{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:DCDLBW6K2Y33KHSMHKIOYPIRDI","short_pith_number":"pith:DCDLBW6K","schema_version":"1.0","canonical_sha256":"1886b0dbcad637b51e4c3a90ec3d111a28ea20a602d12feab6ca7087fd8e89da","source":{"kind":"arxiv","id":"2507.18512","version":1},"attestation_state":"computed","paper":{"title":"Explaining How Visual, Textual and Multimodal Encoders Share Concepts","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Cl\\'ement Cornet, Herv\\'e Le Borgne, Romaric Besan\\c{c}on","submitted_at":"2025-07-24T15:33:31Z","abstract_excerpt":"Sparse autoencoders (SAEs) have emerged as a powerful technique for extracting human-interpretable features from neural networks activations. Previous works compared different models based on SAE-derived features but those comparisons have been restricted to models within the same modality. We propose a novel indicator allowing quantitative comparison of models across SAE features, and use it to conduct a comparative study of visual, textual and multimodal encoders. We also propose to quantify the Comparative Sharedness of individual features between different classes of models. With these two"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2507.18512","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CV","submitted_at":"2025-07-24T15:33:31Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"83ecc8c67721ec520fc124c82e575bc2eb63093d54e6b3b68e91a78134704fed","abstract_canon_sha256":"275903ad0a73bb75863662835a6788789de0a1f32c2654d6b582c5ec54d32b73"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:42:49.178070Z","signature_b64":"MxG9MXg0ZajlMqtpNjV+hYplDQTFH/3rlqFrE4oqkXHDBG32raASnFC/DUt95MwNUdabduWxRTWDEJ4KxwWmBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1886b0dbcad637b51e4c3a90ec3d111a28ea20a602d12feab6ca7087fd8e89da","last_reissued_at":"2026-07-05T11:42:49.177616Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:42:49.177616Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Explaining How Visual, Textual and Multimodal Encoders Share Concepts","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Cl\\'ement Cornet, Herv\\'e Le Borgne, Romaric Besan\\c{c}on","submitted_at":"2025-07-24T15:33:31Z","abstract_excerpt":"Sparse autoencoders (SAEs) have emerged as a powerful technique for extracting human-interpretable features from neural networks activations. Previous works compared different models based on SAE-derived features but those comparisons have been restricted to models within the same modality. We propose a novel indicator allowing quantitative comparison of models across SAE features, and use it to conduct a comparative study of visual, textual and multimodal encoders. We also propose to quantify the Comparative Sharedness of individual features between different classes of models. With these two"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.18512","kind":"arxiv","version":1},"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/2507.18512/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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2507.18512","created_at":"2026-07-05T11:42:49.177670+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.18512v1","created_at":"2026-07-05T11:42:49.177670+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.18512","created_at":"2026-07-05T11:42:49.177670+00:00"},{"alias_kind":"pith_short_12","alias_value":"DCDLBW6K2Y33","created_at":"2026-07-05T11:42:49.177670+00:00"},{"alias_kind":"pith_short_16","alias_value":"DCDLBW6K2Y33KHSM","created_at":"2026-07-05T11:42:49.177670+00:00"},{"alias_kind":"pith_short_8","alias_value":"DCDLBW6K","created_at":"2026-07-05T11:42:49.177670+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.09928","citing_title":"Multimodal Model Diffing for Feature Discovery and Control","ref_index":15,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DCDLBW6K2Y33KHSMHKIOYPIRDI","json":"https://pith.science/pith/DCDLBW6K2Y33KHSMHKIOYPIRDI.json","graph_json":"https://pith.science/api/pith-number/DCDLBW6K2Y33KHSMHKIOYPIRDI/graph.json","events_json":"https://pith.science/api/pith-number/DCDLBW6K2Y33KHSMHKIOYPIRDI/events.json","paper":"https://pith.science/paper/DCDLBW6K"},"agent_actions":{"view_html":"https://pith.science/pith/DCDLBW6K2Y33KHSMHKIOYPIRDI","download_json":"https://pith.science/pith/DCDLBW6K2Y33KHSMHKIOYPIRDI.json","view_paper":"https://pith.science/paper/DCDLBW6K","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.18512&json=true","fetch_graph":"https://pith.science/api/pith-number/DCDLBW6K2Y33KHSMHKIOYPIRDI/graph.json","fetch_events":"https://pith.science/api/pith-number/DCDLBW6K2Y33KHSMHKIOYPIRDI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DCDLBW6K2Y33KHSMHKIOYPIRDI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DCDLBW6K2Y33KHSMHKIOYPIRDI/action/storage_attestation","attest_author":"https://pith.science/pith/DCDLBW6K2Y33KHSMHKIOYPIRDI/action/author_attestation","sign_citation":"https://pith.science/pith/DCDLBW6K2Y33KHSMHKIOYPIRDI/action/citation_signature","submit_replication":"https://pith.science/pith/DCDLBW6K2Y33KHSMHKIOYPIRDI/action/replication_record"}},"created_at":"2026-07-05T11:42:49.177670+00:00","updated_at":"2026-07-05T11:42:49.177670+00:00"}