{"id":"W4394838778","doi":"10.1145/3618260.3649784","title":"On the Power of Interactive Proofs for Learning","year":2024,"lang":"en","type":"preprint","venue":"","topic":"Machine Learning and Algorithms","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"Qualcomm (Canada); Simon Fraser University","funders":"Engineering and Physical Sciences Research Council; Natural Sciences and Engineering Research Council of Canada; University of Warwick; UK Research and Innovation","keywords":"Mathematical proof; Computer science; Power (physics); Mathematics","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":[],"consensus_categories":[],"category_scores_codex":[0.0004708125,0.0001520473,0.0001756254,0.00008950432,0.00005886688,0.0001760585,0.0008354182,0.00007703446,0.00006411278],"category_scores_gemma":[0.0003040899,0.00008524935,0.0001721354,0.00009394484,0.00002257529,0.00002663221,0.001553257,0.00122788,0.00005159604],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002189243,"about_ca_system_score_gemma":0.00007772243,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00003038465,"about_ca_topic_score_gemma":9.938664e-7,"domain_scores_codex":[0.999003,0.00009287321,0.0001717291,0.0004112776,0.0001827396,0.0001383866],"domain_scores_gemma":[0.9985905,0.0007195924,0.0001316655,0.0004493218,0.00008560511,0.00002334291],"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":[0.00002267824,0.00008747413,0.00001950893,0.0002756491,0.0001996537,0.000005343345,0.007338754,0.01510234,0.00005651843,0.8611919,0.01043243,0.1052678],"study_design_scores_gemma":[0.00008115629,0.0002942014,0.0000405053,0.0003938511,0.00001113249,0.000002521379,0.0001122498,0.7810865,0.001181223,0.2091836,0.007422256,0.0001908111],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.008810925,0.0001281177,0.9051898,0.01093519,0.002239658,0.0007190141,0.00000581836,0.0003742585,0.07159724],"genre_scores_gemma":[0.964532,0.000002038287,0.02064819,0.0002370576,0.00008930227,0.00009949525,0.000003167314,0.00001756078,0.01437123],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.955721,"threshold_uncertainty_score":0.5334597,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01367677723551281,"score_gpt":0.2914392437634967,"score_spread":0.2777624665279839,"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."}}