{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2024:HZTMZMWGYLDZARKJM5RF2SZ5QP","short_pith_number":"pith:HZTMZMWG","canonical_record":{"source":{"id":"2412.02237","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-12-03T08:05:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b8498acf7fac8f54aaf28ed76d6f579f8a6b612a923e4bb694b99ef5c5ebe041","abstract_canon_sha256":"4c796eed92c88355a4cb82598e4792eb0bf70b44cbd3a1d0cafa94cf2b805271"},"schema_version":"1.0"},"canonical_sha256":"3e66ccb2c6c2c790454967625d4b3d83cdc8fb936b042b81b5ecfdf0200cc493","source":{"kind":"arxiv","id":"2412.02237","version":3},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2412.02237","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"arxiv_version","alias_value":"2412.02237v3","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.02237","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"pith_short_12","alias_value":"HZTMZMWGYLDZ","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"pith_short_16","alias_value":"HZTMZMWGYLDZARKJ","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"pith_short_8","alias_value":"HZTMZMWG","created_at":"2026-07-05T10:18:45Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2024:HZTMZMWGYLDZARKJM5RF2SZ5QP","target":"record","payload":{"canonical_record":{"source":{"id":"2412.02237","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-12-03T08:05:56Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b8498acf7fac8f54aaf28ed76d6f579f8a6b612a923e4bb694b99ef5c5ebe041","abstract_canon_sha256":"4c796eed92c88355a4cb82598e4792eb0bf70b44cbd3a1d0cafa94cf2b805271"},"schema_version":"1.0"},"canonical_sha256":"3e66ccb2c6c2c790454967625d4b3d83cdc8fb936b042b81b5ecfdf0200cc493","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:18:45.984171Z","signature_b64":"aZnj6v+JSzvLn/s04QuKtIvr6qNLR3x2bwmSjzRSuWd0xhuNsW1ZUPLn/Cnuh/cEKBpPQgPPmgTHxdIVUO2DDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3e66ccb2c6c2c790454967625d4b3d83cdc8fb936b042b81b5ecfdf0200cc493","last_reissued_at":"2026-07-05T10:18:45.983650Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:18:45.983650Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2412.02237","source_version":3,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:18:45Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"O3UML47xxCozrUfEZbPF+bKOt7Unc19jcWqx8TEu9KgRBWHg593yXIhLc7GFLeJJPMM4I/EDJnGyAGOThOgYBQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-20T02:40:17.710871Z"},"content_sha256":"5d68c3f4de44d3c9e61011b19f3d7be980606db09e18876a4bb9a6a29c2bb21a","schema_version":"1.0","event_id":"sha256:5d68c3f4de44d3c9e61011b19f3d7be980606db09e18876a4bb9a6a29c2bb21a"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2024:HZTMZMWGYLDZARKJM5RF2SZ5QP","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Cross-Attention Head Position Patterns Can Align with Human Visual Concepts in Text-to-Image Generative Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Dongnam Byun, Jangwon Suh, Jungmin Ko, Jungwon Park, Wonjong Rhee","submitted_at":"2024-12-03T08:05:56Z","abstract_excerpt":"Recent text-to-image diffusion models leverage cross-attention layers, which have been effectively utilized to enhance a range of visual generative tasks. However, our understanding of cross-attention layers remains somewhat limited. In this study, we introduce a mechanistic interpretability approach for diffusion models by constructing Head Relevance Vectors (HRVs) that align with human-specified visual concepts. An HRV for a given visual concept has a length equal to the total number of cross-attention heads, with each element representing the importance of the corresponding head for the giv"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.02237","kind":"arxiv","version":3},"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/2412.02237/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":"2026-07-05T10:18:45Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"+gti2fzBHrAonbHfTYiWiM/Z8KeYH6yIrMRy2K8MKlsnG4pl2AN2/bOKlrV5bRMuIKGNu8h9+ePARm0XKiOIAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-20T02:40:17.711399Z"},"content_sha256":"a9a104a4f2fdcb86f9190c61946f800cc744b3cdd4b268c7f8543e0aa8309284","schema_version":"1.0","event_id":"sha256:a9a104a4f2fdcb86f9190c61946f800cc744b3cdd4b268c7f8543e0aa8309284"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/HZTMZMWGYLDZARKJM5RF2SZ5QP/bundle.json","state_url":"https://pith.science/pith/HZTMZMWGYLDZARKJM5RF2SZ5QP/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/HZTMZMWGYLDZARKJM5RF2SZ5QP/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":"2026-08-20T02:40:17Z","links":{"resolver":"https://pith.science/pith/HZTMZMWGYLDZARKJM5RF2SZ5QP","bundle":"https://pith.science/pith/HZTMZMWGYLDZARKJM5RF2SZ5QP/bundle.json","state":"https://pith.science/pith/HZTMZMWGYLDZARKJM5RF2SZ5QP/state.json","well_known_bundle":"https://pith.science/.well-known/pith/HZTMZMWGYLDZARKJM5RF2SZ5QP/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2024:HZTMZMWGYLDZARKJM5RF2SZ5QP","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":"4c796eed92c88355a4cb82598e4792eb0bf70b44cbd3a1d0cafa94cf2b805271","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-12-03T08:05:56Z","title_canon_sha256":"b8498acf7fac8f54aaf28ed76d6f579f8a6b612a923e4bb694b99ef5c5ebe041"},"schema_version":"1.0","source":{"id":"2412.02237","kind":"arxiv","version":3}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2412.02237","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"arxiv_version","alias_value":"2412.02237v3","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.02237","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"pith_short_12","alias_value":"HZTMZMWGYLDZ","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"pith_short_16","alias_value":"HZTMZMWGYLDZARKJ","created_at":"2026-07-05T10:18:45Z"},{"alias_kind":"pith_short_8","alias_value":"HZTMZMWG","created_at":"2026-07-05T10:18:45Z"}],"graph_snapshots":[{"event_id":"sha256:a9a104a4f2fdcb86f9190c61946f800cc744b3cdd4b268c7f8543e0aa8309284","target":"graph","created_at":"2026-07-05T10:18:45Z","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/2412.02237/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Recent text-to-image diffusion models leverage cross-attention layers, which have been effectively utilized to enhance a range of visual generative tasks. However, our understanding of cross-attention layers remains somewhat limited. In this study, we introduce a mechanistic interpretability approach for diffusion models by constructing Head Relevance Vectors (HRVs) that align with human-specified visual concepts. An HRV for a given visual concept has a length equal to the total number of cross-attention heads, with each element representing the importance of the corresponding head for the giv","authors_text":"Dongnam Byun, Jangwon Suh, Jungmin Ko, Jungwon Park, Wonjong Rhee","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-12-03T08:05:56Z","title":"Cross-Attention Head Position Patterns Can Align with Human Visual Concepts in Text-to-Image Generative Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.02237","kind":"arxiv","version":3},"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":"sha256:5d68c3f4de44d3c9e61011b19f3d7be980606db09e18876a4bb9a6a29c2bb21a","target":"record","created_at":"2026-07-05T10:18:45Z","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":"4c796eed92c88355a4cb82598e4792eb0bf70b44cbd3a1d0cafa94cf2b805271","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-12-03T08:05:56Z","title_canon_sha256":"b8498acf7fac8f54aaf28ed76d6f579f8a6b612a923e4bb694b99ef5c5ebe041"},"schema_version":"1.0","source":{"id":"2412.02237","kind":"arxiv","version":3}},"canonical_sha256":"3e66ccb2c6c2c790454967625d4b3d83cdc8fb936b042b81b5ecfdf0200cc493","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3e66ccb2c6c2c790454967625d4b3d83cdc8fb936b042b81b5ecfdf0200cc493","first_computed_at":"2026-07-05T10:18:45.983650Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:18:45.983650Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"aZnj6v+JSzvLn/s04QuKtIvr6qNLR3x2bwmSjzRSuWd0xhuNsW1ZUPLn/Cnuh/cEKBpPQgPPmgTHxdIVUO2DDQ==","signature_status":"signed_v1","signed_at":"2026-07-05T10:18:45.984171Z","signed_message":"canonical_sha256_bytes"},"source_id":"2412.02237","source_kind":"arxiv","source_version":3}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5d68c3f4de44d3c9e61011b19f3d7be980606db09e18876a4bb9a6a29c2bb21a","sha256:a9a104a4f2fdcb86f9190c61946f800cc744b3cdd4b268c7f8543e0aa8309284"],"state_sha256":"b9fb500b864fdafb37842e2aec0691b3f23cbe192b530644e37dd0d90588d37a"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"TdEC+2mP/seJLth4Ry01977tpsuJXwElZPm7t7NQe5g3lhl4VN4U7ZUmRzXPIcK8oC4FKvOZr2DnNS9RwAjWBQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-20T02:40:17.716396Z","bundle_sha256":"9679c5f2af27163274424cee6604c7b1b0a1e94f561eae680d75d5a795b0d170"}}