{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:I7KEYSKBR2IMAE6SWVK7JDOSPM","short_pith_number":"pith:I7KEYSKB","schema_version":"1.0","canonical_sha256":"47d44c49418e90c013d2b555f48dd27b1a483c310bd5405ec44aeb2ef9ca3cdb","source":{"kind":"arxiv","id":"2411.16725","version":3},"attestation_state":"computed","paper":{"title":"$\\textit{Revelio}$: Interpreting and leveraging semantic information in diffusion models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Dahye Kim, Deepti Ghadiyaram, Xavier Thomas","submitted_at":"2024-11-23T03:54:22Z","abstract_excerpt":"We study $\\textit{how}$ rich visual semantic information is represented within various layers and denoising timesteps of different diffusion architectures. We uncover monosemantic interpretable features by leveraging k-sparse autoencoders (k-SAE). We substantiate our mechanistic interpretations via transfer learning using light-weight classifiers on off-the-shelf diffusion models' features. On $4$ datasets, we demonstrate the effectiveness of diffusion features for representation learning. We provide an in-depth analysis of how different diffusion architectures, pre-training datasets, and lang"},"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":"2411.16725","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-11-23T03:54:22Z","cross_cats_sorted":[],"title_canon_sha256":"b770546a054c1bccce2c00567c8582da515af66cbd347b1eef570fbf8e2764d2","abstract_canon_sha256":"546adf9c67956467cce5fb45e0e05f84bb4b215ef35860dc418545149690a130"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:47:15.955277Z","signature_b64":"OTecgCrtkVEeQDvLEOiOy5nasO88hVqZmcP5nEHjWMgVuExZ06p/uJpPVdH8t10izQoldS8fjrFpRHOKONC5Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"47d44c49418e90c013d2b555f48dd27b1a483c310bd5405ec44aeb2ef9ca3cdb","last_reissued_at":"2026-07-05T11:47:15.954738Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:47:15.954738Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"$\\textit{Revelio}$: Interpreting and leveraging semantic information in diffusion models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Dahye Kim, Deepti Ghadiyaram, Xavier Thomas","submitted_at":"2024-11-23T03:54:22Z","abstract_excerpt":"We study $\\textit{how}$ rich visual semantic information is represented within various layers and denoising timesteps of different diffusion architectures. We uncover monosemantic interpretable features by leveraging k-sparse autoencoders (k-SAE). We substantiate our mechanistic interpretations via transfer learning using light-weight classifiers on off-the-shelf diffusion models' features. On $4$ datasets, we demonstrate the effectiveness of diffusion features for representation learning. We provide an in-depth analysis of how different diffusion architectures, pre-training datasets, and lang"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2411.16725","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/2411.16725/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":"2411.16725","created_at":"2026-07-05T11:47:15.954807+00:00"},{"alias_kind":"arxiv_version","alias_value":"2411.16725v3","created_at":"2026-07-05T11:47:15.954807+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2411.16725","created_at":"2026-07-05T11:47:15.954807+00:00"},{"alias_kind":"pith_short_12","alias_value":"I7KEYSKBR2IM","created_at":"2026-07-05T11:47:15.954807+00:00"},{"alias_kind":"pith_short_16","alias_value":"I7KEYSKBR2IMAE6S","created_at":"2026-07-05T11:47:15.954807+00:00"},{"alias_kind":"pith_short_8","alias_value":"I7KEYSKB","created_at":"2026-07-05T11:47:15.954807+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/I7KEYSKBR2IMAE6SWVK7JDOSPM","json":"https://pith.science/pith/I7KEYSKBR2IMAE6SWVK7JDOSPM.json","graph_json":"https://pith.science/api/pith-number/I7KEYSKBR2IMAE6SWVK7JDOSPM/graph.json","events_json":"https://pith.science/api/pith-number/I7KEYSKBR2IMAE6SWVK7JDOSPM/events.json","paper":"https://pith.science/paper/I7KEYSKB"},"agent_actions":{"view_html":"https://pith.science/pith/I7KEYSKBR2IMAE6SWVK7JDOSPM","download_json":"https://pith.science/pith/I7KEYSKBR2IMAE6SWVK7JDOSPM.json","view_paper":"https://pith.science/paper/I7KEYSKB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2411.16725&json=true","fetch_graph":"https://pith.science/api/pith-number/I7KEYSKBR2IMAE6SWVK7JDOSPM/graph.json","fetch_events":"https://pith.science/api/pith-number/I7KEYSKBR2IMAE6SWVK7JDOSPM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/I7KEYSKBR2IMAE6SWVK7JDOSPM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/I7KEYSKBR2IMAE6SWVK7JDOSPM/action/storage_attestation","attest_author":"https://pith.science/pith/I7KEYSKBR2IMAE6SWVK7JDOSPM/action/author_attestation","sign_citation":"https://pith.science/pith/I7KEYSKBR2IMAE6SWVK7JDOSPM/action/citation_signature","submit_replication":"https://pith.science/pith/I7KEYSKBR2IMAE6SWVK7JDOSPM/action/replication_record"}},"created_at":"2026-07-05T11:47:15.954807+00:00","updated_at":"2026-07-05T11:47:15.954807+00:00"}