{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:R2YT36PB3IKM5BESTWTQSKQ52Q","short_pith_number":"pith:R2YT36PB","schema_version":"1.0","canonical_sha256":"8eb13df9e1da14ce84929da7092a1dd4186debc9608cf20caedd6dda8e323d16","source":{"kind":"arxiv","id":"2408.03561","version":1},"attestation_state":"computed","paper":{"title":"MPC-Minimized Secure LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Dacheng Li, Deevashwer Rathee, Hao Zhang, Ion Stoica, Raluca Popa","submitted_at":"2024-08-07T05:50:17Z","abstract_excerpt":"Many inference services based on large language models (LLMs) pose a privacy concern, either revealing user prompts to the service or the proprietary weights to the user. Secure inference offers a solution to this problem through secure multi-party computation (MPC), however, it is still impractical for modern LLM workload due to the large overhead imposed by MPC. To address this overhead, we propose Marill, a framework that adapts LLM fine-tuning to minimize MPC usage during secure inference. Marill introduces high-level architectural changes during fine-tuning that significantly reduce the n"},"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":"2408.03561","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2024-08-07T05:50:17Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"e2159760c2b133d8fd11da12c7bc7a00378d3df072b3707ac04210ea315ef10b","abstract_canon_sha256":"4421ae345a34cd91ac37d1cfdca0a643ecfaf715824e822bd698c7afc77f4478"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:53:00.221601Z","signature_b64":"masHXi6gqm/zka389QXT+Bd1YVJ2UjQhQrWyW/uGBX64p2b/8Vubtam9jFtp+vTFM1TdSbTuIzq001S0vNotAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8eb13df9e1da14ce84929da7092a1dd4186debc9608cf20caedd6dda8e323d16","last_reissued_at":"2026-07-05T08:53:00.221257Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:53:00.221257Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MPC-Minimized Secure LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Dacheng Li, Deevashwer Rathee, Hao Zhang, Ion Stoica, Raluca Popa","submitted_at":"2024-08-07T05:50:17Z","abstract_excerpt":"Many inference services based on large language models (LLMs) pose a privacy concern, either revealing user prompts to the service or the proprietary weights to the user. Secure inference offers a solution to this problem through secure multi-party computation (MPC), however, it is still impractical for modern LLM workload due to the large overhead imposed by MPC. To address this overhead, we propose Marill, a framework that adapts LLM fine-tuning to minimize MPC usage during secure inference. Marill introduces high-level architectural changes during fine-tuning that significantly reduce the n"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2408.03561","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/2408.03561/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":"2408.03561","created_at":"2026-07-05T08:53:00.221311+00:00"},{"alias_kind":"arxiv_version","alias_value":"2408.03561v1","created_at":"2026-07-05T08:53:00.221311+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2408.03561","created_at":"2026-07-05T08:53:00.221311+00:00"},{"alias_kind":"pith_short_12","alias_value":"R2YT36PB3IKM","created_at":"2026-07-05T08:53:00.221311+00:00"},{"alias_kind":"pith_short_16","alias_value":"R2YT36PB3IKM5BES","created_at":"2026-07-05T08:53:00.221311+00:00"},{"alias_kind":"pith_short_8","alias_value":"R2YT36PB","created_at":"2026-07-05T08:53:00.221311+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.31208","citing_title":"Probing Memorization of Tabular In-Context Learning","ref_index":82,"is_internal_anchor":false},{"citing_arxiv_id":"2510.23883","citing_title":"Agentic AI Security: Threats, Defenses, Evaluation, and Open Challenges","ref_index":159,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03213","citing_title":"When Agents Handle Secrets: A Survey of Confidential Computing for Agentic AI","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03213","citing_title":"When Agents Handle Secrets: A Survey of Confidential Computing for Agentic AI","ref_index":21,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/R2YT36PB3IKM5BESTWTQSKQ52Q","json":"https://pith.science/pith/R2YT36PB3IKM5BESTWTQSKQ52Q.json","graph_json":"https://pith.science/api/pith-number/R2YT36PB3IKM5BESTWTQSKQ52Q/graph.json","events_json":"https://pith.science/api/pith-number/R2YT36PB3IKM5BESTWTQSKQ52Q/events.json","paper":"https://pith.science/paper/R2YT36PB"},"agent_actions":{"view_html":"https://pith.science/pith/R2YT36PB3IKM5BESTWTQSKQ52Q","download_json":"https://pith.science/pith/R2YT36PB3IKM5BESTWTQSKQ52Q.json","view_paper":"https://pith.science/paper/R2YT36PB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2408.03561&json=true","fetch_graph":"https://pith.science/api/pith-number/R2YT36PB3IKM5BESTWTQSKQ52Q/graph.json","fetch_events":"https://pith.science/api/pith-number/R2YT36PB3IKM5BESTWTQSKQ52Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/R2YT36PB3IKM5BESTWTQSKQ52Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/R2YT36PB3IKM5BESTWTQSKQ52Q/action/storage_attestation","attest_author":"https://pith.science/pith/R2YT36PB3IKM5BESTWTQSKQ52Q/action/author_attestation","sign_citation":"https://pith.science/pith/R2YT36PB3IKM5BESTWTQSKQ52Q/action/citation_signature","submit_replication":"https://pith.science/pith/R2YT36PB3IKM5BESTWTQSKQ52Q/action/replication_record"}},"created_at":"2026-07-05T08:53:00.221311+00:00","updated_at":"2026-07-05T08:53:00.221311+00:00"}