{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BI774JRQE7DVCKZ62D4BCLOW2Q","short_pith_number":"pith:BI774JRQ","schema_version":"1.0","canonical_sha256":"0a3ffe263027c7512b3ed0f8112dd6d42c06f485abb6aefcf2340ed2109c50e0","source":{"kind":"arxiv","id":"2504.18547","version":2},"attestation_state":"computed","paper":{"title":"Low-Bit Integerization of Vision Transformers using Operand Reordering for Efficient Hardware","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CV","cs.SY","eess.SY"],"primary_cat":"cs.LG","authors_text":"Ching-Yi Lin, Sahil Shah","submitted_at":"2025-04-11T16:09:54Z","abstract_excerpt":"Pre-trained vision transformers have achieved remarkable performance across various visual tasks but suffer from expensive computational and memory costs. While model quantization reduces memory usage by lowering precision, these models still incur significant computational overhead due to the dequantization before matrix operations. In this work, we analyze the computation graph and propose an integerization process based on operation reordering. Specifically, the process delays dequantization until after matrix operations. This enables integerized matrix multiplication and linear module by d"},"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":"2504.18547","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-11T16:09:54Z","cross_cats_sorted":["cs.CV","cs.SY","eess.SY"],"title_canon_sha256":"5589755698bfa606b74630858767c493d1c2437f93cf715c719df0eb192b6934","abstract_canon_sha256":"d0a86e64419df5d7b094a876276ff8750f35b43552c826d97996485a0294a02e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:48:25.495341Z","signature_b64":"zlEp1zsUv9HEA/c3/BuTkf0v4d76raKL66JTal+RMvmym+Tn9Kh3B6zd2oOFar1X9NybfJYMyOC70Dg5b9ScCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0a3ffe263027c7512b3ed0f8112dd6d42c06f485abb6aefcf2340ed2109c50e0","last_reissued_at":"2026-07-05T11:48:25.494853Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:48:25.494853Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Low-Bit Integerization of Vision Transformers using Operand Reordering for Efficient Hardware","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.CV","cs.SY","eess.SY"],"primary_cat":"cs.LG","authors_text":"Ching-Yi Lin, Sahil Shah","submitted_at":"2025-04-11T16:09:54Z","abstract_excerpt":"Pre-trained vision transformers have achieved remarkable performance across various visual tasks but suffer from expensive computational and memory costs. While model quantization reduces memory usage by lowering precision, these models still incur significant computational overhead due to the dequantization before matrix operations. In this work, we analyze the computation graph and propose an integerization process based on operation reordering. Specifically, the process delays dequantization until after matrix operations. This enables integerized matrix multiplication and linear module by d"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.18547","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/2504.18547/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":"2504.18547","created_at":"2026-07-05T11:48:25.494910+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.18547v2","created_at":"2026-07-05T11:48:25.494910+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.18547","created_at":"2026-07-05T11:48:25.494910+00:00"},{"alias_kind":"pith_short_12","alias_value":"BI774JRQE7DV","created_at":"2026-07-05T11:48:25.494910+00:00"},{"alias_kind":"pith_short_16","alias_value":"BI774JRQE7DVCKZ6","created_at":"2026-07-05T11:48:25.494910+00:00"},{"alias_kind":"pith_short_8","alias_value":"BI774JRQ","created_at":"2026-07-05T11:48:25.494910+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.20334","citing_title":"Systolic Array-based Architecture for Low-Bit Integerized Vision Transformers","ref_index":6,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BI774JRQE7DVCKZ62D4BCLOW2Q","json":"https://pith.science/pith/BI774JRQE7DVCKZ62D4BCLOW2Q.json","graph_json":"https://pith.science/api/pith-number/BI774JRQE7DVCKZ62D4BCLOW2Q/graph.json","events_json":"https://pith.science/api/pith-number/BI774JRQE7DVCKZ62D4BCLOW2Q/events.json","paper":"https://pith.science/paper/BI774JRQ"},"agent_actions":{"view_html":"https://pith.science/pith/BI774JRQE7DVCKZ62D4BCLOW2Q","download_json":"https://pith.science/pith/BI774JRQE7DVCKZ62D4BCLOW2Q.json","view_paper":"https://pith.science/paper/BI774JRQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.18547&json=true","fetch_graph":"https://pith.science/api/pith-number/BI774JRQE7DVCKZ62D4BCLOW2Q/graph.json","fetch_events":"https://pith.science/api/pith-number/BI774JRQE7DVCKZ62D4BCLOW2Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BI774JRQE7DVCKZ62D4BCLOW2Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BI774JRQE7DVCKZ62D4BCLOW2Q/action/storage_attestation","attest_author":"https://pith.science/pith/BI774JRQE7DVCKZ62D4BCLOW2Q/action/author_attestation","sign_citation":"https://pith.science/pith/BI774JRQE7DVCKZ62D4BCLOW2Q/action/citation_signature","submit_replication":"https://pith.science/pith/BI774JRQE7DVCKZ62D4BCLOW2Q/action/replication_record"}},"created_at":"2026-07-05T11:48:25.494910+00:00","updated_at":"2026-07-05T11:48:25.494910+00:00"}