{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:UN3IU4H4SU53IC3L7PMLRI5PWD","short_pith_number":"pith:UN3IU4H4","schema_version":"1.0","canonical_sha256":"a3768a70fc953bb40b6bfbd8b8a3afb0c7291457775bb819da2cc7f2469855d1","source":{"kind":"arxiv","id":"2401.04578","version":2},"attestation_state":"computed","paper":{"title":"Effective pruning of web-scale datasets based on complexity of concept clusters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Amro Abbas, Ari S. Morcos, Evgenia Rusak, Kamalika Chaudhuri, Kushal Tirumala, Wieland Brendel","submitted_at":"2024-01-09T14:32:24Z","abstract_excerpt":"Utilizing massive web-scale datasets has led to unprecedented performance gains in machine learning models, but also imposes outlandish compute requirements for their training. In order to improve training and data efficiency, we here push the limits of pruning large-scale multimodal datasets for training CLIP-style models. Today's most effective pruning method on ImageNet clusters data samples into separate concepts according to their embedding and prunes away the most prototypical samples. We scale this approach to LAION and improve it by noting that the pruning rate should be concept-specif"},"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":"2401.04578","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2024-01-09T14:32:24Z","cross_cats_sorted":[],"title_canon_sha256":"b5ed905473b8d03e04611c02feed1fc0f2328106b90d4625ce12cb04d8e59b69","abstract_canon_sha256":"e0416724f712c7676d0054e8bd50fdb36b4d81381317a027d4b374ca7a61d0c9"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:54:54.139144Z","signature_b64":"unp3pMkfQvq/LpBFlw9OurutAeEL9IFkwuOP/o7siGZuvDXxxWIdFcaL70ubK5vUx1/GvXE6yA6V3foaJkdXBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a3768a70fc953bb40b6bfbd8b8a3afb0c7291457775bb819da2cc7f2469855d1","last_reissued_at":"2026-07-05T07:54:54.138715Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:54:54.138715Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Effective pruning of web-scale datasets based on complexity of concept clusters","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Amro Abbas, Ari S. Morcos, Evgenia Rusak, Kamalika Chaudhuri, Kushal Tirumala, Wieland Brendel","submitted_at":"2024-01-09T14:32:24Z","abstract_excerpt":"Utilizing massive web-scale datasets has led to unprecedented performance gains in machine learning models, but also imposes outlandish compute requirements for their training. In order to improve training and data efficiency, we here push the limits of pruning large-scale multimodal datasets for training CLIP-style models. Today's most effective pruning method on ImageNet clusters data samples into separate concepts according to their embedding and prunes away the most prototypical samples. We scale this approach to LAION and improve it by noting that the pruning rate should be concept-specif"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2401.04578","kind":"arxiv","version":2},"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/2401.04578/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":"2401.04578","created_at":"2026-07-05T07:54:54.138771+00:00"},{"alias_kind":"arxiv_version","alias_value":"2401.04578v2","created_at":"2026-07-05T07:54:54.138771+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2401.04578","created_at":"2026-07-05T07:54:54.138771+00:00"},{"alias_kind":"pith_short_12","alias_value":"UN3IU4H4SU53","created_at":"2026-07-05T07:54:54.138771+00:00"},{"alias_kind":"pith_short_16","alias_value":"UN3IU4H4SU53IC3L","created_at":"2026-07-05T07:54:54.138771+00:00"},{"alias_kind":"pith_short_8","alias_value":"UN3IU4H4","created_at":"2026-07-05T07:54:54.138771+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.28551","citing_title":"DataComp-VLM: Improved Open Datasets for Vision-Language Models","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2606.28551","citing_title":"DataComp-VLM: Improved Open Datasets for Vision-Language Models","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08398","citing_title":"Exploring and Exploiting Stability in Latent Flow Matching","ref_index":1,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08398","citing_title":"Exploring and Exploiting Stability in Latent Flow Matching","ref_index":1,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UN3IU4H4SU53IC3L7PMLRI5PWD","json":"https://pith.science/pith/UN3IU4H4SU53IC3L7PMLRI5PWD.json","graph_json":"https://pith.science/api/pith-number/UN3IU4H4SU53IC3L7PMLRI5PWD/graph.json","events_json":"https://pith.science/api/pith-number/UN3IU4H4SU53IC3L7PMLRI5PWD/events.json","paper":"https://pith.science/paper/UN3IU4H4"},"agent_actions":{"view_html":"https://pith.science/pith/UN3IU4H4SU53IC3L7PMLRI5PWD","download_json":"https://pith.science/pith/UN3IU4H4SU53IC3L7PMLRI5PWD.json","view_paper":"https://pith.science/paper/UN3IU4H4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2401.04578&json=true","fetch_graph":"https://pith.science/api/pith-number/UN3IU4H4SU53IC3L7PMLRI5PWD/graph.json","fetch_events":"https://pith.science/api/pith-number/UN3IU4H4SU53IC3L7PMLRI5PWD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UN3IU4H4SU53IC3L7PMLRI5PWD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UN3IU4H4SU53IC3L7PMLRI5PWD/action/storage_attestation","attest_author":"https://pith.science/pith/UN3IU4H4SU53IC3L7PMLRI5PWD/action/author_attestation","sign_citation":"https://pith.science/pith/UN3IU4H4SU53IC3L7PMLRI5PWD/action/citation_signature","submit_replication":"https://pith.science/pith/UN3IU4H4SU53IC3L7PMLRI5PWD/action/replication_record"}},"created_at":"2026-07-05T07:54:54.138771+00:00","updated_at":"2026-07-05T07:54:54.138771+00:00"}