{"id":"W4283792669","doi":"10.1609/aaai.v36i10.21386","title":"Supervising Model Attention with Human Explanations for Robust Natural Language Inference","year":2022,"lang":"en","type":"article","venue":"Proceedings of the AAAI Conference on Artificial Intelligence","topic":"Topic Modeling","field":"Computer Science","cited_by":30,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"Azrieli Foundation","keywords":"Computer science; Premise; Inference; Artificial intelligence; Natural (archaeology); Task (project management); Punctuation; Natural language; Natural language processing; Focus (optics); Language model; Machine learning; Linguistics","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":[0.004664932,0.001371195,0.001050211,0.0009251531,0.0005692014,0.001465681,0.002746435,0.002148946,0.003851],"category_scores_gemma":[0.02448626,0.0008161674,0.001004936,0.0007061411,0.001368965,0.003612457,0.002663397,0.004356109,0.001352232],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001389325,"about_ca_system_score_gemma":0.001938292,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007318926,"about_ca_topic_score_gemma":0.01483963,"domain_scores_codex":[0.9974688,0.00124798,0.00008910045,0.0006963563,0.0003472985,0.0001504321],"domain_scores_gemma":[0.9877903,0.008216796,0.0006062749,0.002294918,0.0007854195,0.0003062196],"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":[0.0008731686,0.0006842437,0.01005807,0.0004563309,0.0003399057,0.0003348166,0.001735377,0.2975597,0.02552943,0.01794756,0.01526073,0.6292207],"study_design_scores_gemma":[0.00005852709,0.00007492128,0.000906023,0.00002530657,0.00004341636,0.00007947381,0.00007258608,0.9644585,0.004664514,0.02693322,0.00265881,0.00002469937],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"methods","genre_scores_codex":[0.05218602,0.0008409338,0.934971,0.001128422,0.0001020803,0.0001589778,0.0002047253,0.008078294,0.002329669],"genre_scores_gemma":[0.7772051,0.0002346528,0.2182654,0.0007872792,0.0001443161,0.000176498,0.0006178156,0.0004642016,0.00210468],"genre_candidate":"methods","genre_consensus":"methods","teacher_disagreement_score":0.007318926,"threshold_uncertainty_score":0.02467078,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1128407143945447,"score_gpt":0.3157759058673034,"score_spread":0.2029351914727586,"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."}}