{"id":"W7115903308","doi":"10.3929/ethz-c-000790212","title":"Gumbel Counterfactual Generation From Language Models","year":2025,"lang":"en","type":"other","venue":"Repository for Publications and Research Data (ETH Zurich)","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Institute for Catastrophic Loss Reduction","keywords":"Counterfactual conditional; Counterfactual thinking; Gumbel distribution; Hindsight bias; Causal model; Representation (politics); Language model; Causality (physics)","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","scholarly_communication"],"consensus_categories":[],"category_scores_codex":[0.001949888,0.0003646175,0.0004183845,0.001313559,0.0008066698,0.001195812,0.002116529,0.0005765958,0.0003534705],"category_scores_gemma":[0.0007917525,0.0003362494,0.0000688493,0.0008052011,0.0003669908,0.0009099573,0.001345505,0.0006871693,0.0001964443],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002200651,"about_ca_system_score_gemma":0.0009325793,"about_ca_topic_candidate":true,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.008370872,"about_ca_topic_score_gemma":0.001199852,"domain_scores_codex":[0.9956315,0.0005279891,0.0004984627,0.001646557,0.001046366,0.000649159],"domain_scores_gemma":[0.9937041,0.0005988996,0.0002745093,0.004346879,0.0007917942,0.0002838068],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"not_applicable","study_design_gemma":"not_applicable","study_design_scores_codex":[0.0000195799,0.0001820793,0.00001854805,0.00008563547,0.0002909036,0.000002219662,0.0001459691,0.00000749865,0.003374813,0.01060092,0.9839512,0.001320675],"study_design_scores_gemma":[0.0004775803,0.00003789386,0.00004982148,0.0001240187,0.0001020304,0.000005487434,0.0001832056,0.02186716,0.000341377,0.0004102537,0.9760653,0.0003358822],"study_design_candidate":"not_applicable","study_design_consensus":"not_applicable","genre_codex":"other","genre_gemma":"other","genre_scores_codex":[0.0002349257,0.0230759,0.03292315,0.001467737,0.0007197617,0.004406027,0.134836,0.001026111,0.8013104],"genre_scores_gemma":[0.002710819,0.0008008913,0.007764922,0.00004020875,0.002032584,0.0007791764,0.08054196,0.0006468576,0.9046826],"genre_candidate":"other","genre_consensus":"other","teacher_disagreement_score":0.1033721,"threshold_uncertainty_score":0.9999089,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.237836221524231,"score_gpt":0.4374304829550625,"score_spread":0.1995942614308315,"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."}}