{"id":"W4415428196","doi":"10.3233/faia251188","title":"PoT-PTQ:Two-Step Power-of-Two Post-Training for LLMs","year":2025,"lang":"","type":"book-chapter","venue":"Frontiers in artificial intelligence and applications","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Mila - Quebec Artificial Intelligence Institute; Université de Montréal; McGill University","funders":"","keywords":"Quantization (signal processing); Inference; Software deployment; Integer (computer science); Floating point; Natural language; Language model","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.00170056,0.001731297,0.001215486,0.0007818003,0.0008523129,0.00166517,0.003889421,0.001990644,0.01850654],"category_scores_gemma":[0.01289704,0.0008959037,0.0009962787,0.0008758778,0.001118632,0.004214957,0.00359949,0.004242193,0.005803547],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001084938,"about_ca_system_score_gemma":0.002236665,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005238566,"about_ca_topic_score_gemma":0.01076232,"domain_scores_codex":[0.9988266,0.0002229139,0.000102006,0.0002876193,0.0004228692,0.0001380652],"domain_scores_gemma":[0.9976161,0.001084607,0.0001283944,0.0005765748,0.0004832082,0.000111182],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0006323069,0.0002569587,0.001412959,0.0003703599,0.0001230749,0.0003390606,0.0003080539,0.1138948,0.02148302,0.01835391,0.02781408,0.8150115],"study_design_scores_gemma":[0.00007364334,0.0001107021,0.0002786582,0.00002852842,0.00001985161,0.00009769751,0.00005272733,0.9664951,0.01218554,0.01541373,0.005218764,0.00002503016],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.009033522,0.0003364411,0.9722151,0.0003706212,0.0002109995,0.0001641684,0.0002697217,0.01575452,0.001644935],"genre_scores_gemma":[0.216961,0.0003091239,0.7673351,0.001058248,0.0001704527,0.0006099688,0.00214803,0.002592591,0.008815517],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01850654,"threshold_uncertainty_score":0.06191051,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04209750365547548,"score_gpt":0.3277798479616582,"score_spread":0.2856823443061827,"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."}}