{"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":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow"],"consensus_categories":[],"category_scores_codex":[0.001580307,0.0008817364,0.000866979,0.0009395281,0.001238646,0.0008143023,0.002454299,0.0005137273,0.00003226486],"category_scores_gemma":[0.0000550897,0.0008253857,0.000802573,0.001538203,0.0001885661,0.001599131,0.001097708,0.00064439,0.000003876931],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001652001,"about_ca_system_score_gemma":0.0003906823,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0000858927,"about_ca_topic_score_gemma":0.00004956518,"domain_scores_codex":[0.994409,0.0002740709,0.00109722,0.001742766,0.0005788251,0.001898105],"domain_scores_gemma":[0.9966915,0.0004795111,0.0003132953,0.001745383,0.0005210986,0.0002492396],"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":[0.000240344,0.001447554,0.0003898117,0.001075051,0.0004439112,0.0002011875,0.002758053,0.01008544,0.000830614,0.09705073,0.006520039,0.8789573],"study_design_scores_gemma":[0.001964241,0.000161856,0.00004184073,0.000735087,0.00009729175,0.000006132489,0.0002167643,0.9339399,0.02286882,0.03207467,0.006981281,0.0009120908],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.0001600043,0.0006702293,0.9885968,0.001707429,0.001229699,0.001736373,0.0004778532,0.001074405,0.004347191],"genre_scores_gemma":[0.1961103,0.00009935991,0.7957004,0.004373609,0.0001355096,0.0002848606,0.0000926412,0.00005022756,0.003153117],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.9238545,"threshold_uncertainty_score":0.9994197,"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."}}