{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FVPBNSBSCWKLOHDASC46PTZXEI","short_pith_number":"pith:FVPBNSBS","schema_version":"1.0","canonical_sha256":"2d5e16c8321594b71c6090b9e7cf372213f10ef0b29e40b73447a8fa8646306c","source":{"kind":"arxiv","id":"2403.17726","version":4},"attestation_state":"computed","paper":{"title":"Tiny Models are the Computational Saver for Large Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Antoine Frapp\\'e, Barry Cardiff, Benoit Larras, Deepu John, Qingyuan Wang","submitted_at":"2024-03-26T14:14:30Z","abstract_excerpt":"This paper introduces TinySaver, an early-exit-like dynamic model compression approach which employs tiny models to substitute large models adaptively. Distinct from traditional compression techniques, dynamic methods like TinySaver can leverage the difficulty differences to allow certain inputs to complete their inference processes early, thereby conserving computational resources. Most existing early exit designs are implemented by attaching additional network branches to the model's backbone. Our study, however, reveals that completely independent tiny models can replace a substantial porti"},"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":"2403.17726","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2024-03-26T14:14:30Z","cross_cats_sorted":[],"title_canon_sha256":"3ca7ae80e37501e2a2f7f8793c51e347c76c63a0ac1127bf9a4bc337e4f59a24","abstract_canon_sha256":"8fc6f2de823657f078a66e7aeb7bb382c5ab43b66278f818657f0b9916f3e252"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:59:45.162805Z","signature_b64":"dY8k2/8odHS5+gMdW61QFxpHFZxiN+s3z4r4XtChtd1tey6PQb4Y1+WnvA3ZachytTOm+TRDqaEFclFnF6FgCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2d5e16c8321594b71c6090b9e7cf372213f10ef0b29e40b73447a8fa8646306c","last_reissued_at":"2026-07-05T09:59:45.162344Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:59:45.162344Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Tiny Models are the Computational Saver for Large Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Antoine Frapp\\'e, Barry Cardiff, Benoit Larras, Deepu John, Qingyuan Wang","submitted_at":"2024-03-26T14:14:30Z","abstract_excerpt":"This paper introduces TinySaver, an early-exit-like dynamic model compression approach which employs tiny models to substitute large models adaptively. Distinct from traditional compression techniques, dynamic methods like TinySaver can leverage the difficulty differences to allow certain inputs to complete their inference processes early, thereby conserving computational resources. Most existing early exit designs are implemented by attaching additional network branches to the model's backbone. Our study, however, reveals that completely independent tiny models can replace a substantial porti"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.17726","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2403.17726/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":"2403.17726","created_at":"2026-07-05T09:59:45.162412+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.17726v4","created_at":"2026-07-05T09:59:45.162412+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.17726","created_at":"2026-07-05T09:59:45.162412+00:00"},{"alias_kind":"pith_short_12","alias_value":"FVPBNSBSCWKL","created_at":"2026-07-05T09:59:45.162412+00:00"},{"alias_kind":"pith_short_16","alias_value":"FVPBNSBSCWKLOHDA","created_at":"2026-07-05T09:59:45.162412+00:00"},{"alias_kind":"pith_short_8","alias_value":"FVPBNSBS","created_at":"2026-07-05T09:59:45.162412+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2412.00724","citing_title":"AdaScale: Dynamic Context-aware DNN Scaling via Automated Adaptation Loop on Mobile Devices","ref_index":51,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FVPBNSBSCWKLOHDASC46PTZXEI","json":"https://pith.science/pith/FVPBNSBSCWKLOHDASC46PTZXEI.json","graph_json":"https://pith.science/api/pith-number/FVPBNSBSCWKLOHDASC46PTZXEI/graph.json","events_json":"https://pith.science/api/pith-number/FVPBNSBSCWKLOHDASC46PTZXEI/events.json","paper":"https://pith.science/paper/FVPBNSBS"},"agent_actions":{"view_html":"https://pith.science/pith/FVPBNSBSCWKLOHDASC46PTZXEI","download_json":"https://pith.science/pith/FVPBNSBSCWKLOHDASC46PTZXEI.json","view_paper":"https://pith.science/paper/FVPBNSBS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.17726&json=true","fetch_graph":"https://pith.science/api/pith-number/FVPBNSBSCWKLOHDASC46PTZXEI/graph.json","fetch_events":"https://pith.science/api/pith-number/FVPBNSBSCWKLOHDASC46PTZXEI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FVPBNSBSCWKLOHDASC46PTZXEI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FVPBNSBSCWKLOHDASC46PTZXEI/action/storage_attestation","attest_author":"https://pith.science/pith/FVPBNSBSCWKLOHDASC46PTZXEI/action/author_attestation","sign_citation":"https://pith.science/pith/FVPBNSBSCWKLOHDASC46PTZXEI/action/citation_signature","submit_replication":"https://pith.science/pith/FVPBNSBSCWKLOHDASC46PTZXEI/action/replication_record"}},"created_at":"2026-07-05T09:59:45.162412+00:00","updated_at":"2026-07-05T09:59:45.162412+00:00"}