv","version":5},"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/2406.14393/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":"2026-07-05T10:51:23Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LA59d+95CNh73B9jH8VFctBOQj2viA3mo8g8UIU/RQNivjH1t9Bth+EVaigPvLDd25q7hqEPvvFzUSh5L1TnBg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-05T07:13:34.369400Z"},"content_sha256":"27e29effad86aee5eab245c482798f5f44534b610a4877f2de63539f8328b47a","schema_version":"1.0","event_id":"sha256:27e29effad86aee5eab245c482798f5f44534b610a4877f2de63539f8328b47a"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/PV7FUORUXPF6UTL22FKJHFMOM4/bundle.json","state_url":"https://pith.science/pith/PV7FUORUXPF6UTL22FKJHFMOM4/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/PV7FUORUXPF6UTL22FKJHFMOM4/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":"2026-08-05T07:13:34Z","links":{"resolver":"https://pith.science/pith/PV7FUORUXPF6UTL22FKJHFMOM4","bundle":"https://pith.science/pith/PV7FUORUXPF6UTL22FKJHFMOM4/bundle.json","state":"https://pith.science/pith/PV7FUORUXPF6UTL22FKJHFMOM4/state.json","well_known_bundle":"https://pith.science/.well-known/pith/PV7FUORUXPF6UTL22FKJHFMOM4/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:PV7FUORUXPF6UTL22FKJHFMOM4","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":"46d2ccb7dec9b5aecb43af69910fbd1552a174443131f1809ded55e6f22f88d5","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-20T15:12:27Z","title_canon_sha256":"6dec27b904ea45aa3a41146713abd7d708c48459cbfda4303f6241d7b744f661"},"schema_version":"1.0","source":{"id":"2406.14393","kind":"arxiv","version":5}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2406.14393","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"arxiv_version","alias_value":"2406.14393v5","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.14393","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"pith_short_12","alias_value":"PV7FUORUXPF6","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"pith_short_16","alias_value":"PV7FUORUXPF6UTL2","created_at":"2026-07-05T10:51:23Z"},{"alias_kind":"pith_short_8","alias_value":"PV7FUORU","created_at":"2026-07-05T10:51:23Z"}],"graph_snapshots":[{"event_id":"sha256:27e29effad86aee5eab245c482798f5f44534b610a4877f2de63539f8328b47a","target":"graph","created_at":"2026-07-05T10:51:23Z","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/2406.14393/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"The widespread adoption of large language models (LLMs) has raised concerns about their safety and reliability, particularly regarding their vulnerability to adversarial attacks. In this paper, we propose a novel perspective that attributes this vulnerability to reward misspecification during the alignment process. This misspecification occurs when the reward function fails to accurately capture the intended behavior, leading to misaligned model outputs. We introduce a metric ReGap to quantify the extent of reward misspecification and demonstrate its effectiveness and robustness in detecting h","authors_text":"Jiahui Gao, Lei Li, Lingpeng Kong, Qi Liu, Zhenguo Li, Zhihui Xie","cross_cats":["cs.CL"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-20T15:12:27Z","title":"Jailbreaking as a Reward Misspecification Problem"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.14393","kind":"arxiv","version":5},"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":"sha256:8405d326a918ce6cabe264958c9b43d9fb7ba9442fca04ac89b7b3d57b184ef0","target":"record","created_at":"2026-07-05T10:51:23Z","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":"46d2ccb7dec9b5aecb43af69910fbd1552a174443131f1809ded55e6f22f88d5","cross_cats_sorted":["cs.CL"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-06-20T15:12:27Z","title_canon_sha256":"6dec27b904ea45aa3a41146713abd7d708c48459cbfda4303f6241d7b744f661"},"schema_version":"1.0","source":{"id":"2406.14393","kind":"arxiv","version":5}},"canonical_sha256":"7d7e5a3a34bbcbea4d7ad15493958e671fa87d98128063eb058b764cd99a9acd","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"7d7e5a3a34bbcbea4d7ad15493958e671fa87d98128063eb058b764cd99a9acd","first_computed_at":"2026-07-05T10:51:23.405581Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:51:23.405581Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"89FQPJh5+kkPotux1Oi83OQhfSDgmJKti9uEXf1Gu0KJAStLGD2ahiFPh4k8zETfcbgDCMdZ180T+Hr/XHdWBA==","signature_status":"signed_v1","signed_at":"2026-07-05T10:51:23.406073Z","signed_message":"canonical_sha256_bytes"},"source_id":"2406.14393","source_kind":"arxiv","source_version":5}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:8405d326a918ce6cabe264958c9b43d9fb7ba9442fca04ac89b7b3d57b184ef0","sha256:27e29effad86aee5eab245c482798f5f44534b610a4877f2de63539f8328b47a"],"state_sha256":"0858bb1f433a84b1c3dda062dc5d0c8d8c926200f99955230a65b348ebffc2a3"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"6V6+WXH1/Fd0UL3wcf+dvHHeufHm7DLvOZKL2tW5sc15e7/cjP6NG5dLzBYMRVLRsqGp+cEVsmCyjZj42z0XAQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-05T07:13:34.376923Z","bundle_sha256":"2df1d9491270d7d34caf75649bc7836fc3ca6de3699b19b174cd360a75a359e3"}}