{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:VVBDNYOY6CXNOMKTSYOXIQTA22","short_pith_number":"pith:VVBDNYOY","schema_version":"1.0","canonical_sha256":"ad4236e1d8f0aed73153961d744260d69963cb5a7e8342e1bec1d605f684e3a2","source":{"kind":"arxiv","id":"2505.22988","version":3},"attestation_state":"computed","paper":{"title":"Model-Preserving Adaptive Rounding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Albert Tseng, Christopher De Sa, Zhaofeng Sun","submitted_at":"2025-05-29T01:53:00Z","abstract_excerpt":"The goal of quantization is to produce a compressed model whose output distribution is as close to the original model's as possible. To do this tractably, most quantization algorithms minimize the immediate activation error of each layer as a proxy for the end-to-end error. However, this ignores the effect of future layers, making it a poor proxy. In this work, we introduce Yet Another Quantization Algorithm (YAQA), an adaptive rounding algorithm that directly considers the error at the network's output. YAQA introduces a series of theoretical results that culminate in the first end-to-end err"},"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":"2505.22988","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-29T01:53:00Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"c3cfd9142197ad61adf7b37cad59b17e7a6f13a9e4d871a99afa0286ee610cb4","abstract_canon_sha256":"871f9c0982794ae5f2c88d4cd149c3cf8fded122abc3d053b0ff57d7800292eb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-04T01:08:29.355171Z","signature_b64":"nKOcY09l9OJnZ6jvRWJZqdU5VAexI76O0sFI0uCzldjCEdBoGbBdeDeBEoAANAc6gWyd8nt5ZDWHpgL8Rt2VDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ad4236e1d8f0aed73153961d744260d69963cb5a7e8342e1bec1d605f684e3a2","last_reissued_at":"2026-06-04T01:08:29.354604Z","signature_status":"signed_v1","first_computed_at":"2026-06-04T01:08:29.354604Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Model-Preserving Adaptive Rounding","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Albert Tseng, Christopher De Sa, Zhaofeng Sun","submitted_at":"2025-05-29T01:53:00Z","abstract_excerpt":"The goal of quantization is to produce a compressed model whose output distribution is as close to the original model's as possible. To do this tractably, most quantization algorithms minimize the immediate activation error of each layer as a proxy for the end-to-end error. However, this ignores the effect of future layers, making it a poor proxy. In this work, we introduce Yet Another Quantization Algorithm (YAQA), an adaptive rounding algorithm that directly considers the error at the network's output. YAQA introduces a series of theoretical results that culminate in the first end-to-end err"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.22988","kind":"arxiv","version":3},"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/2505.22988/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":"2505.22988","created_at":"2026-06-04T01:08:29.354664+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.22988v3","created_at":"2026-06-04T01:08:29.354664+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.22988","created_at":"2026-06-04T01:08:29.354664+00:00"},{"alias_kind":"pith_short_12","alias_value":"VVBDNYOY6CXN","created_at":"2026-06-04T01:08:29.354664+00:00"},{"alias_kind":"pith_short_16","alias_value":"VVBDNYOY6CXNOMKT","created_at":"2026-06-04T01:08:29.354664+00:00"},{"alias_kind":"pith_short_8","alias_value":"VVBDNYOY","created_at":"2026-06-04T01:08:29.354664+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":5,"sample":[{"citing_arxiv_id":"2605.15626","citing_title":"IO-SVD: Input-Output Whitened SVD for Adaptive-Rank LLM Compression","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2601.17187","citing_title":"High-Rate Quantized Matrix Multiplication I","ref_index":27,"is_internal_anchor":true},{"citing_arxiv_id":"2602.05902","citing_title":"CoreQ: Learning-Free Mismatch Correction and Successive Rounding for Quantization","ref_index":16,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13768","citing_title":"High-Rate Quantized Matrix Multiplication II","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2605.02404","citing_title":"Statistically-Lossless Quantization of Large Language Models","ref_index":9,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VVBDNYOY6CXNOMKTSYOXIQTA22","json":"https://pith.science/pith/VVBDNYOY6CXNOMKTSYOXIQTA22.json","graph_json":"https://pith.science/api/pith-number/VVBDNYOY6CXNOMKTSYOXIQTA22/graph.json","events_json":"https://pith.science/api/pith-number/VVBDNYOY6CXNOMKTSYOXIQTA22/events.json","paper":"https://pith.science/paper/VVBDNYOY"},"agent_actions":{"view_html":"https://pith.science/pith/VVBDNYOY6CXNOMKTSYOXIQTA22","download_json":"https://pith.science/pith/VVBDNYOY6CXNOMKTSYOXIQTA22.json","view_paper":"https://pith.science/paper/VVBDNYOY","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.22988&json=true","fetch_graph":"https://pith.science/api/pith-number/VVBDNYOY6CXNOMKTSYOXIQTA22/graph.json","fetch_events":"https://pith.science/api/pith-number/VVBDNYOY6CXNOMKTSYOXIQTA22/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VVBDNYOY6CXNOMKTSYOXIQTA22/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VVBDNYOY6CXNOMKTSYOXIQTA22/action/storage_attestation","attest_author":"https://pith.science/pith/VVBDNYOY6CXNOMKTSYOXIQTA22/action/author_attestation","sign_citation":"https://pith.science/pith/VVBDNYOY6CXNOMKTSYOXIQTA22/action/citation_signature","submit_replication":"https://pith.science/pith/VVBDNYOY6CXNOMKTSYOXIQTA22/action/replication_record"}},"created_at":"2026-06-04T01:08:29.354664+00:00","updated_at":"2026-06-04T01:08:29.354664+00:00"}