{"id":"W4405907170","doi":"10.1109/tbdata.2024.3524105","title":"Data Reconstruction and Protection in Federated Learning for Fine-Tuning Large Language Models","year":2024,"lang":"en","type":"article","venue":"IEEE Transactions on Big Data","topic":"Privacy-Preserving Technologies in Data","field":"Computer Science","cited_by":8,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"","keywords":"Computer science","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["open_science"],"consensus_categories":[],"category_scores_codex":[0.0008273374,0.0001522764,0.000142085,0.0003846398,0.0002840781,0.0004612429,0.0071091,0.0001389039,0.000004423011],"category_scores_gemma":[0.0006915885,0.0001552873,0.00001636385,0.0006817625,0.00004139216,0.002972092,0.001521301,0.0006026044,0.000007715513],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.000062603,"about_ca_system_score_gemma":0.00007716245,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001492319,"about_ca_topic_score_gemma":0.001024292,"domain_scores_codex":[0.9982369,0.00007811147,0.0002403467,0.001010403,0.0001536466,0.0002805927],"domain_scores_gemma":[0.9943577,0.0001640009,0.00004490859,0.005371486,0.00002532367,0.00003659121],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.00001874007,0.00004366702,0.000002933124,0.00008996377,0.0000309801,0.00001282257,0.00009605926,0.001086456,0.002357146,0.00008727182,0.00248789,0.9936861],"study_design_scores_gemma":[0.0002991791,0.00004937504,0.000005657585,0.0002021324,0.00001182494,0.00006012686,0.00008695509,0.9908066,0.00205463,0.004101853,0.002155692,0.0001659424],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.003811039,0.0002594869,0.9890424,0.003079685,0.00101176,0.0004212527,0.001356901,0.0009837119,0.00003374496],"genre_scores_gemma":[0.9003171,0.0001881552,0.09880587,0.00002669356,0.0000608746,0.00008115815,0.0004289974,0.0000253457,0.00006577233],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9935201,"threshold_uncertainty_score":0.9982629,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1394083824470668,"score_gpt":0.3164687905736269,"score_spread":0.1770604081265601,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}