{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:YFMVAVKXTL3M6RIBZEULHRBL26","short_pith_number":"pith:YFMVAVKX","schema_version":"1.0","canonical_sha256":"c1595055579af6cf4501c928b3c42bd7a1a1252c7e6a10087bb3224ac2ad9de4","source":{"kind":"arxiv","id":"2106.01548","version":3},"attestation_state":"computed","paper":{"title":"When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Boqing Gong, Cho-Jui Hsieh, Xiangning Chen","submitted_at":"2021-06-03T02:08:03Z","abstract_excerpt":"Vision Transformers (ViTs) and MLPs signal further efforts on replacing hand-wired features or inductive biases with general-purpose neural architectures. Existing works empower the models by massive data, such as large-scale pre-training and/or repeated strong data augmentations, and still report optimization-related problems (e.g., sensitivity to initialization and learning rates). Hence, this paper investigates ViTs and MLP-Mixers from the lens of loss geometry, intending to improve the models' data efficiency at training and generalization at inference. Visualization and Hessian reveal ext"},"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":"2106.01548","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2021-06-03T02:08:03Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"8104df67f42352939eee2a6dfe0ab9c1fae28617c4d10a2ec3169eaea0764ee5","abstract_canon_sha256":"767d237127caf9e5c3eaf62d7755cf4c75a6922e35b3e7e2402f7fe7abd2f191"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:04:39.415955Z","signature_b64":"OzTq7y2+TCLCdeAt7UhkYmvHNlZfsMd/juw1EyZ0KOpLnDOlcAIxJEYHz1duU9x9hcMmwQky8xB+Fwv5ut+pCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c1595055579af6cf4501c928b3c42bd7a1a1252c7e6a10087bb3224ac2ad9de4","last_reissued_at":"2026-07-05T04:04:39.414689Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:04:39.414689Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When Vision Transformers Outperform ResNets without Pre-training or Strong Data Augmentations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Boqing Gong, Cho-Jui Hsieh, Xiangning Chen","submitted_at":"2021-06-03T02:08:03Z","abstract_excerpt":"Vision Transformers (ViTs) and MLPs signal further efforts on replacing hand-wired features or inductive biases with general-purpose neural architectures. Existing works empower the models by massive data, such as large-scale pre-training and/or repeated strong data augmentations, and still report optimization-related problems (e.g., sensitivity to initialization and learning rates). Hence, this paper investigates ViTs and MLP-Mixers from the lens of loss geometry, intending to improve the models' data efficiency at training and generalization at inference. Visualization and Hessian reveal ext"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.01548","kind":"arxiv","version":3},"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/2106.01548/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":"2106.01548","created_at":"2026-07-05T04:04:39.414753+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.01548v3","created_at":"2026-07-05T04:04:39.414753+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.01548","created_at":"2026-07-05T04:04:39.414753+00:00"},{"alias_kind":"pith_short_12","alias_value":"YFMVAVKXTL3M","created_at":"2026-07-05T04:04:39.414753+00:00"},{"alias_kind":"pith_short_16","alias_value":"YFMVAVKXTL3M6RIB","created_at":"2026-07-05T04:04:39.414753+00:00"},{"alias_kind":"pith_short_8","alias_value":"YFMVAVKX","created_at":"2026-07-05T04:04:39.414753+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09125","citing_title":"Unveiling Privacy Risks in Multi-modal Large Language Models: Task-specific Vulnerabilities and Mitigation Challenges","ref_index":3,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28662","citing_title":"Closed-Form Steepest Descent Direction toward Flat Minima: Reducing Upper Bounds on the Loss Hessian Eigenspectrum in Neural Networks","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14200","citing_title":"How to Scale Mixture-of-Experts: From muP to the Maximally Scale-Stable Parameterization","ref_index":106,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10202","citing_title":"Wolkowicz-Styan Upper Bound on the Hessian Eigenspectrum for Cross-Entropy Loss in Nonlinear Smooth Neural Networks","ref_index":40,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YFMVAVKXTL3M6RIBZEULHRBL26","json":"https://pith.science/pith/YFMVAVKXTL3M6RIBZEULHRBL26.json","graph_json":"https://pith.science/api/pith-number/YFMVAVKXTL3M6RIBZEULHRBL26/graph.json","events_json":"https://pith.science/api/pith-number/YFMVAVKXTL3M6RIBZEULHRBL26/events.json","paper":"https://pith.science/paper/YFMVAVKX"},"agent_actions":{"view_html":"https://pith.science/pith/YFMVAVKXTL3M6RIBZEULHRBL26","download_json":"https://pith.science/pith/YFMVAVKXTL3M6RIBZEULHRBL26.json","view_paper":"https://pith.science/paper/YFMVAVKX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.01548&json=true","fetch_graph":"https://pith.science/api/pith-number/YFMVAVKXTL3M6RIBZEULHRBL26/graph.json","fetch_events":"https://pith.science/api/pith-number/YFMVAVKXTL3M6RIBZEULHRBL26/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YFMVAVKXTL3M6RIBZEULHRBL26/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YFMVAVKXTL3M6RIBZEULHRBL26/action/storage_attestation","attest_author":"https://pith.science/pith/YFMVAVKXTL3M6RIBZEULHRBL26/action/author_attestation","sign_citation":"https://pith.science/pith/YFMVAVKXTL3M6RIBZEULHRBL26/action/citation_signature","submit_replication":"https://pith.science/pith/YFMVAVKXTL3M6RIBZEULHRBL26/action/replication_record"}},"created_at":"2026-07-05T04:04:39.414753+00:00","updated_at":"2026-07-05T04:04:39.414753+00:00"}