{"id":"W7109952297","doi":"10.4230/lipics.sat.2025.32","title":"Efficient Certified Reasoning for Binarized Neural Networks","year":2025,"lang":"en","type":"article","venue":"DROPS (Schloss Dagstuhl – Leibniz Center for Informatics)","topic":"Adversarial Robustness in Machine Learning","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of Toronto","funders":"Natural Sciences and Engineering Research Council of Canada; Alliance de recherche numérique du Canada; Vetenskapsrådet; National Supercomputing Centre Singapore; University of Toronto; Innovation, Science and Economic Development Canada","keywords":"Speedup; Soundness; Scalability; Robustness (evolution); Benchmark (surveying); Artificial neural network; Formal verification; Pipeline (software)","routes":{"ca_aff":true,"ca_fund":true,"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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.00114142,0.0005099279,0.0006295111,0.0004269354,0.0009284656,0.0007235872,0.001994599,0.000310121,0.00000900178],"category_scores_gemma":[0.0006214112,0.0004772112,0.0004185789,0.0008143734,0.0001229627,0.0007365599,0.0009864218,0.0006755429,0.00001559468],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002221071,"about_ca_system_score_gemma":0.0001215367,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00001666849,"about_ca_topic_score_gemma":0.000005168952,"domain_scores_codex":[0.9965488,0.00008268856,0.001261501,0.0005095837,0.0004303195,0.001167124],"domain_scores_gemma":[0.9971133,0.0006497031,0.0005522046,0.001111266,0.0003838621,0.0001896022],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0004282192,0.0002448413,0.002194815,0.0005141114,0.0002399576,0.000005291958,0.003714351,0.8361199,0.00003271707,0.1184371,0.004034066,0.03403458],"study_design_scores_gemma":[0.003879363,0.0001382988,0.0003229236,0.0001722642,0.00005112297,0.00001310049,0.0002957073,0.9802649,0.00008235986,0.0006670188,0.01364237,0.0004705685],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02208037,0.00006671309,0.96881,0.0009026215,0.003648118,0.001790865,0.00004293884,0.0005179594,0.002140455],"genre_scores_gemma":[0.8356664,0.000004981406,0.1612586,0.001619387,0.0002589482,0.0003509451,0.0001836644,0.00005171905,0.0006054282],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.813586,"threshold_uncertainty_score":0.999768,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01107840787201233,"score_gpt":0.2723201104293548,"score_spread":0.2612417025573425,"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."}}