Search Results (2)

The pursuit of universal, symmetric semantic representations within large language models (LLMs) faces a fundamental challenge: the inherent asymmetry of natural languages. Different languages exhibit vast disparities in syntactic structures, lexical choices, and cultural nuances, making the creation of a truly shared, symmetric embedding space a non-trivial task. This paper aims to address this critical problem by introducing a novel framework to forge robust and symmetric multilingual sentence embeddings. Our approach, named DACL (Dynamic Asymmetric Contrastive Learning), is anchored in two powerful asymmetric learning paradigms: Contrastive Learning and Dynamic Curriculum Learning (DCL). We extend Contrastive Learning to the multilingual context, where it asymmetrically treats semantically equivalent sentences from different languages (positive pairs) and sentences with distinct meanings (negative pairs) to enforce semantic symmetry in the target embedding space. To further refine this process, we incorporate Dynamic Curriculum Learning, which introduces a second layer of asymmetry by dynamically scheduling training instances from easy to hard. This dual-asymmetric strategy enables the model to progressively master complex cross-lingual relationships, starting with more obvious semantic equivalences and advancing to subtler ones. Our comprehensive experiments on benchmark cross-lingual tasks, including sentence retrieval and cross-lingual classification (XNLI, PAWS-X, MLDoc, MARC), demonstrate that DACL significantly outperforms a wide range of established baselines. The results validate our dual-asymmetric framework as a highly effective approach for forging robust multilingual embeddings, particularly excelling in tasks involving complex linguistic asymmetries. Ultimately, this work contributes a novel dual-asymmetric learning framework that effectively leverages linguistic asymmetry to achieve robust semantic symmetry across languages. It offers valuable insights for developing more capable, fair, and interpretable multilingual LLMs, emphasizing that deliberately leveraging asymmetry in the learning process is a highly effective strategy. Full article

Background: Episodic memory (EM) is particularly sensitive to pathological conditions and aging. In a neurocognitive context, the paired-associate learning (PAL) paradigm, which requires participants to learn and recall associations between stimuli, has been used to measure EM. The present study aimed to explore whether functional near-infrared spectroscopy (fNIRS) can be employed to determine cortical activity underlying encoding and retrieval. Moreover, we examined whether and how different aspects of task (i.e., novelty, difficulty) affects those cortical activities. Methods: Twenty-two male college students (age: M = 20.55, SD = 1.62) underwent a face-name PAL paradigm under 40-channel fNIRS covering fronto-parietal and middle occipital regions. Results: A decreased activity during encoding in a broad network encompassing the bilateral frontal cortex (Brodmann areas 9, 11, 45, and 46) was observed during the encoding, while an increased activity in the left orbitofrontal cortex (Brodmann area 11) was observed during the retrieval. Increased HbO concentration in the superior parietal cortices and decreased HbO concentration in the inferior parietal cortices were observed during encoding while dominant activation of left PFC was found during retrieval only. Higher task difficulty was associated with greater neural activity in the bilateral prefrontal cortex and higher task novelty was associated with greater activation in occipital regions. Conclusion: Combining the PAL paradigm with fNIRS provided the means to differentiate neural activity characterising encoding and retrieval. Therefore, the fNIRS may have the potential to complete EM assessments in clinical settings. Full article