{"id":"W4416036024","doi":"10.18653/v1/2025.emnlp-main.1057","title":"Distribution Prompting: Understanding the Expressivity of Language Models Through the Next-Token Distributions They Can Produce","year":2025,"lang":"","type":"article","venue":"","topic":"Language and cultural evolution","field":"Social Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Natural Sciences and Engineering Research Council of Canada; Canadian Institute for Advanced Research","keywords":"Expressivity; Distribution (mathematics); Language model; Statistical model; Gamma distribution; Probability distribution","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":[0.004639626,0.0008687158,0.0004905709,0.0005222734,0.0004106853,0.001888306,0.00123892,0.001156798,0.003630151],"category_scores_gemma":[0.0352692,0.0006718629,0.000780947,0.0004078252,0.001995354,0.006330892,0.001426592,0.002609587,0.0007813609],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001118071,"about_ca_system_score_gemma":0.0007923372,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.001882168,"about_ca_topic_score_gemma":0.00171611,"domain_scores_codex":[0.9981707,0.00108509,0.00006721715,0.0004104649,0.0001854959,0.00008091236],"domain_scores_gemma":[0.9836992,0.01328068,0.0008056909,0.001439765,0.0005462595,0.000228447],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0004053499,0.0001302236,0.01118635,0.0004748823,0.0001344235,0.0004180929,0.002641534,0.534446,0.02105534,0.2923447,0.002724206,0.1340389],"study_design_scores_gemma":[0.00003510385,0.00008875055,0.00118532,0.00003828488,0.00001892646,0.0001189536,0.0001611439,0.7625108,0.00460026,0.2293126,0.001895588,0.00003428727],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.0359567,0.0001263059,0.9601502,0.0006410423,0.0000235545,0.00003913596,0.0002245205,0.0007924548,0.002045937],"genre_scores_gemma":[0.7918704,0.0002554804,0.2045892,0.0003047117,0.00004110943,0.000229476,0.0004008335,0.0003395346,0.001969396],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.004639626,"threshold_uncertainty_score":0.02453697,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.06713235982875414,"score_gpt":0.3194304568513414,"score_spread":0.2522980970225872,"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."}}