{"id":"W7135077784","doi":"10.1109/cait68620.2025.11424747","title":"Enhancing Explainability for Exchange Rate Forecasting via Self-Reflective Reinforcement Learning","year":2025,"lang":"","type":"article","venue":"","topic":"Stock Market Forecasting Methods","field":"Decision Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"McGill University","funders":"Natural Science Foundation of Guangdong Province","keywords":"Reinforcement learning; Coherence (philosophical gambling strategy); Construct (python library); Relevance (law); Feature (linguistics); Reinforcement","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch","metaepi_narrow","sts","insufficient_payload"],"consensus_categories":["metaresearch"],"category_scores_codex":[0.05906986,0.001071588,0.001669729,0.001538963,0.003208851,0.001023194,0.00168338,0.0005583446,0.002170207],"category_scores_gemma":[0.1011455,0.0009692854,0.0009293872,0.00463223,0.0003054677,0.001108942,0.001766335,0.001100372,0.0000757395],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001742514,"about_ca_system_score_gemma":0.001223288,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0001989076,"about_ca_topic_score_gemma":0.0002525771,"domain_scores_codex":[0.9848375,0.004291074,0.003828549,0.00305526,0.001639093,0.002348541],"domain_scores_gemma":[0.9462026,0.04700622,0.001646877,0.001554305,0.003157555,0.000432491],"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":[0.002025102,0.0002927978,0.004038321,0.001296863,0.0006288167,0.00001378218,0.01299318,0.01686258,0.004328332,0.002530758,0.00211299,0.9528765],"study_design_scores_gemma":[0.002711606,0.001451605,0.0009753518,0.0006987932,0.0003711577,0.00001518135,0.005733402,0.8541553,0.04598355,0.04791149,0.03877986,0.00121272],"study_design_candidate":"design_other","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.02477025,0.0004083315,0.8717353,0.000638508,0.003962968,0.004126602,0.000004303945,0.0003953307,0.09395842],"genre_scores_gemma":[0.7142618,0.00004004664,0.2245579,0.0006147975,0.0004030632,0.0009588868,0.000006971206,0.00007936652,0.05907722],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.9516638,"threshold_uncertainty_score":0.9992757,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.09319439548135743,"score_gpt":0.4029594884895322,"score_spread":0.3097650930081747,"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."}}