{"id":"W7104481854","doi":"10.71781/10885","title":"Towards efficient large language models : training low-bitwidth variants and low-rank decomposition of pretrained models","year":2024,"lang":"en","type":"dissertation","venue":"Open MIND","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Oak Ridge National Laboratory; Office of Science; Canadian Institute for Advanced Research; Canada Excellence Research Chairs, Government of Canada; U.S. Department of Energy","keywords":"Training set; Statistical analysis; Context (archaeology)","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001355269,0.002151285,0.001452073,0.0008132062,0.0003949941,0.001826325,0.002000582,0.001745922,0.003962906],"category_scores_gemma":[0.006398468,0.0009402068,0.001686281,0.0008652358,0.0006304483,0.002867498,0.001307747,0.003946149,0.002931063],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008591053,"about_ca_system_score_gemma":0.001385785,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01167953,"about_ca_topic_score_gemma":0.01866518,"domain_scores_codex":[0.9992464,0.0002390599,0.00004732712,0.000218063,0.0001365275,0.0001127617],"domain_scores_gemma":[0.9978876,0.001184952,0.0001271894,0.0003745818,0.0003217118,0.0001041056],"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":[0.0004456111,0.0002500933,0.00132514,0.0002269495,0.0001988677,0.0001812567,0.0001935077,0.6283007,0.01509701,0.005909624,0.007107676,0.3407635],"study_design_scores_gemma":[0.000009064453,0.00003547896,0.00008953601,0.0000091265,0.00001066082,0.00001503652,0.00001596421,0.9962998,0.001235083,0.001874352,0.0004001343,0.000005711876],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.0558015,0.00138979,0.9338637,0.0005325566,0.0001671728,0.00007827466,0.0004768292,0.005686584,0.002003651],"genre_scores_gemma":[0.4956739,0.001070469,0.485441,0.0007717178,0.0002188674,0.0003686253,0.004145579,0.001199295,0.01111061],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01167953,"threshold_uncertainty_score":0.0232231,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03243441090778062,"score_gpt":0.3186131654673854,"score_spread":0.2861787545596048,"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."}}