{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:25SB56GMBDVJLSY7HLJ53X5QMD","short_pith_number":"pith:25SB56GM","canonical_record":{"source":{"id":"2505.18313","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-23T19:17:55Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"396a3dbf9f5d39e2685e048f4168122fd0619c2c1cd9d389634c517ee6439c42","abstract_canon_sha256":"ad2d155b3ecbb8b7827f29fba2ed89c20fd522757f2060a22cc1b6c902f58d11"},"schema_version":"1.0"},"canonical_sha256":"d7641ef8cc08ea95cb1f3ad3dddfb060fe340ca6fa1982e25144ac0b94788a07","source":{"kind":"arxiv","id":"2505.18313","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.18313","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"arxiv_version","alias_value":"2505.18313v1","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.18313","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"pith_short_12","alias_value":"25SB56GMBDVJ","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"pith_short_16","alias_value":"25SB56GMBDVJLSY7","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"pith_short_8","alias_value":"25SB56GM","created_at":"2026-07-05T11:08:44Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:25SB56GMBDVJLSY7HLJ53X5QMD","target":"record","payload":{"canonical_record":{"source":{"id":"2505.18313","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-23T19:17:55Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"396a3dbf9f5d39e2685e048f4168122fd0619c2c1cd9d389634c517ee6439c42","abstract_canon_sha256":"ad2d155b3ecbb8b7827f29fba2ed89c20fd522757f2060a22cc1b6c902f58d11"},"schema_version":"1.0"},"canonical_sha256":"d7641ef8cc08ea95cb1f3ad3dddfb060fe340ca6fa1982e25144ac0b94788a07","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:08:44.544265Z","signature_b64":"dkI08UrrfwS+YVZ0/Y6HW97jMOvwXc+OZjXp1SUNsr+e7F8C49H/yn1MFd7d0RQNu2BULothO5YuMP+MhZ4DAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d7641ef8cc08ea95cb1f3ad3dddfb060fe340ca6fa1982e25144ac0b94788a07","last_reissued_at":"2026-07-05T11:08:44.543791Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:08:44.543791Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2505.18313","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":"2026-07-05T11:08:44Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"/++zXVOdajCKkhVGSzY3j3HYEbJWlUBSrPWqvhi4eLB+KTw9/ecrP4+8YrEi5ptJ8uMjtF4bPFVFMHZYmGu1Aw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-14T13:47:14.050257Z"},"content_sha256":"ede887a6358774943bee1670b712e0671fbc9c49d13f5f68d4cce414bd6bbfea","schema_version":"1.0","event_id":"sha256:ede887a6358774943bee1670b712e0671fbc9c49d13f5f68d4cce414bd6bbfea"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:25SB56GMBDVJLSY7HLJ53X5QMD","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"PLUMAGE: Probabilistic Low rank Unbiased Min Variance Gradient Estimator for Efficient Large Model Training","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Daniel Soudry, Matan Haroush","submitted_at":"2025-05-23T19:17:55Z","abstract_excerpt":"Accelerator memory and networking constraints have emerged as dominant bottlenecks when training large language models LLMs with billions of parameters. Existing low rank gradient estimators such as GaLoRE and FLORA compress gradients and optimizer tensors by projecting weight gradients onto a rank r subspace, enabling LLM training on consumer hardware. Yet, these methods are either biased or subject to high estimator variance. Moreover, the optimizer state based on the first and second moments estimates expressed in the previous subspace becomes misaligned whenever the projection is updated, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.18313","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/2505.18313/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":"2026-07-05T11:08:44Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"LnKLh2plrdynu093ijDVdoegAzx4EI827sLeE3CMv5TgwrN1rVzWaSGhIAsClVCNU7pyuphAdmFifPpkv69QCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-14T13:47:14.050828Z"},"content_sha256":"6c47c12734f4e7948ae619ada4ce6cc98e862e67c71703904e34613c83072dd2","schema_version":"1.0","event_id":"sha256:6c47c12734f4e7948ae619ada4ce6cc98e862e67c71703904e34613c83072dd2"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/25SB56GMBDVJLSY7HLJ53X5QMD/bundle.json","state_url":"https://pith.science/pith/25SB56GMBDVJLSY7HLJ53X5QMD/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/25SB56GMBDVJLSY7HLJ53X5QMD/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":"2026-08-14T13:47:14Z","links":{"resolver":"https://pith.science/pith/25SB56GMBDVJLSY7HLJ53X5QMD","bundle":"https://pith.science/pith/25SB56GMBDVJLSY7HLJ53X5QMD/bundle.json","state":"https://pith.science/pith/25SB56GMBDVJLSY7HLJ53X5QMD/state.json","well_known_bundle":"https://pith.science/.well-known/pith/25SB56GMBDVJLSY7HLJ53X5QMD/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:25SB56GMBDVJLSY7HLJ53X5QMD","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":"ad2d155b3ecbb8b7827f29fba2ed89c20fd522757f2060a22cc1b6c902f58d11","cross_cats_sorted":["stat.ML"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-23T19:17:55Z","title_canon_sha256":"396a3dbf9f5d39e2685e048f4168122fd0619c2c1cd9d389634c517ee6439c42"},"schema_version":"1.0","source":{"id":"2505.18313","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2505.18313","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"arxiv_version","alias_value":"2505.18313v1","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.18313","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"pith_short_12","alias_value":"25SB56GMBDVJ","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"pith_short_16","alias_value":"25SB56GMBDVJLSY7","created_at":"2026-07-05T11:08:44Z"},{"alias_kind":"pith_short_8","alias_value":"25SB56GM","created_at":"2026-07-05T11:08:44Z"}],"graph_snapshots":[{"event_id":"sha256:6c47c12734f4e7948ae619ada4ce6cc98e862e67c71703904e34613c83072dd2","target":"graph","created_at":"2026-07-05T11:08:44Z","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/2505.18313/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Accelerator memory and networking constraints have emerged as dominant bottlenecks when training large language models LLMs with billions of parameters. Existing low rank gradient estimators such as GaLoRE and FLORA compress gradients and optimizer tensors by projecting weight gradients onto a rank r subspace, enabling LLM training on consumer hardware. Yet, these methods are either biased or subject to high estimator variance. Moreover, the optimizer state based on the first and second moments estimates expressed in the previous subspace becomes misaligned whenever the projection is updated, ","authors_text":"Daniel Soudry, Matan Haroush","cross_cats":["stat.ML"],"headline":"","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-23T19:17:55Z","title":"PLUMAGE: Probabilistic Low rank Unbiased Min Variance Gradient Estimator for Efficient Large Model Training"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.18313","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:ede887a6358774943bee1670b712e0671fbc9c49d13f5f68d4cce414bd6bbfea","target":"record","created_at":"2026-07-05T11:08:44Z","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":"ad2d155b3ecbb8b7827f29fba2ed89c20fd522757f2060a22cc1b6c902f58d11","cross_cats_sorted":["stat.ML"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-23T19:17:55Z","title_canon_sha256":"396a3dbf9f5d39e2685e048f4168122fd0619c2c1cd9d389634c517ee6439c42"},"schema_version":"1.0","source":{"id":"2505.18313","kind":"arxiv","version":1}},"canonical_sha256":"d7641ef8cc08ea95cb1f3ad3dddfb060fe340ca6fa1982e25144ac0b94788a07","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"d7641ef8cc08ea95cb1f3ad3dddfb060fe340ca6fa1982e25144ac0b94788a07","first_computed_at":"2026-07-05T11:08:44.543791Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:08:44.543791Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"dkI08UrrfwS+YVZ0/Y6HW97jMOvwXc+OZjXp1SUNsr+e7F8C49H/yn1MFd7d0RQNu2BULothO5YuMP+MhZ4DAg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:08:44.544265Z","signed_message":"canonical_sha256_bytes"},"source_id":"2505.18313","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:ede887a6358774943bee1670b712e0671fbc9c49d13f5f68d4cce414bd6bbfea","sha256:6c47c12734f4e7948ae619ada4ce6cc98e862e67c71703904e34613c83072dd2"],"state_sha256":"d3d912e858fcb67d174f5f845e1749a22f0d5a800785afab363274f954869e11"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1ICSFFUGaXvB1cPVsPd+AzQpr7AgSuRCbJUdVXQ8TtA1spIMln5SOLjGVE2HjNFvKPtgMLoW83Mw3UoZAh8sBg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-14T13:47:14.063486Z","bundle_sha256":"595577cee039c12eb25bd6e7613b96f1b522da684a9a0c215e4563613212a284"}}