Search for: All records

The long-standing one-to-many problem of gold standard responses in open-domain dialogue systems presents challenges for automatic evaluation metrics. Though prior works have demonstrated some success by applying powerful Large Language Models (LLMs), existing approaches still struggle with the one-to-many problem, and exhibit subpar performance in domain-specific scenarios. We assume the commonsense reasoning biases within LLMs may hinder their performance in domain-specific evaluations. To address both issues, we propose a novel framework SLIDE (Small and Large Integrated for Dialogue Evaluation), that leverages both a small, specialized model (SLM), and LLMs for the evaluation of open-domain dialogues. Our approach introduces several techniques: (1) Contrastive learning to differentiate between robust and non-robust response embeddings; (2) A novel metric for semantic sensitivity that combines embedding cosine distances with similarity learned through neural networks, and (3) A strategy for incorporating the evaluation results from both the SLM and LLMs. Our empirical results demonstrate that our approach achieves state-of-the-art performance in both the classification and evaluation tasks, and additionally, the SLIDE evaluator exhibits a better correlation with human judgments. 

Graph neural networks (GNNs) are proficient machine learning models in handling irregularly structured data. Nevertheless, their generic formulation falls short when applied to the analysis of brain connectomes in Alzheimer’s Disease (AD), necessitating the incorporation of domain-specific knowledge to achieve optimal model performance. The integration of AD-related expertise into GNNs presents a significant challenge. Current methodologies reliant on manual design often demand substantial expertise from external domain specialists to guide the development of novel models, thereby consuming considerable time and resources. To mitigate the need for manual curation, this paper introduces a novel self-guided knowledge-infused multimodal GNN to autonomously integrate domain knowledge into the model development process. We propose to conceptualize existing domain knowledge as natural language, and devise a specialized multimodal GNN framework tailored to leverage this uncurated knowledge to direct the learning of the GNN submodule, thereby enhancing its efficacy and improving prediction interpretability. To assess the effectiveness of our framework, we compile a comprehensive literature dataset comprising recent peer-reviewed publications on AD. By integrating this literature dataset with several real-world AD datasets, our experimental results illustrate the effectiveness of the proposed method in extracting curated knowledge and offering explanations on graphs for domain-specific applications. Furthermore, our approach successfully utilizes the extracted information to enhance the performance of the GNN. 

The MRI-derived brain network serves as a pivotal instrument in elucidating both the structural and functional aspects of the brain, encompassing the ramifications of diseases and developmental processes. However, prevailing methodologies, often focusing on synchronous BOLD signals from functional MRI (fMRI), may not capture directional influences among brain regions and rarely tackle temporal functional dynamics. In this study, we first construct the brain-effective network via the dynamic causal model. Subsequently, we introduce an interpretable graph learning framework termed Spatio-Temporal Embedding ODE (STE-ODE). This framework incorporates specifically designed directed node embedding layers, aiming at capturing the dynamic interplay between structural and effective networks via an ordinary differential equation (ODE) model, which characterizes spatial-temporal brain dynamics. Our framework is validated on several clinical phenotype prediction tasks using two independent publicly available datasets (HCP and OASIS). The experimental results clearly demonstrate the advantages of our model compared to several state-of-the-art methods. 

Multimodal medical image synthesis is an important task. Previous efforts mainly focus on the task domain of medical image synthesis using the complete source data and have achieved great success. However, data collection with completeness in real life might be prohibitive due to high expenses or other difficulties, particularly in brain imaging studies. In this paper, we address the challenging and important problem of medical image synthesis from incomplete multimodal data sources. We propose to learn the modal-wise representations and synthesize the targets accordingly. Particularly, a surrogate sampler is derived to generate the target representations from incomplete observations, based on which an interpretable attention-redistribution network is designed. The experimental results synthesizing PET images from MRI images demonstrate that the proposed method can solve different missing data scenarios and outperforms related baselines consistently. 

The modeling of the interaction between brain structure and function using deep learning techniques has yielded remarkable success in identifying potential biomarkers for different clinical phenotypes and brain diseases. However, most existing studies focus on one-way mapping, either projecting brain function to brain structure or inversely. This type of unidirectional mapping approach is limited by the fact that it treats the mapping as a one-way task and neglects the intrinsic unity between these two modalities. Moreover, when dealing with the same biological brain, mapping from structure to function and from function to structure yields dissimilar outcomes, highlighting the likelihood of bias in one-way mapping. To address this issue, we propose a novel bidirectional mapping model, named Bidirectional Mapping with Contrastive Learning (BMCL), to reduce the bias between these two unidirectional mappings via ROI-level contrastive learning. We evaluate our framework on clinical phenotype and neurodegenerative disease predictions using two publicly available datasets (HCP and OASIS). Our results demonstrate the superiority of BMCL compared to several state-of-the-art methods. 

This Letter reports the first measurement of the oscillation amplitude and frequency of reactor antineutrinos at Daya Bay via neutron capture on hydrogen using 1958 days of data. With over 3.6 million signal candidates, an optimized candidate selection, improved treatment of backgrounds and efficiencies, refined energy calibration, and an energy response model for the capture-on-hydrogen sensitive region, the relative ${\overline{\nu }}_e$rates and energy spectra variation among the near and far detectors gives ${\sin }^{2}2{\theta }_{13}=0.0759_{-0.0049}^{+0.0050}$and $\mathrm{\Delta }{m}_{32}^2=\left(2.72_{-0.15}^{+0.14}\right)\times {10}^{-3}{\mathrm{eV}}^2$assuming the normal neutrino mass ordering, and $\mathrm{\Delta }{m}_{32}^2=(-2.83_{-0.14}^{+0.15})\times {10}^{-3}{\mathrm{eV}}^2$for the inverted neutrino mass ordering. This estimate of ${\sin }^{2}2{\theta }_{13}$is consistent with and essentially independent from the one obtained using the capture-on-gadolinium sample at Daya Bay. The combination of these two results yields ${\sin }^{2}2{\theta }_{13}=0.0833\pm 0.0022$, which represents an 8% relative improvement in precision regarding the Daya Bay full 3158-day capture-on-gadolinium result. Published by the American Physical Society2024