← 研踪 PaperTrailACL Anthology · ACL 2026:acl:2026.acl-srw.30 · 记录 #1
Debiasing Logical Fallacy Detection for Real-World Robustness via Counterfactually Augmented Data
来源日期（仅到月份） 2026-07
来源记录日期 2026-07
雷达观测 2026-11-01 21:44
摘要
Logical fallacy detection models frequentlyover-flag valid reasoning due to reliance onsurface-level spurious correlations. We in-troduce 703 LLM-generated CounterfactuallyAugmented Data (CAD) pairs—minimally dif-ferentiated valid and fallacious arguments—todebias models through targeted augmentation.Fine-tuning DeBERTa-v3-large on CoCoLoFaaugmented with these pairs yields marginalin-distribution improvement (+0.4% F1) butsubstantial out-of-distribution robustness: 58%relative reduction in false positive rate (64%→ 26.7%) on a 300-sample Reddit-sourcedevaluation set. While recent LLMs (Llama-3.1-8B, Llama-3.3-70B) achieve high perfor-mance under optimized prompts (F1 90–94%),they degrade severely under simple human-like prompts (F1 63–72%, FPR 54–74%).Our lightweight, prompt-invariant approachachieves competitive robustness (F1 85.9%,FPR 26.7%) across all prompting regimes with-out prompt engineering, making it stable forproduction deployment with unpredictable userinput. The dataset and model are publicly re-leased.
论文发现出处
这些页面提供论文线索，不代表同行评审或质量认证。
观测到的版本
平台记录号代表本站观测到的元数据变化，不是作者修订号。来源日期仅到月份或年份时，按该月或该年首日存档。
记录 #1 · 2026-07 · 最新观测版本
Debiasing Logical Fallacy Detection for Real-World Robustness via Counterfactually Augmented Data
Logical fallacy detection models frequentlyover-flag valid reasoning due to reliance onsurface-level spurious correlations. We in-troduce 703 LLM-generated CounterfactuallyAugmented Data (CAD) pairs—minimally dif-ferentiated valid and fallacious arguments—todebias models through targeted augmentation.Fine-tuning DeBERTa-v3-large on CoCoLoFaaugmented with these pairs yields marginalin-distribution improvement (+0.4% F1) butsubstantial out-of-distribution robustness: 58%relative reduction in false positive rate (64%→ 26.7%) on a 300-sample Reddit-sourcedevaluation set. While recent LLMs (Llama-3.1-8B, Llama-3.3-70B) achieve high perfor-mance under optimized prompts (F1 90–94%),they degrade severely under simple human-like prompts (F1 63–72%, FPR 54–74%).Our lightweight, prompt-invariant approachachieves competitive robustness (F1 85.9%,FPR 26.7%) across all prompting regimes with-out prompt engineering, making it stable forproduction deployment with unpredictable userinput. The dataset and model are publicly re-leased.
查看此版本解读 →查看此版本原文 ↗