{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.000727296,0.0006100782,0.0009055335,0.0009210352,0.0004722018,0.0003002194,0.001771422,0.0004487216,0.00002996242],"category_scores_gemma":[0.0001608191,0.0006723875,0.0002833072,0.0006016786,0.0007323641,0.0003972212,0.0004900789,0.0007436178,0.00000877481],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001266297,"about_ca_system_score_gemma":0.0004273249,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001117467,"about_ca_topic_score_gemma":0.00009391519,"domain_scores_codex":[0.9960785,0.00004218586,0.001527203,0.001364863,0.0003601373,0.0006271487],"domain_scores_gemma":[0.9970208,0.0003736844,0.0007300645,0.001040492,0.0006700943,0.0001648567],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"theoretical_or_conceptual","study_design_scores_codex":[0.00002912021,0.00004576857,0.000009896251,0.00008988396,0.00003024459,0.000001631415,0.0007099936,0.00003175568,0.0002426819,0.5303595,0.0001650119,0.4682845],"study_design_scores_gemma":[0.00007992154,0.0001904889,0.000001509525,0.0007677564,0.00009233239,0.000005659528,0.0009479673,0.03895332,0.01210451,0.9301513,0.0159904,0.0007147865],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":"theoretical_or_conceptual","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.00001947395,0.009551677,0.9784984,0.0009274019,0.0007634329,0.002450646,0.0002474153,0.0001617314,0.007379797],"genre_scores_gemma":[0.0469334,0.0006064546,0.9478708,0.0004329988,0.000260682,0.0007385283,0.00008322629,0.00005353989,0.003020378],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.4675697,"threshold_uncertainty_score":0.9995728,"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."}}