{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:XRXTHHTXBJOUQMCESSSOEA7FJ6","short_pith_number":"pith:XRXTHHTX","schema_version":"1.0","canonical_sha256":"bc6f339e770a5d48304494a4e203e54f862914d6df220bce9bd43ba529bbbf5d","source":{"kind":"arxiv","id":"2607.10116","version":1},"attestation_state":"computed","paper":{"title":"When Data Imbalance Helps: Robust Generalization Through Shortcut Saturation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Cheng-Ting Chou, Duc Binh Hoang","submitted_at":"2026-07-11T04:35:04Z","abstract_excerpt":"We study robust generalization under spurious correlations: tasks where a shortcut feature is correlated with the true label in training but anti-correlated in an adversarial held-out split. Varying the spurious ratio $r$ (the fraction of training examples where shortcut = true label) and model capacity, we find a counterintuitive result: data imbalance promotes generalization in sufficiently capable models. On a synthetic task where the true label is sum parity of an integer sequence and the shortcut is the parity of the maximum-valued element, a 2-layer, 2-head transformer generalized (reach"},"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":"2607.10116","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-07-11T04:35:04Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"9171f2bc0905fb84da576f2c0176f4f206cc5a3c257d4dff4e44b03d22afdb48","abstract_canon_sha256":"b405898188a29287e4450a0ce2f9aed48c67c4c83a11fd5617c03e17f766a162"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-14T01:20:26.033266Z","signature_b64":"Ip5JkFE/IHFMZUUGyEEJsymlW93U3KG7fXRmjh53yTFTfFCBOw/tUmikU5jfPmrwh34K85TT5LXxjWZ7Wy7oCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"bc6f339e770a5d48304494a4e203e54f862914d6df220bce9bd43ba529bbbf5d","last_reissued_at":"2026-07-14T01:20:26.032284Z","signature_status":"signed_v1","first_computed_at":"2026-07-14T01:20:26.032284Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When Data Imbalance Helps: Robust Generalization Through Shortcut Saturation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Cheng-Ting Chou, Duc Binh Hoang","submitted_at":"2026-07-11T04:35:04Z","abstract_excerpt":"We study robust generalization under spurious correlations: tasks where a shortcut feature is correlated with the true label in training but anti-correlated in an adversarial held-out split. Varying the spurious ratio $r$ (the fraction of training examples where shortcut = true label) and model capacity, we find a counterintuitive result: data imbalance promotes generalization in sufficiently capable models. On a synthetic task where the true label is sum parity of an integer sequence and the shortcut is the parity of the maximum-valued element, a 2-layer, 2-head transformer generalized (reach"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.10116","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/2607.10116/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":"2607.10116","created_at":"2026-07-14T01:20:26.032873+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.10116v1","created_at":"2026-07-14T01:20:26.032873+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.10116","created_at":"2026-07-14T01:20:26.032873+00:00"},{"alias_kind":"pith_short_12","alias_value":"XRXTHHTXBJOU","created_at":"2026-07-14T01:20:26.032873+00:00"},{"alias_kind":"pith_short_16","alias_value":"XRXTHHTXBJOUQMCE","created_at":"2026-07-14T01:20:26.032873+00:00"},{"alias_kind":"pith_short_8","alias_value":"XRXTHHTX","created_at":"2026-07-14T01:20:26.032873+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XRXTHHTXBJOUQMCESSSOEA7FJ6","json":"https://pith.science/pith/XRXTHHTXBJOUQMCESSSOEA7FJ6.json","graph_json":"https://pith.science/api/pith-number/XRXTHHTXBJOUQMCESSSOEA7FJ6/graph.json","events_json":"https://pith.science/api/pith-number/XRXTHHTXBJOUQMCESSSOEA7FJ6/events.json","paper":"https://pith.science/paper/XRXTHHTX"},"agent_actions":{"view_html":"https://pith.science/pith/XRXTHHTXBJOUQMCESSSOEA7FJ6","download_json":"https://pith.science/pith/XRXTHHTXBJOUQMCESSSOEA7FJ6.json","view_paper":"https://pith.science/paper/XRXTHHTX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.10116&json=true","fetch_graph":"https://pith.science/api/pith-number/XRXTHHTXBJOUQMCESSSOEA7FJ6/graph.json","fetch_events":"https://pith.science/api/pith-number/XRXTHHTXBJOUQMCESSSOEA7FJ6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XRXTHHTXBJOUQMCESSSOEA7FJ6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XRXTHHTXBJOUQMCESSSOEA7FJ6/action/storage_attestation","attest_author":"https://pith.science/pith/XRXTHHTXBJOUQMCESSSOEA7FJ6/action/author_attestation","sign_citation":"https://pith.science/pith/XRXTHHTXBJOUQMCESSSOEA7FJ6/action/citation_signature","submit_replication":"https://pith.science/pith/XRXTHHTXBJOUQMCESSSOEA7FJ6/action/replication_record"}},"created_at":"2026-07-14T01:20:26.032873+00:00","updated_at":"2026-07-14T01:20:26.032873+00:00"}