{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:QEIQQNIMJSXHJVBPBEFTEB3AKC","short_pith_number":"pith:QEIQQNIM","schema_version":"1.0","canonical_sha256":"811108350c4cae74d42f090b32076050b8d2904ee0306f143f738dcd7db254e9","source":{"kind":"arxiv","id":"2010.16248","version":1},"attestation_state":"computed","paper":{"title":"Accordion: Adaptive Gradient Communication via Critical Learning Regime Identification","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dimitris Papailiopoulos, Hongyi Wang, Kangwook Lee, Saurabh Agarwal, Shivaram Venkataraman","submitted_at":"2020-10-29T16:41:44Z","abstract_excerpt":"Distributed model training suffers from communication bottlenecks due to frequent model updates transmitted across compute nodes. To alleviate these bottlenecks, practitioners use gradient compression techniques like sparsification, quantization, or low-rank updates. The techniques usually require choosing a static compression ratio, often requiring users to balance the trade-off between model accuracy and per-iteration speedup. In this work, we show that such performance degradation due to choosing a high compression ratio is not fundamental. An adaptive compression strategy can reduce commun"},"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":"2010.16248","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-29T16:41:44Z","cross_cats_sorted":[],"title_canon_sha256":"4d6c467c4705cd1545503fa88b828a6f178ffd479f6d1720252adda273d29460","abstract_canon_sha256":"04d2c7dbd65c591d1d0ebd0a1cf838823e4f0b559c31497b9039c887638461ea"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:47:50.761220Z","signature_b64":"hrUo+V0h9omC/CEz8DkVo5lunPvrH/uJa7uymXg5i/lnP5HXzd0qrISqyw7ORUEeen4b9VXMzLHrLCDwzUI6CQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"811108350c4cae74d42f090b32076050b8d2904ee0306f143f738dcd7db254e9","last_reissued_at":"2026-07-05T01:47:50.760837Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:47:50.760837Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Accordion: Adaptive Gradient Communication via Critical Learning Regime Identification","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dimitris Papailiopoulos, Hongyi Wang, Kangwook Lee, Saurabh Agarwal, Shivaram Venkataraman","submitted_at":"2020-10-29T16:41:44Z","abstract_excerpt":"Distributed model training suffers from communication bottlenecks due to frequent model updates transmitted across compute nodes. To alleviate these bottlenecks, practitioners use gradient compression techniques like sparsification, quantization, or low-rank updates. The techniques usually require choosing a static compression ratio, often requiring users to balance the trade-off between model accuracy and per-iteration speedup. In this work, we show that such performance degradation due to choosing a high compression ratio is not fundamental. An adaptive compression strategy can reduce commun"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.16248","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/2010.16248/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":"2010.16248","created_at":"2026-07-05T01:47:50.760903+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.16248v1","created_at":"2026-07-05T01:47:50.760903+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.16248","created_at":"2026-07-05T01:47:50.760903+00:00"},{"alias_kind":"pith_short_12","alias_value":"QEIQQNIMJSXH","created_at":"2026-07-05T01:47:50.760903+00:00"},{"alias_kind":"pith_short_16","alias_value":"QEIQQNIMJSXHJVBP","created_at":"2026-07-05T01:47:50.760903+00:00"},{"alias_kind":"pith_short_8","alias_value":"QEIQQNIM","created_at":"2026-07-05T01:47:50.760903+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.01989","citing_title":"DBLP: Phase-Aware Bounded-Loss Transport for Burst-Resilient Distributed ML Training","ref_index":19,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QEIQQNIMJSXHJVBPBEFTEB3AKC","json":"https://pith.science/pith/QEIQQNIMJSXHJVBPBEFTEB3AKC.json","graph_json":"https://pith.science/api/pith-number/QEIQQNIMJSXHJVBPBEFTEB3AKC/graph.json","events_json":"https://pith.science/api/pith-number/QEIQQNIMJSXHJVBPBEFTEB3AKC/events.json","paper":"https://pith.science/paper/QEIQQNIM"},"agent_actions":{"view_html":"https://pith.science/pith/QEIQQNIMJSXHJVBPBEFTEB3AKC","download_json":"https://pith.science/pith/QEIQQNIMJSXHJVBPBEFTEB3AKC.json","view_paper":"https://pith.science/paper/QEIQQNIM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.16248&json=true","fetch_graph":"https://pith.science/api/pith-number/QEIQQNIMJSXHJVBPBEFTEB3AKC/graph.json","fetch_events":"https://pith.science/api/pith-number/QEIQQNIMJSXHJVBPBEFTEB3AKC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QEIQQNIMJSXHJVBPBEFTEB3AKC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QEIQQNIMJSXHJVBPBEFTEB3AKC/action/storage_attestation","attest_author":"https://pith.science/pith/QEIQQNIMJSXHJVBPBEFTEB3AKC/action/author_attestation","sign_citation":"https://pith.science/pith/QEIQQNIMJSXHJVBPBEFTEB3AKC/action/citation_signature","submit_replication":"https://pith.science/pith/QEIQQNIMJSXHJVBPBEFTEB3AKC/action/replication_record"}},"created_at":"2026-07-05T01:47:50.760903+00:00","updated_at":"2026-07-05T01:47:50.760903+00:00"}