COLING 2025 Papers

Code search aims to quickly locate target code snippets from databases using natural language queries, which promotes code reusability. Existing methods can effectively obtain aligned token-level and query word-level features. However, these studies usually represent the semantics of code and query by averaging the features of each token and word respectively, which makes it difficult to accurately capture the code details that are closely related to the query. To address this issue, we propose a fine-grained code search model that consists of a cross-modal encoder, a mapping layer, and a classification layer. Specifically, we utilize a pre-trained model, GraphCodeBERT, in the cross-modal encoder to align features. In the mapping layer, we introduce a co-attention network to capture the fine-grained interactions between code and query, ensuring a model can precisely identify key code segments relevant to the query. Finally, in the classification layer, we incorporate instruction learning techniques that leverage contextual reasoning to improve the accuracy of query-code matching. Experimental results show that our proposed model significantly outperforms existing methods across multiple programming language datasets.

Large Language Models (LLMs) have demonstrated impressive capabilities across a wide range of tasks. However, their proficiency and reliability in the specialized domain of financial data analysis, particularly focusing on data-driven thinking, remain uncertain. To bridge this gap, we introduce FinDABench, a comprehensive benchmark designed to evaluate the financial data analysis capabilities of LLMs within this context. The benchmark comprises 15,200 training instances and 8,900 test instances, all meticulously crafted by human experts. FinDABench assesses LLMs across three dimensions: 1) Core Ability, evaluating the models’ ability to perform financial indicator calculation and corporate sentiment risk assessment; 2) Analytical Ability, determining the models’ ability to quickly comprehend textual information and analyze abnormal financial reports; and 3) Technical Ability, examining the models’ use of technical knowledge to address real-world data analysis challenges involving analysis generation and charts visualization from multiple perspectives. We will release FinDABench, and the evaluation scripts at https://github.com/xxx. FinDABench aims to provide a measure for in-depth analysis of LLM abilities and foster the advancement of LLMs in the field of financial data analysis.

Large Language Models (LLMs) present massive inherent knowledge and superior semantic comprehension capability, which have revolutionized various tasks in natural language processing. Despite their success, a critical gap remains in enabling LLMs to perform knowledge graph completion (KGC). Empirical evidence suggests that LLMs consistently perform worse than conventional KGC approaches, even through sophisticated prompt design or tailored instruction-tuning. Fundamentally, applying LLMs on KGC introduces several critical challenges, including a vast set of entity candidates, hallucination issue of LLMs, and under-exploitation of the graph structure. To address these challenges, we propose a novel instruction-tuning-based method, namely FtG. Specifically, we present a filter-then-generate paradigm and formulate the KGC task into a multiple-choice question format. In this way, we can harness the capability of LLMs while mitigating the issue casused by hallucinations. Moreover, we devise a flexible ego-graph serialization prompt and employ a structure-text adapter to couple structure and text information in a contextualized manner. Experimental results demonstrate that FtG achieves substantial performance gain compared to existing state-of-the-art methods. The instruction dataset and code are available at https://github.com/LB0828/FtG.

Named-entity recognition (NER) is a task that typically requires large annotated datasets, which limits its applicability across domains with varying entity definitions. This paper addresses few-shot NER, aiming to transfer knowledge to new domains with minimal supervision. Unlike previous approaches that rely solely on limited annotated data, we propose a weakly-supervised algorithm that combines small labeled datasets with large amounts of unlabeled data. Our method extends the k-means algorithm with label supervision, cluster size constraints, and domain-specific discriminative subspace selection. This unified framework achieves state-of-the-art results in few-shot NER, demonstrating its effectiveness in leveraging unlabeled data and adapting to domain-specific challenges.

Recent advancements in Large Language Models and Retrieval-Augmented Generation have boosted interest in domain-specific question-answering for enterprise products. However, AI Assistants often face challenges in multi-product QA settings, requiring accurate responses across diverse domains. Existing multi-domain RAG-QA approaches either query all domains indiscriminately, increasing computational costs and LLM hallucinations, or rely on rigid resource selection, which can limit search results. We introduce MKP-QA, a novel multi-product knowledge-augmented QA framework with probabilistic federated search across domains and relevant knowledge. This method enhances multi-domain search quality by aggregating query-domain and query-passage probabilistic relevance. To address the lack of suitable benchmarks for multi-product QAs, we also present new datasets focused on three Adobe products: Adobe Experience Platform, Target, and Customer Journey Analytics. Our experiments show that MKP-QA significantly boosts multi-product RAG-QA performance in terms of both retrieval accuracy and response quality.

Federated learning-based Named Entity Recognition (FNER) has attracted widespread attention through decentralized training on local clients. However, most FNER models assume that entity types are pre-fixed, so in practical applications, local clients constantly receive new entity types without enough storage to access old entity types, resulting in severe forgetting on previously learned knowledge. In addition, new clients collecting only new entity types may join the global training of FNER irregularly, further exacerbating catastrophic forgetting. To overcome the above challenges, we propose a Forgetting-Subdued Learning (FSL) model which solves the forgetting problem on old entity types from both intra-client and inter-client two aspects. Specifically, for intra-client aspect, we propose a prototype-guided adaptive pseudo labeling and a prototypical relation distillation loss to surmount catastrophic forgetting of old entity types with semantic shift. Furthermore, for inter-client aspect, we propose a task transfer detector. It can identify the arrival of new entity types that are protected by privacy and store the latest old global model for relation distillation. Qualitative experiments have shown that our model has made significant improvements compared to several baseline methods.

Recent research in federated large language models (LLMs) has primarily focused on enabling clients to fine-tune their locally deployed homogeneous LLMs collaboratively or on transferring knowledge from server-based LLMs to small language models (SLMs) at downstream clients. However, a significant gap remains in the simultaneous mutual enhancement of both the server’s LLM and clients’ SLMs. To bridge this gap, we propose FedMKT, a parameter-efficient federated mutual knowledge transfer framework for large and small language models. This framework is designed to adaptively transfer knowledge from the server’s LLM to clients’ SLMs while concurrently enhancing the LLM with clients’ unique domain insights. We facilitate token alignment using minimum edit distance (MinED) and then selective mutual knowledge transfer between client-side SLMs and a server-side LLM, aiming to collectively enhance their performance. Through extensive experiments across three distinct scenarios, we evaluate the effectiveness of FedMKT by utilizing diverse public LLMs and SLMs on a variety of NLP text generation tasks. Empirical results demonstrate that FedMKT simultaneously boosts the performance of both LLMs and SLMs. Our code has been contributed to the FATE open-source project and is now publicly accessible at https://github.com/FederatedAI/FATE-LLM/tree/main/python/fate_llm/algo/fedmkt

Cross-domain sequential recommendation (CSR) has garnered significant attention. Current federated frameworks for CSR leverage information across multiple domains but often rely on user alignment, which increases communication costs and privacy risks. In this work, we propose FedCSR, a novel federated cross-domain sequential recommendation framework that eliminates the need for user alignment between platforms. FedCSR fully utilizes cross-domain knowledge to address the key challenges related to data heterogeneity both inter- and intra-platform. To tackle the heterogeneity of data patterns between platforms, we introduce Model Contrastive Learning (MCL) to reduce the gap between local and global models. Additionally, we design Sequence Contrastive Learning (SCL) to address the heterogeneity of user preferences across different domains within a platform by employing tailored sequence augmentation techniques. Extensive experiments conducted on multiple real-world datasets demonstrate that FedCSR achieves superior performance compared to existing baseline methods.

We introduce FarExStance, a new dataset for explainable stance detection in Farsi. Each instance in this dataset contains a claim, the stance of an article or social media post towards that claim, and an extractive explanation which provides evidence for the stance label. We compare the performance of a fine-tuned multilingual RoBERTa model to several large language models in zero-shot, few-shot, and parameter-efficient fine-tuned settings on our new dataset. On stance detection, the most accurate models are the fine-tuned RoBERTa model, the LLM Aya-23-8B which has been fine-tuned using parameter-efficient fine-tuning, and few-shot Claude-3.5-Sonnet. Regarding the quality of the explanations, our automatic evaluation metrics indicate that few-shot GPT-4o generates the most coherent explanations, while our human evaluation reveals that the best Overall Explanation Score (OES) belongs to few-shot Claude-3.5-Sonnet. The fine-tuned Aya-32-8B model produced explanations most closely aligned with the reference explanations.

KG-based fact verification verifies the truthfulness of claims by retrieving evidence graphs from the knowledge graph. The *faithful inference chains*, which are precise relation paths between the mentioned entities and evidence entities, retrieve precise evidence graphs addressing poor performance and weak logic for fact verification. Due to the diversity of relation paths, existing methods rarely extract faithful inference chains. To alleviate these issues, we propose Multi-view Heterogeneous Graph with Causal Intervention (MHGCI): (i) We construct a Multi-view Heterogeneous Graph enhancing relation path extraction from the view of different mentioned entities. (ii) We propose a self-optimizing causal intervention model to generate assistant entities mitigating the out-of-distribution problem caused by counterfactual relations. (iii) We propose a grounding method to extract evidence graphs from the KG by faithful inference chains. Experiments on the public KG-based fact verification dataset FactKG demonstrate that our model provides precise evidence graphs and achieves state-of-the-art performance.

Language models encode extensive factual knowledge within their parameters. The accurate assessment of this knowledge is crucial for understanding and improving these models. In the literature, factual knowledge assessment often relies on cloze sentences, which can lead to erroneous conclusions due to the complexity of natural language (out-of-subject continuations, the existence of many correct answers and the several ways of expressing them). In this paper, we introduce a new interpretable knowledge assessment method that mitigates these issues by leveraging distractors—incorrect but plausible alternatives to the correct answer. We propose several strategies for retrieving distractors and determine the most effective one through experimentation. Our method is evaluated against existing approaches, demonstrating solid alignment with human judgment and stronger robustness to verbalization artifacts. The code and data to reproduce our experiments are available on GitHub.

Factual consistency is an important quality in dialogue summarization. Large language model (LLM)-based automatic text summarization models generate more factually consistent summaries compared to those by smaller pretrained language models, but they face deployment challenges in real-world applications due to privacy or resource constraints. In this paper, we investigate the use of symbolic knowledge distillation to improve the factual consistency of smaller pretrained models for dialogue summarization. We employ zero-shot learning to extract symbolic knowledge from LLMs, generating both factually consistent (positive) and inconsistent (negative) summaries. We then apply two contrastive learning objectives on these summaries to enhance smaller summarization models. Experiments with BART, PEGASUS, and Flan-T5 indicate that our approach surpasses strong baselines that rely on complex data augmentation strategies. Our approach demonstrates improved factual consistency while preserving coherence, fluency, and relevance, as verified by both automatic evaluation metrics and human assessments. We provide access to the data and code to facilitate future research.

Despite the massive success of fine-tuning Pre-trained Language Models (PLMs), they remain susceptible to out-of-distribution input. Dataset cartography is a simple yet effective dual-model approach that improves the robustness of fine-tuned PLMs. It involves fine-tuning a model on the original training set (i.e. reference model), selecting a subset of important training instances based on the training dynamics, % of the reference model, and fine-tuning again only on these selected examples (i.e. main model). However, this approach requires fine-tuning the same model twice, which is computationally expensive for large PLMs. In this paper, we show that 1) training dynamics are highly transferable across model sizes and pre-training methods, and that 2) fine-tuning main models using these selected training instances achieves higher training efficiency than empirical risk minimization (ERM). Building on these observations, we propose a novel fine-tuning approach: Fine-Tuning by transFerring Training dynamics (FTFT). Compared with dataset cartography, FTFT uses more efficient reference models and aggressive early stopping. FTFT achieves robustness improvements over ERM while lowering the training cost by up to ~50%

In this work, we propose Few Shot Domain Adapting Graph (FS-DAG), a scalable and efficient model architecture for visually rich document understanding (VRDU) in few-shot settings. FS-DAG leverages domain-specific and language/vision specific backbones within a modular framework to adapt to diverse document types with minimal data. The model is robust to practical challenges such as handling OCR errors, misspellings, and domain shifts, which are critical in real-world deployments. FS-DAG is highly performant with less than 90M parameters, making it well-suited for complex real-world applications for Information Extraction (IE) tasks where computational resources are limited. We demonstrate FS-DAG’s capability through extensive experiments for information extraction task, showing significant improvements in convergence speed and performance compared to state-of-the-art methods. Additionally, this work highlights the ongoing progress in developing smaller, more efficient models that do not compromise on performance.

When carefully optimized by human experts, naive prompts can significantly enhance the task performance of large language models (LLMs). However, such expert-driven prompt optimizations are resource-intensive. To address this, some studies have proposed Automatic Prompt Optimization (APO), which refines naive prompts according to task outputs from in-box testing models, utilizing advanced LLMs (e.g., GPT-4) in an ad-hoc way. Although effective, current approaches face challenges in generalization and privacy risks. To overcome these limitations, we have developed the first large-scale Prompt Optimization Preference (POP) dataset, fine-tuned offline local LLM-based optimizers, and conducted fairly evaluations across various downstream models. Our method, named Free-from Instruction-oriented Prompt Optimization (FIPO), allows precise optimization of the core task instructions in naive prompts in a model-agnostic manner. FIPO uses a modular APO template that dynamically incorporates the naive task instructions, optional instruction responses, and optional ground truth to produce refined prompts. The POP dataset is meticulously constructed using advanced LLMs, undergoing rigorous cross-validation by human experts and analytical models. By leveraging insights from this dataset, along with Tulu2 models and diverse fine-tuning strategies, we validate the efficacy of the FIPO framework across five public benchmarks and six testing models. Our dataset and codes are available at: https://github.com/LuJunru/FIPO_Project.

Recent developments in Natural Language Processing (NLP) for argument mining offer new opportunities to analyze the argumentative units (AUs) in student essays. These advancements can be leveraged to provide automatically generated feedback and exercises for students engaging in online argumentative essay writing practice. Writing standards for both native English speakers (L1) and English-as-a-foreign-language (L2) learners require students to understand formal essay structures and different AUs. To address this need, we developed FEAT-writing (Feedback and Exercises for Argumentative Training in writing), an interactive system that provides students with automatically generated exercises and distinct feedback on their argumentative writing. In a preliminary evaluation involving 346 students, we assessed the impact of six different automated feedback types on essay quality, with results showing general improvements in writing after receiving feedback from the system.

From ice cream flavors to climate change, people exhibit a wide array of opinions on various topics, and understanding the rationale for these opinions can promote healthy discussion and consensus among them. As such, it can be valuable for a large language model (LLM), particularly as an AI assistant, to be able to empathize with or even explain these various standpoints. In this work, we hypothesize that different topic stances often manifest correlations that can be used to extrapolate to topics with unknown opinions. We explore various prompting and fine-tuning methods to improve an LLM’s ability to (a) extrapolate from opinions on known topics to unknown ones and (b) support their extrapolation with reasoning. Our findings suggest that LLMs possess inherent knowledge from training data about these opinion correlations, and with minimal data, the similarities between human opinions and model-extrapolated opinions can be improved by more than 50%. Furthermore, LLM can generate the reasoning process behind their extrapolation of opinions.

The volume of academic articles is increasing rapidly, reflecting the growing emphasis on research and scholarship across different science disciplines. This rapid growth necessitates the development of tools for more efficient and rapid understanding of these articles. Clear and well-defined Research Questions (RQs) in research articles can help guide scholarly inquiries. However, many academic studies lack a proper definition of RQs in their articles. This research addresses this gap by presenting a comprehensive framework for the systematic extraction, detection, and generation of RQs from scientific articles. The extraction component uses a set of regular expressions to identify articles containing well-defined RQs. The detection component aims to identify more complex RQs in articles, beyond those captured by the rule-based extraction method. The RQ generation focuses on creating RQs for articles that lack them. We integrate all these components to build a pipeline to extract RQs or generate them based on the articles’ full text. We evaluate the performance of the designed pipeline on a set of metrics designed to assess the quality of RQs. Our results indicate that the proposed pipeline can reliably detect RQs and generate high-quality ones.

Recent advancements have integrated the execution process and feedback into the training of large language models for code generation, demonstrating enhanced model performance. However, current methods amalgamate erroneous code with feedback and the final correct code as target sentences, inadvertently increasing the probability of generating both correct and incorrect code during inference. While multiple iterations of feedback can eventually yield the correct answer, this iterative process is cumbersome and time-consuming for users who prefer immediate accurate results. To address this challenge, we propose ConCoder, a contrastive learning-based code generation model with execution feedback. This approach enables the model to efficiently produce accurate code from the outset while rectifying and optimizing the incorrect code. Furthermore, our training emphasizes learning from the causes of errors, allowing the model to understand and avoid mistakes. Through extensive experiments, ConCoder demonstrates significant improvements in generating accurate code and understanding error correction, paving the way for more reliable code generation models.