{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:XR4TXNMRKNTPKA57DFRSKOIOVA","short_pith_number":"pith:XR4TXNMR","schema_version":"1.0","canonical_sha256":"bc793bb5915366f503bf196325390ea81e68e8013fb4adf61b448699717b9ea1","source":{"kind":"arxiv","id":"2505.15970","version":1},"attestation_state":"computed","paper":{"title":"Analyzing Hierarchical Structure in Vision Models with Sparse Autoencoders","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Changbai Li, Matthew Lyle Olson, Musashi Hinck, Neale Ratzlaff, Phillip Howard, Shao-Yen Tseng, Vasudev Lal","submitted_at":"2025-05-21T19:38:48Z","abstract_excerpt":"The ImageNet hierarchy provides a structured taxonomy of object categories, offering a valuable lens through which to analyze the representations learned by deep vision models. In this work, we conduct a comprehensive analysis of how vision models encode the ImageNet hierarchy, leveraging Sparse Autoencoders (SAEs) to probe their internal representations. SAEs have been widely used as an explanation tool for large language models (LLMs), where they enable the discovery of semantically meaningful features. Here, we extend their use to vision models to investigate whether learned representations"},"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":"2505.15970","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-05-21T19:38:48Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"ddb0f28461ba8ecaa2f572140ed02dc25a135e1a35f68eaa72892a75009f1e96","abstract_canon_sha256":"89acfd1127d0da02576f268c9482061db5616cf7b2e534fff8d66918073af759"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:07:27.904814Z","signature_b64":"5cpGYleaGp03JG4KGfHgiEt/lE2/ipVQMLbrqZZqucJ+AmlWkjICLtDGw6ZvnKyHo2P0ZRWSwVCMOqDj6lSaBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bc793bb5915366f503bf196325390ea81e68e8013fb4adf61b448699717b9ea1","last_reissued_at":"2026-07-05T11:07:27.904235Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:07:27.904235Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Analyzing Hierarchical Structure in Vision Models with Sparse Autoencoders","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Changbai Li, Matthew Lyle Olson, Musashi Hinck, Neale Ratzlaff, Phillip Howard, Shao-Yen Tseng, Vasudev Lal","submitted_at":"2025-05-21T19:38:48Z","abstract_excerpt":"The ImageNet hierarchy provides a structured taxonomy of object categories, offering a valuable lens through which to analyze the representations learned by deep vision models. In this work, we conduct a comprehensive analysis of how vision models encode the ImageNet hierarchy, leveraging Sparse Autoencoders (SAEs) to probe their internal representations. SAEs have been widely used as an explanation tool for large language models (LLMs), where they enable the discovery of semantically meaningful features. Here, we extend their use to vision models to investigate whether learned representations"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.15970","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/2505.15970/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":"2505.15970","created_at":"2026-07-05T11:07:27.904295+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.15970v1","created_at":"2026-07-05T11:07:27.904295+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.15970","created_at":"2026-07-05T11:07:27.904295+00:00"},{"alias_kind":"pith_short_12","alias_value":"XR4TXNMRKNTP","created_at":"2026-07-05T11:07:27.904295+00:00"},{"alias_kind":"pith_short_16","alias_value":"XR4TXNMRKNTPKA57","created_at":"2026-07-05T11:07:27.904295+00:00"},{"alias_kind":"pith_short_8","alias_value":"XR4TXNMR","created_at":"2026-07-05T11:07:27.904295+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/XR4TXNMRKNTPKA57DFRSKOIOVA","json":"https://pith.science/pith/XR4TXNMRKNTPKA57DFRSKOIOVA.json","graph_json":"https://pith.science/api/pith-number/XR4TXNMRKNTPKA57DFRSKOIOVA/graph.json","events_json":"https://pith.science/api/pith-number/XR4TXNMRKNTPKA57DFRSKOIOVA/events.json","paper":"https://pith.science/paper/XR4TXNMR"},"agent_actions":{"view_html":"https://pith.science/pith/XR4TXNMRKNTPKA57DFRSKOIOVA","download_json":"https://pith.science/pith/XR4TXNMRKNTPKA57DFRSKOIOVA.json","view_paper":"https://pith.science/paper/XR4TXNMR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.15970&json=true","fetch_graph":"https://pith.science/api/pith-number/XR4TXNMRKNTPKA57DFRSKOIOVA/graph.json","fetch_events":"https://pith.science/api/pith-number/XR4TXNMRKNTPKA57DFRSKOIOVA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XR4TXNMRKNTPKA57DFRSKOIOVA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XR4TXNMRKNTPKA57DFRSKOIOVA/action/storage_attestation","attest_author":"https://pith.science/pith/XR4TXNMRKNTPKA57DFRSKOIOVA/action/author_attestation","sign_citation":"https://pith.science/pith/XR4TXNMRKNTPKA57DFRSKOIOVA/action/citation_signature","submit_replication":"https://pith.science/pith/XR4TXNMRKNTPKA57DFRSKOIOVA/action/replication_record"}},"created_at":"2026-07-05T11:07:27.904295+00:00","updated_at":"2026-07-05T11:07:27.904295+00:00"}