Yanming Sun

Han Wu, Yanming Sun, Yunhe Yang et al.

Multimodal sentiment analysis (MSA) leverages information fusion from diverse modalities (e.g., text, audio, visual) to enhance sentiment prediction. However, simple fusion techniques often fail to account for variations in modality quality, such as those that are noisy, missing, or semantically conflicting. This oversight leads to suboptimal performance, especially in discerning subtle emotional nuances. To mitigate this limitation, we introduce a simple yet efficient \textbf{A}daptive \textbf{G}ated \textbf{F}usion \textbf{N}etwork that adaptively adjusts feature weights via a dual gate fusion mechanism based on information entropy and modality importance. This mechanism mitigates the influence of noisy modalities and prioritizes informative cues following unimodal encoding and cross-modal interaction. Experiments on CMU-MOSI and CMU-MOSEI show that AGFN significantly outperforms strong baselines in accuracy, effectively discerning subtle emotions with robust performance. Visualization analysis of feature representations demonstrates that AGFN enhances generalization by learning from a broader feature distribution, achieved by reducing the correlation between feature location and prediction error, thereby decreasing reliance on specific locations and creating more robust multimodal feature representations.

Yanming Sun, Runzhe Zhan, Chi Seng Cheang et al.

\textbf{RE}trieval-\textbf{A}ugmented \textbf{L}LM-based \textbf{M}achine \textbf{T}ranslation (REAL-MT) shows promise for knowledge-intensive tasks like idiomatic translation, but its reliability under noisy retrieval contexts remains poorly understood despite this being a common challenge in real-world deployment. To address this gap, we propose a noise synthesis framework and new metrics to evaluate the robustness of REAL-MT systematically. Using this framework, we instantiate REAL-MT with Qwen-series models, including standard LLMs and large reasoning models (LRMs) with enhanced reasoning, and evaluate their performance on idiomatic translation across high-, medium-, and low-resource language pairs under synthesized noise. Our results show that low-resource language pairs, which rely more heavily on retrieved context, degrade more severely under noise than high-resource ones and often produce nonsensical translations. Although LRMs possess enhanced reasoning capabilities, they show no improvement in error correction and are even more susceptible to noise, tending to rationalize incorrect contexts. We find that this stems from an attention shift away from the source idiom to noisy content, while confidence increases despite declining accuracy, indicating poor calibration. To mitigate these issues, we investigate training-free and fine-tuning strategies, which improve robustness at the cost of performance in clean contexts, revealing a fundamental trade-off. Our findings highlight the limitations of current approaches, underscoring the need for self-verifying integration mechanisms.

Chi Seng Cheang, Hou Pong Chan, Derek F. Wong et al.

Recent pre-trained language models (PLMs) achieve promising results in existing abstractive summarization datasets. However, existing summarization benchmarks overlap in time with the standard pre-training corpora and finetuning datasets. Hence, the strong performance of PLMs may rely on the parametric knowledge that is memorized during pre-training and fine-tuning. Moreover, the knowledge memorized by PLMs may quickly become outdated, which affects the generalization performance of PLMs on future data. In this work, we propose TempoSum, a novel benchmark that contains data samples from 2010 to 2022, to understand the temporal generalization ability of abstractive summarization models. Through extensive human evaluation, we show that parametric knowledge stored in summarization models significantly affects the faithfulness of the generated summaries on future data. Moreover, existing faithfulness enhancement methods cannot reliably improve the faithfulness of summarization models on future data. Finally, we discuss several recommendations to the research community on how to evaluate and improve the temporal generalization capability of text summarization models.

Yanming Sun, Chunyan Wang

The work presented in this paper is to propose a reliable high-quality system of Convolutional Neural Network (CNN) for brain tumor segmentation with a low computation requirement. The system consists of a CNN for the main processing for the segmentation, a pre-CNN block for data reduction and post-CNN refinement block. The unique CNN consists of 7 convolution layers involving only 108 kernels and 20308 trainable parameters. It is custom-designed, following the proposed paradigm of ASCNN (application specific CNN), to perform mono-modality and cross-modality feature extraction, tumor localization and pixel classification. Each layer fits the task assigned to it, by means of (i) appropriate normalization applied to its input data, (ii) correct convolution modes for the assigned task, and (iii) suitable nonlinear transformation to optimize the convolution results. In this specific design context, the number of kernels in each of the 7 layers is made to be just-sufficient for its task, instead of exponentially growing over the layers, to increase information density and to reduce randomness in the processing. The proposed activation function Full-ReLU helps to halve the number of kernels in convolution layers of high-pass filtering without degrading processing quality. A large number of experiments with BRATS2018 dataset have been conducted to measure the processing quality and reproducibility of the proposed system. The results demonstrate that the system reproduces reliably almost the same output to the same input after retraining. The mean dice scores for enhancing tumor, whole tumor and tumor core are 77.2%, 89.2% and 76.3%, respectively. The simple structure and reliable high processing quality of the proposed system will facilitate its implementation and medical applications.