Deep Research (DR) requires LLM agents to autonomously perform multi-step information seeking, processing, and reasoning to generate comprehensive reports. In contrast to existing studies that mainly focus on unstructured web content, a more challenging DR task should additionally utilize structured knowledge to provide a solid data foundation, facilitate quantitative computation, and lead to in-depth analyses. In this paper, we refer to this novel task as Knowledgeable Deep Research (KDR), which requires DR agents to generate reports with both structured and unstructured knowledge. Furthermore, we propose the Hybrid Knowledge Analysis framework (HKA), a multi-agent architecture that reasons over both kinds of knowledge and integrates the texts, figures, and tables into coherent multimodal reports. The key design is the Structured Knowledge Analyzer, which utilizes both coding and vision-language models to produce figures, tables, and corresponding insights. To support systematic evaluation, we construct KDR-Bench, which covers 9 domains, includes 41 expert-level questions, and incorporates a large number of structured knowledge resources (e.g., 1,252 tables). We further annotate the main conclusions and key points for each question and propose three categories of evaluation metrics including general-purpose, knowledge-centric, and vision-enhanced ones. Experimental results demonstrate that HKA consistently outperforms most existing DR agents on general-purpose and knowledge-centric metrics, and even surpasses the Gemini DR agent on vision-enhanced metrics, highlighting its effectiveness in deep, structure-aware knowledge analysis. Finally, we hope this work can serve as a new foundation for structured knowledge analysis in DR agents and facilitate future multimodal DR studies.

Large Language Models (LLMs) have demonstrated strong adaptation to unseen tasks. However, their performance in Information Extraction (IE) under unseen schemas remains limited. Actually, IE requires both general abilities for understanding natural language and semantic concepts, and specialized abilities for aligning extracted information to various human-defined schemas. Training an LLM jointly on multiple schemas, or adapting it to a specific schema, often results in performance drops on datasets with other schemas, especially when conflicts arise between schemas. We refer to this phenomenon as the schema alignment tax in this paper. To alleviate this, we propose a schema-adaptive three-phase framework, Identify–Conceptualize–Align (ICA), which enables LLMs to focus on general abilities such as identifying entity and trigger spans and assigning corresponding concepts to them, while delegating schema-specific alignment to lightweight models. Specifically, in the Identification phase, we train an LLM to identify entity and trigger spans on multiple datasets, with cross-dataset annotation to boost span recall. In the Conceptualization phase, the LLM is used to assign semantic concepts to each span. In the Alignment phase, we train different lightweight alignment models to map these concepts to different human-defined schemas. The first two phases are fully reusable across tasks, so adapting to a new schema requires retraining only the alignment model. We evaluate ICA on entity recognition and event detection on 26 commonly adopted datasets with diverse schemas. Experimental results show that our method not only surpasses state-of-the-art approaches under supervised settings, achieving an average F1 improvement of 1.6%, but also attains a remarkable 11.5% average F1 gain on NER and ED in the 10-shot setting.

Jizhao Zhu, Akang Shi, Zixuan Li, Long Bai, Xiaolong Jin, Jiafeng Guo, Xueqi Cheng. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Event Relation Extraction (ERE) aims to extract various types of relations between different events within texts. Although Large Language Models (LLMs) have demonstrated impressive capabilities in many natural language processing tasks, existing ERE methods based on LLMs still face three key challenges: (1) Time Inefficiency: The existing pairwise method of combining events and determining their relations is time-consuming for LLMs. (2) Low Coverage: When dealing with numerous events in a document, the limited generation length of fine-tuned LLMs restricts the coverage of their extraction results. (3) Lack of Rationale: Essential rationales concerning the results that could enhance the reasoning ability of the model are overlooked. To address these challenges, we propose LLMERE, an LLM-based approach with rationales for the ERE task. LLMERE transforms ERE into a question-and-answer task that may have multiple answers. By extracting all events related to a specified event at once, LLMERE reduces time complexity from O(n^2) to O(n), compared to the pairwise method. Subsequently, LLMERE enhances the coverage of extraction results by employing a partitioning strategy that highlights only a portion of the events in the document at a time. In addition to the extracted results, LLMERE is also required to generate corresponding rationales/reasons behind them, in terms of event coreference information or transitive chains of event relations. Experimental results on three widely used datasets show that LLMERE achieves significant improvements over baseline methods.

Developing a general-purpose system that can extract events with massive types is a longstanding target in Event Extraction (EE).In doing so, the basic challenge comes from the absence of an efficient and effective annotation framework to construct the corresponding datasets.In this paper, we propose an LLM-based collaborative annotation framework.Through collaboration among multiple LLMs and a subsequent voting process, it refines annotations of triggers from distant supervision and then carries out argument annotation.Finally, we create EEMT, the largest EE dataset to date, featuring over 200,000 samples, 3,465 event types, and 6,297 role types.Evaluation on the human-annotated test set demonstrates that the proposed framework achieves the F1 scores of 90.1% and 85.3% for event detection and argument extraction, strongly validating its effectiveness.Besides, to alleviate the excessively long prompts caused by massive types, we propose an LLM-based Partitioning method for EE called LLM-PEE.It first recalls candidate event types and then splits them into multiple partitions for LLMs to extract.After fine-tuning on the EEMT training set, the distilled LLM-PEE with 7B parameters outperforms state-of-the-art methods by 5.4% and 6.1% in event detection and argument extraction.Besides, it also surpasses mainstream LLMs by 12.9% on the unseen datasets, which strongly demonstrates the event diversity of the EEMT dataset and the generalization capabilities of the LLM-PEE method.

Large language models (LLMs) have recently demonstrated remarkable performance across various Natual Language Processing tasks. In the field of multi-hop reasoning, the Chain-of-thought (CoT) prompt method has emerged as a paradigm, using curated stepwise reasoning demonstrations to enhance LLM's ability to reason and produce coherent rational pathways. To ensure the accuracy, reliability, and traceability of the generated answers, many studies have incorporated information retrieval (IR) to provide LLMs with external knowledge. However, existing CoT with IR methods decomposes questions into sub-questions based on a single compositionality type, which limits their effectiveness for questions involving multiple compositionality types. Additionally, these methods suffer from inefficient retrieval, as complex questions often contain abundant information, leading to the retrieval of irrelevant information inconsistent with the query's intent. In this work, we propose a novel question decomposition framework called TRQA for multi-hop question answering, which addresses these limitations. Our framework introduces a reasoning tree (RT) to represent the structure of complex questions. It consists of four components: the Reasoning Tree Constructor (RTC), the Question Generator (QG), the Retrieval and LLM Interaction Module (RAIL), and the Answer Aggregation Module (AAM). Specifically, the RTC predicts diverse sub-question structures to construct the reasoning tree, allowing a more comprehensive representation of complex questions. The QG generates sub-questions for leaf-node in the reasoning tree, and we explore two methods for QG: prompt-based and T5-based approaches. The IR module retrieves documents aligned with sub-questions, while the LLM formulates answers based on the retrieved information. Finally, the AAM aggregates answers along the reason tree, producing a definitive response from bottom to top.

Zixuan Li, Yutao Zeng, Yuxin Zuo, Weicheng Ren, Wenxuan Liu, Miao Su, Yucan Guo, Yantao Liu, Xiang Li, Zhilei Hu, Long Bai, Wei Li, Yidan Liu, Pan Yang, Xiaolong Jin, Jiafeng Guo, Xueqi Cheng. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Temporal Knowledge Graph (TKG), which characterizes temporally evolving facts in the form of (subject, relation, object, timestamp), has attracted much attention recently. TKG reasoning aims to predict future facts based on given historical ones. However, existing TKG reasoning models are unable to abstain from predictions they are uncertain, which will inevitably bring risks in real-world applications. Thus, in this paper, we propose an abstention mechanism for TKG reasoning, which helps the existing models make selective, instead of indiscriminate, predictions. Specifically, we develop a confidence estimator, called Confidence Estimator with History (CEHis), to enable the existing TKG reasoning models to first estimate their confidence in making predictions, and then abstain from those with low confidence. To do so, CEHis takes two kinds of information into consideration, namely, the certainty of the current prediction and the accuracy of historical predictions. Experiments with representative TKG reasoning models on two benchmark datasets demonstrate the effectiveness of the proposed CEHis.

Nested Event Extraction (NEE) aims to extract complex event structures where an event contains other events as its arguments recursively. Nested events involve a kind of Pivot Elements (PEs) that simultaneously act as arguments of outer-nest events and as triggers of inner-nest events, and thus connect them into nested structures. This special characteristic of PEs brings challenges to existing NEE methods, as they cannot well cope with the dual identities of PEs. Therefore, this paper proposes a new model, called PerNee, which extracts nested events mainly based on recognizing PEs. Specifically, PerNee first recognizes the triggers of both inner-nest and outer-nest events and further recognizes the PEs via classifying the relation type between trigger pairs. The model uses prompt learning to incorporate information from both event types and argument roles for better trigger and argument representations to improve NEE performance. Since existing NEE datasets (e.g., Genia11) are limited to specific domains and contain a narrow range of event types with nested structures, we systematically categorize nested events in the generic domain and construct a new NEE dataset, called ACE2005-Nest. Experimental results demonstrate that PerNee consistently achieves state-of-the-art performance on ACE2005-Nest, Genia11, and Genia13. The ACE2005-Nest dataset and the code of the PerNee model are available at https://github.com/waysonren/PerNee.

Event detection is one of the fundamental tasks in information extraction and knowledge graph. However, a realistic event detection system often needs to deal with new event classes constantly. These new classes usually have only a few labeled instances as it is time-consuming and labor-intensive to annotate a large number of unlabeled instances. Therefore, this paper proposes a new task, called class-incremental few-shot event detection. Nevertheless, there are two problems (i.e., old knowledge forgetting and new class overfitting) in this task. To solve these problems, this paper further presents a novel knowledge distillation and prompt learning based method, called Prompt-KD. Specifically, to reduce the forgetting issue about old knowledge, Prompt-KD develops an attention based multi-teacher knowledge distillation framework, where the ancestor teacher model pre-trained on base classes is reused in all learning sessions, and the father teacher model derives the current student model via adaptation. On the other hand, in order to cope with the few-shot learning scenario and alleviate the corresponding new class overfitting problem, Prompt-KD is also equipped with a prompt learning mechanism. Extensive experiments on two benchmark datasets, i.e., FewEvent and MAVEN, demonstrate the state-of-the-art performance of Prompt-KD.

Conventional Knowledge Graph Reasoning (KGR) models learn the embeddings of KG components over the structure of KGs, but their performances are limited when the KGs are severely incomplete.Recent LLM-enhanced KGR models input KG structural information into LLMs.However, they require fine-tuning on open-source LLMs and are not applicable to closed-source LLMs.Therefore, in this paper, to leverage the knowledge in LLMs without fine-tuning to assist and enhance conventional KGR models, we propose a new three-stage pipeline, including knowledge alignment, KG reasoning and entity reranking.Specifically, in the alignment stage, we propose three strategies to align the knowledge in LLMs to the KG schema by explicitly associating unconnected nodes with semantic relations.Based on the enriched KGs, we train structure-aware KGR models to integrate aligned knowledge to original knowledge existing in KGs.In the reranking stage, after obtaining the results of KGR models, we rerank the top-scored entities with LLMs to recall correct answers further.Experiments show our pipeline can enhance the KGR performance in both incomplete and general situations.

Script is a kind of structured knowledge extracted from texts, which contains a sequence of events. Based on such knowledge, script event prediction aims to predict the subsequent event. To do so, two aspects should be considered for events, namely, event description (i.e., what the events should contain) and event encoding (i.e., how they should be encoded). Most existing methods describe an event by a verb together with a few core arguments (i.e., subject, object, and indirect object), which are not precise enough. In addition, existing event encoders are limited to a fixed number of arguments, which are not flexible enough to deal with extra information. Thus, in this paper, we propose the Rich Event Prediction (REP) framework for script event prediction. Fundamentally, it is based on the proposed rich event description, which enriches the existing ones with three kinds of important information, namely, the senses of verbs, extra semantic roles, and types of participants. REP contains an event extractor to extract such information from texts. Based on the extracted rich information, a predictor then selects the most probable subsequent event. The core component of the predictor is a transformer-based event encoder that integrates the above information flexibly. Experimental results on the widely used Gigaword Corpus show the effectiveness of the proposed framework.

Zhilei Hu, Zixuan Li, Xiaolong Jin, Long Bai, Saiping Guan, Jiafeng Guo, Xueqi Cheng. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Legal case retrieval, which aims to find relevant cases based on a short case description, serves as an important part of modern legal systems. Despite the success of existing retrieval methods based on Pretrained Language Models, there are still two issues in legal case retrieval that have not been well considered before. First, existing methods underestimate the semantics associations among legal elements, e.g., law articles and crimes, which played an essential role in legal case retrieval. These methods only adopt the pre-training language model to encode the whole legal case, instead of distinguishing different legal elements in the legal case. They randomly split a legal case into different segments, which may break the completeness of each legal element. Second, due to the difficulty in annotating the relevant labels of similar cases, legal case retrieval inevitably faces the problem of lacking training data. In this paper, we propose a counterfactual graph learning framework for legal case retrieval. Concretely, to overcome the above challenges, we transform the legal case document into a graph and model the semantics of the legal elements through a graph neural network. To alleviate the low resource and learn the causal relationship between the semantics of legal elements and relevance, a counterfactual data generator is designed to augment counterfactual data and enhance legal case representation. Extensive experiments based on two publicly available legal benchmarks demonstrate that our CFGL-LCR can significantly outperform previous state-of-the-art methods in legal case retrieval.

Besides entity-centric knowledge, usually organized as Knowledge Graph (KG), events are also an essential kind of knowledge in the world, which trigger the spring up of event-centric knowledge representation form like Event KG (EKG). It plays an increasingly important role in many downstream applications, such as search, question-answering, recommendation, financial quantitative investments, and text generation. This paper provides a comprehensive survey of EKG from history, ontology, instance, and application views. Specifically, to characterize EKG thoroughly, we focus on its history, definitions, schema induction, acquisition, related representative graphs/systems, and applications. The development processes and trends are studied therein. We further summarize prospective directions to facilitate future research on EKG.

Complex question generation over knowledge bases (KB) aims to generate natural language questions involving multiple KB relations or functional constraints. Existing methods train one encoder-decoder-based model to fit all questions. However, such a one-size-fits-all strategy may not perform well since complex questions exhibit an uneven distribution in many dimensions, such as question types, involved KB relations, and query structures, resulting in insufficient learning for long-tailed samples under different dimensions. To address this problem, we propose a meta-learning framework for complex question generation. The meta-trained generator can acquire universal and transferable meta-knowledge and quickly adapt to long-tailed samples through a few most related training samples. To retrieve similar samples for each input query, we design a self-supervised graph retriever to learn distributed representations for samples, and contrastive learning is leveraged to improve the learned representations. We conduct experiments on both WebQuestionsSP and ComplexWebQuestion, and results on long-tailed samples of different dimensions have been significantly improved, which demonstrates the effectiveness of the proposed framework.

Scripts are structured sequences of events together with the participants, which are extracted from the texts.Script event prediction aims to predict the subsequent event given the historical events in the script.Two kinds of information facilitate this task, namely, the event-level information and the script-level information.At the event level, existing studies view an event as a verb with its participants, while neglecting other useful properties, such as the state of the participants.At the script level, most existing studies only consider a single event sequence corresponding to one common protagonist.In this paper, we propose a Transformer-based model, called M-CPredictor, which integrates deep event-level and script-level information for script event prediction.At the event level, MCPredictor utilizes the rich information in the text to obtain more comprehensive event semantic representations.At the script-level, it considers multiple event sequences corresponding to different participants of the subsequent event.The experimental results on the widely-used New York Times corpus demonstrate the effectiveness and superiority of the proposed model.

Distantly Supervised Relation Extraction (DSRE) has been widely studied, since it can automatically extract relations from very large corpora. However, existing DSRE methods only use little semantic information about entities, such as the information of entity type. Thus, in this paper, we propose a method for integrating entity type information into a neural network based DSRE model. It also adopts two attention mechanisms, namely, sentence attention and type attention. The former selects the representative sentences for a sentence bag, while the latter selects appropriate type information for entities. Experimental comparison with existing methods on a benchmark dataset demonstrates its merits.

The dependency relation between words in the sentence is critical for the relation extraction. Existing methods often utilize the dependencies accompanied with various pruning strategies, thus suffer from the loss of detailed semantic information.In order to exploit dependency structure more effectively, we propose a novel bidirectional dependency-guided attention model. The main idea is to use a top-down attention as well as a bottom-up attention to fully capture the dependencies from different granularity. Specifically, the bottom-up attention aims to model the local semantics from the subtree of each node, while the top-down attention is to model the global semantics from the ancestor nodes. Moreover, we employ a label embedding component to attend the contextual features, which are extracted by the dependency-guided attention. Overall, the proposed model is fully attention-based which make it easy for parallel computing. Experiment results on TACRED dataset and SemEval 2010 Task 8 dataset show that our model outperforms existing dependency based models as well as the powerful pretraining model. Moreover, the proposed model achieves the state-of-the-art performance on TACRED dataset.

Knowledge Graph Question Answering aims to automatically answer natural language questions via well-structured relation information between entities stored in knowledge graphs. When faced with a complex question with compositional semantics, query graph generation is a practical semantic parsing-based method. But existing works rely on heuristic rules with limited coverage, making them impractical on more complex questions. This paper proposes a Director-Actor-Critic framework to overcome these challenges. Through options over a Markov Decision Process, query graph generation is formulated as a hierarchical decision problem. The Director determines which types of triples the query graph needs, the Actor generates corresponding triples by choosing nodes and edges, and the Critic calculates the semantic similarity between the generated triples and the given questions. Moreover, to train from weak supervision, we base the framework on hierarchical Reinforcement Learning with intrinsic motivation. To accelerate the training process, we pre-train the Critic with high-reward trajectories generated by hand-crafted rules, and leverage curriculum learning to gradually increase the complexity of questions during query graph generation. Extensive experiments conducted over widely-used benchmark datasets demonstrate the effectiveness of the proposed framework.