用专家模型乘积方法降低自然语言推理中的数据偏差，提升模型真实推理能力。
Product-of-Experts Training Reduces Dataset Artifacts in Natural Language Inference
- 通过专家模型乘积机制，抑制过度自信的错误样本。
- 准确率仅下降0.2%（89.30%→89.10%），但偏差依赖降低4.71%。
- 适合关注模型公平性与可解释性的研究人员使用。
神经网络自然语言推理模型容易过拟合数据集中的伪相关特征而非真正推理。仅基于假设的模型在SNLI上达到57.7%准确率，表明存在强烈虚假关联；基线模型38.6%的错误由这些伪相关导致。本文提出专家模型乘积（PoE）训练方法，通过降低偏见模型过度自信样本的权重来减少偏差。该方法几乎保持原准确率（89.10% vs. 89.30%），同时将偏差依赖降低4.71%（偏差一致性从49.85%降至45%）。消融实验显示λ=1.5时在去偏与精度间取得最佳平衡。行为测试仍发现模型在否定和数值推理方面存在缺陷。
原文摘要 · Abstract (English)
Neural NLI models overfit dataset artifacts instead of truly reasoning. A hypothesis-only model gets 57.7% in SNLI, showing strong spurious correlations, and 38.6% of the baseline errors are the result of these artifacts. We propose Product-of-Experts (PoE) training, which downweights examples where biased models are overconfident. PoE nearly preserves accuracy (89.10% vs. 89.30%) while cutting bias reliance by 4.71% (bias agreement 49.85% to 45%). An ablation finds lambda = 1.5 that best balances debiasing and accuracy. Behavioral tests still reveal issues with negation and numerical reasoning.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。