{"id":"W4379523158","doi":"10.21428/594757db.63abb0f0","title":"Multihop Factual Claim Verification Using Natural Language Prompts","year":2023,"lang":"en","type":"article","venue":"","topic":"Topic Modeling","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"York University","funders":"","keywords":"Computer science; Generalization; Artificial intelligence; Task (project management); Natural language processing; Machine learning; Domain (mathematical analysis); Natural language","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":[0.004639635,0.001153725,0.001118788,0.001575794,0.0007700672,0.001680478,0.001986086,0.001830219,0.004353326],"category_scores_gemma":[0.03676346,0.0004313822,0.0008879822,0.0008219327,0.0009153857,0.005419618,0.002806157,0.002989837,0.001869153],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0007611278,"about_ca_system_score_gemma":0.001825761,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0008625182,"about_ca_topic_score_gemma":0.001572083,"domain_scores_codex":[0.9957468,0.001358122,0.0002824331,0.001441228,0.0009879189,0.0001834734],"domain_scores_gemma":[0.9708736,0.01927085,0.002508543,0.004175844,0.002539854,0.0006313832],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"not_applicable","study_design_scores_codex":[0.001725116,0.0009730908,0.01210262,0.001168978,0.0001226004,0.00206014,0.001735827,0.04955304,0.06869192,0.02041014,0.01388071,0.8275759],"study_design_scores_gemma":[0.0001792489,0.0005697769,0.005115626,0.0001223685,0.00006371769,0.001229251,0.0006029995,0.8556314,0.05302177,0.0675408,0.01581661,0.0001062809],"study_design_candidate":"not_applicable","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.1144143,0.0006132799,0.8584461,0.0008229806,0.0002170747,0.0003822127,0.001395842,0.02127347,0.002434737],"genre_scores_gemma":[0.6829724,0.0002100303,0.3104277,0.0002502271,0.0001247826,0.0002050824,0.003065967,0.0003305176,0.002413255],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.004639635,"threshold_uncertainty_score":0.02453709,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.04638585074594859,"score_gpt":0.309028633505941,"score_spread":0.2626427827599924,"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."}}