{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:GKNC4X4UYBCVVNUYGKXK2VUJ2Z","short_pith_number":"pith:GKNC4X4U","canonical_record":{"source":{"id":"2504.18053","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-25T03:54:24Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"14ef71b780e0506f8a64dd01279194dbb90c7c136ae38f0dd060b2d05058adad","abstract_canon_sha256":"0129ffff3508ca8724ca716105f1796b89c87d73233abe5e682a167664d2b2ad"},"schema_version":"1.0"},"canonical_sha256":"329a2e5f94c0455ab69832aead5689d660347c30293f384c6fcb907248f2f523","source":{"kind":"arxiv","id":"2504.18053","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.18053","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"arxiv_version","alias_value":"2504.18053v2","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.18053","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"pith_short_12","alias_value":"GKNC4X4UYBCV","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"pith_short_16","alias_value":"GKNC4X4UYBCVVNUY","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"pith_short_8","alias_value":"GKNC4X4U","created_at":"2026-07-05T11:16:33Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:GKNC4X4UYBCVVNUYGKXK2VUJ2Z","target":"record","payload":{"canonical_record":{"source":{"id":"2504.18053","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-25T03:54:24Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"14ef71b780e0506f8a64dd01279194dbb90c7c136ae38f0dd060b2d05058adad","abstract_canon_sha256":"0129ffff3508ca8724ca716105f1796b89c87d73233abe5e682a167664d2b2ad"},"schema_version":"1.0"},"canonical_sha256":"329a2e5f94c0455ab69832aead5689d660347c30293f384c6fcb907248f2f523","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:16:33.456875Z","signature_b64":"03+d+ntyJxL+eCz0qpWOn0F5KtOEUkhTfIPFOooe0PwXqGlGn94jx5KUF4E4QuJW1wAObrHEwAjtbPyWd3hrDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"329a2e5f94c0455ab69832aead5689d660347c30293f384c6fcb907248f2f523","last_reissued_at":"2026-07-05T11:16:33.456321Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:16:33.456321Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2504.18053","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":"2026-07-05T11:16:33Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Wvj0tPwuDBm3082AbQbz2CNVCqLprj52mytvzBDbHV2rxtfEQo4HF5vw/p69xYHVqsjcypg3xPuJSLvTeeHsAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T12:39:09.635023Z"},"content_sha256":"b75e1c53f49e554b93a98a7918f5440a58fd614263c514fc85b785ac5042788a","schema_version":"1.0","event_id":"sha256:b75e1c53f49e554b93a98a7918f5440a58fd614263c514fc85b785ac5042788a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:GKNC4X4UYBCVVNUYGKXK2VUJ2Z","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"DREAM: Disentangling Risks to Enhance Safety Alignment in Multimodal Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Chenchen Jing, Hangyu Guo, Hui Huang, Jiaheng Liu, Jianke Zhu, Jianyu Liu, Jihao Gu, Ranjie Duan, Shilong Li, Xiao Zhang, Xingwei Qu, Xingyuan Bu, Yanan Wu, Yancheng He, Yangguang Li, Yingshui Tan, Yucheng Wang","submitted_at":"2025-04-25T03:54:24Z","abstract_excerpt":"Multimodal Large Language Models (MLLMs) pose unique safety challenges due to their integration of visual and textual data, thereby introducing new dimensions of potential attacks and complex risk combinations. In this paper, we begin with a detailed analysis aimed at disentangling risks through step-by-step reasoning within multimodal inputs. We find that systematic multimodal risk disentanglement substantially enhances the risk awareness of MLLMs. Via leveraging the strong discriminative abilities of multimodal risk disentanglement, we further introduce \\textbf{DREAM} (\\textit{\\textbf{D}isen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.18053","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/2504.18053/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":"2026-07-05T11:16:33Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qIFJQ1RXFWrEgw/IGoGRk2267B1CB/opV5zWpsYEM6vbDX489xyI4vSvxMgyelSZj1YLSjnEFVHKFG67mkJNDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-10T12:39:09.635572Z"},"content_sha256":"ec2a19667ec64bd56f012a318cf53972c8e1db6c70e6b3e1ec2cb87e708794d5","schema_version":"1.0","event_id":"sha256:ec2a19667ec64bd56f012a318cf53972c8e1db6c70e6b3e1ec2cb87e708794d5"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/GKNC4X4UYBCVVNUYGKXK2VUJ2Z/bundle.json","state_url":"https://pith.science/pith/GKNC4X4UYBCVVNUYGKXK2VUJ2Z/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/GKNC4X4UYBCVVNUYGKXK2VUJ2Z/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":"2026-08-10T12:39:09Z","links":{"resolver":"https://pith.science/pith/GKNC4X4UYBCVVNUYGKXK2VUJ2Z","bundle":"https://pith.science/pith/GKNC4X4UYBCVVNUYGKXK2VUJ2Z/bundle.json","state":"https://pith.science/pith/GKNC4X4UYBCVVNUYGKXK2VUJ2Z/state.json","well_known_bundle":"https://pith.science/.well-known/pith/GKNC4X4UYBCVVNUYGKXK2VUJ2Z/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:GKNC4X4UYBCVVNUYGKXK2VUJ2Z","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":"0129ffff3508ca8724ca716105f1796b89c87d73233abe5e682a167664d2b2ad","cross_cats_sorted":["cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-25T03:54:24Z","title_canon_sha256":"14ef71b780e0506f8a64dd01279194dbb90c7c136ae38f0dd060b2d05058adad"},"schema_version":"1.0","source":{"id":"2504.18053","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2504.18053","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"arxiv_version","alias_value":"2504.18053v2","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.18053","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"pith_short_12","alias_value":"GKNC4X4UYBCV","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"pith_short_16","alias_value":"GKNC4X4UYBCVVNUY","created_at":"2026-07-05T11:16:33Z"},{"alias_kind":"pith_short_8","alias_value":"GKNC4X4U","created_at":"2026-07-05T11:16:33Z"}],"graph_snapshots":[{"event_id":"sha256:ec2a19667ec64bd56f012a318cf53972c8e1db6c70e6b3e1ec2cb87e708794d5","target":"graph","created_at":"2026-07-05T11:16:33Z","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/2504.18053/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Multimodal Large Language Models (MLLMs) pose unique safety challenges due to their integration of visual and textual data, thereby introducing new dimensions of potential attacks and complex risk combinations. In this paper, we begin with a detailed analysis aimed at disentangling risks through step-by-step reasoning within multimodal inputs. We find that systematic multimodal risk disentanglement substantially enhances the risk awareness of MLLMs. Via leveraging the strong discriminative abilities of multimodal risk disentanglement, we further introduce \\textbf{DREAM} (\\textit{\\textbf{D}isen","authors_text":"Chenchen Jing, Hangyu Guo, Hui Huang, Jiaheng Liu, Jianke Zhu, Jianyu Liu, Jihao Gu, Ranjie Duan, Shilong Li, Xiao Zhang, Xingwei Qu, Xingyuan Bu, Yanan Wu, Yancheng He, Yangguang Li, Yingshui Tan, Yucheng Wang","cross_cats":["cs.CV"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-25T03:54:24Z","title":"DREAM: Disentangling Risks to Enhance Safety Alignment in Multimodal Large Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.18053","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:b75e1c53f49e554b93a98a7918f5440a58fd614263c514fc85b785ac5042788a","target":"record","created_at":"2026-07-05T11:16:33Z","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":"0129ffff3508ca8724ca716105f1796b89c87d73233abe5e682a167664d2b2ad","cross_cats_sorted":["cs.CV"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-04-25T03:54:24Z","title_canon_sha256":"14ef71b780e0506f8a64dd01279194dbb90c7c136ae38f0dd060b2d05058adad"},"schema_version":"1.0","source":{"id":"2504.18053","kind":"arxiv","version":2}},"canonical_sha256":"329a2e5f94c0455ab69832aead5689d660347c30293f384c6fcb907248f2f523","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"329a2e5f94c0455ab69832aead5689d660347c30293f384c6fcb907248f2f523","first_computed_at":"2026-07-05T11:16:33.456321Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:16:33.456321Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"03+d+ntyJxL+eCz0qpWOn0F5KtOEUkhTfIPFOooe0PwXqGlGn94jx5KUF4E4QuJW1wAObrHEwAjtbPyWd3hrDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T11:16:33.456875Z","signed_message":"canonical_sha256_bytes"},"source_id":"2504.18053","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:b75e1c53f49e554b93a98a7918f5440a58fd614263c514fc85b785ac5042788a","sha256:ec2a19667ec64bd56f012a318cf53972c8e1db6c70e6b3e1ec2cb87e708794d5"],"state_sha256":"12faacac584fdb14eb70d74b5ab89b6ae65fba0cc9ce986d9d4ee45fabae0f1e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"ILGzKY8gYTz0S8cuGeoHnm2sHpH08Z3270rLejtUM4NLvWnHN6Yy3wTXcgbbawBizih513WIH1W+cbsRxc0pBw==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-10T12:39:09.640861Z","bundle_sha256":"eb548c27db6c31a8608561f7e8f8554c828348732d6e2f1fc2691ec98b67f18f"}}