{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:FI2QGB4SC2WGXE6PSWLUC7SD2K","short_pith_number":"pith:FI2QGB4S","canonical_record":{"source":{"id":"2502.20566","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-27T22:08:08Z","cross_cats_sorted":[],"title_canon_sha256":"73d70a2097a4081a56c461d26c7f3eefde94ba1270ecdf24a7949e2aeb23cea2","abstract_canon_sha256":"3b341e4d813fdc07785035adeec8ccabf9d81fd7606977bb4589de9c1aa11e56"},"schema_version":"1.0"},"canonical_sha256":"2a3503079216ac6b93cf9597417e43d2b533cf340b4a93a0dd81c353e22e4154","source":{"kind":"arxiv","id":"2502.20566","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.20566","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"arxiv_version","alias_value":"2502.20566v1","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.20566","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"pith_short_12","alias_value":"FI2QGB4SC2WG","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"pith_short_16","alias_value":"FI2QGB4SC2WGXE6P","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"pith_short_8","alias_value":"FI2QGB4S","created_at":"2026-07-05T10:21:35Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:FI2QGB4SC2WGXE6PSWLUC7SD2K","target":"record","payload":{"canonical_record":{"source":{"id":"2502.20566","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-27T22:08:08Z","cross_cats_sorted":[],"title_canon_sha256":"73d70a2097a4081a56c461d26c7f3eefde94ba1270ecdf24a7949e2aeb23cea2","abstract_canon_sha256":"3b341e4d813fdc07785035adeec8ccabf9d81fd7606977bb4589de9c1aa11e56"},"schema_version":"1.0"},"canonical_sha256":"2a3503079216ac6b93cf9597417e43d2b533cf340b4a93a0dd81c353e22e4154","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:21:35.423795Z","signature_b64":"DtT1Xxc8aMkrc0q+knUaFgEBeBGETcFJn+CK8SqSpkkXKRLLxe73XGe3rndzHaCV4jOZE0eQipLMdPKOvmeqBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2a3503079216ac6b93cf9597417e43d2b533cf340b4a93a0dd81c353e22e4154","last_reissued_at":"2026-07-05T10:21:35.422908Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:21:35.422908Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2502.20566","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":"2026-07-05T10:21:35Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"rt8OzxzYsbxpEL+pf9J+Ctm53N7EhrCZLB36YSx46S4BNjL+JYOWTiK64N3XeixmSP0Zdq0qkqsuA/f6qKBZBA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T01:07:44.999069Z"},"content_sha256":"f8b0f565fa3f79a346cbbd53713732b0b14f59a04b643685c37909cc96bc6a5f","schema_version":"1.0","event_id":"sha256:f8b0f565fa3f79a346cbbd53713732b0b14f59a04b643685c37909cc96bc6a5f"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:FI2QGB4SC2WGXE6PSWLUC7SD2K","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Stochastic Rounding for LLM Training: Theory and Practice","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Kaan Ozkara, Tao Yu, Youngsuk Park","submitted_at":"2025-02-27T22:08:08Z","abstract_excerpt":"As the parameters of Large Language Models (LLMs) have scaled to hundreds of billions, the demand for efficient training methods -- balancing faster computation and reduced memory usage without sacrificing accuracy -- has become more critical than ever. In recent years, various mixed precision strategies, which involve different precision levels for optimization components, have been proposed to increase training speed with minimal accuracy degradation. However, these strategies often require manual adjustments and lack theoretical justification. In this work, we leverage stochastic rounding ("},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.20566","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/2502.20566/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":"2026-07-05T10:21:35Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"1wtQeo7YR7n9uh3KRC9pcieQlkyTPDR1VsEGPzwDxBR+qtljiIieSW0gZ5EAIN6OC0ZRJE4DBjzuOh7uNMu0Bg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-04T01:07:45.000169Z"},"content_sha256":"007efafe1aedcef76b86a630f53a90f2c27d91d277936238174d9a0c5b9f4b09","schema_version":"1.0","event_id":"sha256:007efafe1aedcef76b86a630f53a90f2c27d91d277936238174d9a0c5b9f4b09"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/FI2QGB4SC2WGXE6PSWLUC7SD2K/bundle.json","state_url":"https://pith.science/pith/FI2QGB4SC2WGXE6PSWLUC7SD2K/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/FI2QGB4SC2WGXE6PSWLUC7SD2K/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":"2026-08-04T01:07:45Z","links":{"resolver":"https://pith.science/pith/FI2QGB4SC2WGXE6PSWLUC7SD2K","bundle":"https://pith.science/pith/FI2QGB4SC2WGXE6PSWLUC7SD2K/bundle.json","state":"https://pith.science/pith/FI2QGB4SC2WGXE6PSWLUC7SD2K/state.json","well_known_bundle":"https://pith.science/.well-known/pith/FI2QGB4SC2WGXE6PSWLUC7SD2K/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:FI2QGB4SC2WGXE6PSWLUC7SD2K","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":"3b341e4d813fdc07785035adeec8ccabf9d81fd7606977bb4589de9c1aa11e56","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-27T22:08:08Z","title_canon_sha256":"73d70a2097a4081a56c461d26c7f3eefde94ba1270ecdf24a7949e2aeb23cea2"},"schema_version":"1.0","source":{"id":"2502.20566","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.20566","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"arxiv_version","alias_value":"2502.20566v1","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.20566","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"pith_short_12","alias_value":"FI2QGB4SC2WG","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"pith_short_16","alias_value":"FI2QGB4SC2WGXE6P","created_at":"2026-07-05T10:21:35Z"},{"alias_kind":"pith_short_8","alias_value":"FI2QGB4S","created_at":"2026-07-05T10:21:35Z"}],"graph_snapshots":[{"event_id":"sha256:007efafe1aedcef76b86a630f53a90f2c27d91d277936238174d9a0c5b9f4b09","target":"graph","created_at":"2026-07-05T10:21:35Z","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/2502.20566/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"As the parameters of Large Language Models (LLMs) have scaled to hundreds of billions, the demand for efficient training methods -- balancing faster computation and reduced memory usage without sacrificing accuracy -- has become more critical than ever. In recent years, various mixed precision strategies, which involve different precision levels for optimization components, have been proposed to increase training speed with minimal accuracy degradation. However, these strategies often require manual adjustments and lack theoretical justification. In this work, we leverage stochastic rounding (","authors_text":"Kaan Ozkara, Tao Yu, Youngsuk Park","cross_cats":[],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-27T22:08:08Z","title":"Stochastic Rounding for LLM Training: Theory and Practice"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.20566","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:f8b0f565fa3f79a346cbbd53713732b0b14f59a04b643685c37909cc96bc6a5f","target":"record","created_at":"2026-07-05T10:21:35Z","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":"3b341e4d813fdc07785035adeec8ccabf9d81fd7606977bb4589de9c1aa11e56","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-02-27T22:08:08Z","title_canon_sha256":"73d70a2097a4081a56c461d26c7f3eefde94ba1270ecdf24a7949e2aeb23cea2"},"schema_version":"1.0","source":{"id":"2502.20566","kind":"arxiv","version":1}},"canonical_sha256":"2a3503079216ac6b93cf9597417e43d2b533cf340b4a93a0dd81c353e22e4154","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"2a3503079216ac6b93cf9597417e43d2b533cf340b4a93a0dd81c353e22e4154","first_computed_at":"2026-07-05T10:21:35.422908Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:21:35.422908Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"DtT1Xxc8aMkrc0q+knUaFgEBeBGETcFJn+CK8SqSpkkXKRLLxe73XGe3rndzHaCV4jOZE0eQipLMdPKOvmeqBw==","signature_status":"signed_v1","signed_at":"2026-07-05T10:21:35.423795Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.20566","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:f8b0f565fa3f79a346cbbd53713732b0b14f59a04b643685c37909cc96bc6a5f","sha256:007efafe1aedcef76b86a630f53a90f2c27d91d277936238174d9a0c5b9f4b09"],"state_sha256":"492ef2304ef8a6664a15aa7e6ecf5388d02bf0c4e6b1adc856be9bdd44a7797e"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"oNzebCVQMDHw8QJxVJzRMP23xvzJY4AeGoWvfAl6j3mXIGtsxI54Hmg7QthNNxGdpSQti2JQMCrj7vW+iDLABg==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-04T01:07:45.006300Z","bundle_sha256":"ad3f711effe1f9e88b7dfda86b2275a08ee9957e307b5410fa8bd275c059e377"}}