Cross-Domain Generalization Through Memorization: A Study of Nearest Neighbors in Neural Duplicate Question Detection

大型语言模型的马赛克记忆

The Mosaic Memory of Large Language Models

本文提出LLM的“马赛克记忆”现象，证明模型会整合相似序列信息且记忆主要是句法性的，模糊重复普遍存在并影响隐私与评估。

Generalization through Memorization: Nearest Neighbor Language Models

We introduce $k$NN-LMs, which extend a pre-trained neural language model (LM) by linearly interpolating it with a $k$-nearest neighbors ($k$NN) model. The nearest neighbors are computed according to distance in the pre-trained LM embedding space, and can be drawn from any text collection, including the original LM training data. Applying this augmentation to a strong Wikitext-103 LM, with neighbors drawn from the original training set, our $k$NN-LM achieves a new state-of-the-art perplexity of 15.79 - a 2.9 point improvement with no additional training. We also show that this approach has impl…

反事实影响作为分布量

Counterfactual Influence as a Distributional Quantity

本文将反事实影响视为分布量来研究模型记忆现象，发现仅看自影响会因近似重复样本的存在而严重低估实际风险，完整影响分布能更好地捕捉训练数据间的复杂交互。

Quantifying Memorization Across Neural Language Models

Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an ex…

Question and Answer Test-Train Overlap in Open-Domain Question Answering Datasets

Ideally Open-Domain Question Answering models should exhibit a number of competencies, ranging from simply memorizing questions seen at training time, to answering novel question formulations with answers seen during training, to generalizing to completely novel questions with novel answers. However, single aggregated test set scores do not show the full picture of what capabilities models truly have. In this work, we perform a detailed study of the test sets of three popular open-domain benchmark datasets with respect to these competencies. We find that 60-70% of test-time answers are also pr…

通过记忆进行推理：最近邻知识图谱嵌入

Reasoning Through Memorization: Nearest Neighbor Knowledge Graph Embeddings

提出kNN-KGE方法，通过预训练语言模型结合k近邻线性插值实体分布，显式记忆罕见或新实体，显著提升归纳与直推式链接预测及低资源场景下的性能。