{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:36WJYWZHRZIQJYXLMAIPXVVBAV","short_pith_number":"pith:36WJYWZH","schema_version":"1.0","canonical_sha256":"dfac9c5b278e5104e2eb6010fbd6a105746fdb70e937e4d7a8cd24331b0f98e1","source":{"kind":"arxiv","id":"2404.06453","version":1},"attestation_state":"computed","paper":{"title":"PURE: Turning Polysemantic Neurons Into Pure Features by Identifying Relevant Circuits","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Erblina Purelku, Johanna Vielhaben, Maximilian Dreyer, Sebastian Lapuschkin, Wojciech Samek","submitted_at":"2024-04-09T16:54:19Z","abstract_excerpt":"The field of mechanistic interpretability aims to study the role of individual neurons in Deep Neural Networks. Single neurons, however, have the capability to act polysemantically and encode for multiple (unrelated) features, which renders their interpretation difficult. We present a method for disentangling polysemanticity of any Deep Neural Network by decomposing a polysemantic neuron into multiple monosemantic \"virtual\" neurons. This is achieved by identifying the relevant sub-graph (\"circuit\") for each \"pure\" feature. We demonstrate how our approach allows us to find and disentangle vario"},"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":"2404.06453","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-04-09T16:54:19Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"38c595f24247fff6365e63722cbe3d976d2557d6a247957400545ff463192617","abstract_canon_sha256":"e9d4eb9bbe9e0994ab98c5e9653c851e586e615a94cfdb0820a953ee9b250321"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:06:09.296894Z","signature_b64":"/Wf8q+rqyrPA9IId/kEDajVfBiyjYnHJwFEwjL413nnV2f0s71wfdvpLDzCjaQq/SZBFkoPO0bk8HU/bHV/IBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dfac9c5b278e5104e2eb6010fbd6a105746fdb70e937e4d7a8cd24331b0f98e1","last_reissued_at":"2026-07-05T08:06:09.296384Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:06:09.296384Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PURE: Turning Polysemantic Neurons Into Pure Features by Identifying Relevant Circuits","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Erblina Purelku, Johanna Vielhaben, Maximilian Dreyer, Sebastian Lapuschkin, Wojciech Samek","submitted_at":"2024-04-09T16:54:19Z","abstract_excerpt":"The field of mechanistic interpretability aims to study the role of individual neurons in Deep Neural Networks. Single neurons, however, have the capability to act polysemantically and encode for multiple (unrelated) features, which renders their interpretation difficult. We present a method for disentangling polysemanticity of any Deep Neural Network by decomposing a polysemantic neuron into multiple monosemantic \"virtual\" neurons. This is achieved by identifying the relevant sub-graph (\"circuit\") for each \"pure\" feature. We demonstrate how our approach allows us to find and disentangle vario"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2404.06453","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/2404.06453/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":"2404.06453","created_at":"2026-07-05T08:06:09.296461+00:00"},{"alias_kind":"arxiv_version","alias_value":"2404.06453v1","created_at":"2026-07-05T08:06:09.296461+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2404.06453","created_at":"2026-07-05T08:06:09.296461+00:00"},{"alias_kind":"pith_short_12","alias_value":"36WJYWZHRZIQ","created_at":"2026-07-05T08:06:09.296461+00:00"},{"alias_kind":"pith_short_16","alias_value":"36WJYWZHRZIQJYXL","created_at":"2026-07-05T08:06:09.296461+00:00"},{"alias_kind":"pith_short_8","alias_value":"36WJYWZH","created_at":"2026-07-05T08:06:09.296461+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20431","citing_title":"Sparsity, Superposition, and Forgetting: A Mechanistic Study of Representation Retention in Continual Learning","ref_index":10,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/36WJYWZHRZIQJYXLMAIPXVVBAV","json":"https://pith.science/pith/36WJYWZHRZIQJYXLMAIPXVVBAV.json","graph_json":"https://pith.science/api/pith-number/36WJYWZHRZIQJYXLMAIPXVVBAV/graph.json","events_json":"https://pith.science/api/pith-number/36WJYWZHRZIQJYXLMAIPXVVBAV/events.json","paper":"https://pith.science/paper/36WJYWZH"},"agent_actions":{"view_html":"https://pith.science/pith/36WJYWZHRZIQJYXLMAIPXVVBAV","download_json":"https://pith.science/pith/36WJYWZHRZIQJYXLMAIPXVVBAV.json","view_paper":"https://pith.science/paper/36WJYWZH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2404.06453&json=true","fetch_graph":"https://pith.science/api/pith-number/36WJYWZHRZIQJYXLMAIPXVVBAV/graph.json","fetch_events":"https://pith.science/api/pith-number/36WJYWZHRZIQJYXLMAIPXVVBAV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/36WJYWZHRZIQJYXLMAIPXVVBAV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/36WJYWZHRZIQJYXLMAIPXVVBAV/action/storage_attestation","attest_author":"https://pith.science/pith/36WJYWZHRZIQJYXLMAIPXVVBAV/action/author_attestation","sign_citation":"https://pith.science/pith/36WJYWZHRZIQJYXLMAIPXVVBAV/action/citation_signature","submit_replication":"https://pith.science/pith/36WJYWZHRZIQJYXLMAIPXVVBAV/action/replication_record"}},"created_at":"2026-07-05T08:06:09.296461+00:00","updated_at":"2026-07-05T08:06:09.296461+00:00"}