{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:QP7LUMWYCKTMPCZW7TXYN3PK3J","short_pith_number":"pith:QP7LUMWY","schema_version":"1.0","canonical_sha256":"83feba32d812a6c78b36fcef86edeada414dab007d4df081fb21eff64483590c","source":{"kind":"arxiv","id":"1803.06959","version":4},"attestation_state":"computed","paper":{"title":"On the importance of single directions for generalization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.NE"],"primary_cat":"stat.ML","authors_text":"Ari S. Morcos, David G.T. Barrett, Matthew Botvinick, Neil C. Rabinowitz","submitted_at":"2018-03-19T14:42:19Z","abstract_excerpt":"Despite their ability to memorize large datasets, deep neural networks often achieve good generalization performance. However, the differences between the learned solutions of networks which generalize and those which do not remain unclear. Additionally, the tuning properties of single directions (defined as the activation of a single unit or some linear combination of units in response to some input) have been highlighted, but their importance has not been evaluated. Here, we connect these lines of inquiry to demonstrate that a network's reliance on single directions is a good predictor of it"},"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":"1803.06959","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"stat.ML","submitted_at":"2018-03-19T14:42:19Z","cross_cats_sorted":["cs.AI","cs.LG","cs.NE"],"title_canon_sha256":"8b95dd196131cf27c28375505e7f3f128074ed4e0102f03f2e39932e0a2643c9","abstract_canon_sha256":"b518928f5e0b8262776d91c0cc094f66043cc70d5e3a15b982ce3a6efec8394f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:15:27.872884Z","signature_b64":"sKtjVGYYSYxk/h4GjDGHs2byL+z2uypkHNPnfXYcUsdSu439LSJOPgtfxwyc8l94R4zHJm/L6Bj7KSGIIqAwCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"83feba32d812a6c78b36fcef86edeada414dab007d4df081fb21eff64483590c","last_reissued_at":"2026-05-18T00:15:27.872180Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:15:27.872180Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"On the importance of single directions for generalization","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","cs.NE"],"primary_cat":"stat.ML","authors_text":"Ari S. Morcos, David G.T. Barrett, Matthew Botvinick, Neil C. Rabinowitz","submitted_at":"2018-03-19T14:42:19Z","abstract_excerpt":"Despite their ability to memorize large datasets, deep neural networks often achieve good generalization performance. However, the differences between the learned solutions of networks which generalize and those which do not remain unclear. Additionally, the tuning properties of single directions (defined as the activation of a single unit or some linear combination of units in response to some input) have been highlighted, but their importance has not been evaluated. Here, we connect these lines of inquiry to demonstrate that a network's reliance on single directions is a good predictor of it"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1803.06959","kind":"arxiv","version":4},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"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":"1803.06959","created_at":"2026-05-18T00:15:27.872282+00:00"},{"alias_kind":"arxiv_version","alias_value":"1803.06959v4","created_at":"2026-05-18T00:15:27.872282+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1803.06959","created_at":"2026-05-18T00:15:27.872282+00:00"},{"alias_kind":"pith_short_12","alias_value":"QP7LUMWYCKTM","created_at":"2026-05-18T12:32:46.962924+00:00"},{"alias_kind":"pith_short_16","alias_value":"QP7LUMWYCKTMPCZW","created_at":"2026-05-18T12:32:46.962924+00:00"},{"alias_kind":"pith_short_8","alias_value":"QP7LUMWY","created_at":"2026-05-18T12:32:46.962924+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":3,"sample":[{"citing_arxiv_id":"1907.06916","citing_title":"Single-bit-per-weight deep convolutional neural networks without batch-normalization layers for embedded systems","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"1907.11879","citing_title":"Multi-task Self-Supervised Learning for Human Activity Detection","ref_index":45,"is_internal_anchor":true},{"citing_arxiv_id":"2502.06809","citing_title":"Neurons Speak in Ranges: Breaking Free from Discrete Neuronal Attribution","ref_index":18,"is_internal_anchor":true},{"citing_arxiv_id":"2209.10652","citing_title":"Toy Models of Superposition","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10627","citing_title":"Computational Lesions in Multilingual Language Models Separate Shared and Language-specific Brain Alignment","ref_index":84,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QP7LUMWYCKTMPCZW7TXYN3PK3J","json":"https://pith.science/pith/QP7LUMWYCKTMPCZW7TXYN3PK3J.json","graph_json":"https://pith.science/api/pith-number/QP7LUMWYCKTMPCZW7TXYN3PK3J/graph.json","events_json":"https://pith.science/api/pith-number/QP7LUMWYCKTMPCZW7TXYN3PK3J/events.json","paper":"https://pith.science/paper/QP7LUMWY"},"agent_actions":{"view_html":"https://pith.science/pith/QP7LUMWYCKTMPCZW7TXYN3PK3J","download_json":"https://pith.science/pith/QP7LUMWYCKTMPCZW7TXYN3PK3J.json","view_paper":"https://pith.science/paper/QP7LUMWY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1803.06959&json=true","fetch_graph":"https://pith.science/api/pith-number/QP7LUMWYCKTMPCZW7TXYN3PK3J/graph.json","fetch_events":"https://pith.science/api/pith-number/QP7LUMWYCKTMPCZW7TXYN3PK3J/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QP7LUMWYCKTMPCZW7TXYN3PK3J/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QP7LUMWYCKTMPCZW7TXYN3PK3J/action/storage_attestation","attest_author":"https://pith.science/pith/QP7LUMWYCKTMPCZW7TXYN3PK3J/action/author_attestation","sign_citation":"https://pith.science/pith/QP7LUMWYCKTMPCZW7TXYN3PK3J/action/citation_signature","submit_replication":"https://pith.science/pith/QP7LUMWYCKTMPCZW7TXYN3PK3J/action/replication_record"}},"created_at":"2026-05-18T00:15:27.872282+00:00","updated_at":"2026-05-18T00:15:27.872282+00:00"}