The homogeneity and heterogeneity across modalities are critical factors that influence multimodal fusion. In Multimodal Sentiment Analysis (MSA), the inherent textual information within the audio modality induces cross-modality homogeneity with the text modality. Conversely, the mutual independence between text and vision modalities results in their cross-modal heterogeneity. Although existing disentangle-based methods achieve notable performance gains by separating modality features into distinct subspaces, they overlook the characteristics of cross-modality heterogeneity and homogeneity among different modalities. To this end, we propose a novel Modality-aware Disentangle and Fusion (MDF) framework to investigate the role of core modality features. Specifically, we first use text as the anchor to disentangle the audio modality and extract its unique modality-specific features, thereby establishing cross-modal heterogeneity among text, audio, and vision. We then introduce a Cross-Modality Heterogeneity Enhancement (CHE) module to refine these features, further reinforcing their heterogeneous characteristics. Finally, a Modality Adaptive Weighting (MAW) module is employed to dynamically assign weights to the text, sound, and vision modalities based on their potential contributions to sentiment prediction, achieving a more effective multimodal representation for MSA. Experimental evaluations on different benchmarks demonstrate MDF's superiority, with extensive ablation studies confirming its effectiveness.

Retrieval-augmented generation (RAG) enhances the reasoning capabilities of large language models (LLMs) by incorporating external knowledge. Among available sources, knowledge graphs (KGs) offer a structured and reliable foundation for factual information, making them increasingly popular in efforts to improve reasoning faithfulness in RAG. Most existing KG-based RAG methods rely on LLMs to extract knowledge from KGs. However, these approaches often require costly fine-tuning and struggle to navigate deep graph structures, limiting their effectiveness in multi-hop reasoning tasks. To address these challenges, we propose Stepwise Contrastive Reasoning (SCR), a lightweight framework that integrates graph structure and textual context for efficient and interpretable RAG over KGs. SCR combines relational message passing layers to encode KG entities with a Transformer encoder for processing question text. It decomposes reasoning into a series of alignment steps. At each step, SCR compares the current topic entity and its neighbors with the question representation, selecting the most relevant entity as the next topic entity. The question is then updated with this entity's textual description. This process continues until the selected entity no longer changes, indicating that the answer entity has been reached. Through stepwise alignment, SCR enables compact models to perform faithful and interpretable reasoning over large-scale KGs. Extensive experiments on several widely used KGQA benchmarks demonstrate that SCR not only achieves state-of-the-art performance but also effectively boosts the capabilities of smaller language models to match those of LLMs.

Recent research reveals that a minority of high-entropy tokens significantly influence the reasoning quality of large language models (LLMs). Inspired by this, we propose Prototype Entropy Alignment (PEA), a reinforcement learning framework that models effective reasoning not as a single path but as a collection of learnable "entropy signatures." PEA identifies these signatures by clustering expert trajectories' uncertainty patterns into a diverse and continuously updated set of prototypes. The model is then rewarded for aligning its own reasoning process with these evolving targets, creating a self-improvement loop. Instead of replacing traditional outcome-based rewards, PEA provides a complementary, process-oriented signal. Our experiments show that this synergy is crucial: PEA substantially boosts performance on creative and general reasoning tasks and, when combined with outcome rewards, achieves SOTA results on structured tasks such as mathematics. By rewarding alignment with diverse and evolving reasoning structures, PEA offers a robust, verifier-free pathway to enhance reasoning's adaptability.

This paper presents FDC-Ground, a reinforcement learning framework that addresses the high-cost, low-signal challenge of GUI grounding training. The framework introduces two core contributions: (1) the Exponentially Decayed Distance Reward (EDDR), which provides resolution-robust and continuous feedback for position predictions, and (2) the Fact-Aligned Dynamic Completions Pruning (FDC-Pruning) strategy, which selectively retains completions whose advantage signs align with factual correctness, thereby reducing computational overhead while enhancing gradient quality and training stability. Using only 3.2K training samples and a single epoch, our 7B-parameter model achieves 88.3% and 91.0% accuracy on ScreenSpot and ScreenSpot-v2, outperforming several RL-based models such as UIShift and SE-GUI. Our 3B-parameter model based on Qwen2.5-VL-3B surpasses its original performance by +26.6%, demonstrating the effectiveness of our reward design and pruning strategy under low-resource conditions. Furthermore, the proposed FDC-Pruning strategy achieves a 1.18× training speedup and a +5.9% accuracy improvement over standard GRPO, and expanding the exploration space to 4× yields an additional +10.5% gain, confirming both the scalability and the training efficiency of our approach. These findings highlight that combining EDDR with FDC-Pruning offers a practical path toward scalable and efficient RL-based GUI grounding, even in low-resource settings.

Guanran Luo, Wentao Qiu, Wanru Zhao, Wenhan Lv, Zhongquan Jian, Meihong Wang, Qingqiang Wu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Aspect Sentiment Quad Prediction (ASQP) is the most complex subtask of Aspect-based Sentiment Analysis (ABSA), aiming to predict all sentiment quadruples within the given sentence. Due to the complexity of sentence syntaxes and the diversity of sentiment expressions, generative methods gradually become the mainstream approach in ASQP. However, existing generative models are constrained in the effectiveness of demonstrations. Semantically similar demonstrations help in judging sentiment categories and polarities but may confuse the model in recognizing aspect and opinion terms, which are more related to sentence syntaxes. To this end, we first develop Syn2Vec, a method for calculating syntactic vectors to support the retrieval of syntactically similar demonstrations. Then, we propose Syntactic and Semantic Similarity Retrieval Prompting (SimRP) to construct effective prompts by retrieving the most related demonstrations that are syntactically and semantically similar. With these related demonstrations, pre-trained generative models, especially Large Language Models (LLMs), can fully release their potential to recognize sentiment quadruples. Extensive experiments in Supervised Fine-Tuning (SFT) and In-context Learning (ICL) paradigms demonstrate the effectiveness of SimRP. Furthermore, we find that LLMs' capabilities in ASQP are severely underestimated by biased data annotations and the exact matching metric. We propose a novel constituent subtree-based fuzzy metric for more accurate and rational quadruple recognition.

Retrieval-Augmented Generation (RAG) mitigates the hallucination issues of large language models (LLMs) by integrating external knowledge.For abstractive questions involving multistep reasoning, knowledge from multiple sections is often required.To address this issue, recent research has introduced recursive summarization, which constructs a hierarchical summary tree by clustering text chunks, integrating information from various parts of the document to provide evidence for abstractive questions.However, summary trees often contain a large number of redundant summary nodes, which not only increase construction time but may also negatively impact question answering.Moreover, recursive summarization is not suitable for all types of questions.We introduce DTCRS, a method that dynamically generates summary trees based on document structure and query semantics.DTCRS determines whether a summary tree is necessary by analyzing the question type.It then decomposes the question and uses the embeddings of subquestions as initial cluster centers, reducing redundant summaries while improving the relevance between summaries and the question.Our approach significantly reduces summary tree construction time and achieves substantial improvements across three QA tasks.Additionally, we investigate the applicability of recursive summarization to different question types, providing valuable insights for future research.

Prior studies on Aspect-level Sentiment Classification (ALSC) emphasize modeling interrelationships among aspects and contexts but overlook the crucial role of aspects themselves as essential domain knowledge. To this end, we propose AGCL, a novel Aspect Graph Construction and Learning method, aimed at furnishing the model with finely tuned aspect information to bolster its task-understanding ability. AGCL’s pivotal innovations reside in Aspect Graph Construction (AGC) and Aspect Graph Learning (AGL), where AGC harnesses intrinsic aspect connections to construct the domain aspect graph, and then AGL iteratively updates the introduced aspect graph to enhance its domain expertise, making it more suitable for the ALSC task. Hence, this domain aspect graph can serve as a bridge connecting unseen aspects with seen aspects, thereby enhancing the model’s generalization capability. Experiment results on three widely used datasets demonstrate the significance of aspect information for ALSC and highlight AGL’s superiority in aspect learning, surpassing state-of-the-art baselines greatly. Code is available at https://github.com/jian-projects/agcl.

Using mobile phones for photography and videography has become increasingly prevalent. However, novice users often struggle with mastering professional photography techniques and lack the necessary skills for precise camera movement. To address this issue, we propose ARVideoCam, a video shooting guidance method based on augmented reality (AR) designed to assist novice users in accurately controlling camera movements and learning photography skills through imitating high-quality video shooting. ARVideoCam consists of three layers: the extraction layer, the analysis layer, and the AR layer. The extraction layer of the model captures the subject's motion in the sample video, and the analysis layer converts the subject's motion into camera motion, which is then transformed into AR-based guidance in the AR layer. The user study results indicate that AR guidance significantly enhances novice users' ability to imitate sample videos across different types of camera motion, resulting in more efficient and precise camera movements during shooting. This study offers valuable inspiration for the research on AR-assisted video shooting and enhancing video shooting skills among novice users.

Recognizing emotions in dialogues is vital for effective human-computer interaction, yet remains a challenging task in Natural Language Processing (NLP). Previous studies in Emotion Recognition in Conversation (ERC) have primarily focused on contextual features, while overlooking the importance of emotional features in emotion recognition. To address this gap, we focus on the role of emotional features in ERC and propose a novel method, Emotional Knowledge Self-Distillation (EmoKSD1), to enhance the model’s emotional sensitivity. In EmoKSD, utterances are enriched with implicit ⟨mask⟩ tokens to represent conveyed emotions, allowing the distillation of emotional knowledge from explicit emotional tokens to implicit ⟨mask⟩ tokens, thereby enhancing the model’s ability to perceive subtle emotions within the dialogue. Through thorough evaluations on two public ERC datasets (i.e., IEMOCAP and MELD) using proposed coarse-grained utterance distillation and fine-grained token distillation techniques, EmoKSD demonstrates superior performance compared to existing methods, highlighting the significance of emotional features in ERC.

Pre-trained Language Models (PLMs) have achieved remarkable performance in various Natural Language Processing (NLP) tasks, including Aspect-based Sentiment Analysis (ABSA). Therefore, numerous ABSA models based on PLMs have been proposed, primarily focusing on module design to exploit the inherent connections between aspects and contexts. However, the core factor driving performance improvements, the PLM’s powerful semantic understanding capabilities, has not been fully considered, raising the question of how to further unlock their potential for downstream tasks. To this end, we introduce a novel training strategy, called CCL1, which integrates the strengths of Curriculum Learning (CurL) and Contrastive Learning (ConL) to facilitate the learning of robust feature representations. For the ABSA task, we use aspect similarities to develop the CurL strategy, grouping samples with similar aspects into batches. This allows ConL to learn more robust representations by providing related samples within each batch. The superiority of CCL is demonstrated through extensive experiments on two public ABSA datasets, with ablation studies validating the effectiveness of combining CurL and ConL in enhancing aspect understanding.

This research proposes the METORIE dataset1, a novel resource designed to improve the reasoning capabilities of large language models (LLMs), such as LLaMA3 and GLM4, in information extraction (IE) tasks. The METORIE dataset is derived from brain teasers that incorporate complex logical and metaphorical elements and is designed to train LLMs to navigate intricate reasoning paths and interpret layered expressions. Our findings demonstrate that the METORIE dataset markedly enhances LLMs’ performance across both general and specialized IE tasks. The results of fine-tuning with the METORIE dataset, mixed with a small number of IE datasets, are close to, if not exceeding, those of LLMs of the same parametric size on IE tasks using much larger datasets. Through controlled experiments, we establish that metaphors of medium complexity optimize IE performance, while higher complexities tend to overstretch LLMs’ inference limits. METORIE-fine-tuned LLMs also demonstrate exceptional performance in legal and medical domains, suggesting that enhanced metaphor understanding and logical deduction are key to improving LLMs’ adaptability and efficiency in vertical domains.

Enhancing the precision of segmenting coronary atherosclerotic plaques from CT Angiography (CTA) images is pivotal for advanced Coronary Atherosclerosis Analysis (CAA), which distinctively relies on the analysis of vessel cross-section images reconstructed via Curved Planar Reformation. This task presents significant challenges due to the indistinct boundaries and structures of plaques and blood vessels, leading to the inadequate performance of current deep learning models, compounded by the inherent difficulty in annotating such complex data. To address these issues, we propose a novel dual-consistency semi-supervised framework that integrates Intra-frame Topological Consistency (ITC) and Cross-frame Topological Consistency (CTC) to leverage labeled and unlabeled data. ITC employs a dual-task network for simultaneous segmentation mask and Skeleton-aware Distance Transform (SDT) prediction, achieving similar prediction of topology structure through consistency constraint without additional annotations. Meanwhile, CTC utilizes an unsupervised estimator for analyzing pixel flow between skeletons and boundaries of adjacent frames, ensuring spatial continuity. Experiments on two CTA datasets show that our method surpasses existing semi-supervised methods and approaches the performance of supervised methods on CAA. In addition, our method also performs better than other methods on the ACDC dataset, demonstrating its generalization.

Deep neural networks (DNNs) face substantial challenges in Long-Tail Visual Recognition (LTVR) due to the inherent class imbalances in real-world data distributions. The Mixture of Experts (MoE) framework has emerged as a promising approach to addressing these issues. However, in MoE systems, experts are typically trained to optimize a collective objective, often neglecting the individual optimality of each expert. This individual optimality usually contributes to the overall performance, as the goals of different experts are not mutually exclusive. We propose the Independent and Collaborative Learning (ICL) framework to optimize each expert independently while ensuring global optimality. First, Diverse Optimization Learning (DOL) is introduced to enhance expert diversity and individual performance. Then, we conceptualize experts as parallel circuit branches and introduce Competition and Collaboration Learning (CoL). Competition Learning amplifies the gradients of better-performing experts to preserve individual optimality, and Collaboration Learning encourages collaboration through mutual distillation to enhance optimal knowledge sharing. ICL achieves state-of-the-art accuracy in experiments on CIFAR-100/10-LT, ImageNet-LT, and iNaturalist 2018, respectively. Our code is available at https://github.com/PolarisLight/ICL.

Virtual exhibition is the exhibition whose venue is cyberspace. Providing personalized interactions to users in virtual exhibitions can significantly improve the user experience, however, this is challenging due to the complexity of receiving user information and interacting with users in a timely manner. We propose ExhibChat, a system that integrates role-playing agents based on large language models into virtual exhibitions, which allows the agents to interact with users correctly and strongly in real time. We confirm the feasibility and effectiveness of ExhibChat through user experiments, and the user experience process well evaluates ExhibChat. We summarize some considerations for using role-playing agents in virtual exhibitions and clarify the focus of future research and application.

In an emotional conversation, emotions are causally transmitted among communication participants, constituting a fundamental conversational feature that can facilitate the comprehension of intricate changes in emotional states during the conversation and contribute to neutralizing emotional semantic bias in utterance caused by the absence of modality information. Therefore, emotional transition (ET) plays a crucial role in the task of Emotion Recognition in Conversation (ERC) that has not received sufficient attention in current research. In light of this, an Emotional Transition-based Emotion Recognizer (EmoTrans) is proposed in this paper. Specifically, we concatenate the most recent utterances with their corresponding speakers to construct the model input, known as samples, each with several placeholders to implicitly express the emotions of contextual utterances. Based on these placeholders, two components are developed to make the model sensitive to emotions and effectively capture the ET features in the sample. Furthermore, an ET-based Contrastive Learning (CL) is developed to compact the representation space, making the model achieve more robust sample representations. We conducted exhaustive experiments on four widely used datasets and obtained competitive experimental results, especially, new state-of-the-art results obtained on MELD and IEMOCAP, demonstrating the superiority of EmoTrans.

Computerized Cognitive Behavioral Therapy (CCBT) has been widely proven effective in alleviating negative emotions such as depression and anxiety. To address the lack of real-time feedback, we explored how to add appropriate interactions from the user experience perspective. By analyzing user context, we propose a transformation framework from CBT to interactive CCBT (iCCBT), which makes online interventions more effective. According to the framework, we designed and implemented an iCCBT platform for young adults. Through a 9-day controlled experiment on 28 college students, we conclude that interaction can effectively improve users’ knowledge retention, learning efficiency, satisfaction, and willingness to use. This study demonstrates methods and results for enhancing the interactivity of CCBT. The design thinking of the transformation framework can be extended to other online psychotherapy.

Effective extraction and integration of valuable contextual information is the core of models for the Emotion Recognition in Conversation (ERC) task. However, a significant amount of irrelevant information is inevitably introduced when integrating long-range contextual information, perplexing the model greatly and resulting in incorrect emotion identification. To this end, we proposed a Conversation Clique-based Model (CCM), designed to extract the most efficacious contextual information to bolster the semantic quality of utterances. Specifically, we devise an utterance spatial relationship module (SpaRel) to explicitly model structural-level correlations among utterances by using GAT, and an emotion temporal relationship module (TemRel) to implicitly capture the emotion sequence constraints by employing HMM. We conduct extensive experiments on the publicly available MELD dataset, and the experimental results indicate the effectiveness of our proposed model, achieving new state-of-the-art results.

Three-dimensional reconstruction from multiview images is considered as a longstanding problem in computer vision and graphics. In order to achieve high-fidelity geometry and appearance of 3-D scenes, this article proposes a novel geometric object learning method for multiview reconstruction withfuzzy set theory. We establish anew neural 3D reconstruction theoretical framecalled neural fuzzy geometric representation (NeuFG), which is a special type of implicit representation of geometric scene that only takes value in [0, 1]. NeuFG is essentially a volume image, and thus can be visualized directly with the conventional volume rendering technique. Extensive experiments on two public datasets, i.e., DTU and BlendedMVS, show that our method has the ability of accurately reconstructing complex shapes with vivid geometric details, without the requirement of mask supervision. Both qualitative and quantitative comparisons demonstrate that the proposed method has superior performance over the state-of-the-art neural scene representation methods. The code will be released on GitHub soon.

Detecting pneumonia, especially coronavirus disease 2019 (COVID-19), from chest X-ray (CXR) images is one of the most effective ways for disease diagnosis and patient triage. The application of deep neural networks (DNNs) for CXR image classification is limited due to the small sample size of the well-curated data. To tackle this problem, this article proposes a distance transformation-based deep forest framework with hybrid-feature fusion (DTDF-HFF) for accurate CXR image classification. In our proposed method, hybrid features of CXR images are extracted in two ways: hand-crafted feature extraction and multigrained scanning. Different types of features are fed into different classifiers in the same layer of the deep forest (DF), and the prediction vector obtained at each layer is transformed to form distance vector based on a self-adaptive scheme. The distance vectors obtained by different classifiers are fused and concatenated with the original features, then input into the corresponding classifier at the next layer. The cascade grows until DTDF-HFF can no longer gain benefits from the new layer. We compare the proposed method with other methods on the public CXR datasets, and the experimental results show that the proposed method can achieve state-of-the art (SOTA) performance. The code will be made publicly available at https://github.com/hongqq/DTDF-HFF.

Knowledge graph completion plays a crucial role in downstream applications.However, existing methods tend to only rely on the structure or textual information, resulting in suboptimal model performance.Moreover, recent attempts to leverage pre-trained language models to complete knowledge graphs have proved unsatisfactory.To overcome these limitations, we propose a novel model that combines structural embedding and semantic information of the knowledge graph.Compared with previous works based on pre-trained language models, our model can better use the implicit knowledge of pre-trained language models by using relation templates, entity definitions, and learnable tokens.Furthermore, our model employs a multihead attention mechanism to transform the embedding semantic space of entities and relations obtained from the knowledge graph embedding model, thereby enhancing their expressiveness and unifying the semantic space of both types of information.Finally, we utilize convolutional neural networks to extract features from the matrices created by combining these two types of information for link prediction and triplet classification tasks.Empirical evaluations on two knowledge graph completion datasets demonstrate that our model is effective for both tasks.

Despite the development of computer vision techniques, the micro-expression (ME) recognition task still remains a great challenge because MEs have very low intensity and short duration. However, the ME recognition is of great significance since it provides important clues for real affective states detection. This paper proposes a novel Block Division Convolutional Network (BDCNN) with the implicit deep features augmentation. In detail, BDCNN learns from four optical flow features computed by the onset and apex frames of each video. It innovatively divides each image into a set of small blocks in the deep learning model, then the convolution and pooling operations are performed on these small blocks in sequence. To handle the small sample size problem in the micro-expression data, this study uses the improved implicit semantic data augmentation algorithm in the deep features space. Experiments are conducted on three publicly available databases, viz, CASME II, SMIC, and SAMM. Experimental results show that our model outperforms the state-of-the-art methods by attaining the accuracy of 84.32% and F1-score of 82.13% on the 3-class datasets, and the accuracy of 81.82% and F1-score of 75.46% on the 5-class datasets, respectively. Our source code is publicly available for non-commercial or research use athttps://github.com/MLDMXM2017/BDCNN.

MOTIVATION: Polypharmacy is the combined use of drugs for the treatment of diseases. However, it often shows a high risk of side effects. Due to unnecessary interactions of combined drugs, the side effects of polypharmacy increase the risk of disease and even lead to death. Thus, obtaining abundant and comprehensive information on the side effects of polypharmacy is a vital task in the healthcare industry. Early traditional methods used machine learning techniques to predict side effects. However, they often make costly efforts to extract features of drugs for prediction. Later, several methods based on knowledge graphs are proposed. They are reported to outperform traditional methods. However, they still show limited performance by failing to model complex relations of side effects among drugs. RESULTS: To resolve the above problems, we propose a novel model by further incorporating complex relations of side effects into knowledge graph embeddings. Our model can translate and transmit multidirectional semantics with fewer parameters, leading to better scalability in large-scale knowledge graphs. Experimental evaluation shows that our model outperforms state-of-the-art models in terms of the average area under the ROC and precision-recall curves. AVAILABILITY AND IMPLEMENTATION: Code and data are available at: https://github.com/galaxysunwen/MSTE-master.

Hui Jiang, Chulun Zhou, Fandong Meng, Biao Zhang, Jie Zhou, Degen Huang, Qingqiang Wu, Jinsong Su. Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2021.

Cardiac medical image segmentation plays an important role in the diagnosis and clinical treatment of cardiovascular diseases. However, due to the variability of cardiac anatomy and the ambiguity between cardiac substructures, it is still difficult to quickly segment the entire heart from medical images. Most of the current researches utilize neural network structure to perform whole heart segmentation. Although good segmentation accuracy has been achieved, it usually requires a long training time. This paper aims to build a new whole heart segmentation model based on Deep Forest, called Multi-Resolution Deep Forest Framework(MRDFF), which performs segmentation through two stages. In the first stage, the heart region is located by rough binary classification, and similarity screening is used to reduce redundancy. The second stage subdivides the heart substructures based on the results of the first stage and uses multi-scale fusion to achieve high segmentation accuracy. The experimental results conducted on the public data set MM-WHS show that under the same training data and configuration, our model can be trained in only 4.5 hours, which is about 1/2 of the training time of neural network models, and can reach the accuracy not lower than neural network models, which shows the feasibility and efficiency of our model. The code will be made publicly available at https://github.com/xufeixf/MRDFF.

Identification of gas-oil reservoir is always important but rather difficult in global gas-oil exploration. It is of the great significance to improve the accuracy of reservoir recognition. Seismic exploration is one of the most valuable methods of gas-oil exploration, and the huge amounts of seismic attribute data can be useful for gas-oil exploration. One limitation of the Generative Topographic Mapping (GTM) algorithm is that it cannot determine the classifications of the data points with close probabilities accurately, and it would be more likely to result in confused clarification and fuzzy boundary. To overcome the limitation, an advanced GTM algorithm with Euclidean Distance (GTM-ED) is proposed in this paper, and we use Euclidean Distance to compute the distance from the edge points to the neighbor centroids, and classify it to the closet class to avoid the problems of confused classification. And then the GTM-ED algorithm is used in the research of reservoir identification model, experiments are made with actual seismic data set. First of all, the GTM algorithm is discussed, and then the GTM-ED algorithm is introduced. And afterwards, many experiments are made. In the experiments, the log data and geological data are selected as the labels, and the comparison and analysis are made through three aspects, including relative criteria, absolute criteria, and run-time, and then the results of each model are visualized. The experimental results indicate that the GTM-ED can achieve better results in reservoir clustering and unknown reservoir identification. And in the actual application, the visualization of the GTM-ED can behave better than the GTM in showing the geological characteristics of paleochannel, the string of beads-like reservoirs and linear lava.

The pivot language approach for statistical machine translation (SMT) is a good method to break the resource bottleneck for certain language pairs. However, in the implementation of conventional approaches, pivot-side context information is far from fully utilized, resulting in erroneous estimations of translation probabilities. In this study, we propose two topic-aware pivot language approaches to use different levels of pivot-side context. The first method takes advantage of document-level context by assuming that the bridged phrase pairs should be similar in the document-level topic distributions. The second method focuses on the effect of local context. Central to this approach are that the phrase sense can be reflected by local context in the form of probabilistic topics, and that bridged phrase pairs should be compatible in the latent sense distributions. Then, we build an interpolated model bringing the above methods together to further enhance the system performance. Experimental results on French-Spanish and French-German translations using English as the pivot language demonstrate the effectiveness of topic-based context in pivot-based SMT.