Complexity of symbolic representation in working memory of Transformer correlates with the complexity of a task

Even though Transformers are extensively used for Natural Language Processing tasks, especially for machine translation, they lack an explicit memory to store key concepts of processed texts. This paper explores the properties of the content of symbolic working memory added to the Transformer model decoder. Such working memory enhances the quality of model predictions in machine translation task and works as a neural-symbolic representation of information that is important for the model to make correct translations. The study of memory content revealed that translated text keywords are stored in the working memory, pointing to the relevance of memory content to the processed text. Also, the diversity of tokens and parts of speech stored in memory correlates with the complexity of the corpora for machine translation task.     

Overestimation and contraction biases of depth information stored in working memory depend on spatial configuration

Depth perception is essential for effective interaction with the environment. Although the accuracy of depth perception has been studied extensively, it is unclear how accurate the depth information is stored in working memory. In this study, we investigated the accuracy and systematic biases of depth representation by a delayed estimation task. The memory array consisted of items presented at various stereoscopic depth positions, and the participants were instructed to estimate the depth position of one target item after a retention interval. We examined the effect of spatial configuration by comparing the memory performance in the whole-display condition where non-target memory items were present during retrieval with that in the single-display condition where non-target memory items were absent. In the single-display condition, we found an overestimation bias that the depth estimates were farther than the corresponding depth positions defined by disparity, and a contraction bias that the stored depth positions near the observer were overestimated and those far from the observer were underestimated. The magnitude of these biases increased with the number of to-be-stored items. However, in the whole-display condition, the overestimation bias was corrected and the contraction bias did not increase with the number of to-be-stored items. Our findings suggested that the number of to-be-stored items could affect the accuracy of depth working memory, and its effect depended crucially on whether the information of spatial configuration of memory display was available at the retrieval stage.     

The update of semantic memories in amnestic mild cognitive impairment

It is still controversial to what extent neocortical consolidated memories are susceptible of change by processes of reconsolidation and transformation throughout experience, and whether the medial temporal lobes are necessary for this update of semantic consolidated memories, as they are for episodic remembering. We hypothesize that patients with amnestic mild cognitive impairment (aMCI) who have deficits in episodic memory may also have difficulties in updating information on added new features of objects. Sixteen participants with aMCI and 20 healthy control participants performed a semantic word-to-picture task, in which they were asked to identify as belonging to a given semantic category NEW objects, that have incorporated novel features, as well as OLD items, semantically and visually SIMILAR items and UNRELATED items. Patients with aMCI made a greater percentage of errors than healthy controls. Participants globally made greater percentages of errors in difficult types of items, namely NEW and SIMILAR, as compared to easier ones, OLD and UNRELATED. Importantly, an item by diagnostic group interaction effect was observed, and post hoc analysis showed that patients with aMCI made a higher percentage of errors than controls in NEW items only. In conclusion, patients with aMCI had a particular difficulty in identifying the NEW items of the word-to-picture task as compared to the control participants, supporting the concept of a flexible and dynamic conceptual knowledge system, involving the update of semantic memories and the integration of new attributes in a constant transformation process, which is impaired in these patients.     

Reconstructing maps from text

Previous research has demonstrated that Distributional Semantic Models (DSMs) are capable of reconstructing maps from news corpora (Louwerse & Zwaan, 2009) and novels (Louwerse & Benesh, 2012). The capacity for reproducing maps is surprising since DSMs notoriously lack perceptual grounding. In this paper we investigate the statistical sources required in language to infer maps, and the resulting constraints placed on mechanisms of semantic representation. Study 1 brings word co-occurrence under experimental control to demonstrate that standard DSMs cannot reproduce maps when word co-occurrence is uniform. Specifically, standard DSMs require that direct co-occurrences between city names in a corpus mirror the proximity between the city locations in the map in order to successfully reconstruct the spatial map. Study 2 presents an instance-based DSM that is capable of reconstructing maps independent of the frequency of co-occurrence of city names.     

客体运动方向的视觉工作记忆容量

记忆动态场景中多客体的特征和时空信息是人类重要的认知活动。目前有关视觉工作记忆的研究虽然广泛探讨了对视觉信息的存储容量及机制,然而所采用的刺激材料均呈现于静态场景中,且不包含运动信息。而有关多客体追踪的研究只关注动态场景中多客体信息的实时更新,而不涉及一段时间内对客体信息的保持。本研究结合视觉工作记忆领域的变化觉察范式和多客体追踪范式,以独立运动的客体为刺激材料,探讨多个客体的运动方向信息在工作记忆中的存储容量问题。结果表明,在工作记忆中能够存储大约3个客体的运动方向信息。     

儿童运动视觉表象操作水平的发展及影响因素

采用计算机软件控制实验,以暴露运动一段行程后进入遮蔽的光点为刺激物,光点以三种不同运动速度、三种不同运动行距组合呈现,要求儿童判断光点到达目标位置时按键反应。结果表明：5—19岁儿童运动视觉表象操作水平的发展存在5一8岁、11—14岁两个加速期;客体运动速度对运动视觉表象操作绩效有显著影响,5、8、11岁组快速较中速下．5．8岁组中速较慢速下操作准确性均有显著降低;客体运动行距对运动视觉表象操作绩效也有显著的影响．5、8、11岁组右目标位置下较中目标位置下操作准确性显著降低;性别对运动视觉表象操作水平无显著影响。     

Visuospatial and verbal memory in mental arithmetic

Working memory allows complex information to be remembered and manipulated over short periods of time. Correlations between working memory and mathematics achievement have been shown across the lifespan. However, only a few studies have examined the potentially distinct contributions of domain-specific visuospatial and verbal working memory resources in mental arithmetic computation. Here we aimed to fill this gap in a series of six experiments pairing addition and subtraction tasks with verbal and visuospatial working memory and interference tasks. In general, we found higher levels of interference between mental arithmetic and visuospatial working memory tasks than between mental arithmetic and verbal working memory tasks. Additionally, we found that interference that matched the working memory domain of the task (e.g., verbal task with verbal interference) lowered working memory performance more than mismatched interference (verbal task with visuospatial interference). Findings suggest that mental arithmetic relies on domain-specific working memory resources.     

运动视觉中计时行为的控制操作理论

运动视觉中的计时行为分为两类 :拦截行为和制动行为。“环境—活动者系统”(EAS)中的信息决定着计量行为的控制操作。本文阐明了信息和EAS的特性 ,通过对 tau-margin在拦截行为中的作用以及视觉变量τ对于制动操作重要性的分析 ,论述了运动视觉中计时行为的控制操作理论。     

物体颜色对情景记忆的影响

本研究采用ERP技术和学习-再认范式考察视觉输入的颜色和记忆中的物体颜色知识一致性对情景记忆编码和提取的影响。结果显示, 在对物体图片进行编码时颜色不一致图片诱发更大N400, 而提取阶段则对颜色一致图片有更多熟悉性加工(实验1); 物体名称能够更快激活典型颜色知识, 对颜色一致名称有更多细节回想(LPC更正) (实验2)。实验结果表明, 颜色一致促进知觉水平的记忆编码, 而阻碍语义水平编码。同时, 颜色一致促进物体图片提取(知觉水平)中的熟悉性和回想过程; 而对物体名称提取(概念水平)的促进作用仅表现在回想上。此外, 颜色与物体名称有着密切联系, 同样影响物体的语义表征, 支持激活-扩散模型理论。本研究从知觉和概念水平上为揭示物体颜色在联结记忆中的作用提供了证据。