{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:4ERCPANKWGDOYFTDIHBCQBDQWS","short_pith_number":"pith:4ERCPANK","schema_version":"1.0","canonical_sha256":"e1222781aab186ec166341c2280470b4915dfc1c7e84a503a1c3aa316432994e","source":{"kind":"arxiv","id":"2412.12101","version":1},"attestation_state":"computed","paper":{"title":"InterPLM: Discovering Interpretable Features in Protein Language Models via Sparse Autoencoders","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","q-bio.QM"],"primary_cat":"q-bio.BM","authors_text":"Elana Simon, James Zou","submitted_at":"2024-11-13T18:51:21Z","abstract_excerpt":"Protein language models (PLMs) have demonstrated remarkable success in protein modeling and design, yet their internal mechanisms for predicting structure and function remain poorly understood. Here we present a systematic approach to extract and analyze interpretable features from PLMs using sparse autoencoders (SAEs). By training SAEs on embeddings from the PLM ESM-2, we identify up to 2,548 human-interpretable latent features per layer that strongly correlate with up to 143 known biological concepts such as binding sites, structural motifs, and functional domains. In contrast, examining ind"},"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":"2412.12101","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"q-bio.BM","submitted_at":"2024-11-13T18:51:21Z","cross_cats_sorted":["cs.AI","cs.LG","q-bio.QM"],"title_canon_sha256":"afe7f8484c323d15ad50d6aa943939bae4134a3863b2cc2c1315dc49e93d7bfa","abstract_canon_sha256":"80bada6eeda4c5a1aa756ad955111754d5ed4fdf0f9dc3340c3b16b9a7e5e25a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:50:09.688722Z","signature_b64":"hmZxTwiWEsXhx0S94h2YTLsjw6gJddgf7LbhPZS6artPbZr0E2m/r06vVDwvbA50bh4m9V5NYkrDSNmFe529BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e1222781aab186ec166341c2280470b4915dfc1c7e84a503a1c3aa316432994e","last_reissued_at":"2026-07-05T09:50:09.688252Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:50:09.688252Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"InterPLM: Discovering Interpretable Features in Protein Language Models via Sparse Autoencoders","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","q-bio.QM"],"primary_cat":"q-bio.BM","authors_text":"Elana Simon, James Zou","submitted_at":"2024-11-13T18:51:21Z","abstract_excerpt":"Protein language models (PLMs) have demonstrated remarkable success in protein modeling and design, yet their internal mechanisms for predicting structure and function remain poorly understood. Here we present a systematic approach to extract and analyze interpretable features from PLMs using sparse autoencoders (SAEs). By training SAEs on embeddings from the PLM ESM-2, we identify up to 2,548 human-interpretable latent features per layer that strongly correlate with up to 143 known biological concepts such as binding sites, structural motifs, and functional domains. In contrast, examining ind"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.12101","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/2412.12101/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":"2412.12101","created_at":"2026-07-05T09:50:09.688303+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.12101v1","created_at":"2026-07-05T09:50:09.688303+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.12101","created_at":"2026-07-05T09:50:09.688303+00:00"},{"alias_kind":"pith_short_12","alias_value":"4ERCPANKWGDO","created_at":"2026-07-05T09:50:09.688303+00:00"},{"alias_kind":"pith_short_16","alias_value":"4ERCPANKWGDOYFTD","created_at":"2026-07-05T09:50:09.688303+00:00"},{"alias_kind":"pith_short_8","alias_value":"4ERCPANK","created_at":"2026-07-05T09:50:09.688303+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2607.01336","citing_title":"Mechanistic Interpretability and Causal Feature Steering of Neural Quantum States via Sparse Autoencoders","ref_index":51,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/4ERCPANKWGDOYFTDIHBCQBDQWS","json":"https://pith.science/pith/4ERCPANKWGDOYFTDIHBCQBDQWS.json","graph_json":"https://pith.science/api/pith-number/4ERCPANKWGDOYFTDIHBCQBDQWS/graph.json","events_json":"https://pith.science/api/pith-number/4ERCPANKWGDOYFTDIHBCQBDQWS/events.json","paper":"https://pith.science/paper/4ERCPANK"},"agent_actions":{"view_html":"https://pith.science/pith/4ERCPANKWGDOYFTDIHBCQBDQWS","download_json":"https://pith.science/pith/4ERCPANKWGDOYFTDIHBCQBDQWS.json","view_paper":"https://pith.science/paper/4ERCPANK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.12101&json=true","fetch_graph":"https://pith.science/api/pith-number/4ERCPANKWGDOYFTDIHBCQBDQWS/graph.json","fetch_events":"https://pith.science/api/pith-number/4ERCPANKWGDOYFTDIHBCQBDQWS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4ERCPANKWGDOYFTDIHBCQBDQWS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4ERCPANKWGDOYFTDIHBCQBDQWS/action/storage_attestation","attest_author":"https://pith.science/pith/4ERCPANKWGDOYFTDIHBCQBDQWS/action/author_attestation","sign_citation":"https://pith.science/pith/4ERCPANKWGDOYFTDIHBCQBDQWS/action/citation_signature","submit_replication":"https://pith.science/pith/4ERCPANKWGDOYFTDIHBCQBDQWS/action/replication_record"}},"created_at":"2026-07-05T09:50:09.688303+00:00","updated_at":"2026-07-05T09:50:09.688303+00:00"}