Daniel Zhang-Li, Joy Jia Yin Lim, Binglin Liu, Shangqing Tu, Zijun Yao, Hao Peng, Jifan Yu, Haoxuan Li, Zhanxin Hao, Ye He, Zekun Li, Jiangyi Wang, Lei Hou, Bin Xu, Xin Cong, Zhiyuan Liu, Huiqin Liu, Yu Zhang, Juanzi Li. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Personalized Learning Path Planning (PLPP) aims to design adaptive learning paths that align with individual goals. While large language models (LLMs) show potential in personalizing learning experiences, existing approaches often lack mechanisms for goal-aligned planning. We introduce Pxplore, a novel framework for PLPP that integrates a reinforcement-based training paradigm and an LLM-driven educational architecture. We design a structured learner state model and an automated reward function that transforms abstract objectives into computable signals. We train the policy combining supervised fine-tuning (SFT) and Group Relative Policy Optimization (GRPO), and deploy it within a real-world learning platform. Extensive experiments validate Pxplore's effectiveness in producing coherent, personalized, and goal-driven learning paths. We release our code and dataset at https://github.com/Pxplore/pxplore-algo.

Bohan Lyu, Xin Cong, Heyang Yu, Pan Yang, Cheng Qian, Zihe Wang, Yujia Qin, Yining Ye, Yaxi Lu, Chen Qian, Zhong Zhang, Yukun Yan, Yankai Lin, Zhiyuan Liu, Maosong Sun. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Yaxi Lu, Haolun Li, Xin Cong, Zhong Zhang, Yesai Wu, Yankai Lin, Zhiyuan Liu, Fangming Liu, Maosong Sun. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Yu Xia, Jingru Fan, Weize Chen, Siyu Yan, Xin Cong, Zhong Zhang, Yaxi Lu, Yankai Lin, Zhiyuan Liu, Maosong Sun. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Tool learning enables large language models (LLMs) to interact with external tools and APIs, greatly expanding the application scope of LLMs. However, due to the dynamic nature of external environments, these tools and APIs may become outdated over time, preventing LLMs from correctly invoking tools. Existing research primarily focuses on static environments and overlooks this issue, limiting the adaptability of LLMs in real-world applications. In this paper, we propose ToolEVO, a novel framework designed to enhance the adaptive and reflective capabilities of LLMs against tool variability. By leveraging Monte Carlo Tree Search, ToolEVO facilitates active exploration and interaction of LLMs within dynamic environments, allowing for autonomous self-reflection and self-updating of tool usage based on environmental feedback. Additionally, we introduce ToolQA-D, a benchmark specifically designed to evaluate the impact of tool variability. Extensive experiments demonstrate the effectiveness and stability of our approach, highlighting the importance of adaptability to tool variability for effective tool learning.

Recent advancements in large language models (LLMs) have driven a revolutionary paradigm shift in process automation from Robotic Process Automation to Agentic Process Automation by automating the workflow orchestration procedure based on LLMs. However, existing LLMs (even the advanced OpenAI GPT-4o) are confined to achieving satisfactory capability in workflow orchestration. To address this limitation, we present WorkflowLLM, a data-centric framework elaborately designed to enhance the capability of LLMs in workflow orchestration. It first constructs a large-scale fine-tuning dataset WorkflowBench with 106, 763 samples, covering 1, 503 APIs from 83 applications across 28 categories. Specifically, the construction process can be divided into three phases: (1) Data Collection: we collect real-world workflow data from Apple Shortcuts and RoutineHub, transcribing them into Python-style code. We further equip them with generated hierarchical thought via GPT-4o-mini. (2) Query Expansion: we prompt GPT-4o-mini to generate more task queries to enrich the diversity and complexity of workflows. (3) Workflow Generation: we leverage an annotator model trained on collected data to generate workflows for synthesized queries. Finally, we merge the synthetic samples that pass quality confirmation with the collected samples to obtain the WorkflowBench. Based on WorkflowBench, we fine-tune Llama-3.1-8B to obtain WorkflowLlama. Our experiments show that WorkflowLlama demonstrates a strong capacity to orchestrate complex workflows, while also achieving notable generalization performance on previously unseen APIs. Additionally, WorkflowBench exhibits robust zero-shot generalization capabilities on an out-of-distribution task planning dataset, T-Eval. Our data and code are available at https://github.com/OpenBMB/WorkflowLLM.

Agents powered by large language models have shown remarkable abilities in solving complex tasks. However, most agent systems remain reactive, limiting their effectiveness in scenarios requiring foresight and autonomous decision-making. In this paper, we tackle the challenge of developing proactive agents capable of anticipating and initiating tasks without explicit human instructions. We propose a novel data-driven approach for this problem. Firstly, we collect real-world human activities to generate proactive task predictions. These predictions are then labeled by human annotators as either accepted or rejected. The labeled data is used to train a reward model that simulates human judgment and serves as an automatic evaluator of the proactiveness of LLM agents. Building on this, we develop a comprehensive data generation pipeline to create a diverse dataset, ProactiveBench, containing 6,790 events. Finally, we demonstrate that fine-tuning models with the proposed ProactiveBench can significantly elicit the proactiveness of LLM agents. Experimental results show that our fine-tuned model achieves an F1-Score of 66.47% in proactively offering assistance, outperforming all open-source and close-source models. These results highlight the potential of our method in creating more proactive and effective agent systems, paving the way for future advancements in human-agent collaboration.

With remarkable advancements, large language models (LLMs) have attracted significant efforts to develop LLM-based agents capable of executing intricate multi-step decision-making tasks. Existing approaches predominantly build upon the external performance measure to guide the decision-making process but the reliance on the external performance measure as prior is problematic in real-world scenarios, where such prior may be unavailable, flawed, or even erroneous. For genuine autonomous decision-making for LLM-based agents, it is imperative to develop rationality from their posterior experiences to judge the utility of each decision independently. In this work, we propose RaDAgent (Rational Decision-Making Agent), which fosters the development of its rationality through an iterative framework involving Experience Exploration and Utility Learning. Within this framework, Elo-based Utility Learning is devised to assign Elo scores to individual decision steps to judge their utilities via pairwise comparisons. Consequently, these Elo scores guide the decision-making process to derive optimal outcomes. Experimental results on the Game of 24, WebShop, ToolBench and RestBench datasets demonstrate RaDAgent’s superiority over baselines, achieving about 7.8% improvement on average. Besides, RaDAgent also can reduce costs (ChatGPT API calls), highlighting its effectiveness and efficiency.

Recent efforts to employ large language models (LLMs) as agents have demonstrated promising results in a wide range of multi-step agent tasks. However, existing agents lack an effective experience reuse approach to leverage historical completed tasks. In this paper, we propose a novel experience reuse framework MetaFlowLLM, which constructs a hierarchical experience tree from historically completed tasks. Each node in this experience tree is presented as a MetaFlow which contains static execution workflow and subtask required by agents to complete dynamically. Then, we propose a Hierarchical MetaFlow Merging algorithm to construct the hierarchical experience tree. When accomplishing a new task, MetaFlowLLM can first retrieve the most relevant MetaFlow node from the experience tree and then execute it accordingly. To effectively generate valid MetaFlows from historical data, we further propose a reinforcement learning pipeline to train the MetaFlowGen. Extensive experimental results on AppWorld and WorkBench demonstrate that integrating with MetaFlowLLM, existing agents (e.g., ReAct, Reflexion) can gain substantial performance improvement with reducing execution costs. Notably, MetaFlowLLM achieves an average success rate improvement of 32.3% on AppWorld and 6.2% on WorkBench, respectively.

Chen Qian, Yufan Dang, Jiahao Li, Wei Liu, Zihao Xie, YiFei Wang, Weize Chen, Cheng Yang, Xin Cong, Xiaoyin Che, Zhiyuan Liu, Maosong Sun. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Cheng Qian, Bingxiang He, Zhong Zhuang, Jia Deng, Yujia Qin, Xin Cong, Zhong Zhang, Jie Zhou, Yankai Lin, Zhiyuan Liu, Maosong Sun. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Chen Qian, Wei Liu, Hongzhang Liu, Nuo Chen, Yufan Dang, Jiahao Li, Cheng Yang, Weize Chen, Yusheng Su, Xin Cong, Juyuan Xu, Dahai Li, Zhiyuan Liu, Maosong Sun. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Autonomous agents empowered by Large Language Models (LLMs) have undergone significant improvements, enabling them to generalize across a broad spectrum of tasks. However, in real-world scenarios, cooperation among individuals is often required to enhance the efficiency and effectiveness of task accomplishment. Hence, inspired by human group dynamics, we propose a multi-agent framework AgentVerse that can effectively orchestrate a collaborative group of expert agents as a greater-than-the-sum-of-its-parts system. Our experiments demonstrate that AgentVerse can proficiently deploy multi-agent groups that outperform a single agent. Extensive experiments on text understanding, reasoning, coding, tool utilization, and embodied AI confirm the effectiveness of AgentVerse. Moreover, our analysis of agent interactions within AgentVerse reveals the emergence of specific collaborative behaviors, contributing to heightened group efficiency. We will release our codebase, AgentVerse, to further facilitate multi-agent research.

Despite the advancements of open-source large language models (LLMs), e.g., LLaMA, they remain significantly limited in tool-use capabilities, i.e., using external tools (APIs) to fulfill human instructions. The reason is that current instruction tuning largely focuses on basic language tasks but ignores the tool-use domain. This is in contrast to the excellent tool-use capabilities of state-of-the-art (SOTA) closed-source LLMs, e.g., ChatGPT. To bridge this gap, we introduce ToolLLM, a general tool-use framework encompassing data construction, model training, and evaluation. We first present ToolBench, an instruction-tuning dataset for tool use, which is constructed automatically using ChatGPT. Specifically, the construction can be divided into three stages: (i) API collection: we collect 16,464 real-world RESTful APIs spanning 49 categories from RapidAPI Hub; (ii) instruction generation: we prompt ChatGPT to generate diverse instructions involving these APIs, covering both single-tool and multi-tool scenarios; (iii) solution path annotation: we use ChatGPT to search for a valid solution path (chain of API calls) for each instruction. To enhance the reasoning capabilities of LLMs, we develop a novel depth-first search-based decision tree algorithm. It enables LLMs to evaluate multiple reasoning traces and expand the search space. Moreover, to evaluate the tool-use capabilities of LLMs, we develop an automatic evaluator: ToolEval. Based on ToolBench, we fine-tune LLaMA to obtain an LLM ToolLLaMA, and equip it with a neural API retriever to recommend appropriate APIs for each instruction. Experiments show that ToolLLaMA demonstrates a remarkable ability to execute complex instructions and generalize to unseen APIs, and exhibits comparable performance to ChatGPT. Our ToolLLaMA also demonstrates strong zero-shot generalization ability in an out-of-distribution tool-use dataset: APIBench.

This paper studies the multimodal named entity recognition (MNER) and multimodal relation extraction (MRE), which are important for content analysis and various applications. The core of MNER and MRE lies in incorporating evident visual information to enhance textual semantics, where two issues inherently demand investigations. The first issue is modality-noise, where the task-irrelevant information in each modality may be noises misleading the task prediction. The second issue is modality-gap, where representations from different modalities are inconsistent, preventing from building the semantic alignment between the text and image. To address these issues, we propose a novel method for MNER and MRE byMultiModal representation learning withInformationBottleneck (MMIB). For the first issue, a refinement-regularizer probes the information-bottleneck principle to balance the predictive evidence and noisy information, yielding expressive representations for prediction. For the second issue, an alignment-regularizer is proposed, where a mutual information-based item works in a contrastive manner to regularize the consistent text-image representations. To our best knowledge, we are the first to explore variational IB estimation for MNER and MRE. Experiments show that MMIB achieves the state-of-the-art performances on three public benchmarks.

Cross-domain Named Entity Recognition (NER) transfers knowledge learned from a rich-resource source domain to improve the learning in a low-resource target domain. Most existing works are designed based on the sequence labeling framework, defining entity detection and type prediction as a monolithic process. However, they typically ignore the discrepant transferability of these two sub-tasks: the former locating spans corresponding to entities is largely domain-robust, whereas the latter owns distinct entity types across domains. Combining them into an entangled learning problem may contribute to the complexity of domain transfer. In this work, we propose the novel divide-and-transfer paradigm in which different sub-tasks are learned using separate functional modules for respective cross-domain transfer. To demonstrate the effectiveness of divide-and-transfer, we concretely implement two NER frameworks by applying this paradigm with different cross-domain transfer strategies. Experimental results on 10 different domain pairs show the notable superiority of our proposed frameworks. Experimental analyses indicate that significant advantages of the divide-and-transfer paradigm over prior monolithic ones originate from its better performance on low-resource data and a much greater transferability. It gives us a new insight into cross-domain NER. Our code is available on GitHub. 1

Document-level event factuality identification (DocEFI) is an important task in event knowledge acquisition, which aims to detect whether an event actually occurs or not from the perspective of the document. Unlike the sentence-level task, a document can have multiple sentences with different event factualities, leading to event factuality conflicts in DocEFI. Existing studies attempt to aggregate local event factuality by exploiting document structures, but they mostly consider textual components in the document separately, degrading complicated correlations therein. To address the above issues, this paper proposes a novel approach, namely UR-HAT, to improve DocEFI with uncertain relational hypergraph attention networks. Particularly, we reframe a document graph as a hypergraph, and establish beneficial n-ary correlations among textual nodes with relational hyperedges, which helps to globally consider local factuality features to resolve event factuality conflicts. To better discern the importance of event factuality features, we further represent textual nodes with uncertain Gaussian distributions, and propose novel uncertain relational hypergraph attention networks to refine textual nodes with the document hypergraph. In addition, we select factuality-related keywords as nodes to enrich event factuality features. Experimental results demonstrate the effectiveness of our proposed method, and outperforms previous methods on two widely used benchmark datasets.

Yanzeng Li, Jiangxia Cao, Xin Cong, Zhenyu Zhang, Bowen Yu, Hongsong Zhu, Tingwen Liu. Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2022.

Cross-Domain Sequential Recommendation (CDSR) aims to predict future interactions based on user's historical sequential interactions from multiple domains. Generally, a key challenge of CDSR is how to mine precise cross-domain user preference based on the intra-sequence and inter-sequence item interactions. Existing works first learn single-domain user preference only with intra-sequence item interactions, and then build a transferring module to obtain cross-domain user preference. However, such a pipeline and implicit solution can be severely limited by the bottleneck of the designed transferring module, and ignores to consider inter-sequence item relationships.

Event Causality Identification (ECI), which aims to detect whether a causality relation exists between two given textual events, is an important task for event causality understanding. However, the ECI task ignores crucial event structure and cause-effect causality component information, making it struggle for downstream applications. In this paper, we introduce a novel task, namely Event Causality Extraction (ECE), aiming to extract the cause-effect event causality pairs with their structured event information from plain texts. The ECE task is more challenging since each event can contain multiple event arguments, posing fine-grained correlations between events to decide the cause-effect event pair. Hence, we propose a method with a dual grid tagging scheme to capture the intra- and inter-event argument correlations for ECE. Further, we devise a event type-enhanced model architecture to realize the dual grid tagging scheme. Experiments demonstrate the effectiveness of our method, and extensive analyses point out several future directions for ECE.

Document-level Event Extraction (DEE) is particularly tricky due to the two challenges it poses: scattering-arguments and multi-events. The first challenge means that arguments of one event record could reside in different sentences in the document, while the second one reflects that one document may simultaneously contain multiple such event records. Motivated by humans’ reading cognitive to extract information of interests, in this paper, we propose a method called HRE (Human Reading inspired Extractor for Document Events), where DEE is decomposed into these two iterative stages, rough reading and elaborate reading. Specifically, the first stage browses the document to detect the occurrence of events, and the second stage serves to extract specific event arguments. For each concrete event role, elaborate reading hierarchically works from sentences to characters to locate arguments across sentences, thus the scattering-arguments problem is tackled. Meanwhile, rough reading is explored in a multi-round manner to discover undetected events, thus the multi-events problem is handled. Experiment results show the superiority of HRE over prior competitive methods.

Recommender systems have been widely deployed in many real-world applications, but usually suffer from the long-standing user cold-start problem. As a promising way, Cross-Domain Recommendation (CDR) has attracted a surge of inter-est, which aims to transfer the user preferences observed in the source domain to make recommendations in the target domain. Previous CDR approaches mostly achieve the goal by following the Embedding and Mapping (EMCDR) idea which attempts to learn a mapping function to transfer the pre-trained user repre-sentations (embeddings) from the source domain into the target domain. However, they pre-train the user/item representations independently for each domain, ignoring to consider both domain interactions simultaneously. Therefore, the biased pre-trained representations inevitably involve the domain-specific information which may lead to negative impact to transfer information across domains. In this work, we consider a key point of the CDR task: what information needs to be shared across domains? To achieve the above idea, this paper utilizes the information bottleneck (IB) principle, and proposes a novel approach termed as CDRIB to enforce the representations encoding the domain-shared information. To derive the unbiased representations, we devise two IB regularizers to model the cross-domain/in-domain user-item interactions simultaneously and thereby CDRIB could consider both domain interactions jointly for de-biasing. With an additional contrastive information regularizer, CDRIB can also capture cross-domain user-user correlations. In this way, those regularizers encourage the representations to encode the domain-shared information, which has the capability to make recommendations in both domains directly. To the best of our knowledge, this paper is the first work to capture the domain-shared information for cold-start users via variational information bottleneck. Empirical experiments illustrate that CDRIB outperforms the state-of-the-art approaches on four real-world cross-domain datasets, demonstrating the effectiveness of adopting the information bottleneck for CDR.

Real-world web applications such as Amazon and Netflix often provide services in multiple countries and regions (i.e., markets) around the world. Generally, different markets share similar item sets while containing different amounts of interaction data. Some markets are data-scarce and others are data-rich and leveraging those data from similar and data-rich auxiliary markets could enhance the data-scarce markets. In this paper, we explore multi-market recommendation (MMR), and propose a novel model called M$^3$Rec to improve all markets recommendation simultaneously. Since items play the role to bridge different markets, we argue that mining the similarities among items is the key point of MMR. Our M^3Rec preprocess two global item similarities: intra- and inter- market similarities. Specifically, we first learn the second-order intra-market similarity by adopting linear models with closed-form solutions, and then capture the high-order inter-market similarity by the random walk. Afterward, we incorporate the global item similarities for each local market. We conduct extensive experiments on five public available markets and compare with several state-of-the-art methods. Detailed experimental results demonstrate the effectiveness of our proposed method.

Data sparsity is a long-standing problem in recommender systems. To alleviate it, Cross-Domain Recommendation (CDR) has attracted a surge of interests, which utilizes the rich user-item interaction information from the related source domain to improve the performance on the sparse target domain. Recent CDR approaches pay attention to aggregating the source domain information to generate better user representations for the target domain. However, they focus on designing more powerful interaction encoders to learn both domains simultaneously, but fail to model different user preferences of different domains. Particularly, domain-specific preferences of the source domain usually provide useless information to enhance the performance in the target domain, and directly aggregating the domain-shared and domain-specific information together maybe hurts target domain performance. This work considers a key challenge of CDR: How do we transfer shared information across domains? Grounded in the information theory, we propose DisenCDR, a novel model to disentangle the domain-shared and domain-specific information. To reach our goal, we propose two mutual-information-based disentanglement regularizers. Specifically, an exclusive regularizer aims to enforce the user domain-shared representations and domain-specific representations encoding exclusive information. An information regularizer is to encourage the user domain-shared representations encoding predictive information for both domains. Based on them, we further derive a tractable bound of our disentanglement objective to learn desirable disentangled representations. Extensive experiments show that DisenCDR achieves significant improvements over state-of-the-art baselines on four real-world datasets.

In few-shot relational triple extraction (FS-RTE), one seeks to extract relational triples from plain texts by utilizing only few annotated samples. Recent work first extracts all entities and then classifies their relations. Such an entity-then-relation paradigm ignores the entity discrepancy between relations. To address it, we propose a novel task decomposition strategy, Relation-then-Entity, for FS-RTE. It first detects relations occurred in a sentence and then extracts the corresponding head/tail entities of the detected relations. To instantiate this strategy, we further propose a model, RelATE, which builds a dual-level attention to aggregate relation-relevant information to detect the relation occurrence and utilizes the annotated samples of the detected relations to extract the corresponding head/tail entities. Experimental results show that our model outperforms previous work by an absolute gain (18.98%, 28.85% in F1 in two few-shot settings).

Span extraction, aiming to extract text spans (such as words or phrases) from plain texts, is a fundamental process in Information Extraction.Recent works introduce the label knowledge to enhance the text representation by formalizing the span extraction task into a question answering problem (QA Formalization), which achieves state-of-the-art performance.However, QA Formalization does not fully exploit the label knowledge and suffers from low efficiency in training/inference.To address those problems, we introduce a new paradigm to integrate label knowledge and further propose a novel model to explicitly and efficiently integrate label knowledge into text representations.Specifically, it encodes texts and label annotations independently and then integrates label knowledge into text representation with an elaborate-designed semantics fusion module.We conduct extensive experiments on three typical span extraction tasks: flat NER, nested NER, and event detection.The empirical results show that 1) our method achieves state-of-the-art performance on four benchmarks, and 2) reduces training time and inference time by 76% and 77% on average, respectively, compared with the QA Formalization paradigm.Our code and data are available at https://github.com/ Akeepers/LEAR.