{"id":"W4225999490","doi":"10.1609/aaai.v36i9.21184","title":"Cosine Model Watermarking against Ensemble Distillation","year":2022,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","topic":"Internet Traffic Analysis and Secure E-voting","field":"Computer Science","cited_by":23,"is_retracted":false,"has_abstract":true,"ca_institutions":"McMaster University; Simon Fraser University; Huawei Technologies (Canada)","funders":"","keywords":"Digital watermarking; Distillation; Computer science; Ensemble forecasting; Focus (optics); Task (project management); Image (mathematics); Artificial intelligence; Algorithm; Machine learning; Data mining; Engineering; Chemistry; Chromatography","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":[],"consensus_categories":[],"category_scores_codex":[0.000624919,0.0001883066,0.0002339633,0.0001323224,0.000494834,0.0002290216,0.001841313,0.00003944599,0.00005082887],"category_scores_gemma":[0.00009655303,0.0001477074,0.0001589046,0.0005558004,0.00009398401,0.0002846263,0.0007809202,0.0003482408,0.00001956493],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0000895265,"about_ca_system_score_gemma":0.00006256789,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000008730757,"about_ca_topic_score_gemma":0.000005458436,"domain_scores_codex":[0.9979986,0.00002519792,0.0005396086,0.0004615762,0.0006553856,0.0003196208],"domain_scores_gemma":[0.9989538,0.00004610593,0.0003974751,0.0001863277,0.0003591088,0.00005718628],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0000234423,0.00007914539,0.00002914434,0.00001148626,0.00001349963,5.223735e-7,0.001835566,0.07115491,0.01300028,0.8864733,0.000146433,0.02723229],"study_design_scores_gemma":[0.00001703227,0.00006711512,0.000008413883,0.00005384564,0.000009013591,0.000003042018,0.0003594621,0.938578,0.05068526,0.00995048,0.0001020733,0.0001662641],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.429529,0.00002177742,0.5474309,0.003501721,0.0005592623,0.0003955271,0.000008077957,0.0001701026,0.01838367],"genre_scores_gemma":[0.9974805,0.000006534864,0.00177138,0.0003058589,0.0000479652,0.00002359738,0.000001738784,0.00001141761,0.000351005],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8765228,"threshold_uncertainty_score":0.6023334,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05706464998515177,"score_gpt":0.2672449593726166,"score_spread":0.2101803093874648,"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."}}