{"id":"W4414973481","doi":"10.48550/arxiv.2510.05132","title":"Training Large Language Models To Reason In Parallel With Global Forking Tokens","year":2025,"lang":"en","type":"preprint","venue":"ArXiv.org","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"University of Toronto; Amazon Web Services","keywords":"Set (abstract data type); Matching (statistics); Bipartite graph; Automated reasoning; Diversity (politics); Code (set theory); Training (meteorology)","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":[0.001433042,0.001433263,0.0008839984,0.0006332992,0.0004966728,0.001257132,0.002697029,0.001238421,0.003740369],"category_scores_gemma":[0.006542333,0.0007528777,0.001552698,0.0006442978,0.001089066,0.003234586,0.001516232,0.003764042,0.001816599],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001400341,"about_ca_system_score_gemma":0.002305717,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006992288,"about_ca_topic_score_gemma":0.02540784,"domain_scores_codex":[0.9990985,0.0002211073,0.00004677011,0.0003598856,0.0001638208,0.0001099173],"domain_scores_gemma":[0.9970145,0.001592598,0.0001882973,0.0008101592,0.0002589108,0.0001356044],"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":[0.0003208063,0.0003757608,0.004379383,0.0002426427,0.000156664,0.0001820903,0.0001786038,0.7690142,0.009522674,0.008171098,0.01168037,0.1957757],"study_design_scores_gemma":[0.00001419152,0.00001810092,0.00008612061,0.000003579752,0.000007533809,0.00001130331,0.00001111842,0.9920065,0.001203961,0.006194014,0.0004396662,0.000004013091],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.1308724,0.0006509973,0.8286871,0.001221432,0.0002220642,0.0002296902,0.0008767649,0.03211119,0.005128513],"genre_scores_gemma":[0.6904565,0.0001553371,0.3004623,0.0007433937,0.00009420604,0.0002405711,0.002157977,0.001180043,0.004509578],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.006992288,"threshold_uncertainty_score":0.01390314,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05516510920903404,"score_gpt":0.3201223017253664,"score_spread":0.2649571925163323,"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."}}