{"id":"W7106797679","doi":"10.48448/avtc-fj63","title":"When Punctuation Matters: A Large-Scale Comparison of Prompt Robustness Methods for LLMs","year":2025,"lang":"","type":"other","venue":"Open MIND","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Air Canada","funders":"","keywords":"Robustness (evolution); Punctuation; Robustness testing; Natural language; Generalization","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":[0.01412611,0.003008571,0.001135928,0.002467998,0.001080704,0.003317672,0.003061297,0.002352574,0.004326639],"category_scores_gemma":[0.07361338,0.0006425566,0.001809433,0.0013801,0.001618985,0.006512949,0.004203182,0.004865044,0.004204559],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001739457,"about_ca_system_score_gemma":0.002317092,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.005710059,"about_ca_topic_score_gemma":0.007721953,"domain_scores_codex":[0.9919388,0.004174268,0.0005750947,0.001878604,0.001134162,0.0002991669],"domain_scores_gemma":[0.9657963,0.02281733,0.001240696,0.007143454,0.002291026,0.000711249],"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":[0.00351274,0.0007660333,0.03004112,0.002409209,0.001532263,0.0003975802,0.001420534,0.1837495,0.01256424,0.008294661,0.07618862,0.6791235],"study_design_scores_gemma":[0.0004667164,0.001414698,0.0136796,0.0005704505,0.0005586972,0.0004763157,0.001166848,0.8907899,0.02602738,0.03727818,0.02728676,0.0002843549],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.4190771,0.02704628,0.4261703,0.007130015,0.002456984,0.001029452,0.01041269,0.08656371,0.02011345],"genre_scores_gemma":[0.8620399,0.001862405,0.1078836,0.001812785,0.0003946045,0.0005422913,0.01530516,0.005249039,0.004910223],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01412611,"threshold_uncertainty_score":0.07470691,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.08829493364405619,"score_gpt":0.4412866297029025,"score_spread":0.3529916960588463,"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."}}