论文检索

Haining Wang, Kang He, Bobo Li, Lei Chen, Fei Li, Xu Han, Chong Teng, Donghong Ji

Aspect-based Sentiment Analysis (ABSA) is extensively researched in the NLP community, yet related models face challenges due to data sparsity when shifting to a new domain. Hence, data augmentation for cross-domain ABSA has attracted increasing attention in recent years. However, two key points have been neglected in prior studies: First, target domain unlabeled data are labeled with pseudo labels by the model trained in the source domain with little quality control, leading to inaccuracy and error propagation. Second, the label and text patterns of generated labeled data are monotonous, thus limiting the robustness and generalization ability of trained ABSA models. In this paper, we aim to design a simple yet effective framework to address the above shortages in ABSA data augmentation, called Refining and Synthesis Data Augmentation (RSDA). Our framework roughly includes two steps: First, it refines generated labeled data using a natural language inference (NLI) filter to control data quality. Second, it synthesizes diverse labeled data via novel label composition and paraphrase approaches. We conduct experiments on 4 kinds of ABSA subtasks, and our framework outperforms 7 strong baselines, demonstrating its effectiveness.

Yubao Tang, Ruqing Zhang, Jiafeng Guo, Maarten de Rijke, Yixing Fan, Xueqi Cheng

Generative retrieval uses differentiable search indexes to directly generate relevant document identifiers in response to a query. Recent studies have highlighted the potential of a strong generative retrieval model, trained with carefully crafted pre-training tasks, to enhance downstream retrieval tasks via fine-tuning. However, the full power of pre-training for generative retrieval remains underexploited due to its reliance on pre-defined static document identifiers, which may not align with evolving model parameters. In this work, we introduce BootRet, a bootstrapped pre-training method for generative retrieval that dynamically adjusts document identifiers during pre-training to accommodate the continuing memorization of the corpus. BootRet involves three key training phases: (i) initial identifier generation, (ii) pre-training via corpus indexing and relevance prediction tasks, and (iii) bootstrapping for identifier updates. To facilitate the pre-training phase, we further introduce noisy documents and pseudo-queries, generated by large language models, to resemble semantic connections in both indexing and retrieval tasks. Experimental results demonstrate that BootRet significantly outperforms existing pre-training generative retrieval baselines and performs well even in zero-shot settings.

Tianyi Yan, Fei Wang, James Y. Huang, Wenxuan Zhou, Fan Yin, Aram Galstyan, Wenpeng Yin, Muhao Chen

Instruction tuning has been used as a promising approach to improve the performance of large language models (LLMs) on unseen tasks. However, current LLMs exhibit limited robustness to unseen instructions, generating inconsistent outputs when the same instruction is phrased with slightly varied forms or language styles. This behavior indicates LLMs’ lack of robustness to textual variations and generalizability to unseen instructions, potentially leading to trustworthiness issues. Accordingly, we propose Contrastive Instruction Tuning, which maximizes the similarity between the hidden representations of semantically equivalent instruction-instance pairs while minimizing the similarity between semantically different ones. To facilitate this approach, we augment the existing FLAN collection by paraphrasing task instructions. Experiments on the PromptBench benchmark show that CoIN consistently improves LLMs’ robustness to unseen instructions with variations across character, word, sentence, and semantic levels by an average of +2.5% in accuracy.

Congda Ma, Tianyu Zhao, Manabu Okumura

Due to biases inherently present in data for pre-training, current pre-trained Large Language Models (LLMs) also ubiquitously manifest the same phenomena. Since the bias influences the output from the LLMs across various tasks, the widespread deployment of the LLMs is hampered. We propose a simple method that utilizes structured knowledge to alleviate this issue, aiming to reduce the bias embedded within the LLMs and ensuring they have an encompassing perspective when used in applications. Experimental results indicated that our method has good debiasing ability when applied to existing both autoregressive and masked language models. Additionally, it could ensure that the performances of LLMs on downstream tasks remain uncompromised.Our method outperforms state-of-the-art (SOTA) baselines in the debiasing ability. Importantly, our method obviates the need for training from scratch, thus offering enhanced scalability and cost-effectiveness.

Xin Zhou, Yi Lu, Ruotian Ma, Yujian Wei, Tao Gui, Qi Zhang, Xuanjing Huang

Large language models (LLMs) have shown great potential to empower various domains and are often customized by fine-tuning for the requirements of different applications. However, the powerful learning ability of LLMs not only enables them to learn new tasks but also makes them vulnerable to learning undesired behaviors, such as harmfulness and hallucination, as the fine-tuning data often implicitly or explicitly contains such content. Can we fine-tune LLMs on harmful data without learning harmful behaviors? This paper proposes a controllable training framework to make undesired behaviors unlearnable during the fine-tuning process. Specifically, we introduce security vectors to control the model’s behavior and make it consistent with the undesired behavior. Security vectors are activated during fine-tuning, the consistent behavior makes the model believe that such behavior has already been learned and there is no need for further optimization, while inconsistent data can still be learned. After fine-tuning, security vectors are deactivated to restore the LLM’s normal behavior. Our experiments show that the security vectors can prevent LLM from learning harmful and hallucination behavior while preserving the ability to learn other information.

Huimin Wang, Yutian Zhao, Xian Wu, Yefeng Zheng

Automatic evaluation of natural language generation (NLG) tasks has gained extensive research interests, since it can rapidly assess the performance of large language models (LLMs). However, automatic NLG evaluation struggles with medical QA because it fails to focus on the crucial correctness of medical facts throughout the generated text. To address this, this paper introduces a new data structure, imap, designed to capture key information in questions and answers, enabling evaluators to focus on essential details. The imap comprises three components: Query, Constraint, and Inform, each of which is in the form of term-value pairs to represent medical facts in a structural manner. We then introduce imapScore, which compares the corresponding medical term-value pairs in the imap to score generated texts. We utilize GPT-4 to extract imap from questions, human-annotated answers, and generated responses. To mitigate the diversity in medical terminology for fair term-value pairs comparison, we use a medical knowledge graph to assist GPT-4 in determining matches. To compare imapScore with existing NLG metrics, we establish a new benchmark dataset. The experimental results show that imapScore consistently outperforms state-of-the-art metrics, demonstrating an average improvement of 79.8% in correlation with human scores. Furthermore, incorporating imap into n-gram, embedding, and LLM metrics boosts the base versions, increasing correlation with human scores by averages of 89.9%, 81.7%, and 32.6%, respectively.

YiQiu Guo, Yuchen Yang, Ya Zhang, Yu Wang, Yanfeng Wang

Structured data offers an efficient means of organizing information. Exsisting text-serialization based methods for processing structured data using large language models (LLMs) are not designed to explicitly capture the heterogeneity of structured data. Such methods are suboptimal for LLMs to process structured data, and may lead to large input token size and poor robustness to input perturbation. In this paper, we propose a novel framework called DictLLM, which is an efficient and effective framework for the modeling of medical lab report to deal with the report-assisted diagnosis generation task. DictLLM introduce 1) group positional encoding to maintain the permutation invariance, 2) hierarchical attention bias to capture the inductive bias of structured data, and 3) a optimal transport alignment layer to align the embeddings generated by the dict encoder with the LLM, producing a list of fixed-length virtual tokens. We conduct experiments with multiple LLM models on a large-scale real-world medical lab report dataset for automatic diagnosis generation. The results show that our proposed framework outperforms the baseline methods and few-shot GPT-4 in terms of both Rouge-L and Knowledge F1 score. We also conduct multiple experiments and analyze the scalability and robustness of our proposed framework, demonstrating the superiority of our method in modeling the heterogeneous structure of medical dictionaries data.

Jinwen He, Yujia Gong, Zijin Lin, Cheng’an Wei, Yue Zhao, Kai Chen

Large Language Models (LLMs) have revolutionized various domains with extensive knowledge and creative capabilities. However, a critical issue with LLMs is their tendency to produce outputs that diverge from factual reality. This phenomenon is particularly concerning in sensitive applications such as medical consultation and legal advice, where accuracy is paramount. Inspired by human lie detectors using physiological responses, we introduce the LLM Factoscope, a novel Siamese network-based model that leverages the inner states of LLMs for factual detection. Our investigation reveals distinguishable patterns in LLMs’ inner states when generating factual versus non-factual content. We demonstrate its effectiveness across various architectures, achieving over 96% accuracy on our custom-collected factual detection dataset. Our work opens a new avenue for utilizing LLMs’ inner states for factual detection and encourages further exploration into LLMs’ inner workings for enhanced reliability and transparency.

Yufei Li, Xiao Yu, Yanghong Guo, Yanchi Liu, Haifeng Chen, Cong Liu

Joint entity and relation extraction is a process that identifies entity pairs and their relations using a single model. We focus on the problem of joint extraction in distantly-labeled data, whose labels are generated by aligning entity mentions with the corresponding entity and relation tags using a knowledge base (KB). One key challenge is the presence of noisy labels arising from both incorrect entity and relation annotations, which significantly impairs the quality of supervised learning. Existing approaches, either considering only one source of noise or making decisions using external knowledge, cannot well-utilize significant information in the training data. We propose DENRL, a generalizable framework that 1) incorporates a lightweight transformer backbone into a sequence labeling scheme for joint tagging, and 2) employs a noise-robust framework that regularizes the tagging model with significant relation patterns and entity-relation dependencies, then iteratively self-adapts to instances with less noise from both sources. Surprisingly, experiments on two benchmark datasets show that DENRL, using merely its own parametric distribution and simple data-driven heuristics, outperforms strong baselines by a large margin with better interpretability.

Yong Xie, Karan Aggarwal, Aitzaz Ahmad

Large language models (LLMs) have demonstrated remarkable open-domain capabilities. LLMs tailored for a domain are typically trained entirely on domain corpus to excel at handling domain-specific tasks. In this work, we explore an alternative strategy of continual pre-training as a means to develop domain-specific LLMs over an existing open-domain LLM. We introduce FinPythia-6.9B, developed through domain-adaptive continual pre-training on the financial domain.Continual pre-trained FinPythia showcases consistent improvements on financial tasks over the original foundational model. We further explore simple but effective data selection strategies for continual pre-training. Our data selection strategies outperform vanilla continual pre-training’s performance with just 10% of corpus size and cost, without any degradation on open-domain standard tasks. Our work proposes an alternative solution to building domain-specific LLMs cost-effectively.

Feng Zhang, Wei Chen, Fei Ding, Meng Gao, Tengjiao Wang, Jiahui Yao, Jiabin Zheng

Intent detection aims to identify user goals from utterances, and is a ubiquitous step towards the satisfaction of user desired needs in many interaction systems. As dynamic and varied intents arise, models that are capable of identifying new intents promptly are required. However, existing studies usually fine-tune discriminative models on the specific defined intent classes, precluding them from being directly adopted to new intent domains. In this paper, we introduce a generative pre-trained intent model that can recognize new intents from different domains in low-resource scenarios. We reformulate intent detection into a generation task and design descriptive and regularized instructions to guide the model effectively to detect new intents in open domains with no parameter updates. To validate the proposed method, we introduce a new intent detection benchmark, including the Meta-Intent Dataset and three types of representative evaluation settings. We conduct extensive experiments which demonstrate that our method outperforms a range of strong baselines that needs further fine-tuning or domain-specific samples.

Nishant Balepur, Shramay Palta, Rachel Rudinger

Chain-of-thought (COT) prompting can help large language models (LLMs) reason toward correct answers, but its efficacy in reasoning toward incorrect answers is unexplored. This process of elimination (PoE), when used with COT, can enhance self-consistency, interpretability, and tasks such as medical diagnoses of exclusion. Thus, we propose PoE with COT, where LLMs must reason toward incorrect options on multiple-choice questions. We evaluate the ability of GPT-3.5, LLaMA-2, and Falcon to perform PoE with COT on a total of four commonsense and scientific reasoning datasets. We find that the strategy of PoE always underperforms the strategy of choosing the correct answer. The agreement of these strategies is also lower than the self-consistency of each strategy. To study these issues further, we conduct error analyses and give suggestions for future work.

Yiquan Wu, Anlai Zhou, Yuhang Liu, Yifei Liu, Adam Jatowt, Weiming Lu, Jun Xiao, Kun Kuang

In-context learning (ICL) has emerged as a powerful tool for enhancing large language models (LLMs) in addressing downstream tasks. In this paper, we explore the vital task of example selection in ICL by mimicking the human learning process. We propose a Chain-of-Quizzes (CoQ) framework inspired by educational theories such as Bruner’s Spiral Learning and Mastery Learning theory. Specifically, our framework employs the LLMs to answer the quiz (question in the example) to sift ‘good’ examples, combines these examples iteratively with the increasing complexity, and utilizes a final exam to gauge the combined example chains. Our extensive experiments on diverse reasoning datasets show the proposed approach outperforms baseline models. These findings underscore the framework’s potential for future research.

Rishabh Adiga, Lakshmi Subramanian, Varun Chandrasekaran

Pretrained language models (PLMs) have shown remarkable few-shot learning capabilities when provided with properly formatted examples. However, selecting the “best” examples remains an open challenge. We propose a complexity-based prompt selection approach for sequence tagging tasks. This approach avoids the training of a dedicated model for selection of examples, and instead uses certain metrics to align the syntactico-semantic complexity of test sentences and examples. We use both sentence- and word-level metrics to match the complexity of examples to the (test) sentence being considered. Our results demonstrate that our approach extracts greater performance from PLMs: it achieves state-of-the-art performance on few-shot NER, achieving a 5% absolute improvement in F1 score on the CoNLL2003 dataset for GPT-4. We also see large gains of upto 28.85 points (F1/Acc.) in smaller models like GPT-j-6B.

Sirou Chen, Sakiko Yahata, Shuichiro Shimizu, Zhengdong Yang, Yihang Li, Chenhui Chu, Sadao Kurohashi

Emotion plays a crucial role in human conversation. This paper underscores the significance of considering emotion in speech translation. We present the MELD-ST dataset for the emotion-aware speech translation task, comprising English-to-Japanese and English-to-German language pairs. Each language pair includes about 10,000 utterances annotated with emotion labels from the MELD dataset. Baseline experiments using the SeamlessM4T model on the dataset indicate that fine-tuning with emotion labels can enhance translation performance in some settings, highlighting the need for further research in emotion-aware speech translation systems.

Gökçe Uludoğan, Zeynep Balal, Furkan Akkurt, Meliksah Turker, Onur Gungor, Susan Üsküdarlı

The recent advances in natural language processing have predominantly favored well-resourced English-centric models, resulting in a significant gap with low-resource languages. In this work, we introduce TURNA, a language model developed for the low-resource language Turkish and is capable of both natural language understanding and generation tasks.TURNA is pretrained with an encoder-decoder architecture based on the unified framework UL2 with a diverse corpus that we specifically curated for this purpose. We evaluated TURNA with three generation tasks and five understanding tasks for Turkish. The results show that TURNA outperforms several multilingual models in both understanding and generation tasks and competes with monolingual Turkish models in understanding tasks.

Arda Uzunoğlu, Abdulfattah Safa, Gözde Gül Şahin

Recently, there has been growing interest within the community regarding whether large language models are capable of planning or executing plans. However, most prior studies use LLMs to generate high-level plans for simplified scenarios lacking linguistic complexity and domain diversity, limiting analysis of their planning abilities. These setups constrain evaluation methods (e.g., predefined action space), architectural choices (e.g., only generative models), and overlook the linguistic nuances essential for realistic analysis. To tackle this, we present PARADISE, an abductive reasoning task using Q&A format on practical procedural text sourced from wikiHow. It involves tip and warning inference tasks directly associated with goals, excluding intermediary steps, with the aim of testing the ability of the models to infer implicit knowledge of the plan solely from the given goal. Our experiments, utilizing fine-tuned language models and zero-shot prompting, reveal the effectiveness of task-specific small models over large language models in most scenarios. Despite advancements, all models fall short of human performance. Notably, our analysis uncovers intriguing insights, such as variations in model behavior with dropped keywords, struggles of BERT-family and GPT-4 with physical and abstract goals, and the proposed tasks offering valuable prior knowledge for other unseen procedural tasks. The PARADISE dataset and associated resources are publicly available for further research exploration with https://anonymous.4open.science/r/paradise-53BD/README.md.

Xiongtao Zhou, Jie He, Yuhua Ke, Guangyao Zhu, Victor Gutierrez Basulto, Jeff Pan

Multimodal Large Language Models (MLLMs) fine-tuned with multimodal instruction-following data have demonstrated formidable capabilities in multimodal tasks. However, fine-tuning all parameters of MLLMs has become challenging due to the rapid growth of the overall model’s parameters. To address this issue, we study Parameter-Efficient Fine-Tuning (PEFT) methods for MLLMs. We aim to identify effective methods for enhancing performance in scenarios where only a limited number of parameters are trained. This paper conducts empirical studies that employ four widely used PEFT methods to fine-tune the LLM component of open-source MLLMs. We present a comprehensive analysis that encompasses various aspects, including the impact of PEFT methods on various models, parameters and location of PEFT module, fine-tuning data scale, model stability based on PEFT method, MLLM’s generalization, and hallucination. We evaluated four PEFT methods on seven datasets from two different categories, unseen and seen datasets. Across all experiments, we show that the adapter is the best-performing PEFT method in various aspects. At the same time, fine-tuning the connector layers leads to improved performance in most MLLMs.

David Wan, Koustuv Sinha, Srini Iyer, Asli Celikyilmaz, Mohit Bansal, Ramakanth Pasunuru

The impressive generation capabilities of large language models (LLMs) have made it harder to detect the subtle hallucinations they make in abstractive summarization, where generated summaries consist of a blend of correct and incorrect information w.r.t. a given document. Recently-proposed LLM-based evaluation metrics attempt to capture this, but still face challenges: (1) they are biased towards summaries generated from the same underlying LLM, and (2) they lack interpretability, offering only a single score. In this work, we present ACUEval, a metric that leverages the power of LLMs to perform two sub-tasks: decomposing summaries into atomic content units (ACUs), and validating them against the source document. Compared to current strong LLM-based metrics, our two-step evaluation strategy improves correlation with human judgments of faithfulness on three summarization evaluation benchmarks by 3% in balanced accuracy compared to the next-best metric, and also shows reduced preference bias towards LLM-generated summary. Further, we show that errors detected by ACUEval can be used to generate actionable feedback for refining the summary, improving the faithfulness scores by more than 10%.

Raghuveer Peri, Sai Muralidhar Jayanthi, Srikanth Ronanki, Anshu Bhatia, Karel Mundnich, Saket Dingliwal, Nilaksh Das, Zejiang Hou, Goeric Huybrechts, Srikanth Vishnubhotla 等

Integrated Speech and Large Language Models (SLMs) that can follow speech instructions and generate relevant text responses have gained popularity lately. However, the safety and robustness of these models remains largely unclear. In this work, we investigate the potential vulnerabilities of such instruction-following speech-language models to adversarial attacks and jailbreaking. Specifically, we design algorithms that can generate adversarial examples to jailbreak SLMs in both white-box and black-box attack settings without human involvement. Additionally, we propose countermeasures to thwart such jailbreaking attacks. Our models, trained on dialog data with speech instructions, achieve state-of-the-art performance on spoken question-answering task, scoring over 80% on both safety and helpfulness metrics. Despite safety guardrails, experiments on jailbreaking demonstrate the vulnerability of SLMs to adversarial perturbations and transfer attacks, with average attack success rates of 90% and 10% respectively when evaluated on a dataset of carefully designed harmful questions spanning 12 different toxic categories. However, we demonstrate that our proposed countermeasures reduce the attack success significantly.