{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:SV4L6FEY5OHQLS4PWY6K6RJJUJ","short_pith_number":"pith:SV4L6FEY","schema_version":"1.0","canonical_sha256":"9578bf1498eb8f05cb8fb63caf4529a26fe6628813bb3f4f90dbd084a6f58a62","source":{"kind":"arxiv","id":"2201.07703","version":2},"attestation_state":"computed","paper":{"title":"Q-ViT: Fully Differentiable Quantization for Vision Transformer","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jian Cheng, Peisong Wang, Tong Yang, Zhexin Li","submitted_at":"2022-01-19T16:43:17Z","abstract_excerpt":"In this paper, we propose a fully differentiable quantization method for vision transformer (ViT) named as Q-ViT, in which both of the quantization scales and bit-widths are learnable parameters. Specifically, based on our observation that heads in ViT display different quantization robustness, we leverage head-wise bit-width to squeeze the size of Q-ViT while preserving performance. In addition, we propose a novel technique named switchable scale to resolve the convergence problem in the joint training of quantization scales and bit-widths. In this way, Q-ViT pushes the limits of ViT quantiza"},"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":"2201.07703","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CV","submitted_at":"2022-01-19T16:43:17Z","cross_cats_sorted":[],"title_canon_sha256":"7e33914539696760867dec73b23c678a9579dced6bb877499953faf134e14c09","abstract_canon_sha256":"1cbf106a4cec2ed1c828639e8d43c8a4b124dfebaabc30c8dd35bf6ce9eeb9cd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:54:50.395428Z","signature_b64":"vMGqMDYcmG+1SKmoHpGpmZyyEOQW3r0mVzBSpgMP+mNPwdHoTiRC+LEsjuzk0ZyF4NDcYyiNmhK8KtQ3K/AYCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9578bf1498eb8f05cb8fb63caf4529a26fe6628813bb3f4f90dbd084a6f58a62","last_reissued_at":"2026-07-05T04:54:50.394948Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:54:50.394948Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Q-ViT: Fully Differentiable Quantization for Vision Transformer","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Jian Cheng, Peisong Wang, Tong Yang, Zhexin Li","submitted_at":"2022-01-19T16:43:17Z","abstract_excerpt":"In this paper, we propose a fully differentiable quantization method for vision transformer (ViT) named as Q-ViT, in which both of the quantization scales and bit-widths are learnable parameters. Specifically, based on our observation that heads in ViT display different quantization robustness, we leverage head-wise bit-width to squeeze the size of Q-ViT while preserving performance. In addition, we propose a novel technique named switchable scale to resolve the convergence problem in the joint training of quantization scales and bit-widths. In this way, Q-ViT pushes the limits of ViT quantiza"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2201.07703","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/2201.07703/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":"2201.07703","created_at":"2026-07-05T04:54:50.395001+00:00"},{"alias_kind":"arxiv_version","alias_value":"2201.07703v2","created_at":"2026-07-05T04:54:50.395001+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2201.07703","created_at":"2026-07-05T04:54:50.395001+00:00"},{"alias_kind":"pith_short_12","alias_value":"SV4L6FEY5OHQ","created_at":"2026-07-05T04:54:50.395001+00:00"},{"alias_kind":"pith_short_16","alias_value":"SV4L6FEY5OHQLS4P","created_at":"2026-07-05T04:54:50.395001+00:00"},{"alias_kind":"pith_short_8","alias_value":"SV4L6FEY","created_at":"2026-07-05T04:54:50.395001+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.04115","citing_title":"dMX: Differentiable Mixed-Precision Assignment for Low-Precision Floating-Point Formats","ref_index":37,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SV4L6FEY5OHQLS4PWY6K6RJJUJ","json":"https://pith.science/pith/SV4L6FEY5OHQLS4PWY6K6RJJUJ.json","graph_json":"https://pith.science/api/pith-number/SV4L6FEY5OHQLS4PWY6K6RJJUJ/graph.json","events_json":"https://pith.science/api/pith-number/SV4L6FEY5OHQLS4PWY6K6RJJUJ/events.json","paper":"https://pith.science/paper/SV4L6FEY"},"agent_actions":{"view_html":"https://pith.science/pith/SV4L6FEY5OHQLS4PWY6K6RJJUJ","download_json":"https://pith.science/pith/SV4L6FEY5OHQLS4PWY6K6RJJUJ.json","view_paper":"https://pith.science/paper/SV4L6FEY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2201.07703&json=true","fetch_graph":"https://pith.science/api/pith-number/SV4L6FEY5OHQLS4PWY6K6RJJUJ/graph.json","fetch_events":"https://pith.science/api/pith-number/SV4L6FEY5OHQLS4PWY6K6RJJUJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SV4L6FEY5OHQLS4PWY6K6RJJUJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SV4L6FEY5OHQLS4PWY6K6RJJUJ/action/storage_attestation","attest_author":"https://pith.science/pith/SV4L6FEY5OHQLS4PWY6K6RJJUJ/action/author_attestation","sign_citation":"https://pith.science/pith/SV4L6FEY5OHQLS4PWY6K6RJJUJ/action/citation_signature","submit_replication":"https://pith.science/pith/SV4L6FEY5OHQLS4PWY6K6RJJUJ/action/replication_record"}},"created_at":"2026-07-05T04:54:50.395001+00:00","updated_at":"2026-07-05T04:54:50.395001+00:00"}