{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:DGBQOXDIAV5SI35LSAMQ6DTLZE","short_pith_number":"pith:DGBQOXDI","schema_version":"1.0","canonical_sha256":"1983075c68057b246fab90190f0e6bc913e7a6c339e12cb198fcb52516ba78b8","source":{"kind":"arxiv","id":"2005.07647","version":1},"attestation_state":"computed","paper":{"title":"Finding Experts in Transformer Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Luca Zappella, Nicholas Apostoloff, Xavier Suau","submitted_at":"2020-05-15T17:07:02Z","abstract_excerpt":"In this work we study the presence of expert units in pre-trained Transformer Models (TM), and how they impact a model's performance. We define expert units to be neurons that are able to classify a concept with a given average precision, where a concept is represented by a binary set of sentences containing the concept (or not). Leveraging the OneSec dataset (Scarlini et al., 2019), we compile a dataset of 1641 concepts that allows diverse expert units in TM to be discovered. We show that expert units are important in several ways: (1) The presence of expert units is correlated ($r^2=0.833$) "},"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":"2005.07647","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2020-05-15T17:07:02Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"1b0ffd951ba675cfc0b4fe1b24ef204b72c75c3252a96e34cc56c85a244da628","abstract_canon_sha256":"32534eb9c76de22b32ee57d3f3b12fc3dbf71d1e831ab6977b69637fbc9f6734"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:03:04.653238Z","signature_b64":"+c8sh7ad4Z3iaTdDLhHc+fVZI9MuvrSEbKhGSoKfBJtJIskYk+GHf7jPU4hQ34p9KLzfgeAfOa6xGwJ7emI8Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1983075c68057b246fab90190f0e6bc913e7a6c339e12cb198fcb52516ba78b8","last_reissued_at":"2026-07-05T01:03:04.652855Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:03:04.652855Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Finding Experts in Transformer Models","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Luca Zappella, Nicholas Apostoloff, Xavier Suau","submitted_at":"2020-05-15T17:07:02Z","abstract_excerpt":"In this work we study the presence of expert units in pre-trained Transformer Models (TM), and how they impact a model's performance. We define expert units to be neurons that are able to classify a concept with a given average precision, where a concept is represented by a binary set of sentences containing the concept (or not). Leveraging the OneSec dataset (Scarlini et al., 2019), we compile a dataset of 1641 concepts that allows diverse expert units in TM to be discovered. We show that expert units are important in several ways: (1) The presence of expert units is correlated ($r^2=0.833$) "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2005.07647","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/2005.07647/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":"2005.07647","created_at":"2026-07-05T01:03:04.652912+00:00"},{"alias_kind":"arxiv_version","alias_value":"2005.07647v1","created_at":"2026-07-05T01:03:04.652912+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2005.07647","created_at":"2026-07-05T01:03:04.652912+00:00"},{"alias_kind":"pith_short_12","alias_value":"DGBQOXDIAV5S","created_at":"2026-07-05T01:03:04.652912+00:00"},{"alias_kind":"pith_short_16","alias_value":"DGBQOXDIAV5SI35L","created_at":"2026-07-05T01:03:04.652912+00:00"},{"alias_kind":"pith_short_8","alias_value":"DGBQOXDI","created_at":"2026-07-05T01:03:04.652912+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.11802","citing_title":"Psychological Concept Neurons: Can Neural Control Bias Probing and Shift Generation in LLMs?","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DGBQOXDIAV5SI35LSAMQ6DTLZE","json":"https://pith.science/pith/DGBQOXDIAV5SI35LSAMQ6DTLZE.json","graph_json":"https://pith.science/api/pith-number/DGBQOXDIAV5SI35LSAMQ6DTLZE/graph.json","events_json":"https://pith.science/api/pith-number/DGBQOXDIAV5SI35LSAMQ6DTLZE/events.json","paper":"https://pith.science/paper/DGBQOXDI"},"agent_actions":{"view_html":"https://pith.science/pith/DGBQOXDIAV5SI35LSAMQ6DTLZE","download_json":"https://pith.science/pith/DGBQOXDIAV5SI35LSAMQ6DTLZE.json","view_paper":"https://pith.science/paper/DGBQOXDI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2005.07647&json=true","fetch_graph":"https://pith.science/api/pith-number/DGBQOXDIAV5SI35LSAMQ6DTLZE/graph.json","fetch_events":"https://pith.science/api/pith-number/DGBQOXDIAV5SI35LSAMQ6DTLZE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DGBQOXDIAV5SI35LSAMQ6DTLZE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DGBQOXDIAV5SI35LSAMQ6DTLZE/action/storage_attestation","attest_author":"https://pith.science/pith/DGBQOXDIAV5SI35LSAMQ6DTLZE/action/author_attestation","sign_citation":"https://pith.science/pith/DGBQOXDIAV5SI35LSAMQ6DTLZE/action/citation_signature","submit_replication":"https://pith.science/pith/DGBQOXDIAV5SI35LSAMQ6DTLZE/action/replication_record"}},"created_at":"2026-07-05T01:03:04.652912+00:00","updated_at":"2026-07-05T01:03:04.652912+00:00"}