{"id":"W4386978004","doi":"10.48550/arxiv.2309.12314","title":"TinyCLIP: CLIP Distillation via Affinity Mimicking and Weight Inheritance","year":2023,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Multimodal Machine Learning Applications","field":"Computer Science","cited_by":4,"is_retracted":false,"has_abstract":true,"ca_institutions":"St. Stephen's University","funders":"National Natural Science Foundation of China","keywords":"Distillation; Computer science; Inheritance (genetic algorithm); Modal; Artificial intelligence; Feature (linguistics); AKA; Machine learning; Transferability; Code (set theory); Scratch; Image (mathematics); Pattern recognition (psychology); Programming language; 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":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.000277363,0.0002731593,0.0002628705,0.0002034364,0.0002785209,0.0001460306,0.001120038,0.0002395576,0.000008910362],"category_scores_gemma":[0.00008487624,0.0003356242,0.00009592927,0.0006153628,0.0001040756,0.0002578131,0.002069928,0.0007028929,0.0001624737],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001393836,"about_ca_system_score_gemma":0.00006827656,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.000744322,"about_ca_topic_score_gemma":0.0001323896,"domain_scores_codex":[0.9980552,0.0001504215,0.0002078164,0.001204339,0.00009510739,0.0002871345],"domain_scores_gemma":[0.998062,0.0002546744,0.0003018405,0.001130398,0.00010616,0.0001449227],"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":[0.00003533776,0.0001689992,0.3187583,0.000316353,0.0001671346,0.000179482,0.001300282,0.3834242,0.0003626854,0.2788071,0.0003914956,0.01608851],"study_design_scores_gemma":[0.0001803919,0.00001180118,0.1634907,0.00005663206,0.00002425854,0.000002724416,0.000009196989,0.7881238,0.00001783598,0.04743357,0.0003361102,0.0003128789],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4067293,0.00002573766,0.591271,0.0003669369,0.0002842561,0.0002137357,0.000008966738,0.0005420885,0.0005580483],"genre_scores_gemma":[0.9908226,0.0001060176,0.008361423,0.00004890626,0.00009818892,0.000002932722,0.00002561409,0.0000268924,0.0005074625],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.5840933,"threshold_uncertainty_score":0.9999096,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08317322424467298,"score_gpt":0.2121273096551278,"score_spread":0.1289540854104548,"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."}}