{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:QIQ35BLADBXW3Q2XSW6BR73TNX","short_pith_number":"pith:QIQ35BLA","schema_version":"1.0","canonical_sha256":"8221be8560186f6dc35795bc18ff736dfb14e3592b80c17ad8b94cc8ca54638c","source":{"kind":"arxiv","id":"2211.12312","version":1},"attestation_state":"computed","paper":{"title":"Interpreting Neural Networks through the Polytope Lens","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Beren Millidge, Carlos Ram\\'on Guevara, Connor Leahy, Dan Braun, Eric Winsor, Gabriel Alfour, Jacob Merizian, Kip Parker, Lee Sharkey, Leo Grinsztajn, Sid Black","submitted_at":"2022-11-22T15:03:48Z","abstract_excerpt":"Mechanistic interpretability aims to explain what a neural network has learned at a nuts-and-bolts level. What are the fundamental primitives of neural network representations? Previous mechanistic descriptions have used individual neurons or their linear combinations to understand the representations a network has learned. But there are clues that neurons and their linear combinations are not the correct fundamental units of description: directions cannot describe how neural networks use nonlinearities to structure their representations. Moreover, many instances of individual neurons and thei"},"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":"2211.12312","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-11-22T15:03:48Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"4df2f71ba598c9b4a075a763e945d8b52504ef8f7c1f0021d61636293267def4","abstract_canon_sha256":"00bb82f4f432308061037e5a7b301f9dc1ad9a0ab1a1380384f3513f96813e85"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:18:15.682503Z","signature_b64":"mH7smpI3ZONf1EThhjbdMNG+O+1pW0UK/KyrFC7/MhIciPQFnUmpcPmY46S7TJnYmgovZBGQt7CLFfIJx21JBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8221be8560186f6dc35795bc18ff736dfb14e3592b80c17ad8b94cc8ca54638c","last_reissued_at":"2026-07-05T05:18:15.682099Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:18:15.682099Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Interpreting Neural Networks through the Polytope Lens","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Beren Millidge, Carlos Ram\\'on Guevara, Connor Leahy, Dan Braun, Eric Winsor, Gabriel Alfour, Jacob Merizian, Kip Parker, Lee Sharkey, Leo Grinsztajn, Sid Black","submitted_at":"2022-11-22T15:03:48Z","abstract_excerpt":"Mechanistic interpretability aims to explain what a neural network has learned at a nuts-and-bolts level. What are the fundamental primitives of neural network representations? Previous mechanistic descriptions have used individual neurons or their linear combinations to understand the representations a network has learned. But there are clues that neurons and their linear combinations are not the correct fundamental units of description: directions cannot describe how neural networks use nonlinearities to structure their representations. Moreover, many instances of individual neurons and thei"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2211.12312","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/2211.12312/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":"2211.12312","created_at":"2026-07-05T05:18:15.682155+00:00"},{"alias_kind":"arxiv_version","alias_value":"2211.12312v1","created_at":"2026-07-05T05:18:15.682155+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2211.12312","created_at":"2026-07-05T05:18:15.682155+00:00"},{"alias_kind":"pith_short_12","alias_value":"QIQ35BLADBXW","created_at":"2026-07-05T05:18:15.682155+00:00"},{"alias_kind":"pith_short_16","alias_value":"QIQ35BLADBXW3Q2X","created_at":"2026-07-05T05:18:15.682155+00:00"},{"alias_kind":"pith_short_8","alias_value":"QIQ35BLA","created_at":"2026-07-05T05:18:15.682155+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.32008","citing_title":"Surrogate Fidelity: When Can Open LLMs Explain Closed Ones?","ref_index":57,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12671","citing_title":"All Circuits Lead to Rome: Rethinking Functional Anisotropy in Circuit and Sheaf Discovery for LLMs","ref_index":86,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19260","citing_title":"Understanding the Mechanism of Altruism in Large Language Models","ref_index":225,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QIQ35BLADBXW3Q2XSW6BR73TNX","json":"https://pith.science/pith/QIQ35BLADBXW3Q2XSW6BR73TNX.json","graph_json":"https://pith.science/api/pith-number/QIQ35BLADBXW3Q2XSW6BR73TNX/graph.json","events_json":"https://pith.science/api/pith-number/QIQ35BLADBXW3Q2XSW6BR73TNX/events.json","paper":"https://pith.science/paper/QIQ35BLA"},"agent_actions":{"view_html":"https://pith.science/pith/QIQ35BLADBXW3Q2XSW6BR73TNX","download_json":"https://pith.science/pith/QIQ35BLADBXW3Q2XSW6BR73TNX.json","view_paper":"https://pith.science/paper/QIQ35BLA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2211.12312&json=true","fetch_graph":"https://pith.science/api/pith-number/QIQ35BLADBXW3Q2XSW6BR73TNX/graph.json","fetch_events":"https://pith.science/api/pith-number/QIQ35BLADBXW3Q2XSW6BR73TNX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QIQ35BLADBXW3Q2XSW6BR73TNX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QIQ35BLADBXW3Q2XSW6BR73TNX/action/storage_attestation","attest_author":"https://pith.science/pith/QIQ35BLADBXW3Q2XSW6BR73TNX/action/author_attestation","sign_citation":"https://pith.science/pith/QIQ35BLADBXW3Q2XSW6BR73TNX/action/citation_signature","submit_replication":"https://pith.science/pith/QIQ35BLADBXW3Q2XSW6BR73TNX/action/replication_record"}},"created_at":"2026-07-05T05:18:15.682155+00:00","updated_at":"2026-07-05T05:18:15.682155+00:00"}