{"id":"W4416035459","doi":"10.18653/v1/2025.emnlp-main.1567","title":"SimMark: A Robust Sentence-Level Similarity-Based Watermarking Algorithm for Large Language Models","year":2025,"lang":"","type":"article","venue":"","topic":"Advanced Steganography and Watermarking Techniques","field":"Computer Science","cited_by":3,"is_retracted":false,"has_abstract":false,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada","keywords":"Digital watermarking; Robustness (evolution); Scheme (mathematics); Language model; Key (lock); Natural language","routes":{"ca_aff":false,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001333268,0.001004271,0.001169584,0.001942935,0.0006309009,0.001034822,0.001646488,0.001276301,0.005825447],"category_scores_gemma":[0.003672348,0.0004105312,0.0008489443,0.001448974,0.0006003318,0.002444074,0.001987851,0.001458189,0.003959005],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004807498,"about_ca_system_score_gemma":0.0008919716,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001300817,"about_ca_topic_score_gemma":0.00255567,"domain_scores_codex":[0.9989794,0.0001826981,0.00007419678,0.0002135232,0.0004712109,0.00007894765],"domain_scores_gemma":[0.998777,0.0003208025,0.0001269625,0.000362514,0.0003502803,0.00006249362],"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":[0.000742505,0.0002018385,0.0004684925,0.0001952385,0.0001397673,0.0001520736,0.00008385395,0.01285751,0.08475875,0.006532231,0.008978521,0.8848892],"study_design_scores_gemma":[0.0001310672,0.0004142136,0.0008104767,0.00002661953,0.00007915642,0.0004486416,0.00006648256,0.8286873,0.1467596,0.009958453,0.01254825,0.00006968636],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.01340398,0.0005234432,0.9758949,0.0001284981,0.0001490215,0.0001324186,0.0003697983,0.00816225,0.001235701],"genre_scores_gemma":[0.1396531,0.0004173639,0.849007,0.0002514102,0.0001728766,0.0002274453,0.001957981,0.0008477831,0.007465],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.005825447,"threshold_uncertainty_score":0.0194881,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.03820838180341266,"score_gpt":0.2862023075527145,"score_spread":0.2479939257493018,"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."}}