{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2023:CGWJ6SZNOYXIONKDZGJEHFKJIE","short_pith_number":"pith:CGWJ6SZN","canonical_record":{"source":{"id":"2309.13681","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-09-24T16:08:21Z","cross_cats_sorted":[],"title_canon_sha256":"a7c4b85ce015c067c0021879ae634060a46f80b571f8b5014447dac21910a929","abstract_canon_sha256":"ccead3c5b82ea79a6fe9b0669d05f06a81afb861e15422f3e8f9acac1fb67d6c"},"schema_version":"1.0"},"canonical_sha256":"11ac9f4b2d762e873543c99243954941340615474ed747d1221b819941a30a5b","source":{"kind":"arxiv","id":"2309.13681","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2309.13681","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"arxiv_version","alias_value":"2309.13681v1","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.13681","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"pith_short_12","alias_value":"CGWJ6SZNOYXI","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"pith_short_16","alias_value":"CGWJ6SZNOYXIONKD","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"pith_short_8","alias_value":"CGWJ6SZN","created_at":"2026-07-05T06:53:55Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2023:CGWJ6SZNOYXIONKDZGJEHFKJIE","target":"record","payload":{"canonical_record":{"source":{"id":"2309.13681","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-09-24T16:08:21Z","cross_cats_sorted":[],"title_canon_sha256":"a7c4b85ce015c067c0021879ae634060a46f80b571f8b5014447dac21910a929","abstract_canon_sha256":"ccead3c5b82ea79a6fe9b0669d05f06a81afb861e15422f3e8f9acac1fb67d6c"},"schema_version":"1.0"},"canonical_sha256":"11ac9f4b2d762e873543c99243954941340615474ed747d1221b819941a30a5b","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:53:55.841778Z","signature_b64":"4weyoSFFFw2h9BDNIjQBKO8Spc0qSNqHXCt6GB8auzkORNINTeyHgZgTG/RYTdWg8SpfnBl5ZF8M9SJO1YeCAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"11ac9f4b2d762e873543c99243954941340615474ed747d1221b819941a30a5b","last_reissued_at":"2026-07-05T06:53:55.841372Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:53:55.841372Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2309.13681","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T06:53:55Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AVLcun169er4bAfnyvu/IMN61+CDbJ35ISE/QfrqQl4Iz1ZOybma+c8yfGD9oXlK+yMEnM2sEttn/RgUJp4wDg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T17:12:45.224752Z"},"content_sha256":"6b5548d0ab31fe89d1b0f6a590448a7b22fcacdd5274f432f71c8698c8249bef","schema_version":"1.0","event_id":"sha256:6b5548d0ab31fe89d1b0f6a590448a7b22fcacdd5274f432f71c8698c8249bef"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2023:CGWJ6SZNOYXIONKDZGJEHFKJIE","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Accelerating Large Batch Training via Gradient Signal to Noise Ratio (GSNR)","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Guo-qing Jiang, Jinlong Liu, Lin Guo, Wei Lin, Zixiang Ding","submitted_at":"2023-09-24T16:08:21Z","abstract_excerpt":"As models for nature language processing (NLP), computer vision (CV) and recommendation systems (RS) require surging computation, a large number of GPUs/TPUs are paralleled as a large batch (LB) to improve training throughput. However, training such LB tasks often meets large generalization gap and downgrades final precision, which limits enlarging the batch size. In this work, we develop the variance reduced gradient descent technique (VRGD) based on the gradient signal to noise ratio (GSNR) and apply it onto popular optimizers such as SGD/Adam/LARS/LAMB. We carry out a theoretical analysis o"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.13681","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/2309.13681/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T06:53:55Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"cSMza1C+KSHfyoLXrNCvRw5mRvluZGYLcdTLo3NcH+Xm7s7+j9deRkFZhCRDDqo4/+fvEz3itT1cisVxWuzdBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-18T17:12:45.225383Z"},"content_sha256":"6db5f08e5e2cf0f7362dfae6e0375f68126af508d0f6d47625de2a168e99a752","schema_version":"1.0","event_id":"sha256:6db5f08e5e2cf0f7362dfae6e0375f68126af508d0f6d47625de2a168e99a752"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/CGWJ6SZNOYXIONKDZGJEHFKJIE/bundle.json","state_url":"https://pith.science/pith/CGWJ6SZNOYXIONKDZGJEHFKJIE/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/CGWJ6SZNOYXIONKDZGJEHFKJIE/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-18T17:12:45Z","links":{"resolver":"https://pith.science/pith/CGWJ6SZNOYXIONKDZGJEHFKJIE","bundle":"https://pith.science/pith/CGWJ6SZNOYXIONKDZGJEHFKJIE/bundle.json","state":"https://pith.science/pith/CGWJ6SZNOYXIONKDZGJEHFKJIE/state.json","well_known_bundle":"https://pith.science/.well-known/pith/CGWJ6SZNOYXIONKDZGJEHFKJIE/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2023:CGWJ6SZNOYXIONKDZGJEHFKJIE","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"ccead3c5b82ea79a6fe9b0669d05f06a81afb861e15422f3e8f9acac1fb67d6c","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-09-24T16:08:21Z","title_canon_sha256":"a7c4b85ce015c067c0021879ae634060a46f80b571f8b5014447dac21910a929"},"schema_version":"1.0","source":{"id":"2309.13681","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2309.13681","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"arxiv_version","alias_value":"2309.13681v1","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.13681","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"pith_short_12","alias_value":"CGWJ6SZNOYXI","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"pith_short_16","alias_value":"CGWJ6SZNOYXIONKD","created_at":"2026-07-05T06:53:55Z"},{"alias_kind":"pith_short_8","alias_value":"CGWJ6SZN","created_at":"2026-07-05T06:53:55Z"}],"graph_snapshots":[{"event_id":"sha256:6db5f08e5e2cf0f7362dfae6e0375f68126af508d0f6d47625de2a168e99a752","target":"graph","created_at":"2026-07-05T06:53:55Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2309.13681/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"As models for nature language processing (NLP), computer vision (CV) and recommendation systems (RS) require surging computation, a large number of GPUs/TPUs are paralleled as a large batch (LB) to improve training throughput. However, training such LB tasks often meets large generalization gap and downgrades final precision, which limits enlarging the batch size. In this work, we develop the variance reduced gradient descent technique (VRGD) based on the gradient signal to noise ratio (GSNR) and apply it onto popular optimizers such as SGD/Adam/LARS/LAMB. We carry out a theoretical analysis o","authors_text":"Guo-qing Jiang, Jinlong Liu, Lin Guo, Wei Lin, Zixiang Ding","cross_cats":[],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-09-24T16:08:21Z","title":"Accelerating Large Batch Training via Gradient Signal to Noise Ratio (GSNR)"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.13681","kind":"arxiv","version":1},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:6b5548d0ab31fe89d1b0f6a590448a7b22fcacdd5274f432f71c8698c8249bef","target":"record","created_at":"2026-07-05T06:53:55Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"ccead3c5b82ea79a6fe9b0669d05f06a81afb861e15422f3e8f9acac1fb67d6c","cross_cats_sorted":[],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-09-24T16:08:21Z","title_canon_sha256":"a7c4b85ce015c067c0021879ae634060a46f80b571f8b5014447dac21910a929"},"schema_version":"1.0","source":{"id":"2309.13681","kind":"arxiv","version":1}},"canonical_sha256":"11ac9f4b2d762e873543c99243954941340615474ed747d1221b819941a30a5b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"11ac9f4b2d762e873543c99243954941340615474ed747d1221b819941a30a5b","first_computed_at":"2026-07-05T06:53:55.841372Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T06:53:55.841372Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"4weyoSFFFw2h9BDNIjQBKO8Spc0qSNqHXCt6GB8auzkORNINTeyHgZgTG/RYTdWg8SpfnBl5ZF8M9SJO1YeCAQ==","signature_status":"signed_v1","signed_at":"2026-07-05T06:53:55.841778Z","signed_message":"canonical_sha256_bytes"},"source_id":"2309.13681","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:6b5548d0ab31fe89d1b0f6a590448a7b22fcacdd5274f432f71c8698c8249bef","sha256:6db5f08e5e2cf0f7362dfae6e0375f68126af508d0f6d47625de2a168e99a752"],"state_sha256":"a39b7ef6281e6195112dd32381f52480a4c0ad5ed90f8f413300df97095e34ba"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"0WyiWurqIwDAV+viSkS1z08SMoOxCV0Fvgr0etra/u0H+DNpvExhr/EKm3Kzhnz8dg64CE/ZVlCaDZqw1+6sDQ==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-18T17:12:45.230866Z","bundle_sha256":"38a07831aa232e7e62e15e05c74f36476759ee5c09914a6babe2dbf8ad4a853a"}}