{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:A2D2G3T5PXYWFHB26UIYC5K7XS","short_pith_number":"pith:A2D2G3T5","schema_version":"1.0","canonical_sha256":"0687a36e7d7df1629c3af51181755fbc9386de34a7df6ee047ba54ba04ac045c","source":{"kind":"arxiv","id":"2502.12125","version":1},"attestation_state":"computed","paper":{"title":"Hypernym Bias: Unraveling Deep Classifier Training Dynamics through the Lens of Class Hierarchy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Alexander Mullin, Roman Malashin, Valeria Yachnaya","submitted_at":"2025-02-17T18:47:01Z","abstract_excerpt":"We investigate the training dynamics of deep classifiers by examining how hierarchical relationships between classes evolve during training. Through extensive experiments, we argue that the learning process in classification problems can be understood through the lens of label clustering. Specifically, we observe that networks tend to distinguish higher-level (hypernym) categories in the early stages of training, and learn more specific (hyponym) categories later. We introduce a novel framework to track the evolution of the feature manifold during training, revealing how the hierarchy of class"},"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":"2502.12125","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-02-17T18:47:01Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"60094dd2c73c56443f67516705622a4d267ca04c1dafcc4c7abd2d60b61c8d4c","abstract_canon_sha256":"5b2a74afb95518437f5d9433f57423d6f14f82f67b057b3fc991e4ef86424bcd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:39.396492Z","signature_b64":"4rjXqV7DVwiu4sAWLd0CTBf1NphO3ExXjIGiK8VL1TbL7TsVaSdgdh5w82cmWj5yom8AtFquciyu55uC9TzPBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0687a36e7d7df1629c3af51181755fbc9386de34a7df6ee047ba54ba04ac045c","last_reissued_at":"2026-07-05T10:15:39.395991Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:39.395991Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hypernym Bias: Unraveling Deep Classifier Training Dynamics through the Lens of Class Hierarchy","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Alexander Mullin, Roman Malashin, Valeria Yachnaya","submitted_at":"2025-02-17T18:47:01Z","abstract_excerpt":"We investigate the training dynamics of deep classifiers by examining how hierarchical relationships between classes evolve during training. Through extensive experiments, we argue that the learning process in classification problems can be understood through the lens of label clustering. Specifically, we observe that networks tend to distinguish higher-level (hypernym) categories in the early stages of training, and learn more specific (hyponym) categories later. We introduce a novel framework to track the evolution of the feature manifold during training, revealing how the hierarchy of class"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.12125","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/2502.12125/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":"2502.12125","created_at":"2026-07-05T10:15:39.396039+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.12125v1","created_at":"2026-07-05T10:15:39.396039+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.12125","created_at":"2026-07-05T10:15:39.396039+00:00"},{"alias_kind":"pith_short_12","alias_value":"A2D2G3T5PXYW","created_at":"2026-07-05T10:15:39.396039+00:00"},{"alias_kind":"pith_short_16","alias_value":"A2D2G3T5PXYWFHB2","created_at":"2026-07-05T10:15:39.396039+00:00"},{"alias_kind":"pith_short_8","alias_value":"A2D2G3T5","created_at":"2026-07-05T10:15:39.396039+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2504.19592","citing_title":"Neural network task specialization via domain constraining","ref_index":14,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/A2D2G3T5PXYWFHB26UIYC5K7XS","json":"https://pith.science/pith/A2D2G3T5PXYWFHB26UIYC5K7XS.json","graph_json":"https://pith.science/api/pith-number/A2D2G3T5PXYWFHB26UIYC5K7XS/graph.json","events_json":"https://pith.science/api/pith-number/A2D2G3T5PXYWFHB26UIYC5K7XS/events.json","paper":"https://pith.science/paper/A2D2G3T5"},"agent_actions":{"view_html":"https://pith.science/pith/A2D2G3T5PXYWFHB26UIYC5K7XS","download_json":"https://pith.science/pith/A2D2G3T5PXYWFHB26UIYC5K7XS.json","view_paper":"https://pith.science/paper/A2D2G3T5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.12125&json=true","fetch_graph":"https://pith.science/api/pith-number/A2D2G3T5PXYWFHB26UIYC5K7XS/graph.json","fetch_events":"https://pith.science/api/pith-number/A2D2G3T5PXYWFHB26UIYC5K7XS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A2D2G3T5PXYWFHB26UIYC5K7XS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A2D2G3T5PXYWFHB26UIYC5K7XS/action/storage_attestation","attest_author":"https://pith.science/pith/A2D2G3T5PXYWFHB26UIYC5K7XS/action/author_attestation","sign_citation":"https://pith.science/pith/A2D2G3T5PXYWFHB26UIYC5K7XS/action/citation_signature","submit_replication":"https://pith.science/pith/A2D2G3T5PXYWFHB26UIYC5K7XS/action/replication_record"}},"created_at":"2026-07-05T10:15:39.396039+00:00","updated_at":"2026-07-05T10:15:39.396039+00:00"}