{"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":"codex-gemma-dda1882f352a","candidate_categories":["sts"],"consensus_categories":[],"category_scores_codex":[0.001152141,0.000264911,0.0002846656,0.00001665777,0.003485714,0.0003381497,0.0008288636,0.0001831855,0.0001440122],"category_scores_gemma":[0.0005327314,0.0001311123,0.0002390096,0.001019101,0.00102945,0.0008736686,0.0002849636,0.0004175894,0.000004564902],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007868368,"about_ca_system_score_gemma":0.0004478493,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.02219213,"about_ca_topic_score_gemma":0.01225244,"domain_scores_codex":[0.9970386,0.000787661,0.0005069667,0.0004485346,0.0006107448,0.0006074905],"domain_scores_gemma":[0.9983832,0.0003484271,0.0003542605,0.0006252822,0.0002296277,0.00005920538],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"theoretical_or_conceptual","study_design_gemma":"qualitative","study_design_scores_codex":[0.00003068287,0.0001529525,0.0001017392,0.00006199547,0.0001189709,0.00000124977,0.1652717,0.0005350898,0.001930497,0.8257108,0.004758816,0.00132555],"study_design_scores_gemma":[0.0004440958,0.00005384094,0.000941581,0.0003599131,0.0004174057,0.000002367167,0.807169,0.004849017,0.00504393,0.1774126,0.002913175,0.0003930525],"study_design_candidate":"theoretical_or_conceptual","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.4067219,0.007311624,0.317136,0.1080162,0.001745128,0.005725139,0.001613748,0.0003369184,0.1513933],"genre_scores_gemma":[0.9957873,0.0002680233,0.00002771177,0.0001669073,0.0002439394,0.00006758161,0.0001512121,0.000007230413,0.003280156],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.6482982,"threshold_uncertainty_score":0.9978116,"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."}}