{"id":"W4407184660","doi":"10.48550/arxiv.2502.01657","title":"Improving Rule-based Reasoning in LLMs using Neurosymbolic Representations","year":2025,"lang":"en","type":"preprint","venue":"arXiv (Cornell University)","topic":"Natural Language Processing Techniques","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Air Force Office of Scientific Research; Natural Sciences and Engineering Research Council of Canada; Canada Research Chairs","keywords":"Rule-based system; Computer science; Artificial intelligence","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":[0.001103989,0.001063462,0.0007142751,0.001000041,0.000460219,0.001927892,0.001825658,0.0009243565,0.005099668],"category_scores_gemma":[0.008459486,0.0004794741,0.001407388,0.0006853463,0.001052012,0.003292383,0.001754732,0.002131972,0.001740216],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001196598,"about_ca_system_score_gemma":0.00160722,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.003108431,"about_ca_topic_score_gemma":0.007432801,"domain_scores_codex":[0.9991845,0.000236155,0.00006554248,0.0002647238,0.000199244,0.00004980559],"domain_scores_gemma":[0.9976326,0.001417938,0.0002209771,0.0004147552,0.0002304927,0.00008324433],"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":[0.0002944798,0.0002610713,0.002393518,0.0005052941,0.0001916491,0.0003326907,0.0003896137,0.2683069,0.0233604,0.04642643,0.0088044,0.6487336],"study_design_scores_gemma":[0.00002292007,0.00003573148,0.000177283,0.00002710764,0.00002304739,0.00004857841,0.00003434117,0.9232209,0.008467612,0.06554753,0.002380909,0.00001391271],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.02174363,0.0003196579,0.9670382,0.0005496584,0.00008260722,0.00009229627,0.0004142867,0.00723093,0.002528591],"genre_scores_gemma":[0.3352453,0.0003504665,0.6578358,0.0004995379,0.00009040318,0.0002245154,0.001210762,0.000625154,0.003918101],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.005099668,"threshold_uncertainty_score":0.01706004,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.05274446377624193,"score_gpt":0.2338925119606467,"score_spread":0.1811480481844047,"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."}}