In multi-task mixture-of-experts parameter-efficient fine-tuning, fixed-rank Low-Rank Adaptation (LoRA) adapters often struggle to accommodate the heterogeneous representation requirements across different tasks and network layers. Existing methods mainly focus on task-level expert selection, layer-wise rank allocation, or post-training parameter compression, making it difficult to jointly characterize low-rank capacity differences across tasks, layers, and modules within a unified framework. To address this issue, we propose Group-Sparse Routing-aware Low-Rank Adaptation (GSR-LoRA), a routing-aware structured rank compression method for multi-task parameter-efficient fine-tuning (PEFT). Specifically, we model low-rank components as selectable rank groups and use a task-dependent router to allocate rank capacity across tasks. Furthermore, we introduce a Group Lasso structured sparsity constraint to suppress redundant rank groups during training, forming a prunable adapter structure. During compression, we select rank groups by jointly considering parameter importance and routing usage signals, and map the router to the compressed rank space to obtain compact LoRA adapters that remain consistent with task routing. Experiments on GLUE show that GSR-LoRA reduces the average adapter rank from 16 to 5.5, removing 65.6% of rank components with only a 0.1-point performance degradation. These results demonstrate its ability to learn compact task-adaptive LoRA adapters while preserving multi-task performance. The code is available at https://github.com/HNLP93/GSR-LoRA

“In recent years, dialogue summarization has emerged as a rapidly growing area of research in natural language processing. Dialogue summarization is challenging due to dispersed key information, redundant expressions, ambiguous topic identification, and difficult content selection.To address these challenges, we propose an innovative approach to dialogue summarization that integrates topic segmentation and graph-structured modeling. Specifically, we first per-form topic segmentation of the dialogue through clustering and quantify the key information in each utterance, thereby capturing the dialogue topics more effectively. Then, a redundancy graph and a keyword graph are constructed to suppress redundant information and extract key content, thereby enhancing the conciseness and coherence of the summary. Evaluations were conducted on the DialogSum, SAMSum, CSDS, and NaturalConv datasets. The experimental results demonstrate that the proposed method significantly outperforms existing benchmark mod-els in terms of summary accuracy and information coverage. The Rouge-1 scores achieved were 48.03%, 53.75%, 60.78%, and 81.48%, respectively, validating its effectiveness in the dialogue summarization task. Our code is available at https://anonymous.4open.science/r/TAG-E64A.”

Nowadays, neural networks play an important role in the task of relation classification. In this paper, we propose a novel attention-based convolutional neural network architecture for this task. Our model makes full use of word embedding, part-of-speech tag embedding and position embedding information. Word level attention mechanism is able to better determine which parts of the sentence are most influential with respect to the two entities of interest. This architecture enables learning some important features from task-specific labeled data, forgoing the need for external knowledge such as explicit dependency structures. Experiments on the SemEval-2010 Task 8 benchmark dataset show that our model achieves better performances than several state-of-the-art neural network models and can achieve a competitive performance just with minimal feature engineering.