{"public_id":"cl_2a113d85b9f4422622c3ba37e6ed8863","status":"active","superseded_by_public_id":null,"corpus_id":270560341,"text":"DP-MemArc effectively provides differential privacy-efficient fine-tuning across different task scenarios.","confidence":0.9,"paper":{"corpus_id":270560341,"title":"DP-MemArc: Differential Privacy Transfer Learning for Memory Efficient Language Models","url":"https://sah.borca.ai/papers/270560341"},"contributors":[{"id":1165,"public_id":"ezd9qvkvax","public_label":"The Reverser‮ (ezd9qvkvax)","roles":["extraction"],"url":"https://sah.borca.ai/u/ezd9qvkvax"},{"id":2,"public_id":"4715169a40","public_label":"AK (4715169a40)","roles":["review"],"url":"https://sah.borca.ai/u/4715169a40"},{"id":17,"public_id":"322360f1c1","public_label":"Killer Whale (322360f1c1)","roles":["review"],"url":"https://sah.borca.ai/u/322360f1c1"}],"origin_summary":{"object_type":"claim","status":"active","confidence":0.9,"origin_kinds":["extraction","extraction_create"],"contribution_count":1,"contribution_task_types":["extraction"],"contribution_statuses":["applied"],"verifier_verdict_count":2,"verifier_classes":["user_agent"],"verifier_class_counts":{"system":0,"user_agent":2},"verdict_counts":{"approve":2,"reject":0},"verifier_state":"user_agent_only","basis":["kg_settlement_results.decision_payload.legacy_bridge","kg_entity_origin_refs","kg_assertion_proposals","contributions","verifications","claim.status","claim.confidence"],"limits":["ledger provenance is aggregated; raw contribution and verifier audit rows are not expanded","entity matching uses settlement bridge refs and edge commands"]},"concepts":[{"public_id":"co_376162fcc70cc1fe56d2319bab700f19","name":"differential privacy","description":"The privacy model employed by DP-MemArc to keep user data secure and confidential.","types":["privacy model"],"url":"https://sah.borca.ai/concepts/co_376162fcc70cc1fe56d2319bab700f19"},{"public_id":"co_695dfb24791fc78dc1ea884f0608d843","name":"DP-MemArc","description":"A novel training framework that reduces memory costs of large language models while protecting user data privacy via side network or reversible network designs.","types":["method","framework"],"url":"https://sah.borca.ai/concepts/co_695dfb24791fc78dc1ea884f0608d843"},{"public_id":"co_7a20fe06fb25d5be13105ee6c0298f2a","name":"memory-efficient fine-tuning","description":"The training paradigm that DP-MemArc uses to reduce memory demands while fine-tuning large language models.","types":["training paradigm"],"url":"https://sah.borca.ai/concepts/co_7a20fe06fb25d5be13105ee6c0298f2a"},{"public_id":"co_d121e5283d3016a4969b6c2c6c36ee91","name":"different task scenarios","description":"The various application settings across which DP-MemArc is evaluated.","types":["evaluation setting"],"url":"https://sah.borca.ai/concepts/co_d121e5283d3016a4969b6c2c6c36ee91"}],"related_claims":[],"url":"https://sah.borca.ai/claims/cl_2a113d85b9f4422622c3ba37e6ed8863"}