Zhengyi Zhao, Shubo Zhang, Yiming Du, Bin Liang, Baojun Wang, Zhongyang Li, Binyang Li, Kam-Fai Wong. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Zhengyi Zhao, Shubo Zhang, Zezhong Wang, Yuxi Zhang, Huimin Wang, Yutian Zhao, Yefeng Zheng, Binyang Li, Kam-Fai Wong, Xian Wu. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

AI applications increasingly run on fast-evolving, heterogeneous hardware to maximize performance, but general-purpose libraries lag in supporting these features. Performance-minded programmers often build custom communication stacks that are fast but error-prone and non-portable. This paper introduces MSCCL++, a design methodology for developing high-performance, portable communication kernels. It provides (1) a low-level, performance-preserving primitive interface that exposes minimal hardware abstractions while hiding the complexities of synchronization and consistency, (2) a higher-level DSL for application developers to implement workload-specific communication algorithms, and (3) a library of efficient algorithms implementing the standard collective API, enabling adoption by users with minimal expertise. Compared to state-of-the-art baselines, MSCCL++ achieves geomean speedups of 1.7× (up to 5.4×) for collective communication and 1.2× (up to 1.38×) for AI inference workloads. MSCCL++ is in production of multiple AI services provided by Microsoft Azure, and has also been adopted by RCCL, the GPU collective communication library maintained by AMD. MSCCL++ is open source and available at https://github.com/microsoft/mscclpp. Our two years of experience with MSCCL++ suggests that its abstractions are robust, enabling support for new hardware features, such as multimem, within weeks of development.

Abstract Large Reasoning Models (LRMs) have demonstrated impressive capabilities in many tasks, yet they struggle with reliably following multiple instructions, either by failing to satisfy individual constraints or by struggling to balance competing constraints simultaneously. We formalize this challenge as the Constraint Adherence Problem (CAP). This paper introduces a novel framework that addresses CAP by representing instructions as a structured knowledge graph of constraints. Our approach, Constraint Relationship Graph Completion (CRGC), explicitly models relationships between constraints, identifies adherence challenges, and discovers “bridge constraints” that help the model better focus on and reconcile requirements. Bridge constraints act as auxiliary instructions that make primary constraints more salient and compatible. Unlike existing approaches that enhance instruction following through general training methods, CRGC specifically improves constraint satisfaction by leveraging the model’s own knowledge to create better pathways for generation. Experiments across three popular instruction following datasets demonstrate that our approach reduces constraint violations by 39% compared to standard prompting while maintaining reasoning abilities of large reasoning models.

Zhengyi Zhao, Shubo Zhang, Zezhong Wang, Huimin Wang, Yutian Zhao, Bin Liang, Yefeng Zheng, Binyang Li, Kam-Fai Wong, Xian Wu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Zhengyi Zhao, Shubo Zhang, Yuxi Zhang, Yanxi Zhao, Yifan Zhang, Zezhong Wang, Huimin Wang, Yutian Zhao, Bin Liang, Yefeng Zheng, Binyang Li, Kam-Fai Wong, Xian Wu. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Conversational humor is the key to capturing dialogue semantics and dialogue comprehension, which is usually generated in multiple modalities, such as linguistic rhetoric (textual modality), exaggerated facial expressions or movements (visual modality), and quirky intonation (acoustic modality). However, existing multimodal corpora for conversation humor are coarse-grained, and the modality is insufficient to support the conversational humor recognition task. This paper designed an annotation scheme for multimodal humor datasets, and constructed a corpus based on a Chinese sitcom for conversational humor recognition, named MUCH. The MUCH corpus consists of 34,804 utterances in total, and 7,079 of them are humorous. We employed both unimodal and multimodal methods to test our MUCH corpus. Experimental results showed that the multimodal approach could achieve 75.94% in terms of F1-score and surpassed the performance of most unimodal methods, which demonstrated that the MUCH corpus was effective for multimodal humor recognition tasks.

Deep learning training on cloud platforms usually follows the tradition of the separation of storage and computing. The training executes on a compute cluster equipped with GPUs/TPUs while reading data from a separate cluster hosting the storage service. To alleviate the potential bottleneck, a training cluster usually leverages its local storage as a cache to reduce the remote IO from the storage cluster. However, existing deep learning schedulers do not manage storage resources thus fail to consider the diverse caching effects across different training jobs. This could degrade scheduling quality significantly.

Early rumor detection is a key challenging task to prevent rumors from spreading widely. Sociological research shows that social bots’ behavior in the early stage has become the main reason for rumors’ wide spread. However, current models do not explicitly distinguish genuine users from social bots, and their failure in identifying rumors timely. Therefore, this paper aims at early rumor detection by accounting for social bots’ behavior, and presents a Social Bot-Aware Graph Neural Network, named SBAG. SBAG firstly pre-trains a multi-layer perception network to capture social bot features, and then constructs multiple graph neural networks by embedding the features to model the early propagation of posts, which is further used to detect rumors. Extensive experiments on three benchmark datasets show that SBAG achieves significant improvements against the baselines and also identifies rumors within 3 hours while maintaining more than 90% accuracy.

With the development of online social media, the fabrication and dissemination of rumors are much easier than before. As a result, automatic rumor detection becomes more urgent for internet governors, and has received great interest from the AI community. Even though many initial successes have been achieved in terms of detection accuracy and timeliness, existing rumor detection methods are still not robust due to thefrequent domain shiftproblem and theinconsistent propagationproblem.Frequent domain shiftrefers that rumors’ topic changes frequently on online social media, so the trained model may expire quickly in real applications.Inconsistent propagationis another practical challenge, referring to the fact that the spread of rumors is often accompanied by misleading comments. Therefore, a vulnerable model may output inconsistent predictions for the same instance. In this paper, we propose a new framework to improve the robustness of existing rumor detection methods, named SIFTER (SubjectiveInFormaTionEnhancedReinforcement learning). To address thefrequent domain shift, SIFTER employs multi-task learning to introduce external knowledge that can explicitly describe rumors, and thus makes models more adaptive across domains. In particular, SIFTER involves a multi-task learning module to capture subjective information, which is found to be discriminative in rumors’ propagations. To address theinconsistent propagation, SIFTER employs reinforcement learning to realize a new training schema tailed for rumor detection, namely sequential training, which can reduce the attendance of noisy comments in the feature extraction process. Experimental results on two benchmark datasets, Twitter and Weibo, validate the effectiveness of SIFTER. After migrating existing methods to SIFTER, their accuracy score is improved up to 9.4% in cross-domain inference, and their reverse ratio decreases by nearly 4% in continuous prediction.

We introduce CHIME, a cross-passage hierarchical memory network for question answering (QA) via text generation.It extends XLNet (Yang et al., 2019) introducing an auxiliary memory module consisting of two components: the context memory collecting cross-passage evidence, and the answer memory working as a buffer continually refining the generated answers.Empirically, we show the efficacy of the proposed architecture in the multi-passage generative QA, outperforming the state-of-the-art baselines with better syntactically well-formed answers and increased precision in addressing the questions of the AmazonQA review dataset.An additional qualitative analysis revealed the interpretability introduced by the memory module.

Attention-based neural models were employed to detect the different aspects and sentiment polarities of the same target in targeted aspectbased sentiment analysis (TABSA).However, existing methods do not specifically pre-train reasonable embeddings for targets and aspects in TABSA.This may result in targets or aspects having the same vector representations in different contexts and losing the contextdependent information.To address this problem, we propose a novel method to refine the embeddings of targets and aspects.Such pivotal embedding refinement utilizes a sparse coefficient vector to adjust the embeddings of target and aspect from the context.Hence the embeddings of targets and aspects can be refined from the highly correlative words instead of using context-independent or randomly initialized vectors.Experiment results on two benchmark datasets show that our approach yields the state-of-the-art performance in TABSA task.

Currently, a cloud database is employed to serve on-line query-intensive applications. It inevitably happens that some cloud data nodes storing hot records are facing high frequent query requests while others are rarely visited or even idle. Therefore, how data are dynamically allocated and migrated at runtime has significant impact on query load distribution and system performance. Existing system adopt centralized approaches, and they face two main challenges: (1) Query load on individual node cannot be always balancing even if the data are fairly distributed; (2) For each node, the dynamic changes of configuration resources cannot be captured during the runtime. To this end, this paper presents an environment-aware market strategy based system, named e-MARS, for reasonable data migration to achieve query load balance in cloud database. In e-MARS, cloud database is modeled as a cloudDB market, while data nodes are regarded as intelligent traders and the query load as commodity. Each trader is aware of its local environmental re-sources, such as computing capacity, disk volume, based on which the trader itself decides how to trade the query load and migrates the corresponding data. In this way the cloudDB market will achieve equilibrium. Experiments are conducted on the real communication data, and e-MARS significantly enhances the efficiency. Compared with HBase Balancer, more than 65% improvement is achieved in terms of query response time.

Sea ice model is a typical high performance computing problem. CPU and GPU based parallel method has been proposed to accelerate the simulation process, but it is still hard to meet the large-scale calculation demand due to the compute-intensive nature of the model. Sunway TaihuLight supercomputer use the SW26010 processor as its computing unit and achieves high performance for large-scale scientific computing. In this paper we present a programming framework (PFSI.sw) for sea ice model algorithms based on Sunway many-core processor. Based on this framework, programmer can exploit the parallelism of existing sea ice model algorithms and achieve good performance. Several strategies are introduced to this framework, data dividing, data transfer as well as the load balance are the main aspects we currently concerned. This framework has been implemented and tested with two sea ice model algorithms by using real world dataset on Sunway many-core processors. The experiment demonstrates comparable performance to the traditional parallel implementation on Sunway many-core processor and our framework improves the performance up to 40%.

Microblog has become a major platform for information about real-world events.Automatically discovering realworld events from microblog has attracted the attention of many researchers.However, most of existing work ignore the importance of emotion information for event detection.We argue that people's emotional reactions immediately reflect the occurring of real-world events and should be important for event detection.In this study, we focus on the problem of communityrelated event detection by community emotions.To address the problem, we propose a novel framework which include the following three key components: microblog emotion classification, community emotion aggregation and community emotion burst detection.We evaluate our approach on real microblog data sets.Experimental results demonstrate the effectiveness of the proposed framework.

Extensive experiments have validated the effectiveness of the corpus-based method for classifying the word's sentiment polarity.However, no work is done for comparing different corpora in the polarity classification task.Nowadays, Twitter has aggregated huge amount of data that are full of people's sentiments.In this paper, we empirically evaluate the performance of different corpora in sentiment similarity measurement, which is the fundamental task for word polarity classification.Experiment results show that the Twitter data can achieve a much better performance than the Google, Web1T and Wikipedia based methods.