{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:VFH6I4MBICB7KXBSUBWM34M3LU","short_pith_number":"pith:VFH6I4MB","schema_version":"1.0","canonical_sha256":"a94fe471814083f55c32a06ccdf19b5d0114985833b8286d650947942c1abff9","source":{"kind":"arxiv","id":"2408.00657","version":2},"attestation_state":"computed","paper":{"title":"Disentangling Dense Embeddings with Sparse Autoencoders","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Charles O'Neill, Christine Ye, John F. Wu, Kartheik Iyer","submitted_at":"2024-08-01T15:46:22Z","abstract_excerpt":"Sparse autoencoders (SAEs) have shown promise in extracting interpretable features from complex neural networks. We present one of the first applications of SAEs to dense text embeddings from large language models, demonstrating their effectiveness in disentangling semantic concepts. By training SAEs on embeddings of over 420,000 scientific paper abstracts from computer science and astronomy, we show that the resulting sparse representations maintain semantic fidelity while offering interpretability. We analyse these learned features, exploring their behaviour across different model capacities"},"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":"2408.00657","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-08-01T15:46:22Z","cross_cats_sorted":[],"title_canon_sha256":"a78d34b05522ec319d12a1dbbf4b5259cfac5190bf7f875189e11321e0d38209","abstract_canon_sha256":"f51914393ca7f0abb8157fe99d80b1836704421255d620f0e26ca292b7150aac"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:51:57.250301Z","signature_b64":"1XjaomA0EoT7r9qo36f/wxtA+MT92c753S8oNr9szhwVxzS9pLuoXxPcEtqhnS3ibjyakamxuGsjKkaNwBdjBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a94fe471814083f55c32a06ccdf19b5d0114985833b8286d650947942c1abff9","last_reissued_at":"2026-07-05T08:51:57.249881Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:51:57.249881Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Disentangling Dense Embeddings with Sparse Autoencoders","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Charles O'Neill, Christine Ye, John F. Wu, Kartheik Iyer","submitted_at":"2024-08-01T15:46:22Z","abstract_excerpt":"Sparse autoencoders (SAEs) have shown promise in extracting interpretable features from complex neural networks. We present one of the first applications of SAEs to dense text embeddings from large language models, demonstrating their effectiveness in disentangling semantic concepts. By training SAEs on embeddings of over 420,000 scientific paper abstracts from computer science and astronomy, we show that the resulting sparse representations maintain semantic fidelity while offering interpretability. We analyse these learned features, exploring their behaviour across different model capacities"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.00657","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/2408.00657/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":"2408.00657","created_at":"2026-07-05T08:51:57.249937+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.00657v2","created_at":"2026-07-05T08:51:57.249937+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.00657","created_at":"2026-07-05T08:51:57.249937+00:00"},{"alias_kind":"pith_short_12","alias_value":"VFH6I4MBICB7","created_at":"2026-07-05T08:51:57.249937+00:00"},{"alias_kind":"pith_short_16","alias_value":"VFH6I4MBICB7KXBS","created_at":"2026-07-05T08:51:57.249937+00:00"},{"alias_kind":"pith_short_8","alias_value":"VFH6I4MB","created_at":"2026-07-05T08:51:57.249937+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07007","citing_title":"A Geometric View for Understanding Concept Learning and Neuron Interpretation in Sparse Autoencoders","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22531","citing_title":"Disentanglement Beyond Generative Models with Riemannian ICA","ref_index":63,"is_internal_anchor":false},{"citing_arxiv_id":"2603.26930","citing_title":"In your own words: computationally identifying interpretable themes in free-text survey data","ref_index":50,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VFH6I4MBICB7KXBSUBWM34M3LU","json":"https://pith.science/pith/VFH6I4MBICB7KXBSUBWM34M3LU.json","graph_json":"https://pith.science/api/pith-number/VFH6I4MBICB7KXBSUBWM34M3LU/graph.json","events_json":"https://pith.science/api/pith-number/VFH6I4MBICB7KXBSUBWM34M3LU/events.json","paper":"https://pith.science/paper/VFH6I4MB"},"agent_actions":{"view_html":"https://pith.science/pith/VFH6I4MBICB7KXBSUBWM34M3LU","download_json":"https://pith.science/pith/VFH6I4MBICB7KXBSUBWM34M3LU.json","view_paper":"https://pith.science/paper/VFH6I4MB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.00657&json=true","fetch_graph":"https://pith.science/api/pith-number/VFH6I4MBICB7KXBSUBWM34M3LU/graph.json","fetch_events":"https://pith.science/api/pith-number/VFH6I4MBICB7KXBSUBWM34M3LU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VFH6I4MBICB7KXBSUBWM34M3LU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VFH6I4MBICB7KXBSUBWM34M3LU/action/storage_attestation","attest_author":"https://pith.science/pith/VFH6I4MBICB7KXBSUBWM34M3LU/action/author_attestation","sign_citation":"https://pith.science/pith/VFH6I4MBICB7KXBSUBWM34M3LU/action/citation_signature","submit_replication":"https://pith.science/pith/VFH6I4MBICB7KXBSUBWM34M3LU/action/replication_record"}},"created_at":"2026-07-05T08:51:57.249937+00:00","updated_at":"2026-07-05T08:51:57.249937+00:00"}