{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:7K7YW5EXCPU7B4TXCZNUVRL7YY","short_pith_number":"pith:7K7YW5EX","schema_version":"1.0","canonical_sha256":"fabf8b749713e9f0f277165b4ac57fc603e1e020e491545a492d0a231ccbe220","source":{"kind":"arxiv","id":"2605.28095","version":1},"attestation_state":"computed","paper":{"title":"SiDP: Memory-Efficient Data Parallelism for Offline LLM Inference","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Alan Zhao, Cyril Y. He","submitted_at":"2026-05-27T07:52:03Z","abstract_excerpt":"The rapid adoption of large language models (LLMs) has shifted a substantial portion of inference workloads into throughput-oriented offline regimes, where fully utilizing GPU compute requires large batch sizes. However, existing deployments face a structural tension. Data parallelism (DP) scales throughput well but replicates model weights, leaving limited GPU memory for key-value (KV) cache and constraining batch size. Model parallelism reduces per-device weights, but requires fine-grained synchronization that erodes DP's independence and scheduling flexibility. We present SiDP, a memory-eff"},"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":"2605.28095","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.DC","submitted_at":"2026-05-27T07:52:03Z","cross_cats_sorted":[],"title_canon_sha256":"dd4de984695dad7c90511058c2f21122b68d3efa3068da79d0beb722f30ecff1","abstract_canon_sha256":"8c77cb9eb4a238b08909038e7548f51482c8fcfc141542581a74bdac6fb18726"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T01:04:58.611901Z","signature_b64":"9bJotEunEMnwUN3IbUGKh4cqFKSR0kMnJfKxHukQT5W45u2SaQHPUNkJDDBKSPBpBLYahQAqkgIVMyV12wasDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fabf8b749713e9f0f277165b4ac57fc603e1e020e491545a492d0a231ccbe220","last_reissued_at":"2026-05-28T01:04:58.611379Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T01:04:58.611379Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"SiDP: Memory-Efficient Data Parallelism for Offline LLM Inference","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.DC","authors_text":"Alan Zhao, Cyril Y. He","submitted_at":"2026-05-27T07:52:03Z","abstract_excerpt":"The rapid adoption of large language models (LLMs) has shifted a substantial portion of inference workloads into throughput-oriented offline regimes, where fully utilizing GPU compute requires large batch sizes. However, existing deployments face a structural tension. Data parallelism (DP) scales throughput well but replicates model weights, leaving limited GPU memory for key-value (KV) cache and constraining batch size. Model parallelism reduces per-device weights, but requires fine-grained synchronization that erodes DP's independence and scheduling flexibility. We present SiDP, a memory-eff"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.28095","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/2605.28095/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":"2605.28095","created_at":"2026-05-28T01:04:58.611447+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.28095v1","created_at":"2026-05-28T01:04:58.611447+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.28095","created_at":"2026-05-28T01:04:58.611447+00:00"},{"alias_kind":"pith_short_12","alias_value":"7K7YW5EXCPU7","created_at":"2026-05-28T01:04:58.611447+00:00"},{"alias_kind":"pith_short_16","alias_value":"7K7YW5EXCPU7B4TX","created_at":"2026-05-28T01:04:58.611447+00:00"},{"alias_kind":"pith_short_8","alias_value":"7K7YW5EX","created_at":"2026-05-28T01:04:58.611447+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7K7YW5EXCPU7B4TXCZNUVRL7YY","json":"https://pith.science/pith/7K7YW5EXCPU7B4TXCZNUVRL7YY.json","graph_json":"https://pith.science/api/pith-number/7K7YW5EXCPU7B4TXCZNUVRL7YY/graph.json","events_json":"https://pith.science/api/pith-number/7K7YW5EXCPU7B4TXCZNUVRL7YY/events.json","paper":"https://pith.science/paper/7K7YW5EX"},"agent_actions":{"view_html":"https://pith.science/pith/7K7YW5EXCPU7B4TXCZNUVRL7YY","download_json":"https://pith.science/pith/7K7YW5EXCPU7B4TXCZNUVRL7YY.json","view_paper":"https://pith.science/paper/7K7YW5EX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.28095&json=true","fetch_graph":"https://pith.science/api/pith-number/7K7YW5EXCPU7B4TXCZNUVRL7YY/graph.json","fetch_events":"https://pith.science/api/pith-number/7K7YW5EXCPU7B4TXCZNUVRL7YY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7K7YW5EXCPU7B4TXCZNUVRL7YY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7K7YW5EXCPU7B4TXCZNUVRL7YY/action/storage_attestation","attest_author":"https://pith.science/pith/7K7YW5EXCPU7B4TXCZNUVRL7YY/action/author_attestation","sign_citation":"https://pith.science/pith/7K7YW5EXCPU7B4TXCZNUVRL7YY/action/citation_signature","submit_replication":"https://pith.science/pith/7K7YW5EXCPU7B4TXCZNUVRL7YY/action/replication_record"}},"created_at":"2026-05-28T01:04:58.611447+00:00","updated_at":"2026-05-28T01:04:58.611447+00:00"}