{"id":"W7106856021","doi":"10.48448/4pxd-2w74","title":"Direct Value Optimization: Improving Chain-of-Thought Reasoning in LLMs with Refined Values","year":2025,"lang":"","type":"other","venue":"Open MIND","topic":"","field":"","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Artificial Intelligence in Medicine (Canada)","funders":"","keywords":"Preference; Value (mathematics); Margin (machine learning); Tree (set theory); Coding (social sciences); Key (lock); Outcome (game theory)","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.002632717,0.001177942,0.001227333,0.0007312765,0.0005849835,0.001913619,0.002279363,0.001369118,0.007000109],"category_scores_gemma":[0.01488189,0.0006047693,0.001075953,0.0006121329,0.001856952,0.003613581,0.002906653,0.002634999,0.001321201],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001449242,"about_ca_system_score_gemma":0.00226866,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003847146,"about_ca_topic_score_gemma":0.006332662,"domain_scores_codex":[0.9981747,0.0007929712,0.0000860728,0.0003626909,0.0004053194,0.0001783247],"domain_scores_gemma":[0.9947851,0.003513444,0.0004095417,0.0006663716,0.0004048439,0.0002207465],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.0003399576,0.0002717001,0.00214893,0.0003709501,0.00009192203,0.000268991,0.0004963184,0.6345782,0.005301194,0.08080354,0.005331792,0.2699966],"study_design_scores_gemma":[0.00002483418,0.00003591845,0.00006174309,0.00002264456,0.000008395486,0.00001493234,0.00002072553,0.9478716,0.001209589,0.04992208,0.0007989288,0.00000861732],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.01966601,0.0002662383,0.9743032,0.0004398183,0.00004341035,0.00007048938,0.00009656649,0.001505271,0.003609113],"genre_scores_gemma":[0.6537105,0.0001891761,0.3403617,0.0004726772,0.00006681272,0.0001847681,0.0002814657,0.0005041412,0.004228794],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.007000109,"threshold_uncertainty_score":0.02341771,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.01620426234569054,"score_gpt":0.2835385228520458,"score_spread":0.2673342605063552,"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."}}