Guodong Du, Zitao Fang, Jing Li, Junlin Li, Runhua Jiang, Shuyang Yu, Yifei Guo, Yangneng Chen, Sim Kuan Goh, Ho-Kin Tang, Daojing He, Honghai Liu, Min Zhang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Junlin Li, Guodong Du, Jing Li, Sim Kuan Goh, Wenya Wang, Yequan Wang, Fangming Liu, Ho-Kin Tang, Saleh Alharbi, Daojing He, Min Zhang. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Fine-tuning pre-trained models on targeted datasets enhances task-specific performance but often comes at the expense of generalization.Model merging techniques, which integrate multiple fine-tuned models into a single multi-task model through task arithmetic, offer a promising solution.However, task interference remains a fundamental challenge, leading to performance degradation and suboptimal merged models.Existing approaches largely overlooked the fundamental roles of neurons, their connectivity, and activation, resulting in a merging process and a merged model that does not consider how neurons relay and process information.In this work, we present the first study that relies on neuronal mechanisms for model merging.Specifically, we decomposed task-specific representations into two complementary neuronal subspaces that regulate input sensitivity and task adaptability.Leveraging this decomposition, we introduced NeuroMerging, a novel merging framework developed to mitigate task interference within neuronal subspaces, enabling training-free model fusion across diverse tasks.Through extensive experiments, we demonstrated that NeuroMerging achieved superior performance compared to existing methods on multi-task benchmarks across both natural language and vision domains.Our findings highlighted the importance of aligning neuronal mechanisms in model merging, offering new insights into mitigating task interference and improving knowledge fusion.Our project is available at https://ZzzitaoFang. github.io/projects/NeuroMerging/.

Electroencephalography (EEG) has wide-ranging applications, from clinical diagnosis to brain-computer interfaces (BCIs). With the increasing volume and variety of EEG data, there has been growing interest in establishing foundation models (FMs) to scale up and generalize neural decoding. Despite showing early potential, applying FMs to EEG remains challenging due to substantial inter-subject, inter-task, and inter-condition variability, as well as diverse electrode configurations across recording setups. To tackle these open challenges, we propose **NeurIPT**, a foundation model tailored for diverse EEG-based **Neur**al **I**nterfaces with a **P**re-trained **T**ransformer by capturing both homogeneous and heterogeneous spatio-temporal characteristics inherent in EEG signals. Temporally, we introduce Amplitude-Aware Masked Pretraining (AAMP), masking based on signal amplitude rather than random intervals, to learn robust representations across varying signal intensities beyond local interpolation. Moreover, this temporal representation is enhanced by a progressive Mixture-of-Experts (MoE) architecture, where specialized expert subnetworks are progressively introduced at deeper layers, adapting effectively to the diverse temporal characteristics of EEG signals. Spatially, NeurIPT leverages the 3D physical coordinates of electrodes, enabling effective transfer across varying EEG settings, and develops Intra-Inter Lobe Pooling (IILP) during fine-tuning to efficiently exploit regional brain features. Empirical evaluations across nine downstream BCI datasets, via fine-tuning and training from scratch, demonstrated NeurIPT consistently achieved state-of-the-art performance, highlighting its broad applicability and robust generalization. Our work pushes forward the state of FMs in EEG and offers insights into scalable and generalizable neural information processing systems.

Spiking neural networks (SNNs) have gained prominence for their potential in neuromorphic computing and energy-efficient artificial intelligence, yet optimizing them remains a formidable challenge for gradient-based methods due to their discrete, spike-based computation. This paper attempts to tackle the challenges by introducing Cosine Annealing Differential Evolution (CADE), designed to modulate the mutation factor (F) and crossover rate (CR) of differential evolution (DE) for the SNN model, i.e., Spiking Element Wise (SEW) ResNet. Extensive empirical evaluations were conducted to analyze CADE. CADE showed a balance in exploring and exploiting the search space, resulting in accelerated convergence and improved accuracy compared to existing gradient-based and DE-based methods. Moreover, an initialization method based on a transfer learning setting was developed, pretraining on a source dataset (i.e., CIFAR-10) and fine-tuning the target dataset (i.e., CIFAR-100), to improve population diversity. It was found to further enhance CADE for SNN. Remarkably, CADE elevates the performance of the highest accuracy SEW model by an additional 0.52 percentage points, underscoring its effectiveness in fine-tuning and enhancing SNNs. These findings emphasize the pivotal role of a scheduler for F and CR adjustment, especially for DE-based SNN. Source Code on Github: https://github.com/Tank-Jiang/CADE4SNN.

While fine-tuning pretrained models has become common practice, these models often underperform outside their specific domains. Recently developed model merging techniques enable the direct integration of multiple models, each fine-tuned for distinct tasks, into a single model. This strategy promotes multitasking capabilities without requiring retraining on the original datasets. However, existing methods fall short in addressing potential conflicts and complex correlations between tasks, especially in parameter-level adjustments, posing a challenge in effectively balancing parameter competition across various tasks. This paper introduces an innovative technique named **PCB-Merging** (Parameter Competition Balancing), a *lightweight* and *training-free* technique that adjusts the coefficients of each parameter for effective model merging. PCB-Merging employs intra-balancing to gauge parameter significance within individual tasks and inter-balancing to assess parameter similarities across different tasks. Parameters with low importance scores are dropped, and the remaining ones are rescaled to form the final merged model. We assessed our approach in diverse merging scenarios, including cross-task, cross-domain, and cross-training configurations, as well as out-of-domain generalization. The experimental results reveal that our approach achieves substantial performance enhancements across multiple modalities, domains, model sizes, number of tasks, fine-tuning forms, and large language models, outperforming existing model merging methods.

Darwinian evolution of the biological brain is documented through multiple lines of evidence, although the modes of evolutionary changes remain unclear. Drawing inspiration from the evolved neural systems (e.g., visual cortex), deep learning models have demonstrated superior performance in visual tasks, among others. While the success of training deep neural networks has been relying on back-propagation (BP) and its variants to learn representations from data, BP does not incorporate the evolutionary processes that govern biological neural systems. This work proposes a neural network optimization framework based on evolutionary theory. Specifically, BP-trained deep neural networks for visual recognition tasks obtained from the ending epochs are considered the primordial ancestors (initial population). Subsequently, the population evolved with differential evolution. Extensive experiments are carried out to examine the relationships between Darwinian evolution and neural network optimization, including the correspondence between datasets, environment, models, and living species. The empirical results show that the proposed framework has positive impacts on the network, with reduced over-fitting and an order of magnitude lower time complexity compared to BP. Moreover, the experiments show that the proposed framework performs well on deep neural networks and big datasets.

Neural architecture search (NAS) automates neural network design by using optimization algorithms to navigate architecture spaces, reducing the burden of manual architecture design. While NAS has achieved success, applying it to emerging domains, such as analyzing unstructured 3D point clouds, remains underexplored due to the data lying in non-Euclidean spaces, unlike images. This paper presents Success-History-based Self-adaptive Differential Evolution with a Joint Point Interaction Dimension Search (SHSADE-PIDS), an evolution-ary NAS framework that encodes discrete deep neural network architectures to continuous spaces and performs searches in the continuous spaces for efficient point cloud neural architectures. Comprehensive experiments on challenging 3D segmentation and classification benchmarks demonstrate SHSADE-PIDS's capabilities. It discovered highly efficient architectures with higher accuracy, significantly advancing prior NAS techniques. For segmentation on SemanticKITTI, SHSADE-PIDS attained 64.51% mean IoU using only 0.55M parameters and 4.5GMACs, reducing overhead by over 22-26X versus other top methods. For ModelNet40 classification, it achieved 93.4% accuracy with just 1.31M parameters, surpassing larger models. SHSADE-PIDS provided valuable insights into bridging evolutionary algorithms with neural architecture optimization, particularly for emerging frontiers like point cloud learning.

Summary Unsupervised person re‐identification based on video sequences can be applied to surveillance systems and is attracting much more attention. It aims to spot specific person in other scenes captured by different cameras. This work explores an innovative strategy, namely, learning to cluster unlabeled person in the videos through graph convolutional networks. In this article, we find that the possibility of inter‐frame linkage can be inferred from context. Therefore, a pose‐guided topology linkage clustering framework is proposed. Our framework consists of three modules: (i) a pose‐guided representation module; (ii) a pose‐guided embedding module; (iii) a link prediction module. First, the representation coding alone is performed at the level of relational induction bias, embedding the implicit pose structure information in image features. Then, based on the consideration of the topology relationship between adjacent and cross‐frame, graph convolutional network is introduced to infer the likelihood of linkage between frame nodes. Experiments show that the proposed method demonstrates excellent scalability in addition to being an effective response to person clustering in case of changes, and does not need the number of clusters as a prior.

Summary Hospital readmission prediction is defined as an evaluation task to model the historical medical data to predict whether patients will be readmitted after discharge. In the past few years, many feasible and effective prediction methods have been proposed, however, most of them neglect the imbalanced distribution of medical data, which causes great difficulties in modeling. Thus, we proposed a new hospital readmission prediction method, which utilizes hybrid‐sampling and self‐paced balance learning strategies to solve the class‐imbalance problem. To be specifically, we first employ an interference negative sample deletion strategy to reduce the probability of important majority class samples being deleted. Then, we design a hard positive sample generation strategy to generate more positive samples. Meanwhile, we also introduce a self‐paced balance factor during the oversampling process to improve the similarity between newly generated minority class samples and hard positive samples. Finally, we perform the experiments on six real‐world readmission datasets to indicate the superiority of the proposed method.

Continual learning (CL) desires that the neural network sequentially perform learning tasks from a dynamic data stream without forgetting learned knowledge. To overcome forgetting, a line of work relies on gradient projection to minimize the influence between gradients during optimization. This article focuses on a challenging problem concerning CL:When, how, and where to implement gradient projection to promote CL.Tackling this problem can be divided into two perspectives, namely the gradient direction (when and how) and the area of gradient conflict (where). First, we propose a plug-and-play method UniGrad to tackle the inconsistency of conflicting and nonconflicting gradients during optimization in CL. Second, we explore the interaction mechanism of gradient projection and loss landscape in CL, and further propose a pluggable method UniGrad-FS to improve the CL performance. In short, this work expects to overcome forgetting through an efficient gradient projection at the area where the gradient conflicts are less intense. In essence, the proposed method is a general and pluggable method that can be used in any gradient-based optimizer. For evaluation, we plug UniGrad and UniGrad-FS into two top-performing baselines (WA and MEMO). Our method shows clear improvements, i.e., boosting WA and MEMO by +2.09% and 1.72% in the 20-step of the CIFAR100 benchmark. In addition, we observe performance enhancement on all settings of CIFAR100 and Tiny-ImageNet datasets. Extensive experiments demonstrate the simplicity and effectiveness of the proposed method.

Class imbalance is a common issue in the community of machine learning and data mining. The class-imbalance distribution can make most classical classification algorithms neglect the significance of the minority class and tend toward the majority class. In this article, we propose a label enhancement method to solve the class-imbalance problem in a graph manner, which estimates the numerical label and trains the inductive model simultaneously. It gives a new perspective on the class-imbalance learning based on the numerical label rather than the original logical label. We also present an iterative optimization algorithm and analyze the computation complexity and its convergence. To demonstrate the superiority of the proposed method, several single-label and multilabel datasets are applied in the experiments. The experimental results show that the proposed method achieves a promising performance and outperforms some state-of-the-art single-label and multilabel class-imbalance learning methods.

Summary In recent years, multilabel learning has been applied to a lot of application areas and is yet a challenging task. In multilabel learning, an instance often belongs to multiple class labels simultaneously. The labels usually have correlations with others, and mining label correlations is helpful to enhance the multilabel classification performance. Aiming at increasing the accuracy of prediction, Label embedding (LE) is an important technique, and conducive to extracting label information for multilabel learning. In this paper, we present a novel multilabel learning approach via exploiting label correlations, which can be naturally extended to tackle feature selection problem. First, to obtain the discriminative features shared by all labels, the proposed algorithm learns a latent space by employing deep canonical correlation analysis. Then we exploit label correlations by enforcing predictions on similar labels to be similar, thereby improving the prediction performance. Results on several multiple datasets illustrate that the proposed algorithm has the advantages on multilabel classification and feature selection.