This paper presents Xiaoyu-MT, our submission system for the WMT 2026 Chinese-Southeast Asian Multilingual MT Task, focusing on bidirectional translation between Chinese and seven Southeast Asian languages. To address limited parallel resources and imbalanced multilingual data distributions, we select Gemma-4-12B-it as the base model and construct multilingual training data by combining official WMT resources with additional monolingual and bilingual corpora. Xiaoyu-MT adopts a multi-stage training pipeline consisting of Continual Pre-Training (CPT), Supervised Fine-Tuning (SFT), and Direct Preference Optimization (DPO). During SFT, we introduce a language-aware Prompt encoding mechanism to explicitly model source language, target language, and their directed translation relationships through structured Prompt representations and Prompt compression. DPO further optimizes translation preferences using preference data constructed from reference and model-generated translations. Official evaluation results on WMT 2026 show that Xiaoyu-MT achieves an overall BLEU score of 34.73, a COMET score of 81.87, and a Final Result score of 61.80, demonstrating its effectiveness for multilingual and low-resource machine translation scenarios.

Large Language Models (LLMs) exhibit strong language understanding and instruction-following abilities, making them promising backbones for grammatical error correction (GEC). However, prompt-only correction for low-resource languages still suffers from the over-correction challenge, since LLMs cannot fully understand their internal grammar. To address this issue, we propose an automatic linguistic neuron probing and optimizing approach to improve the low-resource GEC with only a few neuron activations. Concretely, we first utilize multiple low-resource GEC corpora to observe the changes in neurons, thus obtaining a linguistically specific probe matrix to ensure whether the neuron is activated. Then, we construct an extra weight matrix to determine the activation intensity of each neuron. Finally, we combine the probe matrix and the weight matrix to obtain an activation matrix, which is injected into selected model layers during inference to guide the frozen LLM toward more controllable GEC. Experiments on four benchmark datasets show that our approach consistently improves GEC performance, indicating that lightweight sparse activation rescaling is extremely useful for low-resource grammatical knowledge understanding. Detailed analysis demonstrates that our approach can further provide interpretable neuron-level control without large-scale parameter fine-tuning.

Document-level event causality identification (DECI) requires identifying causal relations across a document, where inter-sentence causality is often implicit and supported by weak evidence. Existing methods typically construct event graphs from pairwise predictions, which may overlook weak but valid causal edges, and thus hinder causal message passing. To address these issues, we propose ICGR, an Iterative Causal Graph Refinement framework that iteratively refines event graphs by preserving low-confidence edges and performing message passing on the refined graph to produce a more complete and reliable structure. In each refinement round, the Constrained Edge Completion module identifies missing causal edges, and the Conditional Diffusion module refines low-confidence edges through denoising. Experiments on EventStoryLine and MAVEN-ERE show that ICGR achieves state-of-the-art performance, with notable improvements on inter-sentence causal relations.

Reinforcement learning (RL) excels in reasoning tasks with verifiable rewards, while its adaptation to machine translation (MT) remains challenging due to the lack of unique reward signals under multiple valid translations. Existing RL approaches for MT face either fixed references in supervised settings or the production of homogeneous references leading to mode collapse in unsupervised settings. Both limitations arise from ignoring entropy dynamics in RL-based MT. The core challenge is leveraging entropy for supervision construction and self-evolution. In this paper, we propose an Entropy-Driven Unsupervised RL for MT. Our framework integrates entropy-guided sampling for exploration, confidence-weighted label generation to transcend majority-voting bias, and uncertainty-aware optimization to prioritize high-entropy tokens. These mechanisms allow reward signals to co-evolve with model proficiency beyond fixed references. Experiments across multiple language pairs show our method outperforms supervised and unsupervised baselines by +0.63 and +2.52 average points, respectively. Our code is available at https://github.com/fortunatekiss/URLMT.

Large Language Models (LLMs) enable a single model to support translation across multiple language directions simultaneously, eliminating the need to train a separate model for each direction and substantially improving the performance of multilingual machine translation. However, during multilingual translation instruction tuning, parameter interference among different languages significantly limits the improvement of translation performance and may even cause performance degradation for some languages. To address this issue, we propose MeRP, a multi-expert routing prompt encoding strategy for parameter-efficient fine-tuning (PEFT) in multilingual machine translation tasks. This method encodes representations of source languages through an attention mechanism and dynamically selects expert groups according to translation directions, enabling both the sharing and disentanglement of multilingual semantic representations. As a result, it preserves LLM generalization while reducing training time and improving translation performance. We conduct experiments on five language pairs. The results show that, compared with the baseline models, our method updates only 0.06% of the total model parameters during training while achieving better multilingual translation performance. Notably, our MeRP reduces training time by 46.75% and achieves average gains of +0.62 BLEU points and +0.34 COMET-22 points over the arithmetic mean of the four PEFT baselines.Code is available at https://github.com/wangzhenhan93/MeRP.

Existing research on news summarization primarily focuses on single-language single-document (SLSD), single-language multi-document (SLMD) or cross-language single-document (CLSD). However, in real-world scenarios, news about an international event often involves multiple documents in different languages, i.e., mixed-language multi-document (MLMD). Therefore, summarizing MLMD news is of great significance. However, the lack of datasets for MLMD news summarization has constrained the development of research in this area. To fill this gap, we construct a mixed-language multi-document news summarization dataset (MLMD-news), which contains four different languages and 10,992 source document cluster and target summary pairs. Additionally, we propose a graph-based extract-generate model and benchmark various methods on the MLMD-news dataset and publicly release our dataset and code, aiming to advance research in summarization within MLMD scenarios.

Large language models (LLMs) based Multilingual Knowledge Graph Completion (MKGC) aim to predict missing facts by leveraging LLMs’ multilingual understanding capabilities, improving the completeness of multilingual knowledge graphs (KGs).However, existing MKGC research underutilizes the multilingual capabilities of LLMs and ignores the shareability of cross-lingual knowledge.In this paper, we propose a novel MKGC framework that leverages multilingual shared knowledge to significantly enhance performance through two components: Knowledge-level Grouped Mixture of Experts (KL-GMoE) and Iterative Entity Reranking (IER).KL-GMoE efficiently models shared knowledge, while IER significantly enhances its utilization.To evaluate our framework, we constructed a mKG dataset containing 5 languages and conducted comprehensive comparative experiments with existing state-of-the-art (SOTA) MKGC method.The experimental results demonstrate that our framework achieves improvements of 5.47%, 3.27%, and 1.01% in the Hits@1, Hits@3, and Hits@10 metrics, respectively, compared with SOTA MKGC method.Further experimental analysis revealed the properties of knowledge sharing in settings of unseen and unbalanced languages.We have released the dataset and code for our work on https://github.com/gaoxiaofei07/KL-GMoE.

Generative Information Retrieval is an emerging retrieval paradigm that exhibits remarkable performance in monolingual scenarios. However, applying these methods to multilingual retrieval still encounters two primary challenges, cross-lingual identifier misalignment and identifier inflation. To address these limitations, we propose Multilingual Generative Retrieval via Cross-lingual Semantic Compression (MGR-CSC), a novel framework that unifies semantically equivalent multilingual keywords into shared atoms to align semantics and compresses the identifier space, and we propose a dynamic multi-step constrained decoding strategy during retrieval. MGR-CSC improves cross-lingual alignment by assigning consistent identifiers and enhances decoding efficiency by reducing redundancy. Experiments demonstrate that MGR-CSC achieves outstanding retrieval accuracy, improving by 6.83% on mMarco100k and 4.77% on mNQ320k, while reducing document identifiers length by 74.51% and 78.2%, respectively. We publicly release our dataset and code at https://github.com/simengggg/MGR-CSC

Large language models (LLMs) demonstrate remarkable text generation and syntax parsing capabilities in high-resource languages. However, their performance notably declines in low-resource languages due to memory forgetting stemming from semantic interference across languages. To address this issue, we propose a novel deep hierarchical syntax understanding approach to improve the cross-lingual semantic memory capability of LLMs. First, we design a multi-task joint fine-tuning strategy to implicitly align linguistic knowledge between source and target languages in LLMs, which is leveraged to initially parse the target text. Second, we automatically construct the multilingual dependency label banks based on the statistical structure information from the Universal Dependencies (UD) data. Third, we obtain each label’s memory strength via in-depth analysis of the initial parsing tree and its dependency label bank. Finally, memory strength is further exploited to guide LLMs to learn the linguistic commonalities from multilingual dependency label banks, thus activating the memory ability of weak labels. Experimental results on four benchmark datasets show that our method can dramatically improve the parsing accuracy of all baseline models, leading to new state-of-the-art results. Further analysis reveals that our approach can effectively enhance the weak syntactic label memory cognition of LLMs by combining the advantages of both implicit multi-task fine-tuning and explicit label bank guiding. Our code and dependency label banks are released at https://github.com/Flamelunar/memory_dep.

Sentence representation learning (SRL) aims to learn sentence embeddings that conform to the semantic information of sentences. In recent years, fine-tuning methods based on pre-trained models and contrastive learning frameworks have significantly advanced the quality of sentence representations. However, within the semantic space of SRL models, both word embeddings and sentence representations derived from word embeddings exhibit substantial redundant information, which can adversely affect the precision of sentence representations. Existing approaches predominantly optimize training strategies to alleviate the redundancy problem, lacking fine-grained guidance on reducing redundant representations. This paper proposes a novel approach that dynamically identifies and reduces redundant information from a dimensional perspective, training the SRL model to redistribute semantics on different dimensions, and entailing better sentence representations. Extensive experiments across seven semantic text similarity benchmarks demonstrate the effectiveness and generality of the proposed method. A comprehensive analysis of the experimental results is conducted, and the code/data will be released.

With the strong representational capabilities of pre-trained language models, dependency parsing in resource-rich languages has seen significant advancements. However, the parsing accuracy drops sharply when the model is transferred to low-resource language due to distribution shifts. To alleviate this issue, we propose a representation alignment and adversarial model to filter out useful knowledge from rich-resource language and ignore useless ones. Our proposed model consists of two components, i.e., an alignment network in the input layer for selecting useful language-specific features and an adversarial network in the encoder layer for augmenting the language-invariant contextualized features. Experiments on the benchmark datasets show that our proposed model outperforms RoBERTa-enhanced strong baseline models by 1.37 LAS and 1.34 UAS. Detailed analysis shows that both alignment and adversarial networks are equally important in alleviating the distribution shifts problem and can complement each other. In addition, the comparative experiments demonstrate that both the alignment and adversarial networks can substantially facilitate extracting and utilizing relevant target language features, thereby increasing the adaptation capability of our proposed model.

Large language models (LLMs) have demonstrated remarkable capabilities in comprehensively handling various types of natural language processing (NLP) tasks. However, there are significant differences in the knowledge and abilities required for different tasks. Therefore, it is important to understand whether the same LLM processes different tasks in the same way. Are there specific neurons in a LLM for different tasks? Inspired by neuroscience, this paper pioneers the exploration of whether distinct neurons are activated when a LLM handles different tasks. Compared with current research exploring the neurons of language and knowledge, task-specific neurons present a greater challenge due to their abstractness, diversity, and complexity. To address these challenges, this paper proposes a method for task-specific neuron localization based on Causal Gradient Variation with Special Tokens (CGVST). CGVST identifies task-specific neurons by concentrating on the most significant tokens during task processing, thereby eliminating redundant tokens and minimizing interference from non-essential neurons. Compared to traditional neuron localization methods, our approach can more effectively identify task-specific neurons. We conduct experiments across eight different public tasks. Experiments involving the inhibition and amplification of identified neurons demonstrate that our method can accurately locate task-specific neurons.

Multilingual Knowledge Graph Completion (mKGC) aim at solving queries in different languages by reasoning a tail entity thus improving multilingual knowledge graphs. Previous studies leverage multilingual pretrained language models (PLMs) and the generative paradigm to achieve mKGC. Although multilingual pretrained language models contain extensive knowledge of different languages, its pretraining tasks cannot be directly aligned with the mKGC tasks. Moreover, the majority of KGs and PLMs currently available exhibit a pronounced English-centric bias. This makes it difficult for mKGC to achieve good results, particularly in the context of low-resource languages. To overcome previous problems, this paper introduces global and local knowledge constraints for mKGC. The former is used to constrain the reasoning of answer entities , while the latter is used to enhance the representation of query contexts. The proposed method makes the pretrained model better adapt to the mKGC task. Experimental results on public datasets demonstrate that our method outperforms the previous SOTA on Hits@1 and Hits@10 by an average of 12.32% and 16.03%, which indicates that our proposed method has significant enhancement on mKGC.