{"id":"W4414726848","doi":"10.1145/3737902.3768352","title":"WiP: Efficient Speculative Decoding for AI PCs via Hierarchical N-Gram Retrieval","year":2025,"lang":"en","type":"article","venue":"","topic":"Algorithms and Data Compression","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Advanced Micro Devices (Canada)","funders":"","keywords":"Decoding methods; Speedup; Inference; Look-ahead; Quality (philosophy); Face (sociological concept)","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":[0.0007804493,0.001598298,0.0008515733,0.001049347,0.0008368061,0.001725635,0.002909748,0.00100862,0.005507341],"category_scores_gemma":[0.005790408,0.000564496,0.0007737004,0.001554232,0.0008371881,0.003876783,0.002374607,0.001553817,0.006096178],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0008638992,"about_ca_system_score_gemma":0.003342607,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.01219641,"about_ca_topic_score_gemma":0.02275357,"domain_scores_codex":[0.9990791,0.0001206461,0.00008841511,0.0002593846,0.0003421948,0.0001103272],"domain_scores_gemma":[0.9979483,0.0006327219,0.0001080587,0.0008158842,0.0004101706,0.00008484851],"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":[0.001414772,0.000285103,0.004023951,0.000490054,0.0001779926,0.0007193329,0.0005497903,0.0457654,0.07082906,0.01477695,0.0586198,0.8023477],"study_design_scores_gemma":[0.0001405485,0.0002055978,0.0008174908,0.00003417069,0.00005969041,0.0003077193,0.0001734336,0.8933069,0.06538359,0.01958346,0.01991286,0.00007454296],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.03605781,0.0009887473,0.8566634,0.0004412671,0.000238594,0.0002264102,0.002061299,0.09827887,0.005043644],"genre_scores_gemma":[0.2478732,0.0005033967,0.7281922,0.0006203263,0.0001637432,0.0003019617,0.008855934,0.003185873,0.01030341],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.01219641,"threshold_uncertainty_score":0.02425086,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01182716123524061,"score_gpt":0.2985599656614644,"score_spread":0.2867328044262238,"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."}}