[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-178985-105":59,"doc-detail-178985-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","2024-findings-logical-fallacy-identification-deduction-and-correction","2024 Findings - Logical Fallacy Identification, Deduction, and Correction","","2024 findings analyze how different model families perform on logical fallacy understanding tasks framed as WHAT, WHY, and HOW. The work evaluates sentence-level identification and classification of faulty reasoning, then measures deduction and backward deduction performance under specified fallacy types, and finally tests modification by correcting erroneous statements. Experiments compare fine-tuning settings and multiple datasets across several model sizes, reporting accuracy, deltas, and per-fallacy breakdowns to assess task contributions and error patterns.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/2024-findings-logical-fallacy-identification-deduction-and-correction/178985/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/2024-findings-logical-fallacy-identification-deduction-and-correction/178985.png","ImageObject",300,407,{"name":92,"@type":93},"Emma Wilson","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-24","2026-09-02",true,{"@type":102,"interactionType":103,"userInteractionCount":29},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What do the WHAT, WHY, and HOW tasks cover in the findings?","Question",{"text":112,"@type":113},"WHAT focuses on identifying and classifying whether a sentence contains logical fallacies. WHY evaluates deduction and backward deduction from premises to conclusions (and vice versa). HOW measures modification by correcting the logical fallacy in the sentence.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"Which models are compared and how is performance reported?",{"text":117,"@type":113},"The findings compare models such as LLaMA2 variants, Vicuna variants, ChatGPT, and GPT-4. Performance is reported using accuracy percentages, deltas between conditions (e.g., Origin vs added logic/strategy variants), and additional counts where available.",{"name":119,"@type":110,"acceptedAnswer":120},"How do task removals (e.g., w/o Task1) and fallacy-type analyses help interpret results?",{"text":121,"@type":113},"Ablation rows like w/o Task1 or w/o Task2 show how removing a capability affects accuracy on different datasets. Per fallacy type breakdown tables further reveal which reasoning errors are more or less difficult across tasks, clarifying where models succeed or fail.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},178985,1790794295,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":29,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":144,"read_time":145,"preview_page_count":4},3848291630094,"https://eur-avatar.wpscdn.com/davatar_085a072bc5b1113ac321206ff7593b45","| Singular Proposition | 54 |\n| --- | --- |\n| Particular Proposition | 5 |\n| Universal Proposition | 8 |\n| Affirmative Proposition\u003Cbr>Negative Proposition | 56\u003Cbr>11 |\n| Propositions without Pronouns\u003Cbr>Propositions with Pronouns | 58\u003Cbr>9 |\n| Propositions with Human Subjects\u003Cbr>Propositions with Non-Human Subjects | 52\u003Cbr>15 |\n\n\n| Dimension Task name |  | Task definition |\n| --- | --- | --- |\n| WHAT | Task1: Identification\u003Cbr>Task2: Classification | Identify whether the given sentence has logical fallacy.\u003Cbr>Select the sentence belonging to a certain type of logical fallacy. |\n| WHY | Task3: Deduction\u003Cbr>Task4: Backward Deduction | Derive the conclusion from the premise according to a certain type of logical fallacy. Infer the premise from the conclusion according to a certain type of logical fallacy. |\n| HOW | Task5: Modification | Correct the logical fallacy in the given sentence. |\n\n| Datasets | FT Data | LLaMA2-13B |  | LLaMA2-7B |  | Vicuna-13B |  | Vicuna-7B |  | Orca2-7B |  |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n|  |  | Acc. | ∆% | Acc. | ∆% | Acc. | ∆% | Acc. | ∆% | Acc. | ∆% |\n| LogiQA2.0 | Origin | 45.55 | - | 42.30 | - | 52.74 | - | 47.71 | - | 54.39 | - |\n|  | Origin+LOGIC | 44.66 | -1.95 | 35.62 | -15.79 | 53.37 | 1.19 | 45.10 | -5.47 | 52.93 | -2.68 |\n|  | Origin+LFUD | 47.90 | 5.16 | 43.13 | 1.96 | 55.85 | 5.90 | 47.84 | 0.27 | 56.55 | 3.97 |\n| Reclor | Origin | 47.20 | - | 40.40 | - | 54.40 | - | 49.20 | - | 55.80 | - |\n|  |  |  |  |  |  |  |  |  |  |  |  |\n|  | Origin+LOGIC | 46.20 | -2.12 | 42.20 | 4.46 | 54.00 | -0.74 | 47.80 | -2.85 | 55.80 | 0.00 |\n|  | Origin+LFUD | 50.20 | 6.36 | 46.40 | 14.85 | 57.00 | 4.78 | 51.80 | 5.28 | 58.20 | 4.30 |\n| TaxiNLI | Origin | 68.54 | - | 62.68 | - | 78.91 | - | 77.47 | - | 82.33 | - |\n|  | Origin+LOGIC | 40.60 | -40.76 | 58.80 | -6.19 | 77.92 | -1.25 | 76.18 | -1.67 | 82.18 | -0.18 |\n|  | Origin+LFUD | 73.70 | 7.53 | 67.26 | 7.31 | 79.76 | 1.08 | 77.77 | 0.39 | 84.02 | 2.05 |\n| FOLIO | Origin | 61.76 | - | 50.98 | - | 36.76 | - | 50.49 | - | 72.55 | - |\n|  | Origin+LOGIC | 62.25 | 0.79 | 52.45 | 2.88 | 36.28 | -1.31 | 45.10 | -10.68 | 73.53 | 1.35 |\n|  | Origin+LFUD | 66.18 | 7.16 | 59.31 | 16.34 | 44.61 | 21.35 | 56.37 | 11.65 | 76.47 | 5.40 |\n\n| Models | WHAT |  | WHY |  | HOW\u003Cbr>Task5 |\n| --- | --- | --- | --- | --- | --- |\n|  | Task1 | Task2 | Task3 | Task4 |  |\n| LLaMA2-7B | 46.84 | 47.86 | 24.75 | 29.35 | 40.00 |\n| LLaMA2-13B | 55.76 | 53.61 | 52.36 | 54.48 | 50.00 |\n| Vicuna-7B | 32.96 | 57.71 | 60.70 | 56.59 | 46.00 |\n| Vicuna-13B | 50.76 | 58.46 | 58.33 | 61.32 | 56.00 |\n| ChatGPT | 54.66 | 73.88 | 62.94 | 70.65 | 60.00 |\n| GPT-4 | 86.35 | 86.19 | 78.61 | 85.70 | 88.00 |\n\n\n| Task Category | LogiQA2.0 | Reclor | TaxiNLI | FOLIO |\n| --- | --- | --- | --- | --- |\n| No Tasks | 45.55 | 47.20 | 68.54 | 61.76 |\n| w/o Task1 | 46.69 | 49.80 | 69.53 | 63.73 |\n| w/o Task2 | 45.74 | 48.00 | 69.88 | 65.20 |\n| w/o Task3 | 47.46 | 49.00 | 72.01 | 64.22 |\n| w/o Task4 | 46.44 | 48.80 | 69.28 | 65.20 |\n| All Tasks | 47.90 | 50.20 | 73.70 | 66.18 |\n\n\n| Fallacy Type | LogiQA2.0 | Reclor | TaxiNLI | FOLIO |\n| --- | --- | --- | --- | --- |\n| No Fallacy Data | 45.55 | 47.20 | 68.54 | 61.76 |\n| w/o Faulty Generalization | 46.56 | 49.80 | 71.91 | 64.71 |\n| w/o False Causality | 46.69 | 47.60 | 72.56 | 62.75 |\n| w/o Circular Reasoning | 46.12 | 49.80 | 72.95 | 64.22 |\n| w/o Ad Populum | 46.25 | 47.60 | 72.85 | 64.22 |\n| w/o Ad hominem | 46.95 | 48.60 | 69.53 | 65.20 |\n| w/o Deductive Fallacy | 45.87 | 49.40 | 73.78 | 62.75 |\n| w/o Appeal to Emotion | 47.65 | 49.80 | 69.93 | 63.73 |\n| w/o False Dilemma | 46.12 | 50.00 | 73.10 | 63.24 |\n| w/o Fallacy of Extension | 45.93 | 49.40 | 72.51 | 64.71 |\n| w/o Fallacy of Relevance | 47.65 | 50.20 | 70.92 | 61.27 |\n| w/o Fallacy of Credibility | 47.58 | 48.60 | 72.06 | 62.75 |\n| w/o Intentional Fallacy | 46.88 | 49.80 | 69.48 | 65.69 |\n| All Fallacy Types | 47.90 | 50.20 | 73.70 | 66.18 |\n\n\n| Models | Accuracy","cbCaiiBFjizQcu5f","https://ap.wps.com/l/cbCaiiBFjizQcu5f","pdf",1511596,14,"English","# Task Design and Objectives\n## WHAT: Identification and Classification\n## WHY: Deduction and Backward Deduction\n## HOW: Modification and Correction\n# Experimental Results and Ablations\n## Model and Dataset Accuracy Tables\n## Task and Fallacy Type Breakdown","[{\"question\":\"What do the WHAT, WHY, and HOW tasks cover in the findings?\",\"answer\":\"WHAT focuses on identifying and classifying whether a sentence contains logical fallacies. WHY evaluates deduction and backward deduction from premises to conclusions (and vice versa). HOW measures modification by correcting the logical fallacy in the sentence.\"},{\"question\":\"Which models are compared and how is performance reported?\",\"answer\":\"The findings compare models such as LLaMA2 variants, Vicuna variants, ChatGPT, and GPT-4. Performance is reported using accuracy percentages, deltas between conditions (e.g., Origin vs added logic/strategy variants), and additional counts where available.\"},{\"question\":\"How do task removals (e.g., w/o Task1) and fallacy-type analyses help interpret results?\",\"answer\":\"Ablation rows like w/o Task1 or w/o Task2 show how removing a capability affects accuracy on different datasets. Per fallacy type breakdown tables further reveal which reasoning errors are more or less difficult across tasks, clarifying where models succeed or fail.\"}]","2024 Findings - Logical Fallacy Identification, Deduction, and Correction | PDF",1788334814,35]