Siheng Li, Kejiao Li, Zenan Xu, Guanhua Huang, Kun Li, Haoyuan Wu, Wujiajia, Zihao Zheng, Chenchen Zhang, Kun Shi, Xue Gong, Qi Yi, Ruibin Xiong, Tingqiang Xu, Yuhao Jiang, Jianfeng Yan, Yuyuan Zeng, Guanghui Xu, Jinbao Xue, Zhijiang xu, Zheng Fang, Shuai LI, Qibin Liu, Xiaoxue Li, Zhuoyu Li, Yangyu Tao, Fei Gao, Cheng Jiang, Bochao Wang, Kai Liu, Jianchen Zhu, Wai Lam, Bo Zhou, Di Wang. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Tool-Integrated Reasoning has emerged as a key paradigm to augment Large Language Models (LLMs) with computational capabilities, yet integrating tool-use into long Chainof-Thought (long CoT) remains underexplored, largely due to the scarcity of training data and the challenge of integrating tool-use without compromising the model's intrinsic longchain reasoning.In this paper, we introduce DART (Discovery And Reinforcement of Tool-Integrated Reasoning Chains via Rollout Trees), a reinforcement learning framework that enables spontaneous tool-use during long CoT reasoning without additional human annotation.DART operates by constructing dynamic rollout trees during training to discover valid tool-use opportunities, branching out at promising positions to explore tool-integrated trajectories.Subsequently, a tree-based process advantage estimation identifies and credits specific sub-trajectories where tool invocation positively contributes to the solution, effectively reinforcing these beneficial behaviors during training.Extensive experiments on challenging benchmarks like AIME and GPQA-Diamond demonstrate that DART significantly outperforms existing methods, successfully harmonizing tool execution with long CoT reasoning.

Warning: Please note that this article contains potential harmful or offensive content. This content is only for the evaluating and analysis of LLMs and does not imply any intention to promote criminal activities.

Although great progress has been made for Machine Reading Comprehension (MRC) in English, scaling out to a large number of languages remains a huge challenge due to the lack of large amounts of annotated training data in non-English languages. To address this challenge, some recent efforts of cross-lingual MRC employ machine translation to transfer knowledge from English to other languages, through either explicit alignment or implicit attention. For effective knowledge transition, it is beneficial to leverage both semantic and syntactic information. However, the existing methods fail to explicitly incorporate syntax information in model learning. Consequently, the models are not robust to errors in alignment and noises in attention. In this work, we propose a novel approach, which jointly models the cross-lingual alignment information and the mono-lingual syntax information using a graph. We develop a series of algorithms, including graph construction, learning, and pre-training. The experiments on two benchmark datasets for cross-lingual MRC show that our approach outperforms all strong baselines, which verifies the effectiveness of syntax information for cross-lingual MRC.

The multimodal aspect-based sentiment classification task (mABSC) aims to recognize the sentiment polarities of aspect entities according to the associated textual and visual resources. Inspired by the cross-modal alignment ability of Transformers, some recent mABSC methods have also proposed to use Transformers to discover relevance between aspect entity and cross-modal visual regions, based on which relevant visual regions can be identified and leveraged to help recognize the sentiment polarity of aspect entities. However, due to the limited training data in mABSC tasks, it is found that in the Transformers, the aspect entities are often attended to irrelevant visual information, which obviously will not benefit the sentiment polarity recognition. To address the issue, we seek help from external knowledge, including textual syntax and cross-modal relevancy knowledge. The basic idea is to cut off the irrelevant connections among textual or cross-modal modalities in the Transformer layer using a knowledge-induced matrix. To have the matrix going beyond simply capturing direct relations, a mechanism is developed to enable it to reflect multi-hop relations, followed by a discretization operation to filter out extreme relevancy. Extensive experiments on two public multimodal datasets show that our method outperforms all competing baselines. Further studies demonstrate the effectiveness of each component, and suggest that the introduced external knowledge can instruct the model to learn the correct relevance among textual or cross-modal features, thus benefiting the mABSC task.

A key challenge in video question answering (VideoQA) is how to align textual concepts with the cross-modal visual regions accurately. Existing methods mostly rely on the alignment between individual words and relevant video regions, but individual words are generally not able to capture the complete information of a textual concept, which is often represented by the composition of several words. To address this issue, we propose to build a syntactic dependency tree for each question with an off-the-shelf tool and use it to extract meaningful word compositions (i.e., textual concept). By viewing the words and compositions as nodes and hyperedges, respectively, a hypergraph convolutional network (HCN) is built to learn the representations of textual concepts. Then, to enable cross-modal interaction of relevant concepts from different modalities, an optimal transport (OT) based alignment method is developed to establish the connection between textual concepts and their relevant visual regions. Experimental results on three benchmarks show that our method outperforms all competing baselines. Further analyses demonstrate the effectiveness of each component, and show that our model is good at modeling different levels of semantic compositions and filtering out irrelevant information.

Multimodal abstractive summarization for videos (MAS) requires generating a concise textual summary to describe the highlights of a video according to multimodal resources, in our case, the video content and its transcript. Inspired by the success of the large-scale generative pre-trained language model (GPLM) in generating high-quality textual content (e.g., summary), recent MAS methods have proposed to adapt the GPLM to this task by equipping it with the visual information, which is often obtained through a general-purpose visual feature extractor. However, the generally extracted visual features may overlook some summary-worthy visual information, which impedes model performance. In this work, we propose a novel approach to learning the summary-worthy visual representation that facilitates abstractive summarization. Our method exploits the summary-worthy information from both the cross-modal transcript data and the knowledge that distills from the pseudo summary. Extensive experiments on three public multimodal datasets show that our method outperforms all competing baselines. Furthermore, with the advantages of summary-worthy visual information, our model can have a significant improvement on small datasets or even datasets with limited training data.

Zenan Xu, Daya Guo, Duyu Tang, Qinliang Su, Linjun Shou, Ming Gong, Wanjun Zhong, Xiaojun Quan, Daxin Jiang, Nan Duan. 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.

Fact checking is a challenging task because verifying the truthfulness of a claim requires reasoning about multiple retrievable evidence.In this work, we present a method suitable for reasoning about the semantic-level structure of evidence.Unlike most previous works, which typically represent evidence sentences with either string concatenation or fusing the features of isolated evidence sentences, our approach operates on rich semantic structures of evidence obtained by semantic role labeling.We propose two mechanisms to exploit the structure of evidence while leveraging the advances of pre-trained models like BERT, GPT or XLNet.Specifically, using XLNet as the backbone, we first utilize the graph structure to re-define the relative distances of words, with the intuition that semantically related words should have short distances.Then, we adopt graph convolutional network and graph attention network to propagate and aggregate information from neighboring nodes on the graph.We evaluate our system on FEVER, a benchmark dataset for fact checking, and find that rich structural information is helpful and both our graph-based mechanisms improve the accuracy.Our model is the state-of-the-art system in terms of both official evaluation metrics, namely claim verification accuracy and FEVER score.

Network embedding has recently emerged as a promising technique to embed nodes of a network into low-dimensional vectors.While fairly successful, most existing works focus on the embedding techniques for static networks.But in practice, there are many networks that are evolving over time and hence are dynamic, e.g., the social networks.To address this issue, a high-order spatio-temporal embedding model is developed to track the evolutions of dynamic networks.Specifically, an activeness-aware neighborhood embedding method is first proposed to extract the high-order neighborhood information at each given timestamp.Then, an embedding prediction framework is further developed to capture the temporal correlations, in which the attention mechanism is employed instead of recurrent neural networks (RNNs) for its efficiency in computing and flexibility in modeling.Extensive experiments are conducted on four realworld datasets from three different areas.It is shown that the proposed method outperforms all the baselines by a substantial margin for the tasks of dynamic link prediction and node classification, which demonstrates the effectiveness of the proposed methods on tracking the evolutions of dynamic networks.

Deepfake detection, the task of automatically discriminating machine-generated text, is increasingly critical with recent advances in natural language generative models.Existing approaches to deepfake detection typically represent documents with coarse-grained representations.However, they struggle to capture factual structures of documents, which is a discriminative factor between machinegenerated and human-written text according to our statistical analysis.To address this, we propose a graph-based model that utilizes the factual structure of a document for deepfake detection of text.Our approach represents the factual structure of a given document as an entity graph, which is further utilized to learn sentence representations with a graph neural network.Sentence representations are then composed to a document representation for making predictions, where consistent relations between neighboring sentences are sequentially modeled.Results of experiments on two public deepfake datasets show that our approach significantly improves strong base models built with RoBERTa.Model analysis further indicates that our model can distinguish the difference in the factual structure between machine-generated text and humanwritten text.

Zenan Xu, Qinliang Su, Xiaojun Quan, Weijia Zhang. Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing (EMNLP-IJCNLP). 2019.