{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PMVYEBDGVIB6C2RACTIXJUSO2N","short_pith_number":"pith:PMVYEBDG","schema_version":"1.0","canonical_sha256":"7b2b820466aa03e16a2014d174d24ed3637baca142b3800656b135c9813846a4","source":{"kind":"arxiv","id":"2507.10547","version":1},"attestation_state":"computed","paper":{"title":"Quantize-then-Rectify: Efficient VQ-VAE Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Borui Zhang, Jie Zhou, Jiwen Lu, Qihang Rao, Wenzhao Zheng","submitted_at":"2025-07-14T17:59:41Z","abstract_excerpt":"Visual tokenizers are pivotal in multimodal large models, acting as bridges between continuous inputs and discrete tokens. Nevertheless, training high-compression-rate VQ-VAEs remains computationally demanding, often necessitating thousands of GPU hours. This work demonstrates that a pre-trained VAE can be efficiently transformed into a VQ-VAE by controlling quantization noise within the VAE's tolerance threshold. We present \\textbf{Quantize-then-Rectify (ReVQ)}, a framework leveraging pre-trained VAEs to enable rapid VQ-VAE training with minimal computational overhead. By integrating \\textbf{"},"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":"2507.10547","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2025-07-14T17:59:41Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"4acb1d0ce027d111233337f530ee322fa30f818620e5db414745430a4b688298","abstract_canon_sha256":"dc4e0c3aafaab264f1814a4d343f4eb01c528af773c11d02efd4e6354ce4574d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:36:53.436891Z","signature_b64":"V39vUlbO3fDXLe/N6Hk6L3yu0GyztPya0J/kwu2DA+qfGEQQ+Xi4FsLwBbvRdyvJrTI5NdoJPHqrb/43RBJADQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7b2b820466aa03e16a2014d174d24ed3637baca142b3800656b135c9813846a4","last_reissued_at":"2026-07-05T11:36:53.436363Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:36:53.436363Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Quantize-then-Rectify: Efficient VQ-VAE Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CV","authors_text":"Borui Zhang, Jie Zhou, Jiwen Lu, Qihang Rao, Wenzhao Zheng","submitted_at":"2025-07-14T17:59:41Z","abstract_excerpt":"Visual tokenizers are pivotal in multimodal large models, acting as bridges between continuous inputs and discrete tokens. Nevertheless, training high-compression-rate VQ-VAEs remains computationally demanding, often necessitating thousands of GPU hours. This work demonstrates that a pre-trained VAE can be efficiently transformed into a VQ-VAE by controlling quantization noise within the VAE's tolerance threshold. We present \\textbf{Quantize-then-Rectify (ReVQ)}, a framework leveraging pre-trained VAEs to enable rapid VQ-VAE training with minimal computational overhead. By integrating \\textbf{"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.10547","kind":"arxiv","version":1},"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/2507.10547/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":"2507.10547","created_at":"2026-07-05T11:36:53.436426+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.10547v1","created_at":"2026-07-05T11:36:53.436426+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.10547","created_at":"2026-07-05T11:36:53.436426+00:00"},{"alias_kind":"pith_short_12","alias_value":"PMVYEBDGVIB6","created_at":"2026-07-05T11:36:53.436426+00:00"},{"alias_kind":"pith_short_16","alias_value":"PMVYEBDGVIB6C2RA","created_at":"2026-07-05T11:36:53.436426+00:00"},{"alias_kind":"pith_short_8","alias_value":"PMVYEBDG","created_at":"2026-07-05T11:36:53.436426+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11363","citing_title":"NSVQ: Mitigating Codebook Collapse by Stabilizing Encoder Drift in Vector Quantization","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.04461","citing_title":"ChannelTok: Efficient Flexible-Length Vision Tokenization","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PMVYEBDGVIB6C2RACTIXJUSO2N","json":"https://pith.science/pith/PMVYEBDGVIB6C2RACTIXJUSO2N.json","graph_json":"https://pith.science/api/pith-number/PMVYEBDGVIB6C2RACTIXJUSO2N/graph.json","events_json":"https://pith.science/api/pith-number/PMVYEBDGVIB6C2RACTIXJUSO2N/events.json","paper":"https://pith.science/paper/PMVYEBDG"},"agent_actions":{"view_html":"https://pith.science/pith/PMVYEBDGVIB6C2RACTIXJUSO2N","download_json":"https://pith.science/pith/PMVYEBDGVIB6C2RACTIXJUSO2N.json","view_paper":"https://pith.science/paper/PMVYEBDG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.10547&json=true","fetch_graph":"https://pith.science/api/pith-number/PMVYEBDGVIB6C2RACTIXJUSO2N/graph.json","fetch_events":"https://pith.science/api/pith-number/PMVYEBDGVIB6C2RACTIXJUSO2N/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PMVYEBDGVIB6C2RACTIXJUSO2N/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PMVYEBDGVIB6C2RACTIXJUSO2N/action/storage_attestation","attest_author":"https://pith.science/pith/PMVYEBDGVIB6C2RACTIXJUSO2N/action/author_attestation","sign_citation":"https://pith.science/pith/PMVYEBDGVIB6C2RACTIXJUSO2N/action/citation_signature","submit_replication":"https://pith.science/pith/PMVYEBDGVIB6C2RACTIXJUSO2N/action/replication_record"}},"created_at":"2026-07-05T11:36:53.436426+00:00","updated_at":"2026-07-05T11:36:53.436426+00:00"}