For the WMT26 General Machine Translation Task (Unconstrained Track), we propose TRIVE (Tri-Fusion Verified Translation Agent), a model-agnostic multi-agent translation framework. TRIVE organizes translation as a closed loop of Generation–Fusion–Verification–Correction: it first generates candidates in parallel under three complementary strategies—Faithful, Fluent, and Free—then synthesizes a draft that balances semantic fidelity and natural expression, and finally iteratively refines the output via rule-based validation and large language model evaluation. Because each module is implemented primarily via prompting, TRIVE can be directly overlaid onto diverse backbone models without task-specific fine-tuning. Experiments show that overlaying TRIVE on multiple backbones consistently improves instruction-following translation quality. Finally, we submitted Gemini-3.5-Flash + TRIVE as our primary system to WMT26.

We present Qingqiu-MT-9B, an instruction-following multilingual translation model developed for the constrained track of the WMT26 General Machine Translation shared task. Our model supports translation among 20 languages and covers 22 of the 23 official language pairs. It uses a source-grounded synthesis pipeline with separate LLM-based steps for generating instance-specific translation instructions and their corresponding translations. The resulting corpus covers diverse translation requirements, task complexity, instruction formulations, domains, document lengths, and glossary constraints, reducing reliance on fixed prompt templates and supporting generalization to unseen translation instructions. Our model is based on Qwen3.5-9B and trained through two-stage full-parameter supervised fine-tuning (SFT) followed by reinforcement learning (RL). The first SFT stage performs broad adaptation on a large-scale, medium-quality corpus, while the second refines the model on a smaller, higher-quality set. The RL stage further optimizes the model with Group Relative Policy Optimization (GRPO). During development, we use an LLM-based, rubric-guided contrastive method that scores candidates by dimension and aggregates an overall score to assess translation quality. Experiments show that Qingqiu-MT-9B demonstrates strong generalization and competitiveness among similarly sized models.We release the model at https://huggingface.co/WPS-Qingqiu/Qingqiu-MT-9B.

Empathetic response generation endeavors to empower dialogue systems to perceive speakers’ emotions and generate empathetic responses accordingly. Psychological research demonstrates that emotion, as an essential factor in empathy, encompasses trait emotions, which are static and context-independent, and state emotions, which are dynamic and context-dependent. However, previous studies treat them in isolation, leading to insufficient emotional perception of the context, and subsequently, less effective empathetic expression. To address this problem, we propose Combining Trait and State emotions for Empathetic Response Model (CTSM). Specifically, to sufficiently perceive emotions in dialogue, we first construct and encode trait and state emotion embeddings, and then we further enhance emotional perception capability through an emotion guidance module that guides emotion representation. In addition, we propose a cross-contrastive learning decoder to enhance the model’s empathetic expression capability by aligning trait and state emotions between generated responses and contexts. Both automatic and manual evaluation results demonstrate that CTSM outperforms state-of-the-art baselines and can generate more empathetic responses. Our code is available at https://github.com/wangyufeng-empty/CTSM

Textual backdoor attacks pose significant security threats. Current detection approaches, typically relying on intermediate feature representation or reconstructing potential triggers, are task-specific and less effective beyond sentence classification, struggling with tasks like question answering and named entity recognition. We introduce TABDet (Task-Agnostic Backdoor Detector), a pioneering task-agnostic method for backdoor detection. TABDet leverages final layer logits combined with an efficient pooling technique, enabling unified logit representation across three prominent NLP tasks. TABDet can jointly learn from diverse task-specific models, demonstrating superior detection efficacy over traditional task-specific methods.

Empathetic response generation aims to comprehend the cognitive and emotional states in dialogue utterances and generate proper responses. Psychological theories posit that comprehending emotional and cognitive states necessitates iteratively capturing and understanding associated words across dialogue utterances. However, existing approaches regard dialogue utterances as either a long sequence or independent utterances for comprehension, which are prone to overlook the associated words between them. To address this issue, we propose an Iterative Associative Memory Model (IAMM) for empathetic response generation. Specifically, we employ a novel second-order interaction attention mechanism to iteratively capture vital associated words between dialogue utterances and situations, dialogue history, and a memory module (for storing associated words), thereby accurately and nuancedly comprehending the utterances.We conduct experiments on the Empathetic-Dialogue dataset. Both automatic and human evaluations validate the efficacy of the model. Variant experiments on LLMs also demonstrate that attending to associated words improves empathetic comprehension and expression.

Recent studies have revealed that Backdoor Attacks can threaten the safety of natural language processing (NLP) models. Investigating the strategies of backdoor attacks will help to understand the model’s vulnerability. Most existing textual backdoor attacks focus on generating stealthy triggers or modifying model weights. In this paper, we directly target the interior structure of neural networks and the backdoor mechanism. We propose a novel Trojan Attention Loss (TAL), which enhances the Trojan behavior by directly manipulating the attention patterns. Our loss can be applied to different attacking methods to boost their attack efficacy in terms of attack successful rates and poisoning rates. It applies to not only traditional dirty-label attacks, but also the more challenging clean-label attacks. We validate our method on different backbone models (BERT, RoBERTa, and DistilBERT) and various tasks (Sentiment Analysis, Toxic Detection, and Topic Classification).

Efforts have been made to apply topic seed words to improve the topic interpretability of topic models. However, due to the semantic diversity of natural language, supervisions from seed words could be ambiguous, making it hard to be incorporated into the current neural topic models. In this paper, we propose SeededNTM, a neural topic model enhanced with supervisions from seed words on both word and document levels. We introduce a context-dependency assumption to alleviate the ambiguities with context document information, and an auto-adaptation mechanism to automatically balance between multi-level information. Moreover, an intra-sample consistency regularizer is proposed to deal with noisy supervisions via encouraging perturbation and semantic consistency. Extensive experiments on multiple datasets show that SeededNTM can derive semantically meaningful topics and outperforms the state-of-the-art seeded topic models in terms of topic quality and classification accuracy.