Yan Xiang

AAAI Conference 2026 Conference Paper

Consensus-Aligned Neuron Efficient Fine-Tuning Large Language Models for Multi-Domain Machine Translation

• Shuting Jiang
• Ran Song
• Yuxin Huang
• Yan Xiang
• Yantuan Xian
• Shengxiang Gao
• Zhengtao Yu

Multi-domain machine translation (MDMT) aims to build a unified model capable of translating content across diverse domains. Despite the impressive machine translation capabilities demonstrated by large language models (LLMs), domain adaptation still remains a challenge for LLMs. Existing MDMT methods such as in-context learning and parameter-efficient fine-tuning often suffer from domain shift, parameter interference and limited generalization. In this work, we propose a neuron-efficient fine-tuning framework for MDMT that identifies and updates consensus-aligned neurons within LLMs. These neurons are selected by maximizing the mutual information between neuron behavior and domain features, enabling LLMs to capture both generalizable translation patterns and domain-specific nuances. Our method then fine-tunes LLMs guided by these neurons, effectively mitigating parameter interference and domain-specific overfitting. Comprehensive experiments on three LLMs across ten German-English and Chinese-English translation domains evidence that our method consistently outperforms strong PEFT baselines on both seen and unseen domains, achieving state-of-the-art performance.

EAAI Journal 2025 Journal Article

Corrigendum to “Image–text sentiment analysis based on hierarchical interaction fusion and contrast learning enhanced” [Eng. Appl. Artific. Intellig. 146 (2025) 110262]

• Hongbin Wang
• Qifei Du
• Yan Xiang

EAAI Journal 2025 Journal Article

Image–text sentiment analysis based on hierarchical interaction fusion and contrast learning enhanced

• Hongbin Wang
• Qifei Du
• Yan Xiang

As social media continues to evolve, an increasing number of individuals prefer to publish a variety of content that combines different forms of expression on platforms to convey their feelings. In recent times, the study of sentiment within these diverse media formats has gained significant traction. However, some studies have neglected the multilayered interactions between modalities and underutilized the relationships between data samples and between classes. To address this problem, this paper proposes an image–text sentiment analysis method based on layer interaction fusion and contrast learning enhanced. First, the method uses a multi-layer cross-modal interaction module to emphasize the complementarity between modalities. The correlation information between modalities is deeply mined. Then, feature integration is performed by the multimodal fusion module. In addition, the model introduces a comparative learning task to exploit the relationship between samples and between classes. The emotional features under different classes and the key emotional features of the samples are mined. Finally, sentiment classification is performed to realize image–text sentiment analysis. In order to assess the efficacy of the approach, this paper performs extensive experimental validation on a multimodal sentiment dataset. Compared with several baseline models, the model in this paper achieves certain improvements. For example, on the multi-view sentiment analysis single dataset(MVSA-Single dataset), the F1 score is improved by 1. 90% compared to the baseline model.

EAAI Journal 2025 Journal Article

Simplified syntax-guided domain-shared representation learning for cross-domain aspect-based sentiment analysis

• Jiqun Zhang
• Yan Xiang

Cross-domain aspect-based sentiment analysis (ABSA) aims to identify aspect terms and their sentiment polarities in the target domain using annotated data from the source domain. Previous studies have shown that leveraging syntactic knowledge can partially bridge domain gaps. However, routine syntactic labels are diverse and fail to directly reflect the connection between aspects and opinions. This limitation hinders their ability to assist the model in perceiving aspects-sentiments in different domains. To overcome this challenge, we present a cross-domain ABSA approach based on simplified syntactic-guided representation learning. Initially, we design a dependency syntax simplification strategy that unifies syntactic labels of aspect terms and their corresponding opinion terms. We employ these simplified syntactic labels to guide the model's self-supervised learning process, thereby obtaining the domain-shared representations from unlabeled data in both the source and target domains. Using the domain-shared representation, we train the model on annotated data from the source domain and then apply it to make predictions in the target domain. Experimental results across ten transfer tasks using four public datasets demonstrate that our proposed method consistently outperforms other baseline models.

IROS Conference 2025 Conference Paper

THE-SEAN: A Heart Rate Variation-Inspired Temporally High-Order Event-Based Visual Odometry with Self-Supervised Spiking Event Accumulation Networks

• Chaoran Xiong
• Litao Wei
• Kehui Ma
• Zhen Sun
• Yan Xiang
• Zihan Nan
• Trieu-Kien Truong
• Ling Pei

Event-based visual odometry has recently gained attention for its high accuracy and real-time performance in fast-motion systems. Unlike traditional synchronous estimators that rely on constant-frequency (zero-order) triggers, event-based visual odometry can actively accumulate information to generate temporally high-order estimation triggers. However, existing methods primarily focus on adaptive event representation after estimation triggers, neglecting the decision-making process for efficient temporal triggering itself. This oversight leads to the computational redundancy and noise accumulation. In this paper, we introduce a temporally high-order event-based visual odometry with spiking event accumulation networks (THE-SEAN). To the best of our knowledge, it is the first event-based visual odometry capable of dynamically adjusting its estimation trigger decision in response to motion and environmental changes. Inspired by biological systems that regulate hormone secretion to modulate heart rate, a self-supervised spiking neural network is designed to generate estimation triggers. This spiking network extracts temporal features to produce triggers, with rewards based on block matching points and Fisher information matrix (FIM) trace acquired from the estimator itself. Finally, THE-SEAN is evaluated across several open datasets, thereby demonstrating average improvements of 13% in estimation accuracy, 9% in smoothness, and 38% in triggering efficiency compared to the state-of-the-art methods.