{"id":"W4388212323","doi":"10.1109/issre59848.2023.00052","title":"Resilience Assessment of Large Language Models under Transient Hardware Faults","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":12,"is_retracted":false,"has_abstract":true,"ca_institutions":"University of British Columbia","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Computer science; Transient (computer programming); Reliability (semiconductor); Resilience (materials science); Power (physics)","routes":{"ca_aff":true,"ca_fund":true,"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":[],"consensus_categories":[],"category_scores_codex":[0.0003424587,0.00008586563,0.0001275512,0.0001000638,0.00005009716,0.00002419558,0.0006254296,0.00003788316,0.00002865282],"category_scores_gemma":[0.000006631312,0.00007318991,0.00005455704,0.0004163202,0.00001386597,0.0003232253,0.0001956593,0.00008015883,0.00002026553],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.00002812316,"about_ca_system_score_gemma":0.00007438045,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.00006261642,"about_ca_topic_score_gemma":0.00004513362,"domain_scores_codex":[0.9987476,0.00003694868,0.0002118465,0.0003293278,0.0003981693,0.0002761009],"domain_scores_gemma":[0.9992167,0.00004719618,0.00003901559,0.0005865143,0.00004919694,0.00006140379],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.000001613278,0.00008275636,0.000278315,0.00004800496,0.00001547736,0.00003015852,0.004551604,0.3325685,0.00429264,0.6503204,0.0009075617,0.006903037],"study_design_scores_gemma":[0.0002077662,0.00001897997,0.002901821,0.00001749047,0.000001902939,0.000001446717,0.0005427842,0.9920627,0.001140361,0.002904291,0.0001095169,0.0000909193],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1038601,0.00002268766,0.8856594,0.0007564475,0.0001269196,0.0001025746,0.000005139365,0.0002553837,0.009211324],"genre_scores_gemma":[0.9479448,0.000008022181,0.05019371,0.0002390194,0.00001355849,0.00000827005,0.000002761884,0.000005418513,0.001584399],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8440847,"threshold_uncertainty_score":0.2984598,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03433932826595434,"score_gpt":0.3259301749395708,"score_spread":0.2915908466736165,"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."}}