Authors: Jiaming Shen, Tianqi Liu, Jialu Liu, Zhen Qin, Jay Pavagadhi, Simon Baumgartner, Michael Bendersky
Abstract: The popularity of automated news headline generation has surged with advancements in pre-trained language models. However, these models often suffer from the ``hallucination'' problem, where the generated headline is not fully supported by its source article. Efforts to address this issue have predominantly focused on English, using over-simplistic classification schemes that overlook nuanced hallucination types. In this study, we introduce the first multilingual, fine-grained news headline hallucination detection dataset that contains over 11 thousand pairs in 5 languages, each annotated with detailed hallucination types by experts. We conduct extensive experiments on this dataset under two settings. First, we implement several supervised fine-tuning approaches as preparatory solutions and demonstrate this dataset's challenges and utilities. Second, we test various large language models' in-context learning abilities and propose two novel techniques, language-dependent demonstration selection and coarse-to-fine prompting, to boost the few-shot hallucination detection performance in terms of the example-F1 metric. We release this dataset to foster further research in multilingual, fine-grained headline hallucination detection.
Authors: Sophia Zhi, Roger P. Levy, Stephan C. Meylan
Abstract: One of the many tasks facing the typically-developing child language learner is learning to discriminate between the distinctive sounds that make up words in their native language. Here we investigate whether multimodal information--specifically adult speech coupled with video frames of speakers' faces--benefits a computational model of phonetic learning. We introduce a method for creating high-quality synthetic videos of speakers' faces for an existing audio corpus. Our learning model, when both trained and tested on audiovisual inputs, achieves up to a 8.1% relative improvement on a phoneme discrimination battery compared to a model trained and tested on audio-only input. It also outperforms the audio model by up to 3.9% when both are tested on audio-only data, suggesting that visual information facilitates the acquisition of acoustic distinctions. Visual information is especially beneficial in noisy audio environments, where an audiovisual model closes 67% of the loss in discrimination performance of the audio model in noise relative to a non-noisy environment. These results demonstrate that visual information benefits an ideal learner and illustrate some of the ways that children might be able to leverage visual cues when learning to discriminate speech sounds.
Authors: John Palowitch, Hamidreza Alvari, Mehran Kazemi, Tanvir Amin, Filip Radlinski
Abstract: Web authors frequently embed social media to support and enrich their content, creating the potential to derive web-based, cross-platform social media representations that can enable more effective social media retrieval systems and richer scientific analyses. As step toward such capabilities, we introduce a novel language modeling framework that enables automatic annotation of roles that social media entities play in their embedded web context. Using related communication theory, we liken social media embeddings to quotes, formalize the page context as structured natural language signals, and identify a taxonomy of roles for quotes within the page context. We release SocialQuotes, a new data set built from the Common Crawl of over 32 million social quotes, 8.3k of them with crowdsourced quote annotations. Using SocialQuotes and the accompanying annotations, we provide a role classification case study, showing reasonable performance with modern-day LLMs, and exposing explainable aspects of our framework via page content ablations. We also classify a large batch of un-annotated quotes, revealing interesting cross-domain, cross-platform role distributions on the web.
Authors: Jiaming Shen, Ran Xu, Yennie Jun, Zhen Qin, Tianqi Liu, Carl Yang, Yi Liang, Simon Baumgartner, Michael Bendersky
Abstract: Reward models (RMs) are crucial for aligning large language models (LLMs) with human preferences. They are trained using preference datasets where each example consists of one input prompt, two responses, and a preference label. As curating a high-quality human labeled preference dataset is both time-consuming and expensive, people often rely on existing powerful LLMs for preference label generation. This can potentially introduce noise and impede RM training. In this work, we present RMBoost, a novel synthetic preference data generation paradigm to boost reward model quality. Unlike traditional methods, which generate two responses before obtaining the preference label, RMBoost first generates one response and selects a preference label, followed by generating the second more (or less) preferred response conditioned on the pre-selected preference label and the first response. This approach offers two main advantages. First, RMBoost reduces labeling noise since preference pairs are constructed intentionally. Second, RMBoost facilitates the creation of more diverse responses by incorporating various quality aspects (e.g., helpfulness, relevance, completeness) into the prompts. We conduct extensive experiments across three diverse datasets and demonstrate that RMBoost outperforms other synthetic preference data generation techniques and significantly boosts the performance of four distinct reward models.
Authors: Irwin Deng, Kushagra Dixit, Vivek Gupta, Dan Roth
Abstract: Temporal reasoning over tabular data presents substantial challenges for large language models (LLMs), as evidenced by recent research. In this study, we conduct a comprehensive analysis of temporal datasets to pinpoint the specific limitations of LLMs. Our investigation leads to enhancements in TempTabQA, a dataset specifically designed for tabular temporal question answering. We provide critical insights for improving LLM performance in temporal reasoning tasks with tabular data. Furthermore, we introduce a novel approach, C.L.E.A.R to strengthen LLM capabilities in this domain. Our findings demonstrate that our method significantly improves evidence-based reasoning across various models. Additionally, our experimental results reveal that indirect supervision with auxiliary data substantially boosts model performance in these tasks. This work contributes to a deeper understanding of LLMs' temporal reasoning abilities over tabular data and promotes advancements in their application across diverse fields.
Authors: Davide Savarro, Davide Zago, Stefano Zoia
Abstract: Geolocalization of social media content is the task of determining the geographical location of a user based on textual data, that may show linguistic variations and informal language. In this project, we address the GeoLingIt challenge of geolocalizing tweets written in Italian by leveraging large language models (LLMs). GeoLingIt requires the prediction of both the region and the precise coordinates of the tweet. Our approach involves fine-tuning pre-trained LLMs to simultaneously predict these geolocalization aspects. By integrating innovative methodologies, we enhance the models' ability to understand the nuances of Italian social media text to improve the state-of-the-art in this domain. This work is conducted as part of the Large Language Models course at the Bertinoro International Spring School 2024. We make our code publicly available on GitHub https://github.com/dawoz/geolingit-biss2024.
Authors: Swetha Eppalapally, Daksh Dangi, Chaithra Bhat, Ankita Gupta, Ruiyi Zhang, Shubham Agarwal, Karishma Bagga, Seunghyun Yoon, Nedim Lipka, Ryan A. Rossi, Franck Dernoncourt
Abstract: Question-answering for domain-specific applications has recently attracted much interest due to the latest advancements in large language models (LLMs). However, accurately assessing the performance of these applications remains a challenge, mainly due to the lack of suitable benchmarks that effectively simulate real-world scenarios. To address this challenge, we introduce two product question-answering (QA) datasets focused on Adobe Acrobat and Photoshop products to help evaluate the performance of existing models on domain-specific product QA tasks. Additionally, we propose a novel knowledge-driven RAG-QA framework to enhance the performance of the models in the product QA task. Our experiments demonstrated that inducing domain knowledge through query reformulation allowed for increased retrieval and generative performance when compared to standard RAG-QA methods. This improvement, however, is slight, and thus illustrates the challenge posed by the datasets introduced.
Authors: Yukio Ohsawa, Dingming Xue, Kaira Sekiguchi
Abstract: The senses of words evolve. The sense of the same word may change from today to tomorrow, and multiple senses of the same word may be the result of the evolution of each other, that is, they may be parents and children. If we view Juba as an evolving ecosystem, the paradigm of learning the correct answer, which does not move with the sense of a word, is no longer valid. This paper is a case study that shows that word polysemy is an evolutionary consequence of the modification of Semantic Cells, which has al-ready been presented by the author, by introducing a small amount of diversity in its initial state as an example of analyzing the current set of short sentences. In particular, the analysis of a sentence sequence of 1000 sentences in some order for each of the four senses of the word Spring, collected using Chat GPT, shows that the word acquires the most polysemy monotonically in the analysis when the senses are arranged in the order in which they have evolved. In other words, we present a method for analyzing the dynamism of a word's acquiring polysemy with evolution and, at the same time, a methodology for viewing polysemy from an evolutionary framework rather than a learning-based one.
Authors: Yang Liu, Xiaobin Tian, Zequn Sun, Wei Hu
Abstract: Traditional knowledge graph (KG) completion models learn embeddings to predict missing facts. Recent works attempt to complete KGs in a text-generation manner with large language models (LLMs). However, they need to ground the output of LLMs to KG entities, which inevitably brings errors. In this paper, we present a finetuning framework, DIFT, aiming to unleash the KG completion ability of LLMs and avoid grounding errors. Given an incomplete fact, DIFT employs a lightweight model to obtain candidate entities and finetunes an LLM with discrimination instructions to select the correct one from the given candidates. To improve performance while reducing instruction data, DIFT uses a truncated sampling method to select useful facts for finetuning and injects KG embeddings into the LLM. Extensive experiments on benchmark datasets demonstrate the effectiveness of our proposed framework.
Authors: Chao-Chun Hsu, Erin Bransom, Jenna Sparks, Bailey Kuehl, Chenhao Tan, David Wadden, Lucy Lu Wang, Aakanksha Naik
Abstract: Literature review requires researchers to synthesize a large amount of information and is increasingly challenging as the scientific literature expands. In this work, we investigate the potential of LLMs for producing hierarchical organizations of scientific studies to assist researchers with literature review. We define hierarchical organizations as tree structures where nodes refer to topical categories and every node is linked to the studies assigned to that category. Our naive LLM-based pipeline for hierarchy generation from a set of studies produces promising yet imperfect hierarchies, motivating us to collect CHIME, an expert-curated dataset for this task focused on biomedicine. Given the challenging and time-consuming nature of building hierarchies from scratch, we use a human-in-the-loop process in which experts correct errors (both links between categories and study assignment) in LLM-generated hierarchies. CHIME contains 2,174 LLM-generated hierarchies covering 472 topics, and expert-corrected hierarchies for a subset of 100 topics. Expert corrections allow us to quantify LLM performance, and we find that while they are quite good at generating and organizing categories, their assignment of studies to categories could be improved. We attempt to train a corrector model with human feedback which improves study assignment by 12.6 F1 points. We release our dataset and models to encourage research on developing better assistive tools for literature review.
Authors: Jiaheng Liu, Chenchen Zhang, Jinyang Guo, Yuanxing Zhang, Haoran Que, Ken Deng, Zhiqi Bai, Jie Liu, Ge Zhang, Jiakai Wang, Yanan Wu, Congnan Liu, Wenbo Su, Jiamang Wang, Lin Qu, Bo Zheng
Abstract: Despite the advanced intelligence abilities of large language models (LLMs) in various applications, they still face significant computational and storage demands. Knowledge Distillation (KD) has emerged as an effective strategy to improve the performance of a smaller LLM (i.e., the student model) by transferring knowledge from a high-performing LLM (i.e., the teacher model). Prevailing techniques in LLM distillation typically use a black-box model API to generate high-quality pretrained and aligned datasets, or utilize white-box distillation by altering the loss function to better transfer knowledge from the teacher LLM. However, these methods ignore the knowledge differences between the student and teacher LLMs across domains. This results in excessive focus on domains with minimal performance gaps and insufficient attention to domains with large gaps, reducing overall performance. In this paper, we introduce a new LLM distillation framework called DDK, which dynamically adjusts the composition of the distillation dataset in a smooth manner according to the domain performance differences between the teacher and student models, making the distillation process more stable and effective. Extensive evaluations show that DDK significantly improves the performance of student models, outperforming both continuously pretrained baselines and existing knowledge distillation methods by a large margin.
Authors: Yao-Shun Chuang, Atiquer Rahman Sarkar, Noman Mohammed, Xiaoqian Jiang
Abstract: This study examines integrating EHRs and NLP with large language models (LLMs) to improve healthcare data management and patient care. It focuses on using advanced models to create secure, HIPAA-compliant synthetic patient notes for biomedical research. The study used de-identified and re-identified MIMIC III datasets with GPT-3.5, GPT-4, and Mistral 7B to generate synthetic notes. Text generation employed templates and keyword extraction for contextually relevant notes, with one-shot generation for comparison. Privacy assessment checked PHI occurrence, while text utility was tested using an ICD-9 coding task. Text quality was evaluated with ROUGE and cosine similarity metrics to measure semantic similarity with source notes. Analysis of PHI occurrence and text utility via the ICD-9 coding task showed that the keyword-based method had low risk and good performance. One-shot generation showed the highest PHI exposure and PHI co-occurrence, especially in geographic location and date categories. The Normalized One-shot method achieved the highest classification accuracy. Privacy analysis revealed a critical balance between data utility and privacy protection, influencing future data use and sharing. Re-identified data consistently outperformed de-identified data. This study demonstrates the effectiveness of keyword-based methods in generating privacy-protecting synthetic clinical notes that retain data usability, potentially transforming clinical data-sharing practices. The superior performance of re-identified over de-identified data suggests a shift towards methods that enhance utility and privacy by using dummy PHIs to perplex privacy attacks.
Authors: Yani Huang, Xuefeng Zhang, Richong Zhang, Junfan Chen, Jaein Kim
Abstract: Multi-Modal Entity Alignment aims to discover identical entities across heterogeneous knowledge graphs. While recent studies have delved into fusion paradigms to represent entities holistically, the elimination of features irrelevant to alignment and modal inconsistencies is overlooked, which are caused by inherent differences in multi-modal features. To address these challenges, we propose a novel strategy of progressive modality freezing, called PMF, that focuses on alignmentrelevant features and enhances multi-modal feature fusion. Notably, our approach introduces a pioneering cross-modal association loss to foster modal consistency. Empirical evaluations across nine datasets confirm PMF's superiority, demonstrating stateof-the-art performance and the rationale for freezing modalities. Our code is available at https://github.com/ninibymilk/PMF-MMEA.
Authors: Jinwook Park, Kangil Kim
Abstract: Neural parameterization has significantly advanced unsupervised grammar induction. However, training these models with a traditional likelihood loss for all possible parses exacerbates two issues: 1) $\textit{structural optimization ambiguity}$ that arbitrarily selects one among structurally ambiguous optimal grammars despite the specific preference of gold parses, and 2) $\textit{structural simplicity bias}$ that leads a model to underutilize rules to compose parse trees. These challenges subject unsupervised neural grammar induction (UNGI) to inevitable prediction errors, high variance, and the necessity for extensive grammars to achieve accurate predictions. This paper tackles these issues, offering a comprehensive analysis of their origins. As a solution, we introduce $\textit{sentence-wise parse-focusing}$ to reduce the parse pool per sentence for loss evaluation, using the structural bias from pre-trained parsers on the same dataset. In unsupervised parsing benchmark tests, our method significantly improves performance while effectively reducing variance and bias toward overly simplistic parses. Our research promotes learning more compact, accurate, and consistent explicit grammars, facilitating better interpretability.
Authors: Zhuocheng Gong, Jiahao Liu, Ziyue Wang, Pengfei Wu, Jingang Wang, Xunliang Cai, Dongyan Zhao, Rui Yan
Abstract: Speculative decoding has emerged as a promising technique to accelerate the inference of Large Language Models (LLMs) by employing a small language model to draft a hypothesis sequence, which is then validated by the LLM. The effectiveness of this approach heavily relies on the balance between performance and efficiency of the draft model. In our research, we focus on enhancing the proportion of draft tokens that are accepted to the final output by generating multiple hypotheses instead of just one. This allows the LLM more options to choose from and select the longest sequence that meets its standards. Our analysis reveals that hypotheses produced by the draft model share many common token sequences, suggesting a potential for optimizing computation. Leveraging this observation, we introduce an innovative approach utilizing a directed acyclic graph (DAG) to manage the drafted hypotheses. This structure enables us to efficiently predict and merge recurring token sequences, vastly reducing the computational demands of the draft model. We term this approach Graph-structured Speculative Decoding (GSD). We apply GSD across a range of LLMs, including a 70-billion parameter LLaMA-2 model, and observe a remarkable speedup of 1.73$\times$ to 1.96$\times$, significantly surpassing standard speculative decoding.
Authors: Zhichao Wang (James), Bin Bi (James), Shiva Kumar Pentyala (James), Kiran Ramnath (James), Sougata Chaudhuri (James), Shubham Mehrotra (James), Zixu (James), Zhu (Claire), Xiang-Bo Mao (Claire), Sitaram Asur (Claire), Na (Claire), Cheng
Abstract: With advancements in self-supervised learning, the availability of trillions tokens in a pre-training corpus, instruction fine-tuning, and the development of large Transformers with billions of parameters, large language models (LLMs) are now capable of generating factual and coherent responses to human queries. However, the mixed quality of training data can lead to the generation of undesired responses, presenting a significant challenge. Over the past two years, various methods have been proposed from different perspectives to enhance LLMs, particularly in aligning them with human expectation. Despite these efforts, there has not been a comprehensive survey paper that categorizes and details these approaches. In this work, we aim to address this gap by categorizing these papers into distinct topics and providing detailed explanations of each alignment method, thereby helping readers gain a thorough understanding of the current state of the field.
Authors: Nishanth Madhusudhan, Sathwik Tejaswi Madhusudhan, Vikas Yadav, Masoud Hashemi
Abstract: As Large Language Models (LLMs) achieve remarkable performance across various NLP tasks, their reliability becomes essential for widespread adoption. This paper focuses on Abstention Ability (AA), a critical yet under explored aspect of reliability - the ability of LLMs to refrain from answering questions when they are uncertain or when definitive answer is not possible, while maintaining question-answering (QA) task performance. While previous works have focused on understanding the recollection abilities of LLMs or their ability to identify imponderable/unanswerable questions, we believe there is a need for an effective AA evaluation method. Therefore, we propose a black-box evaluation methodology to examine and understand the AA of LLMs across a variety of multiple-choice QA tasks. We measure AA by rewarding models for abstaining from answering when their predictions are incorrect or when the questions are inherently unanswerable. We investigate three strategies, Strict Prompting, Verbal Confidence Thresholding, and Chain-of-Thought (CoT), to understand their impact on abstention across different LLMs. Our findings reveal that while even state-of-the-art LLMs like GPT-4 struggle with abstention, strategic prompting such as CoT, can significantly enhance this ability. Furthermore, we demonstrate that improving AA also leads to better overall QA task performance, underscoring the importance of evaluating AA in LLMs.
Authors: Jiahuan Li, Shujian Huang, Xinyu Dai, Jiajun Chen
Abstract: Large language models demonstrate reasonable multilingual abilities, despite predominantly English-centric pretraining. However, the spontaneous multilingual alignment in these models is shown to be weak, leading to unsatisfactory cross-lingual transfer and knowledge sharing. Previous works attempt to address this issue by explicitly injecting multilingual alignment information during or after pretraining. Thus for the early stage in pretraining, the alignment is weak for sharing information or knowledge across languages. In this paper, we propose PreAlign, a framework that establishes multilingual alignment prior to language model pretraining. PreAlign injects multilingual alignment by initializing the model to generate similar representations of aligned words and preserves this alignment using a code-switching strategy during pretraining. Extensive experiments in a synthetic English to English-Clone setting demonstrate that PreAlign significantly outperforms standard multilingual joint training in language modeling, zero-shot cross-lingual transfer, and cross-lingual knowledge application. Further experiments in real-world scenarios further validate PreAlign's effectiveness across various model sizes.
Authors: Pin-Jie Lin, Miaoran Zhang, Marius Mosbach, Dietrich Klakow
Abstract: Identifying beneficial tasks to transfer from is a critical step toward successful intermediate-task transfer learning. In this work, we experiment with 130 source-target task combinations and demonstrate that the transfer performance exhibits severe variance across different source tasks and training seeds, highlighting the crucial role of intermediate-task selection in a broader context. We compare four representative task selection methods in a unified setup, focusing on their effectiveness and consistency. Compared to embedding-free methods and text embeddings, task embeddings constructed from fine-tuned weights can better estimate task transferability by improving task prediction scores from 2.59% to 3.96%. Despite their strong performance, we observe that the task embeddings do not consistently demonstrate superiority for tasks requiring reasoning abilities. Furthermore, we introduce a novel method that measures pairwise token similarity using maximum inner product search, leading to the highest performance in task prediction. Our findings suggest that token-wise similarity is better predictive for predicting transferability compared to averaging weights.
Authors: Jingyun Sun, Chengxiao Dai, Zhongze Luo, Yangbo Chang, Yang Li
Abstract: Large Language Models (LLMs) demonstrate substantial potential in delivering legal consultation services to users without a legal background, attributed to their superior text comprehension and generation capabilities. Nonetheless, existing Chinese legal LLMs limit interaction to a single model-user dialogue, unlike the collaborative consultations typical of law firms, where multiple staff members contribute to a single consultation. This limitation prevents an authentic consultation experience. Additionally, extant Chinese legal LLMs suffer from critical limitations: (1) insufficient control over the quality of instruction fine-tuning data; (2) increased model hallucination resulting from users' ambiguous queries; and (3) a reduction in the model's ability to follow instructions over multiple dialogue turns. In response to these challenges, we propose a novel legal dialogue framework that leverages the collaborative capabilities of multiple LLM agents, termed LawLuo. This framework encompasses four agents: a receptionist, a lawyer, a secretary, and a boss, each responsible for different functionalities, collaboratively providing a comprehensive legal consultation to users. Additionally, we constructed two high-quality legal dialogue datasets, KINLED and MURLED, and fine-tuned ChatGLM-3-6b using these datasets. We propose a legal query clarification algorithm called ToLC. Experimental results demonstrate that LawLuo outperforms baseline LLMs, including GPT-4, across three dimensions: lawyer-like language style, the usefulness of legal advice, and the accuracy of legal knowledge. Our code and datasets are available at https://github.com/NEFUJing/LawLuo.
Authors: Yijie Chen, Yijin Liu, Fandong Meng, Jinan Xu, Yufeng Chen, Jie Zhou
Abstract: Gender bias has been a focal point in the study of bias in machine translation and language models. Existing machine translation gender bias evaluations are primarily focused on male and female genders, limiting the scope of the evaluation. To assess gender bias accurately, these studies often rely on calculating the accuracy of gender pronouns or the masculine and feminine attributes of grammatical gender via the stereotypes triggered by occupations or sentiment words ({\em i.e.}, clear positive or negative attitude), which cannot extend to non-binary groups. This study presents a benchmark AmbGIMT (Gender-Inclusive Machine Translation with Ambiguous attitude words), which assesses gender bias beyond binary gender. Meanwhile, we propose a novel process to evaluate gender bias based on the Emotional Attitude Score (EAS), which is used to quantify ambiguous attitude words. In evaluating three recent and effective open-source LLMs and one powerful multilingual translation-specific model, our main observations are: (1) The translation performance within non-binary gender contexts is markedly inferior in terms of translation quality and exhibits more negative attitudes than binary-gender contexts. (2) The analysis experiments indicate that incorporating constraint context in prompts for gender identity terms can substantially reduce translation bias, while the bias remains evident despite the presence of the constraints. The code is publicly available at \url{https://github.com/pppa2019/ambGIMT}.
Authors: Zhiheng Lyu, Kevin Yang, Lingpeng Kong, Daniel Klein
Abstract: While accurately detecting and correcting factual contradictions in language model outputs has become increasingly important as their capabilities improve, doing so is highly challenging. We propose a novel method, FACTTRACK, for tracking atomic facts and addressing factual contradictions. Crucially, FACTTRACK also maintains time-aware validity intervals for each fact, allowing for change over time. At a high level, FACTTRACK consists of a four-step pipeline to update a world state data structure for each new event: (1) decompose the event into directional atomic facts; (2) determine the validity interval of each atomic fact using the world state; (3) detect contradictions with existing facts in the world state; and finally (4) add new facts to the world state and update existing atomic facts. When we apply FACTTRACK to contradiction detection on structured story outlines, we find that FACTTRACK using LLaMA2-7B-Chat substantially outperforms a fair baseline using LLaMA2-7B-Chat, and achieves performance comparable to a GPT4 baseline. Moreover, when using GPT4, FACTTRACK significantly outperforms the GPT4 baseline.
Authors: Rithik Sachdev, Zhong-Qiu Wang, Chao-Han Huck Yang
Abstract: Building upon the strength of modern large language models (LLMs), generative error correction (GEC) has emerged as a promising paradigm that can elevate the performance of modern automatic speech recognition (ASR) systems. One representative approach is to leverage in-context learning to prompt LLMs so that a better hypothesis can be generated by the LLMs based on a carefully-designed prompt and an $N$-best list of hypotheses produced by ASR systems. However, it is yet unknown whether the existing prompts are the most effective ones for the task of post-ASR error correction. In this context, this paper first explores alternative prompts to identify an initial set of effective prompts, and then proposes to employ an evolutionary prompt optimization algorithm to refine the initial prompts. Evaluations results on the CHiME-4 subset of the Task $1$ of the SLT $2024$ GenSEC challenge show the effectiveness and potential of the proposed algorithms.
Authors: MohammadAli SadraeiJavaheri, Ali Moghaddaszadeh, Milad Molazadeh, Fariba Naeiji, Farnaz Aghababaloo, Hamideh Rafiee, Zahra Amirmahani, Tohid Abedini, Fatemeh Zahra Sheikhi, Amirmohammad Salehoof
Abstract: The field of natural language processing (NLP) has seen remarkable advancements, thanks to the power of deep learning and foundation models. Language models, and specifically BERT, have been key players in this progress. In this study, we trained and introduced two new BERT models using Persian data. We put our models to the test, comparing them to seven existing models across 14 diverse Persian natural language understanding (NLU) tasks. The results speak for themselves: our larger model outperforms the competition, showing an average improvement of at least +2.8 points. This highlights the effectiveness and potential of our new BERT models for Persian NLU tasks.
Authors: Ewoenam Kwaku Tokpo, Toon Calders
Abstract: Despite the evolution of language models, they continue to portray harmful societal biases and stereotypes inadvertently learned from training data. These inherent biases often result in detrimental effects in various applications. Counterfactual Data Augmentation (CDA), which seeks to balance demographic attributes in training data, has been a widely adopted approach to mitigate bias in natural language processing. However, many existing CDA approaches rely on word substitution techniques using manually compiled word-pair dictionaries. These techniques often lead to out-of-context substitutions, resulting in potential quality issues. The advancement of model-based techniques, on the other hand, has been challenged by the need for parallel training data. Works in this area resort to manually generated parallel data that are expensive to collect and are consequently limited in scale. This paper proposes FairFlow, an automated approach to generating parallel data for training counterfactual text generator models that limits the need for human intervention. Furthermore, we show that FairFlow significantly overcomes the limitations of dictionary-based word-substitution approaches whilst maintaining good performance.
Authors: Kai Liu, Zhihang Fu, Chao Chen, Wei Zhang, Rongxin Jiang, Fan Zhou, Yaowu Chen, Yue Wu, Jieping Ye
Abstract: When reading long-form text, human cognition is complex and structurized. While large language models (LLMs) process input contexts through a causal and sequential perspective, this approach can potentially limit their ability to handle intricate and complex inputs effectively. To enhance LLM's cognition capability, this paper presents a novel concept of context structurization. Specifically, we transform the plain, unordered contextual sentences into well-ordered and hierarchically structurized elements. By doing so, LLMs can better grasp intricate and extended contexts through precise attention and information-seeking along the organized structures. Extensive evaluations are conducted across various model architectures and sizes (including several 7B- to 72B-size auto-regressive LLMs as well as BERT-like masking models) on a diverse set of NLP tasks (e.g., context-based question-answering, exhaustive hallucination evaluation, and passage-level dense retrieval). Empirical results show consistent and significant performance gains afforded by a single-round structurization. In particular, we boost a 72B-parameter open-source model to achieve comparable performance against GPT-3.5-Turbo as the hallucination evaluator. Besides, we show the feasibility of distilling advanced LLMs' language processing abilities to a smaller yet effective StruXGPT-7B to execute structurization, addressing the practicality of our approach. Code will be made public soon.
Authors: Kenza Benkirane (University College London), Laura Gongas (University College London), Shahar Pelles (University College London), Naomi Fuchs (University College London), Joshua Darmon (University College London), Pontus Stenetorp (University College London), David Ifeoluwa Adelani (University College London), Eduardo Sanchez (University College London, Meta)
Abstract: Recent advancements in massively multilingual machine translation systems have significantly enhanced translation accuracy; however, even the best performing systems still generate hallucinations, severely impacting user trust. Detecting hallucinations in Machine Translation (MT) remains a critical challenge, particularly since existing methods excel with High-Resource Languages (HRLs) but exhibit substantial limitations when applied to Low-Resource Languages (LRLs). This paper evaluates hallucination detection approaches using Large Language Models (LLMs) and semantic similarity within massively multilingual embeddings. Our study spans 16 language directions, covering HRLs, LRLs, with diverse scripts. We find that the choice of model is essential for performance. On average, for HRLs, Llama3-70B outperforms the previous state of the art by as much as 0.16 MCC (Matthews Correlation Coefficient). However, for LRLs we observe that Claude Sonnet outperforms other LLMs on average by 0.03 MCC. The key takeaway from our study is that LLMs can achieve performance comparable or even better than previously proposed models, despite not being explicitly trained for any machine translation task. However, their advantage is less significant for LRLs.
Authors: Julian Schelb, Roberto Ulloa, Andreas Spitz
Abstract: Researchers in the political and social sciences often rely on classification models to analyze trends in information consumption by examining browsing histories of millions of webpages. Automated scalable methods are necessary due to the impracticality of manual labeling. In this paper, we model the detection of topic-related content as a binary classification task and compare the accuracy of fine-tuned pre-trained encoder models against in-context learning strategies. Using only a few hundred annotated data points per topic, we detect content related to three German policies in a database of scraped webpages. We compare multilingual and monolingual models, as well as zero and few-shot approaches, and investigate the impact of negative sampling strategies and the combination of URL & content-based features. Our results show that a small sample of annotated data is sufficient to train an effective classifier. Fine-tuning encoder-based models yields better results than in-context learning. Classifiers using both URL & content-based features perform best, while using URLs alone provides adequate results when content is unavailable.
Authors: Yizhou Chi, Lingjun Mao, Zineng Tang
Abstract: Strategic social deduction games serve as valuable testbeds for evaluating the understanding and inference skills of language models, offering crucial insights into social science, artificial intelligence, and strategic gaming. This paper focuses on creating proxies of human behavior in simulated environments, with \textit{Among Us} utilized as a tool for studying simulated human behavior. The study introduces a text-based game environment, named AmongAgent, that mirrors the dynamics of \textit{Among Us}. Players act as crew members aboard a spaceship, tasked with identifying impostors who are sabotaging the ship and eliminating the crew. Within this environment, the behavior of simulated language agents is analyzed. The experiments involve diverse game sequences featuring different configurations of Crewmates and Impostor personality archetypes. Our work demonstrates that state-of-the-art large language models (LLMs) can effectively grasp the game rules and make decisions based on the current context. This work aims to promote further exploration of LLMs in goal-oriented games with incomplete information and complex action spaces, as these settings offer valuable opportunities to assess language model performance in socially driven scenarios.
Authors: Sean Robertson, Gerald Penn, Ewan Dunbar
Abstract: A long-standing question in automatic speech recognition research is how to attribute errors to the ability of a model to model the acoustics, versus its ability to leverage higher-order context (lexicon, morphology, syntax, semantics). We validate a novel approach which models error rates as a function of relative textual predictability, and yields a single number, $k$, which measures the effect of textual predictability on the recognizer. We use this method to demonstrate that a Wav2Vec 2.0-based model makes greater stronger use of textual context than a hybrid ASR model, in spite of not using an explicit language model, and also use it to shed light on recent results demonstrating poor performance of standard ASR systems on African-American English. We demonstrate that these mostly represent failures of acoustic--phonetic modelling. We show how this approach can be used straightforwardly in diagnosing and improving ASR.
Authors: Ioana Buhnila, Aman Sinha, Mathieu Constant
Abstract: Recent surge in the accessibility of large language models (LLMs) to the general population can lead to untrackable use of such models for medical-related recommendations. Language generation via LLMs models has two key problems: firstly, they are prone to hallucination and therefore, for any medical purpose they require scientific and factual grounding; secondly, LLMs pose tremendous challenge to computational resources due to their gigantic model size. In this work, we introduce pRAGe, a pipeline for Retrieval Augmented Generation and evaluation of medical paraphrases generation using Small Language Models (SLM). We study the effectiveness of SLMs and the impact of external knowledge base for medical paraphrase generation in French.
Authors: Eunseop Yoon, Hee Suk Yoon, SooHwan Eom, Gunsoo Han, Daniel Wontae Nam, Daejin Jo, Kyoung-Woon On, Mark A. Hasegawa-Johnson, Sungwoong Kim, Chang D. Yoo
Abstract: Reinforcement Learning from Human Feedback (RLHF) leverages human preference data to train language models to align more closely with human essence. These human preference data, however, are labeled at the sequence level, creating a mismatch between sequence-level preference labels and tokens, which are autoregressively generated from the language model. Although several recent approaches have tried to provide token-level (i.e., dense) rewards for each individual token, these typically rely on predefined discrete reward values (e.g., positive: +1, negative: -1, neutral: 0), failing to account for varying degrees of preference inherent to each token. To address this limitation, we introduce TLCR (Token-Level Continuous Reward) for RLHF, which incorporates a discriminator trained to distinguish positive and negative tokens, and the confidence of the discriminator is used to assign continuous rewards to each token considering the context. Extensive experiments show that our proposed TLCR leads to consistent performance improvements over previous sequence-level or token-level discrete rewards on open-ended generation benchmarks.
Authors: Giorgos Lysandrou, Roma English Owen, Vanja Popovic, Grant Le Brun, Aryo Pradipta Gema, Beatrice Alex, Elizabeth A. L. Fairley
Abstract: There exists an invisible barrier between healthcare professionals' perception of a patient's clinical experience and the reality. This barrier may be induced by the environment that hinders patients from sharing their experiences openly with healthcare professionals. As patients are observed to discuss and exchange knowledge more candidly on social media, valuable insights can be leveraged from these platforms. However, the abundance of non-patient posts on social media necessitates filtering out such irrelevant content to distinguish the genuine voices of patients, a task we refer to as patient voice classification. In this study, we analyse the importance of linguistic characteristics in accurately classifying patient voices. Our findings underscore the essential role of linguistic and statistical text similarity analysis in identifying common patterns among patient groups. These results allude to even starker differences in the way patients express themselves at a disease level and across various therapeutic domains. Additionally, we fine-tuned a pre-trained Language Model on the combined datasets with similar linguistic patterns, resulting in a highly accurate automatic patient voice classification. Being the pioneering study on the topic, our focus on extracting authentic patient experiences from social media stands as a crucial step towards advancing healthcare standards and fostering a patient-centric approach.
Authors: Yilun Zhou, Caiming Xiong, Silvio Savarese, Chien-Sheng Wu
Abstract: Despite the recent proliferation of large language models (LLMs), their training recipes -- model architecture, pre-training data and optimization algorithm -- are often very similar. This naturally raises the question of the similarity among the resulting models. In this paper, we propose a novel setting, imaginary question answering (IQA), to better understand model similarity. In IQA, we ask one model to generate purely imaginary questions (e.g., on completely made-up concepts in physics) and prompt another model to answer. Surprisingly, despite the total fictionality of these questions, all models can answer each other's questions with remarkable success, suggesting a "shared imagination space" in which these models operate during such hallucinations. We conduct a series of investigations into this phenomenon and discuss implications on model homogeneity, hallucination, and computational creativity.
Authors: Jonathan Hayase, Alisa Liu, Yejin Choi, Sewoong Oh, Noah A. Smith
Abstract: The pretraining data of today's strongest language models is opaque. In particular, little is known about the proportions of various domains or languages represented. In this work, we tackle a task which we call data mixture inference, which aims to uncover the distributional make-up of training data. We introduce a novel attack based on a previously overlooked source of information -- byte-pair encoding (BPE) tokenizers, used by the vast majority of modern language models. Our key insight is that the ordered list of merge rules learned by a BPE tokenizer naturally reveals information about the token frequencies in its training data: the first merge is the most common byte pair, the second is the most common pair after merging the first token, and so on. Given a tokenizer's merge list along with data samples for each category of interest, we formulate a linear program that solves for the proportion of each category in the tokenizer's training set. Importantly, to the extent to which tokenizer training data is representative of the pretraining data, we indirectly learn about the pretraining data. In controlled experiments, we show that our attack recovers mixture ratios with high precision for tokenizers trained on known mixtures of natural languages, programming languages, and data sources. We then apply our approach to off-the-shelf tokenizers released with recent LMs. We confirm much publicly disclosed information about these models, and also make several new inferences: GPT-4o's tokenizer is much more multilingual than its predecessors, training on 39% non-English data; Llama3 extends GPT-3.5's tokenizer primarily for multilingual (48%) use; GPT-3.5's and Claude's tokenizers are trained on predominantly code (~60%). We hope our work sheds light on current design practices for pretraining data, and inspires continued research into data mixture inference for LMs.
Authors: Ricardo Dominguez-Olmedo, Vedant Nanda, Rediet Abebe, Stefan Bechtold, Christoph Engel, Jens Frankenreiter, Krishna Gummadi, Moritz Hardt, Michael Livermore
Abstract: Annotation and classification of legal text are central components of empirical legal research. Traditionally, these tasks are often delegated to trained research assistants. Motivated by the advances in language modeling, empirical legal scholars are increasingly turning to prompting commercial models, hoping that it will alleviate the significant cost of human annotation. Despite growing use, our understanding of how to best utilize large language models for legal tasks remains limited. We conduct a comprehensive study of 260 legal text classification tasks, nearly all new to the machine learning community. Starting from GPT-4 as a baseline, we show that it has non-trivial but highly varied zero-shot accuracy, often exhibiting performance that may be insufficient for legal work. We then demonstrate that a lightly fine-tuned Llama 3 model vastly outperforms GPT-4 on almost all tasks, typically by double-digit percentage points. We find that larger models respond better to fine-tuning than smaller models. A few tens to hundreds of examples suffice to achieve high classification accuracy. Notably, we can fine-tune a single model on all 260 tasks simultaneously at a small loss in accuracy relative to having a separate model for each task. Our work points to a viable alternative to the predominant practice of prompting commercial models. For concrete legal tasks with some available labeled data, researchers are better off using a fine-tuned open-source model.
Authors: Jader Martins Camboim de S\'a, Marcos Da Silveira, C\'edric Pruski
Abstract: Languages continually evolve in response to societal events, resulting in new terms and shifts in meanings. These changes have significant implications for computer applications, including automatic translation and chatbots, making it essential to characterize them accurately. The recent development of LLMs has notably advanced natural language understanding, particularly in sense inference and reasoning. In this paper, we investigate the potential of LLMs in characterizing three types of semantic change: dimension, relation, and orientation. We achieve this by combining LLMs' Chain-of-Thought with rhetorical devices and conducting an experimental assessment of our approach using newly created datasets. Our results highlight the effectiveness of LLMs in capturing and analyzing semantic changes, providing valuable insights to improve computational linguistic applications.
Authors: Rongwu Xu, Yishuo Cai, Zhenhong Zhou, Renjie Gu, Haiqin Weng, Yan Liu, Tianwei Zhang, Wei Xu, Han Qiu
Abstract: The risk of harmful content generated by large language models (LLMs) becomes a critical concern. This paper presents a systematic study on assessing and improving LLMs' capability to perform the task of \textbf{course-correction}, \ie, the model can steer away from generating harmful content autonomously. To start with, we introduce the \textsc{C$^2$-Eval} benchmark for quantitative assessment and analyze 10 popular LLMs, revealing varying proficiency of current safety-tuned LLMs in course-correction. To improve, we propose fine-tuning LLMs with preference learning, emphasizing the preference for timely course-correction. Using an automated pipeline, we create \textsc{C$^2$-Syn}, a synthetic dataset with 750K pairwise preferences, to teach models the concept of timely course-correction through data-driven preference learning. Experiments on 2 LLMs, \textsc{Llama2-Chat 7B} and \textsc{Qwen2 7B}, show that our method effectively enhances course-correction skills without affecting general performance. Additionally, it effectively improves LLMs' safety, particularly in resisting jailbreak attacks.
Authors: Laxmi Pandey, Ke Li, Jinxi Guo, Debjyoti Paul, Arthur Guo, Jay Mahadeokar, Xuedong Zhang
Abstract: Multilingual pretraining for transfer learning significantly boosts the robustness of low-resource monolingual ASR models. This study systematically investigates three main aspects: (a) the impact of transfer learning on model performance during initial training or fine-tuning, (b) the influence of transfer learning across dataset domains and languages, and (c) the effect on rare-word recognition compared to non-rare words. Our finding suggests that RNNT-loss pretraining, followed by monolingual fine-tuning with Minimum Word Error Rate (MinWER) loss, consistently reduces Word Error Rates (WER) across languages like Italian and French. WER Reductions (WERR) reach 36.2% and 42.8% compared to monolingual baselines for MLS and in-house datasets. Out-of-domain pretraining leads to 28% higher WERR than in-domain pretraining. Both rare and non-rare words benefit, with rare words showing greater improvements with out-of-domain pretraining, and non-rare words with in-domain pretraining.
Authors: Yuanwei Wu, Yue Huang, Yixin Liu, Xiang Li, Pan Zhou, Lichao Sun
Abstract: GPT-4V has attracted considerable attention due to its extraordinary capacity for integrating and processing multimodal information. At the same time, its ability of face recognition raises new safety concerns of privacy leakage. Despite researchers' efforts in safety alignment through RLHF or preprocessing filters, vulnerabilities might still be exploited. In our study, we introduce AutoJailbreak, an innovative automatic jailbreak technique inspired by prompt optimization. We leverage Large Language Models (LLMs) for red-teaming to refine the jailbreak prompt and employ weak-to-strong in-context learning prompts to boost efficiency. Furthermore, we present an effective search method that incorporates early stopping to minimize optimization time and token expenditure. Our experiments demonstrate that AutoJailbreak significantly surpasses conventional methods, achieving an Attack Success Rate (ASR) exceeding 95.3\%. This research sheds light on strengthening GPT-4V security, underscoring the potential for LLMs to be exploited in compromising GPT-4V integrity.
Authors: Pedro Ferreira, Wilker Aziz, Ivan Titov
Abstract: Explanation regularisation (ER) has been introduced as a way to guide models to make their predictions in a manner more akin to humans, i.e., making their attributions "plausible". This is achieved by introducing an auxiliary explanation loss, that measures how well the output of an input attribution technique for the model agrees with relevant human-annotated rationales. One positive outcome of using ER appears to be improved performance in out-of-domain (OOD) settings, presumably due to an increased reliance on "plausible" tokens. However, previous work has under-explored the impact of the ER objective on model attributions, in particular when obtained with techniques other than the one used to train ER. In this work, we contribute a study of ER's effectiveness at informing classification decisions on plausible tokens, and the relationship between increased plausibility and robustness to OOD conditions. Through a series of analyses, we find that the connection between ER and the ability of a classifier to rely on plausible features has been overstated and that a stronger reliance on plausible tokens does not seem to be the cause for any perceived OOD improvements.
Authors: Xiaoyue Xu, Qinyuan Ye, Xiang Ren
Abstract: We introduce Lifelong ICL, a problem setting that challenges long-context language models (LMs) to learn from a sequence of language tasks through in-context learning (ICL). We further introduce Task Haystack, an evaluation suite dedicated to assessing and diagnosing how long-context LMs utilizes contexts in Lifelong ICL. When given a task instruction and test inputs, long-context LMs are expected to leverage the relevant demonstrations in the Lifelong ICL prompt, avoid distraction and interference from other tasks, and achieve test accuracies that are not significantly worse than the Single-task ICL baseline. Task Haystack draws inspiration from the widely-adopted "needle-in-a-haystack" (NIAH) evaluation, but presents new and unique challenges. It demands that models (1) utilize the contexts with deeper understanding, rather than resorting to simple copying and pasting; (2) navigate through long streams of evolving topics and tasks, which closely approximates the complexities of real-world usage of long-context LMs. Additionally, Task Haystack inherits the controllability aspect of NIAH, providing model developers with tools and visualizations to identify model vulnerabilities effectively. We benchmark 12 long-context LMs using Task Haystack. We find that state-of-the-art closed models such as GPT-4o still struggle in this setting, failing 15% of the cases on average, while all open-weight models we evaluate further lack behind by a large margin, failing up to 61% of the cases. In our controlled analysis, we identify factors such as distraction and recency bias as contributors to these failure cases. Further, we observe declines in performance when task instructions are paraphrased at test time or when ICL demonstrations are repeated excessively, raising concerns about the robustness, instruction understanding, and true context utilization of current long-context LMs.
Authors: Simen Eide, Arnoldo Frigessi
Abstract: This paper introduces Bayesian Hierarchical Low-Rank Adaption (BoRA), a novel method for finetuning multi-task Large Language Models (LLMs). Current finetuning approaches, such as Low-Rank Adaption (LoRA), perform exeptionally well in reducing training parameters and memory usage but face limitations when applied to multiple similar tasks. Practitioners usually have to choose between training separate models for each task or a single model for all tasks, both of which come with trade-offs in specialization and data utilization. BoRA addresses these trade-offs by leveraging a Bayesian hierarchical model that allows tasks to share information through global hierarchical priors. This enables tasks with limited data to benefit from the overall structure derived from related tasks while allowing tasks with more data to specialize. Our experimental results show that BoRA outperforms both individual and unified model approaches, achieving lower perplexity and better generalization across tasks. This method provides a scalable and efficient solution for multi-task LLM finetuning, with significant practical implications for diverse applications.
Authors: Qiuhong Wei, Ying Cui, Mengwei Ding, Yanqin Wang, Lingling Xiang, Zhengxiong Yao, Ceran Chen, Ying Long, Zhezhen Jin, Ximing Xu
Abstract: Large language models (LLMs) have demonstrated potential applications in medicine, yet data privacy and computational burden limit their deployment in healthcare institutions. Open-source and lightweight versions of LLMs emerge as potential solutions, but their performance, particularly in pediatric settings remains underexplored. In this cross-sectional study, 250 patient consultation questions were randomly selected from a public online medical forum, with 10 questions from each of 25 pediatric departments, spanning from December 1, 2022, to October 30, 2023. Two lightweight open-source LLMs, ChatGLM3-6B and Vicuna-7B, along with a larger-scale model, Vicuna-13B, and the widely-used proprietary ChatGPT-3.5, independently answered these questions in Chinese between November 1, 2023, and November 7, 2023. To assess reproducibility, each inquiry was replicated once. We found that ChatGLM3-6B demonstrated higher accuracy and completeness than Vicuna-13B and Vicuna-7B (P < .001), but all were outperformed by ChatGPT-3.5. ChatGPT-3.5 received the highest ratings in accuracy (65.2%) compared to ChatGLM3-6B (41.2%), Vicuna-13B (11.2%), and Vicuna-7B (4.4%). Similarly, in completeness, ChatGPT-3.5 led (78.4%), followed by ChatGLM3-6B (76.0%), Vicuna-13B (34.8%), and Vicuna-7B (22.0%) in highest ratings. ChatGLM3-6B matched ChatGPT-3.5 in readability, both outperforming Vicuna models (P < .001). In terms of empathy, ChatGPT-3.5 outperformed the lightweight LLMs (P < .001). In safety, all models performed comparably well (P > .05), with over 98.4% of responses being rated as safe. Repetition of inquiries confirmed these findings. In conclusion, Lightweight LLMs demonstrate promising application in pediatric healthcare. However, the observed gap between lightweight and large-scale proprietary LLMs underscores the need for continued development efforts.
Authors: Hanlin Tang, Yang Lin, Jing Lin, Qingsen Han, Shikuan Hong, Yiwu Yao, Gongyi Wang
Abstract: The memory and computational demands of Key-Value (KV) cache present significant challenges for deploying long-context language models. Previous approaches attempt to mitigate this issue by selectively dropping tokens, which irreversibly erases critical information that might be needed for future queries. In this paper, we propose a novel compression technique for KV cache that preserves all token information. Our investigation reveals that: i) Most attention heads primarily focus on the local context; ii) Only a few heads, denoted as retrieval heads, can essentially pay attention to all input tokens. These key observations motivate us to use separate caching strategy for attention heads. Therefore, we propose RazorAttention, a training-free KV cache compression algorithm, which maintains a full cache for these crucial retrieval heads and discards the remote tokens in non-retrieval heads. Furthermore, we introduce a novel mechanism involving a "compensation token" to further recover the information in the dropped tokens. Extensive evaluations across a diverse set of large language models (LLMs) demonstrate that RazorAttention achieves a reduction in KV cache size by over 70% without noticeable impacts on performance. Additionally, RazorAttention is compatible with FlashAttention, rendering it an efficient and plug-and-play solution that enhances LLM inference efficiency without overhead or retraining of the original model.
Authors: Liu Qi, He Yongyi, Lian Defu, Zheng Zhi, Xu Tong, Liu Che, Chen Enhong
Abstract: Multimodal Entity Linking (MEL) is a crucial task that aims at linking ambiguous mentions within multimodal contexts to the referent entities in a multimodal knowledge base, such as Wikipedia. Existing methods focus heavily on using complex mechanisms and extensive model tuning methods to model the multimodal interaction on specific datasets. However, these methods overcomplicate the MEL task and overlook the visual semantic information, which makes them costly and hard to scale. Moreover, these methods can not solve the issues like textual ambiguity, redundancy, and noisy images, which severely degrade their performance. Fortunately, the advent of Large Language Models (LLMs) with robust capabilities in text understanding and reasoning, particularly Multimodal Large Language Models (MLLMs) that can process multimodal inputs, provides new insights into addressing this challenge. However, how to design a universally applicable LLMs-based MEL approach remains a pressing challenge. To this end, we propose UniMEL, a unified framework which establishes a new paradigm to process multimodal entity linking tasks using LLMs. In this framework, we employ LLMs to augment the representation of mentions and entities individually by integrating textual and visual information and refining textual information. Subsequently, we employ the embedding-based method for retrieving and re-ranking candidate entities. Then, with only ~0.26% of the model parameters fine-tuned, LLMs can make the final selection from the candidate entities. Extensive experiments on three public benchmark datasets demonstrate that our solution achieves state-of-the-art performance, and ablation studies verify the effectiveness of all modules. Our code is available at https://anonymous.4open.science/r/UniMEL/.
Authors: Maud Van Lier, Gorka Mu\~noz-Gil
Abstract: The arrival of Large Language Models (LLMs) has stirred up philosophical debates about the possibility of realizing agency in an artificial manner. In this work we contribute to the debate by presenting a theoretical model that can be used as a threshold conception for artificial agents. The model defines agents as systems whose actions and goals are always influenced by a dynamic framework of factors that consists of the agent's accessible history, its adaptive repertoire and its external environment. This framework, in turn, is influenced by the actions that the agent takes and the goals that it forms. We show with the help of the model that state-of-the-art LLMs are not agents yet, but that there are elements to them that suggest a way forward. The paper argues that a combination of the agent architecture presented in Park et al. (2023) together with the use of modules like the Coscientist in Boiko et al. (2023) could potentially be a way to realize agency in an artificial manner. We end the paper by reflecting on the obstacles one might face in building such an artificial agent and by presenting possible directions for future research.
Authors: Fengran Mo, Longxiang Zhao, Kaiyu Huang, Yue Dong, Degen Huang, Jian-Yun Nie
Abstract: Personalized conversational information retrieval (CIR) combines conversational and personalizable elements to satisfy various users' complex information needs through multi-turn interaction based on their backgrounds. The key promise is that the personal textual knowledge base (PTKB) can improve the CIR effectiveness because the retrieval results can be more related to the user's background. However, PTKB is noisy: not every piece of knowledge in PTKB is relevant to the specific query at hand. In this paper, we explore and test several ways to select knowledge from PTKB and use it for query reformulation by using a large language model (LLM). The experimental results show the PTKB might not always improve the search results when used alone, but LLM can help generate a more appropriate personalized query when high-quality guidance is provided.
Authors: Shi Lin, Rongchang Li, Xun Wang, Changting Lin, Wenpeng Xing, Meng Han
Abstract: The rapid development of Large Language Models (LLMs) has brought remarkable generative capabilities across diverse tasks. However, despite the impressive achievements, these models still have numerous security vulnerabilities, particularly when faced with jailbreak attacks. Therefore, by investigating jailbreak attacks, we can uncover hidden weaknesses in LLMs and guide us in developing more robust defense mechanisms to fortify their security. In this paper, we further explore the boundary of jailbreak attacks on LLMs and propose Analyzing-based Jailbreak (ABJ). This effective jailbreak attack method takes advantage of LLMs' growing analyzing and reasoning capability and reveals their underlying vulnerabilities when facing analysis-based tasks. We conduct a detailed evaluation of ABJ across various open-source and closed-source LLMs, which achieves 94.8% Attack Success Rate (ASR) and 1.06 Attack Efficiency (AE) on GPT-4-turbo-0409, demonstrating state-of-the-art attack effectiveness and efficiency. Our research highlights the importance of prioritizing and enhancing the safety of LLMs to mitigate the risks of misuse.
Authors: Yiping Zhang, Yuntao Shou, Tao Meng, Wei Ai, Keqin Li
Abstract: The age estimation task aims to use facial features to predict the age of people and is widely used in public security, marketing, identification, and other fields. However, the features are mainly concentrated in facial keypoints, and existing CNN and Transformer-based methods have inflexibility and redundancy for modeling complex irregular structures. Therefore, this paper proposes a Multi-view Mask Contrastive Learning Graph Convolutional Neural Network (MMCL-GCN) for age estimation. Specifically, the overall structure of the MMCL-GCN network contains a feature extraction stage and an age estimation stage. In the feature extraction stage, we introduce a graph structure to construct face images as input and then design a Multi-view Mask Contrastive Learning (MMCL) mechanism to learn complex structural and semantic information about face images. The learning mechanism employs an asymmetric siamese network architecture, which utilizes an online encoder-decoder structure to reconstruct the missing information from the original graph and utilizes the target encoder to learn latent representations for contrastive learning. Furthermore, to promote the two learning mechanisms better compatible and complementary, we adopt two augmentation strategies and optimize the joint losses. In the age estimation stage, we design a Multi-layer Extreme Learning Machine (ML-IELM) with identity mapping to fully use the features extracted by the online encoder. Then, a classifier and a regressor were constructed based on ML-IELM, which were used to identify the age grouping interval and accurately estimate the final age. Extensive experiments show that MMCL-GCN can effectively reduce the error of age estimation on benchmark datasets such as Adience, MORPH-II, and LAP-2016.
Authors: Blazej Manczak, Eliott Zemour, Eric Lin, Vaikkunth Mugunthan
Abstract: Deploying language models (LMs) necessitates outputs to be both high-quality and compliant with safety guidelines. Although Inference-Time Guardrails (ITG) offer solutions that shift model output distributions towards compliance, we find that current methods struggle in balancing safety with helpfulness. ITG Methods that safely address non-compliant queries exhibit lower helpfulness while those that prioritize helpfulness compromise on safety. We refer to this trade-off as the guardrail tax, analogous to the alignment tax. To address this, we propose PrimeGuard, a novel ITG method that utilizes structured control flow. PrimeGuard routes requests to different self-instantiations of the LM with varying instructions, leveraging its inherent instruction-following capabilities and in-context learning. Our tuning-free approach dynamically compiles system-designer guidelines for each query. We construct and release safe-eval, a diverse red-team safety benchmark. Extensive evaluations demonstrate that PrimeGuard, without fine-tuning, overcomes the guardrail tax by (1) significantly increasing resistance to iterative jailbreak attacks and (2) achieving state-of-the-art results in safety guardrailing while (3) matching helpfulness scores of alignment-tuned models. Extensive evaluations demonstrate that PrimeGuard, without fine-tuning, outperforms all competing baselines and overcomes the guardrail tax by improving the fraction of safe responses from 61% to 97% and increasing average helpfulness scores from 4.17 to 4.29 on the largest models, while reducing attack success rate from 100% to 8%. PrimeGuard implementation is available at https://github.com/dynamofl/PrimeGuard and safe-eval dataset is available at https://huggingface.co/datasets/dynamoai/safe_eval.
URLs: https://github.com/dynamofl/PrimeGuard, https://huggingface.co/datasets/dynamoai/safe_eval.
Authors: Giuseppe Riva, Fabrizia Mantovani, Brenda K. Wiederhold, Antonella Marchetti, Andrea Gaggioli
Abstract: Although LLMs and other artificial intelligence systems demonstrate cognitive skills similar to humans, like concept learning and language acquisition, the way they process information fundamentally differs from biological cognition. To better understand these differences this paper introduces Psychomatics, a multidisciplinary framework bridging cognitive science, linguistics, and computer science. It aims to better understand the high-level functioning of LLMs, focusing specifically on how LLMs acquire, learn, remember, and use information to produce their outputs. To achieve this goal, Psychomatics will rely on a comparative methodology, starting from a theory-driven research question - is the process of language development and use different in humans and LLMs? - drawing parallels between LLMs and biological systems. Our analysis shows how LLMs can map and manipulate complex linguistic patterns in their training data. Moreover, LLMs can follow Grice's Cooperative Principle to provide relevant and informative responses. However, human cognition draws from multiple sources of meaning, including experiential, emotional, and imaginative facets, which transcend mere language processing and are rooted in our social and developmental trajectories. Moreover, current LLMs lack physical embodiment, reducing their ability to make sense of the intricate interplay between perception, action, and cognition that shapes human understanding and expression. Ultimately, Psychomatics holds the potential to yield transformative insights into the nature of language, cognition, and intelligence, both artificial and biological. Moreover, by drawing parallels between LLMs and human cognitive processes, Psychomatics can inform the development of more robust and human-like AI systems.
Authors: Aristeidis Panos, Rahaf Aljundi, Daniel Olmeda Reino, Richard E Turner
Abstract: Vision language models (VLMs) demonstrate impressive capabilities in visual question answering and image captioning, acting as a crucial link between visual and language models. However, existing open-source VLMs heavily rely on pretrained and frozen vision encoders (such as CLIP). Despite CLIP's robustness across diverse domains, it still exhibits non-negligible image understanding errors. These errors propagate to the VLM responses, resulting in sub-optimal performance. In our work, we propose an efficient and robust method for updating vision encoders within VLMs. Our approach selectively and locally updates encoders, leading to substantial performance improvements on data where previous mistakes occurred, while maintaining overall robustness. Furthermore, we demonstrate the effectiveness of our method during continual few-shot updates. Theoretical grounding, generality, and computational efficiency characterize our approach.
Authors: Huiyu Xu, Wenhui Zhang, Zhibo Wang, Feng Xiao, Rui Zheng, Yunhe Feng, Zhongjie Ba, Kui Ren
Abstract: Recently, advanced Large Language Models (LLMs) such as GPT-4 have been integrated into many real-world applications like Code Copilot. These applications have significantly expanded the attack surface of LLMs, exposing them to a variety of threats. Among them, jailbreak attacks that induce toxic responses through jailbreak prompts have raised critical safety concerns. To identify these threats, a growing number of red teaming approaches simulate potential adversarial scenarios by crafting jailbreak prompts to test the target LLM. However, existing red teaming methods do not consider the unique vulnerabilities of LLM in different scenarios, making it difficult to adjust the jailbreak prompts to find context-specific vulnerabilities. Meanwhile, these methods are limited to refining jailbreak templates using a few mutation operations, lacking the automation and scalability to adapt to different scenarios. To enable context-aware and efficient red teaming, we abstract and model existing attacks into a coherent concept called "jailbreak strategy" and propose a multi-agent LLM system named RedAgent that leverages these strategies to generate context-aware jailbreak prompts. By self-reflecting on contextual feedback in an additional memory buffer, RedAgent continuously learns how to leverage these strategies to achieve effective jailbreaks in specific contexts. Extensive experiments demonstrate that our system can jailbreak most black-box LLMs in just five queries, improving the efficiency of existing red teaming methods by two times. Additionally, RedAgent can jailbreak customized LLM applications more efficiently. By generating context-aware jailbreak prompts towards applications on GPTs, we discover 60 severe vulnerabilities of these real-world applications with only two queries per vulnerability. We have reported all found issues and communicated with OpenAI and Meta for bug fixes.
Authors: Fan Bai, Junmo Kang, Gabriel Stanovsky, Dayne Freitag, Mark Dredze, Alan Ritter
Abstract: In this paper, we explore the question of whether large language models can support cost-efficient information extraction from tables. We introduce schema-driven information extraction, a new task that transforms tabular data into structured records following a human-authored schema. To assess various LLM's capabilities on this task, we present a benchmark comprised of tables from four diverse domains: machine learning papers, chemistry literature, material science journals, and webpages. We use this collection of annotated tables to evaluate the ability of open-source and API-based language models to extract information from tables covering diverse domains and data formats. Our experiments demonstrate that surprisingly competitive performance can be achieved without requiring task-specific pipelines or labels, achieving F1 scores ranging from 74.2 to 96.1, while maintaining cost efficiency. Moreover, through detailed ablation studies and analyses, we investigate the factors contributing to model success and validate the practicality of distilling compact models to reduce API reliance.
Authors: Yikun Wang, Rui Zheng, Haoming Li, Qi Zhang, Tao Gui, Fei Liu
Abstract: Customizing LLMs for a specific task involves separating high-quality responses from lower-quality ones. This skill can be developed using supervised fine-tuning with extensive human preference data. However, obtaining a large volume of expert-annotated data is costly for most tasks. In this paper, we explore a novel method to optimize LLMs using ranking metrics. This method trains the model to prioritize the best responses from a pool of candidates created for a particular task. Rather than a traditional full ordering, we advocate for a partial ordering, as achieving consensus on the perfect order of candidate responses can be challenging. Our partial ordering is more robust, less sensitive to noise, and can be achieved with limited human annotations or through heuristic methods. We test our system's improved response generation ability using benchmark datasets, including textual entailment and multi-document question answering. We conduct ablation studies to understand crucial factors, such as how to gather candidate responses for a specific task, determine their most suitable order, and balance supervised fine-tuning with ranking metrics. Our approach, named Rescue, offers a promising avenue for enhancing the response generation and task accuracy of LLMs.
Authors: Qiang Zhang, Keyang Ding, Tianwen Lyv, Xinda Wang, Qingyu Yin, Yiwen Zhang, Jing Yu, Yuhao Wang, Xiaotong Li, Zhuoyi Xiang, Kehua Feng, Xiang Zhuang, Zeyuan Wang, Ming Qin, Mengyao Zhang, Jinlu Zhang, Jiyu Cui, Tao Huang, Pengju Yan, Renjun Xu, Hongyang Chen, Xiaolin Li, Xiaohui Fan, Huabin Xing, Huajun Chen
Abstract: Large Language Models (LLMs) have emerged as a transformative power in enhancing natural language comprehension, representing a significant stride toward artificial general intelligence. The application of LLMs extends beyond conventional linguistic boundaries, encompassing specialized linguistic systems developed within various scientific disciplines. This growing interest has led to the advent of scientific LLMs, a novel subclass specifically engineered for facilitating scientific discovery. As a burgeoning area in the community of AI for Science, scientific LLMs warrant comprehensive exploration. However, a systematic and up-to-date survey introducing them is currently lacking. In this paper, we endeavor to methodically delineate the concept of "scientific language", whilst providing a thorough review of the latest advancements in scientific LLMs. Given the expansive realm of scientific disciplines, our analysis adopts a focused lens, concentrating on the biological and chemical domains. This includes an in-depth examination of LLMs for textual knowledge, small molecules, macromolecular proteins, genomic sequences, and their combinations, analyzing them in terms of model architectures, capabilities, datasets, and evaluation. Finally, we critically examine the prevailing challenges and point out promising research directions along with the advances of LLMs. By offering a comprehensive overview of technical developments in this field, this survey aspires to be an invaluable resource for researchers navigating the intricate landscape of scientific LLMs.
Authors: Tuan-Phong Nguyen, Simon Razniewski, Gerhard Weikum
Abstract: Despite recent progress, large language models (LLMs) still face the challenge of appropriately reacting to the intricacies of social and cultural conventions. This paper presents MANGO, a methodology for distilling high-accuracy, high-recall assertions of cultural knowledge. We judiciously and iteratively prompt LLMs for this purpose from two entry points, concepts and cultures. Outputs are consolidated via clustering and generative summarization. Running the MANGO method with GPT-3.5 as underlying LLM yields 167K high-accuracy assertions for 30K concepts and 11K cultures, surpassing prior resources by a large margin in quality and size. In an extrinsic evaluation for intercultural dialogues, we explore augmenting dialogue systems with cultural knowledge assertions. Notably, despite LLMs inherently possessing cultural knowledge, we find that adding knowledge from MANGO improves the overall quality, specificity, and cultural sensitivity of dialogue responses, as judged by human annotators. Data and code are available for download.
Authors: Min Zhang, Jianfeng He, Taoran Ji, Chang-Tien Lu
Abstract: The fairness and trustworthiness of Large Language Models (LLMs) are receiving increasing attention. Implicit hate speech, which employs indirect language to convey hateful intentions, occupies a significant portion of practice. However, the extent to which LLMs effectively address this issue remains insufficiently examined. This paper delves into the capability of LLMs to detect implicit hate speech (Classification Task) and express confidence in their responses (Calibration Task). Our evaluation meticulously considers various prompt patterns and mainstream uncertainty estimation methods. Our findings highlight that LLMs exhibit two extremes: (1) LLMs display excessive sensitivity towards groups or topics that may cause fairness issues, resulting in misclassifying benign statements as hate speech. (2) LLMs' confidence scores for each method excessively concentrate on a fixed range, remaining unchanged regardless of the dataset's complexity. Consequently, the calibration performance is heavily reliant on primary classification accuracy. These discoveries unveil new limitations of LLMs, underscoring the need for caution when optimizing models to ensure they do not veer towards extremes. This serves as a reminder to carefully consider sensitivity and confidence in the pursuit of model fairness.
Authors: Sayantan Adak, Daivik Agrawal, Animesh Mukherjee, Somak Aditya
Abstract: We investigate the knowledge of object affordances in pre-trained language models (LMs) and pre-trained Vision-Language models (VLMs). A growing body of literature shows that PTLMs fail inconsistently and non-intuitively, demonstrating a lack of reasoning and grounding. To take a first step toward quantifying the effect of grounding (or lack thereof), we curate a novel and comprehensive dataset of object affordances -- Text2Afford, characterized by 15 affordance classes. Unlike affordance datasets collected in vision and language domains, we annotate in-the-wild sentences with objects and affordances. Experimental results reveal that PTLMs exhibit limited reasoning abilities when it comes to uncommon object affordances. We also observe that pre-trained VLMs do not necessarily capture object affordances effectively. Through few-shot fine-tuning, we demonstrate improvement in affordance knowledge in PTLMs and VLMs. Our research contributes a novel dataset for language grounding tasks, and presents insights into LM capabilities, advancing the understanding of object affordances. Codes and data are available at https://github.com/sayantan11995/Affordance
Authors: Lele Cao, Valentin Buchner, Zineb Senane, Fangkai Yang
Abstract: Multimodal Large Language Models (MLLMs) are typically assessed using expensive annotated multimodal benchmarks, which often lag behind the rapidly evolving demands of MLLM evaluation. This paper outlines and validates GenCeption, a novel, annotation-free evaluation method that requires only unimodal data to measure inter-modality semantic coherence and inversely assesses MLLMs' tendency to hallucinate. This approach eliminates the need for costly data annotation, minimizes the risk of training data contamination, results in slower benchmark saturation, and avoids the illusion of emerging abilities. Inspired by the DrawCeption game, GenCeption begins with a non-textual sample and proceeds through iterative description and generation steps. The semantic drift across iterations is quantified using the GC@T metric. Based on the GenCeption method, we establish the MMECeption benchmark for evaluating Vision LLMs (VLLMs), and compare performance of several popular VLLMs and human annotators. Our empirical results validate GenCeption's effectiveness, demonstrating strong correlations with established VLLM benchmarks. VLLMs still significantly lack behind human performance and struggle especially with text-intensive tasks.
Authors: Piotr Nawrot, Adrian {\L}a\'ncucki, Marcin Chochowski, David Tarjan, Edoardo M. Ponti
Abstract: Transformers have emerged as the backbone of large language models (LLMs). However, generation remains inefficient due to the need to store in memory a cache of key-value representations for past tokens, whose size scales linearly with the input sequence length and batch size. As a solution, we propose Dynamic Memory Compression (DMC), a method for online key-value cache compression at inference time. Most importantly, the model learns to apply different compression ratios in different heads and layers. We retrofit pre-trained LLMs such as Llama 2 (7B, 13B and 70B) into DMC Transformers, achieving up to 7x throughput increase during auto-regressive inference on an NVIDIA H100 GPU. DMC is applied via continued pre-training on a negligible percentage of the original data without adding any extra parameters. DMC preserves the original downstream performance with up to 4x cache compression, outperforming up-trained grouped-query attention (GQA) and key-value eviction policies (H$_2$O, TOVA). GQA and DMC can be even combined to obtain compounded gains. Hence, DMC can serve as a drop-in replacement for KV caching in existing LLMs to fit longer contexts and larger batches within any given memory budget.
Authors: Pengda Wang, Zilin Xiao, Hanjie Chen, Frederick L. Oswald
Abstract: Although large language models (LLMs) have demonstrated remarkable proficiency in modeling text and generating human-like text, they may exhibit biases acquired from training data in doing so. Specifically, LLMs may be susceptible to a common cognitive trap in human decision-making called the representativeness heuristic. This is a concept in psychology that refers to judging the likelihood of an event based on how closely it resembles a well-known prototype or typical example, versus considering broader facts or statistical evidence. This research investigates the impact of the representativeness heuristic on LLM reasoning. We created ReHeAT (Representativeness Heuristic AI Testing), a dataset containing a series of problems spanning six common types of representativeness heuristics. Experiments reveal that four LLMs applied to ReHeAT all exhibited representativeness heuristic biases. We further identify that the model's reasoning steps are often incorrectly based on a stereotype rather than on the problem's description. Interestingly, the performance improves when adding a hint in the prompt to remind the model to use its knowledge. This suggests the uniqueness of the representativeness heuristic compared to traditional biases. It can occur even when LLMs possess the correct knowledge while falling into a cognitive trap. This highlights the importance of future research focusing on the representativeness heuristic in model reasoning and decision-making and on developing solutions to address it.
Authors: Siye Wu, Jian Xie, Jiangjie Chen, Tinghui Zhu, Kai Zhang, Yanghua Xiao
Abstract: By leveraging the retrieval of information from external knowledge databases, Large Language Models (LLMs) exhibit enhanced capabilities for accomplishing many knowledge-intensive tasks. However, due to the inherent flaws of current retrieval systems, there might exist irrelevant information within those retrieving top-ranked passages. In this work, we present a comprehensive investigation into the robustness of LLMs to different types of irrelevant information under various conditions. We initially introduce a framework to construct high-quality irrelevant information that ranges from semantically unrelated, partially related, and related to questions. Furthermore, our analysis demonstrates that the constructed irrelevant information not only scores highly on similarity metrics, being highly retrieved by existing systems, but also bears semantic connections to the context. Our investigation reveals that current LLMs still face challenges in discriminating highly semantically related information and can be easily distracted by these irrelevant yet misleading content. Besides, we also find that current solutions for handling irrelevant information have limitations in improving the robustness of LLMs to such distractions. All the resources are available on GitHub at https://github.com/Di-viner/LLM-Robustness-to-Irrelevant-Information.
URLs: https://github.com/Di-viner/LLM-Robustness-to-Irrelevant-Information.
Authors: Constanza Fierro, Reinald Kim Amplayo, Fantine Huot, Nicola De Cao, Joshua Maynez, Shashi Narayan, Mirella Lapata
Abstract: The increasing demand for the deployment of LLMs in information-seeking scenarios has spurred efforts in creating verifiable systems, which generate responses to queries along with supporting evidence. In this paper, we explore the attribution capabilities of plan-based models which have been recently shown to improve the faithfulness, grounding, and controllability of generated text. We conceptualize plans as a sequence of questions which serve as blueprints of the generated content and its organization. We propose two attribution models that utilize different variants of blueprints, an abstractive model where questions are generated from scratch, and an extractive model where questions are copied from the input. Experiments on long-form question-answering show that planning consistently improves attribution quality. Moreover, the citations generated by blueprint models are more accurate compared to those obtained from LLM-based pipelines lacking a planning component.
Authors: Quanyu Long, Yin Wu, Wenya Wang, Sinno Jialin Pan
Abstract: In-context Learning (ICL) has emerged as a powerful capability alongside the development of scaled-up large language models (LLMs). By instructing LLMs using few-shot demonstrative examples, ICL enables them to perform a wide range of tasks without updating millions of parameters. However, the precise contributions of demonstrations towards improving end-task performance have not been thoroughly investigated in recent analytical studies. In this paper, we empirically decompose the overall performance of ICL into three dimensions, label space, format, and discrimination, and we evaluate four general-purpose LLMs across a diverse range of tasks. Counter-intuitively, we find that the demonstrations have a marginal impact on provoking discriminative knowledge of language models. However, ICL exhibits significant efficacy in regulating the label space and format, which helps LLMs respond to desired label words. We then demonstrate that this ability functions similar to detailed instructions for LLMs to follow. We additionally provide an in-depth analysis of the mechanism of retrieval helping with ICL. Our findings demonstrate that retrieving the semantically similar examples notably boosts the model's discriminative capability. However, we also observe a trade-off in selecting good in-context examples regarding label diversity.
Authors: Samee Arif, Sualeha Farid, Awais Athar, Agha Ali Raza
Abstract: This paper introduces UQA, a novel dataset for question answering and text comprehension in Urdu, a low-resource language with over 70 million native speakers. UQA is generated by translating the Stanford Question Answering Dataset (SQuAD2.0), a large-scale English QA dataset, using a technique called EATS (Enclose to Anchor, Translate, Seek), which preserves the answer spans in the translated context paragraphs. The paper describes the process of selecting and evaluating the best translation model among two candidates: Google Translator and Seamless M4T. The paper also benchmarks several state-of-the-art multilingual QA models on UQA, including mBERT, XLM-RoBERTa, and mT5, and reports promising results. For XLM-RoBERTa-XL, we have an F1 score of 85.99 and 74.56 EM. UQA is a valuable resource for developing and testing multilingual NLP systems for Urdu and for enhancing the cross-lingual transferability of existing models. Further, the paper demonstrates the effectiveness of EATS for creating high-quality datasets for other languages and domains. The UQA dataset and the code are publicly available at www.github.com/sameearif/UQA.
Authors: \c{S}\"ukr\"u \"Ozt\"urk, Hacer Yalim Keles
Abstract: This study introduces the continuous Educational Turkish Sign Language (E-TSL) dataset, collected from online Turkish language lessons for 5th, 6th, and 8th grades. The dataset comprises 1,410 videos totaling nearly 24 hours and includes performances from 11 signers. Turkish, an agglutinative language, poses unique challenges for sign language translation, particularly with a vocabulary where 64% are singleton words and 85% are rare words, appearing less than five times. We developed two baseline models to address these challenges: the Pose to Text Transformer (P2T-T) and the Graph Neural Network based Transformer (GNN-T) models. The GNN-T model achieved 19.13% BLEU-1 score and 3.28% BLEU-4 score, presenting a significant challenge compared to existing benchmarks. The P2T-T model, while demonstrating slightly lower performance in BLEU scores, achieved a higher ROUGE-L score of 22.09%. Additionally, we benchmarked our model using the well-known PHOENIX-Weather 2014T dataset to validate our approach.
Authors: Andrew Shin, Kunitake Kaneko
Abstract: Large language models (LLMs) have demonstrated remarkable performances on a wide range of natural language tasks. Yet, LLMs' successes have been largely restricted to tasks concerning words, sentences, or documents, and it remains questionable how much they understand the minimal units of text, namely characters. In this paper, we examine contemporary LLMs regarding their ability to understand character composition of words, and show that most of them fail to reliably carry out even the simple tasks that can be handled by humans with perfection. We analyze their behaviors with comparison to token level performances, and discuss the potential directions for future research.
Authors: Xiou Ge, Ali Mousavi, Edouard Grave, Armand Joulin, Kun Qian, Benjamin Han, Mostafa Arefiyan, Yunyao Li
Abstract: Large Language Models (LLMs) have demonstrated impressive capability in different tasks and are bringing transformative changes to many domains. However, keeping the knowledge in LLMs up-to-date remains a challenge once pretraining is complete. It is thus essential to design effective methods to both update obsolete knowledge and induce new knowledge into LLMs. Existing locate-and-edit knowledge editing (KE) method suffers from two limitations. First, the post-edit LLMs by such methods generally have poor capability in answering complex queries that require multi-hop reasoning. Second, the long run-time of such locate-and-edit methods to perform knowledge edits make it infeasible for large scale KE in practice. In this paper, we explore Parameter-Efficient Fine-Tuning (PEFT) techniques as an alternative for KE. We curate a more comprehensive temporal KE dataset with both knowledge update and knowledge injection examples for KE performance benchmarking. We further probe the effect of fine-tuning on a range of layers in an LLM for the multi-hop QA task. We find that PEFT performs better than locate-and-edit techniques for time-sensitive knowledge edits.
Authors: Daniil Gurgurov, Mareike Hartmann, Simon Ostermann
Abstract: This paper explores the integration of graph knowledge from linguistic ontologies into multilingual Large Language Models (LLMs) using adapters to improve performance for low-resource languages (LRLs) in sentiment analysis (SA) and named entity recognition (NER). Building upon successful parameter-efficient fine-tuning techniques, such as K-ADAPTER and MAD-X, we propose a similar approach for incorporating knowledge from multilingual graphs, connecting concepts in various languages with each other through linguistic relationships, into multilingual LLMs for LRLs. Specifically, we focus on eight LRLs -- Maltese, Bulgarian, Indonesian, Nepali, Javanese, Uyghur, Tibetan, and Sinhala -- and employ language-specific adapters fine-tuned on data extracted from the language-specific section of ConceptNet, aiming to enable knowledge transfer across the languages covered by the knowledge graph. We compare various fine-tuning objectives, including standard Masked Language Modeling (MLM), MLM with full-word masking, and MLM with targeted masking, to analyse their effectiveness in learning and integrating the extracted graph data. Through empirical evaluation on language-specific tasks, we assess how structured graph knowledge affects the performance of multilingual LLMs for LRLs in SA and NER, providing insights into the potential benefits of adapting language models for low-resource scenarios.
Authors: Mokhaiber Dandash, Masoud Asadpour
Abstract: Social media is heading towards more and more personalization, where individuals reveal their beliefs, interests, habits, and activities, simply offering glimpses into their personality traits. This study, explores the correlation between the use of Arabic language on twitter, personality traits and its impact on sentiment analysis. We indicated the personality traits of users based on the information extracted from their profile activities, and the content of their tweets. Our analysis incorporated linguistic features, profile statistics (including gender, age, bio, etc.), as well as additional features like emoticons. To obtain personality data, we crawled the timelines and profiles of users who took the 16personalities test in Arabic on 16personalities.com. Our dataset, "AraPers", comprised 3,250 users who shared their personality results on twitter. We implemented various machine learning techniques, to reveal personality traits and developed a dedicated model for this purpose, achieving a 74.86% accuracy rate with BERT, analysis of this dataset proved that linguistic features, profile features and derived model can be used to differentiate between different personality traits. Furthermore, our findings demonstrated that personality affect sentiment in social media. This research contributes to the ongoing efforts in developing robust understanding of the relation between human behaviour on social media and personality features for real-world applications, such as political discourse analysis, and public opinion tracking.
Authors: Tianjie Ju, Yiting Wang, Xinbei Ma, Pengzhou Cheng, Haodong Zhao, Yulong Wang, Lifeng Liu, Jian Xie, Zhuosheng Zhang, Gongshen Liu
Abstract: The rapid adoption of large language models (LLMs) in multi-agent systems has highlighted their impressive capabilities in various applications, such as collaborative problem-solving and autonomous negotiation. However, the security implications of these LLM-based multi-agent systems have not been thoroughly investigated, particularly concerning the spread of manipulated knowledge. In this paper, we investigate this critical issue by constructing a detailed threat model and a comprehensive simulation environment that mirrors real-world multi-agent deployments in a trusted platform. Subsequently, we propose a novel two-stage attack method involving Persuasiveness Injection and Manipulated Knowledge Injection to systematically explore the potential for manipulated knowledge (i.e., counterfactual and toxic knowledge) spread without explicit prompt manipulation. Our method leverages the inherent vulnerabilities of LLMs in handling world knowledge, which can be exploited by attackers to unconsciously spread fabricated information. Through extensive experiments, we demonstrate that our attack method can successfully induce LLM-based agents to spread both counterfactual and toxic knowledge without degrading their foundational capabilities during agent communication. Furthermore, we show that these manipulations can persist through popular retrieval-augmented generation frameworks, where several benign agents store and retrieve manipulated chat histories for future interactions. This persistence indicates that even after the interaction has ended, the benign agents may continue to be influenced by manipulated knowledge. Our findings reveal significant security risks in LLM-based multi-agent systems, emphasizing the imperative need for robust defenses against manipulated knowledge spread, such as introducing ``guardian'' agents and advanced fact-checking tools.
Authors: David Rau, Shuai Wang, Herv\'e D\'ejean, St\'ephane Clinchant
Abstract: Retrieval-Augmented Generation (RAG) allows overcoming the limited knowledge of LLMs by extending the input with external information. As a consequence, the contextual inputs to the model become much longer which slows down decoding time directly translating to the time a user has to wait for an answer. We address this challenge by presenting COCOM, an effective context compression method, reducing long contexts to only a handful of Context Embeddings speeding up the generation time by a large margin. Our method allows for different compression rates trading off decoding time for answer quality. Compared to earlier methods, COCOM allows for handling multiple contexts more effectively, significantly reducing decoding time for long inputs. Our method demonstrates a speed-up of up to 5.69 $\times$ while achieving higher performance compared to existing efficient context compression methods.
Authors: Usneek Singh, Jos\'e Cambronero, Sumit Gulwani, Aditya Kanade, Anirudh Khatry, Vu Le, Mukul Singh, Gust Verbruggen
Abstract: Large language models (LLMs) can be leveraged to help with writing formulas in spreadsheets, but resources on these formulas are scarce, impacting both the base performance of pre-trained models and limiting the ability to fine-tune them. Given a corpus of formulas, we can use a(nother) model to generate synthetic natural language utterances for fine-tuning. However, it is important to validate whether the NL generated by the LLM is indeed accurate to be beneficial for fine-tuning. In this paper, we provide empirical results on the impact of validating these synthetic training examples with surrogate objectives that evaluate the accuracy of the synthetic annotations. We demonstrate that validation improves performance over raw data across four models (2 open and 2 closed weight). Interestingly, we show that although validation tends to prune more challenging examples, it increases the complexity of problems that models can solve after being fine-tuned on validated data.
Authors: Peiyang Wu, Nan Guo, Xiao Xiao, Wenming Li, Xiaochun Ye, Dongrui Fan
Abstract: Recently, large language models (LLMs) have demonstrated excellent performance in understanding human instructions and generating code, which has inspired researchers to explore the feasibility of generating RTL code with LLMs. However, the existing approaches to fine-tune LLMs on RTL codes typically are conducted on fixed datasets, which do not fully stimulate the capability of LLMs and require large amounts of reference data. To mitigate these issues , we introduce a simple yet effective iterative training paradigm named ITERTL. During each iteration, samples are drawn from the model trained in the previous cycle. Then these new samples are employed for training in this loop. Through this iterative approach, the distribution mismatch between the model and the training samples is reduced. Additionally, the model is thus enabled to explore a broader generative space and receive more comprehensive feedback. Theoretical analyses are conducted to investigate the mechanism of the effectiveness. Experimental results show the model trained through our proposed approach can compete with and even outperform the state-of-the-art (SOTA) open-source model with nearly 37\% reference samples, achieving remarkable 42.9\% and 62.2\% pass@1 rate on two VerilogEval evaluation datasets respectively. While using the same amount of reference samples, our method can achieved a relative improvement of 16.9\% and 12.5\% in pass@1 compared to the non-iterative method. This study facilitates the application of LLMs for generating RTL code in practical scenarios with limited data.
Authors: Youmin Ko, Hyemin Yang, Taeuk Kim, Hyunjoon Kim
Abstract: Fine-tuning pre-trained language models (PLMs) has recently shown a potential to improve knowledge graph completion (KGC). However, most PLM-based methods encode only textual information, neglecting various topological structures of knowledge graphs (KGs). In this paper, we empirically validate the significant relations between the structural properties of KGs and the performance of the PLM-based methods. To leverage the structural knowledge, we propose a Subgraph-Aware Training framework for KGC (SATKGC) that combines (i) subgraph-aware mini-batching to encourage hard negative sampling, and (ii) a new contrastive learning method to focus more on harder entities and harder negative triples in terms of the structural properties. To the best of our knowledge, this is the first study to comprehensively incorporate the structural inductive bias of the subgraphs into fine-tuning PLMs. Extensive experiments on four KGC benchmarks demonstrate the superiority of SATKGC. Our code is available.
Authors: Jinfeng Wei, Xiaofeng Zhang
Abstract: In this work, we introduce DOPRA, a novel approach designed to mitigate hallucinations in multi-modal large language models (MLLMs). Unlike existing solutions that typically involve costly supplementary training data or the integration of external knowledge sources, DOPRA innovatively addresses hallucinations by decoding specific weighted layer penalties and redistribution, offering an economical and effective solution without additional resources. DOPRA is grounded in unique insights into the intrinsic mechanisms controlling hallucinations within MLLMs, especially the models' tendency to over-rely on a subset of summary tokens in the self-attention matrix, neglecting critical image-related information. This phenomenon is particularly pronounced in certain strata. To counteract this over-reliance, DOPRA employs a strategy of weighted overlay penalties and redistribution in specific layers, such as the 12th layer, during the decoding process. Furthermore, DOPRA includes a retrospective allocation process that re-examines the sequence of generated tokens, allowing the algorithm to reallocate token selection to better align with the actual image content, thereby reducing the incidence of hallucinatory descriptions in auto-generated captions. Overall, DOPRA represents a significant step forward in improving the output quality of MLLMs by systematically reducing hallucinations through targeted adjustments during the decoding process.
Authors: Dominik Hintersdorf, Lukas Struppek, Daniel Neider, Kristian Kersting
Abstract: The proliferation of large AI models trained on uncurated, often sensitive web-scraped data has raised significant privacy concerns. One of the concerns is that adversaries can extract information about the training data using privacy attacks. Unfortunately, the task of removing specific information from the models without sacrificing performance is not straightforward and has proven to be challenging. We propose a rather easy yet effective defense based on backdoor attacks to remove private information, such as names and faces of individuals, from vision-language models by fine-tuning them for only a few minutes instead of re-training them from scratch. Specifically, by strategically inserting backdoors into text encoders, we align the embeddings of sensitive phrases with those of neutral terms-"a person" instead of the person's actual name. For image encoders, we map individuals' embeddings to be removed from the model to a universal, anonymous embedding. The results of our extensive experimental evaluation demonstrate the effectiveness of our backdoor-based defense on CLIP by assessing its performance using a specialized privacy attack for zero-shot classifiers. Our approach provides a new "dual-use" perspective on backdoor attacks and presents a promising avenue to enhance the privacy of individuals within models trained on uncurated web-scraped data.
Authors: Haibo Wang, Chenghang Lai, Yixuan Sun, Weifeng Ge
Abstract: Video Question Answering (VideoQA) aims to answer natural language questions based on the information observed in videos. Despite the recent success of Large Multimodal Models (LMMs) in image-language understanding and reasoning, they deal with VideoQA insufficiently, by simply taking uniformly sampled frames as visual inputs, which ignores question-relevant visual clues. Moreover, there are no human annotations for question-critical timestamps in existing VideoQA datasets. In light of this, we propose a novel weakly supervised framework to enforce the LMMs to reason out the answers with question-critical moments as visual inputs. Specifically, we first fuse the question and answer pairs as event descriptions to find multiple keyframes as target moments and pseudo-labels, with the visual-language alignment capability of the CLIP models. With these pseudo-labeled keyframes as additionally weak supervision, we devise a lightweight Gaussian-based Contrastive Grounding (GCG) module. GCG learns multiple Gaussian functions to characterize the temporal structure of the video, and sample question-critical frames as positive moments to be the visual inputs of LMMs. Extensive experiments on several benchmarks verify the effectiveness of our framework, and we achieve substantial improvements compared to previous state-of-the-art methods.
Authors: Iain Xie Weissburg, Mehir Arora, Xinyi Wang, Liangming Pan, William Yang Wang
Abstract: As the number of accepted papers at AI and ML conferences reaches into the thousands, it has become unclear how researchers access and read research publications. In this paper, we investigate the role of social media influencers in enhancing the visibility of machine learning research, particularly the citation counts of papers they share. We have compiled a comprehensive dataset of over 8,000 papers, spanning tweets from December 2018 to October 2023, alongside controls precisely matched by 9 key covariates. Our statistical and causal inference analysis reveals a significant increase in citations for papers endorsed by these influencers, with median citation counts 2-3 times higher than those of the control group. Additionally, the study delves into the geographic, gender, and institutional diversity of highlighted authors. Given these findings, we advocate for a responsible approach to curation, encouraging influencers to uphold the journalistic standard that includes showcasing diverse research topics, authors, and institutions.
Authors: Yihong Tang, Zhaokai Wang, Ao Qu, Yihao Yan, Zhaofeng Wu, Dingyi Zhuang, Jushi Kai, Kebing Hou, Xiaotong Guo, Jinhua Zhao, Zhan Zhao, Wei Ma
Abstract: Citywalk, a recently popular form of urban travel, requires genuine personalization and understanding of fine-grained requests compared to traditional itinerary planning. In this paper, we introduce the novel task of Open-domain Urban Itinerary Planning (OUIP), which generates personalized urban itineraries from user requests in natural language. We then present ITINERA, an OUIP system that integrates spatial optimization with large language models to provide customized urban itineraries based on user needs. This involves decomposing user requests, selecting candidate points of interest (POIs), ordering the POIs based on cluster-aware spatial optimization, and generating the itinerary. Experiments on real-world datasets and the performance of the deployed system demonstrate our system's capacity to deliver personalized and spatially coherent itineraries compared to current solutions.
Authors: Yusu Qian, Haotian Zhang, Yinfei Yang, Zhe Gan
Abstract: The remarkable advancements in Multimodal Large Language Models (MLLMs) have not rendered them immune to challenges, particularly in the context of handling deceptive information in prompts, thus producing hallucinated responses under such conditions. To quantitatively assess this vulnerability, we present MAD-Bench, a carefully curated benchmark that contains 1000 test samples divided into 5 categories, such as non-existent objects, count of objects, and spatial relationship. We provide a comprehensive analysis of popular MLLMs, ranging from GPT-4v, Reka, Gemini-Pro, to open-sourced models, such as LLaVA-NeXT and MiniCPM-Llama3. Empirically, we observe significant performance gaps between GPT-4o and other models; and previous robust instruction-tuned models are not effective on this new benchmark. While GPT-4o achieves 82.82% accuracy on MAD-Bench, the accuracy of any other model in our experiments ranges from 9% to 50%. We further propose a remedy that adds an additional paragraph to the deceptive prompts to encourage models to think twice before answering the question. Surprisingly, this simple method can even double the accuracy; however, the absolute numbers are still too low to be satisfactory. We hope MAD-Bench can serve as a valuable benchmark to stimulate further research to enhance model resilience against deceptive prompts.
Authors: Nigel Fernandez, Alexander Scarlatos, Andrew Lan
Abstract: Automated teaching assistants and chatbots have significant potential to reduce the workload of human instructors, especially for logistics-related question answering, which is important to students yet repetitive for instructors. However, due to privacy concerns, there is a lack of publicly available datasets. We introduce SyllabusQA, an open-source dataset with 63 real course syllabi covering 36 majors, containing 5,078 open-ended course logistics-related question-answer pairs that are diverse in both question types and answer formats. Since many logistics-related questions contain critical information like the date of an exam, it is important to evaluate the factuality of answers. We benchmark several strong baselines on this task, from large language model prompting to retrieval-augmented generation. We introduce Fact-QA, an LLM-based (GPT-4) evaluation metric to evaluate the factuality of predicted answers. We find that despite performing close to humans on traditional metrics of textual similarity, there remains a significant gap between automated approaches and humans in terms of fact precision.
Authors: Prasanna Vijayaraghavan, Jeffrey Frederic Queisser, Sergio Verduzco Flores, Jun Tani
Abstract: Humans excel at applying learned behavior to unlearned situations. A crucial component of this generalization behavior is our ability to compose/decompose a whole into reusable parts, an attribute known as compositionality. One of the fundamental questions in robotics concerns this characteristic. "How can linguistic compositionality be developed concomitantly with sensorimotor skills through associative learning, particularly when individuals only learn partial linguistic compositions and their corresponding sensorimotor patterns?" To address this question, we propose a brain-inspired neural network model that integrates vision, proprioception, and language into a framework of predictive coding and active inference, based on the free-energy principle. The effectiveness and capabilities of this model were assessed through various simulation experiments conducted with a robot arm. Our results show that generalization in learning to unlearned verb-noun compositions, is significantly enhanced when training variations of task composition are increased. We attribute this to self-organized compositional structures in linguistic latent state space being influenced significantly by sensorimotor learning. Ablation studies show that visual attention and working memory are essential to accurately generate visuo-motor sequences to achieve linguistically represented goals. These insights advance our understanding of mechanisms underlying development of compositionality through interactions of linguistic and sensorimotor experience.
Authors: Bin Lei, Yi Zhang, Shan Zuo, Ali Payani, Caiwen Ding
Abstract: Recent advancements in large language models, such as GPT-4, have demonstrated remarkable capabilities in processing standard queries. Despite these advancements, their performance substantially declines in \textbf{advanced mathematical problems requiring complex, multi-step logical reasoning}. To enhance their inferential capabilities, current research has delved into \textit{prompting engineering}, exemplified by methodologies such as the Tree of Thought and Graph of Thought. Nonetheless, these existing approaches encounter two significant limitations. Firstly, their effectiveness in tackling complex mathematical problems is somewhat constrained. Secondly, the necessity to design distinct prompts for individual problems hampers their generalizability. In response to these limitations, this paper introduces the \textit{Multi-Agent System for conditional Mining} (\textbf{MACM}) prompting method. It not only resolves intricate mathematical problems but also demonstrates strong generalization capabilities across various mathematical contexts. With the assistance of MACM, the accuracy of GPT-4 Turbo on the most challenging level five mathematical problems in the MATH dataset increase from $\mathbf{54.68\%} \text{ to } \mathbf{76.73\%}$. The code is available in \url{https://github.com/bin123apple/MACM}.
Authors: Jun Li, Su Hwan Kim, Philip M\"uller, Lina Felsner, Daniel Rueckert, Benedikt Wiestler, Julia A. Schnabel, Cosmin I. Bercea
Abstract: This research explores the integration of language models and unsupervised anomaly detection in medical imaging, addressing two key questions: (1) Can language models enhance the interpretability of anomaly detection maps? and (2) Can anomaly maps improve the generalizability of language models in open-set anomaly detection tasks? To investigate these questions, we introduce a new dataset for multi-image visual question-answering on brain magnetic resonance images encompassing multiple conditions. We propose KQ-Former (Knowledge Querying Transformer), which is designed to optimally align visual and textual information in limited-sample contexts. Our model achieves a 60.81% accuracy on closed questions, covering disease classification and severity across 15 different classes. For open questions, KQ-Former demonstrates a 70% improvement over the baseline with a BLEU-4 score of 0.41, and achieves the highest entailment ratios (up to 71.9%) and lowest contradiction ratios (down to 10.0%) among various natural language inference models. Furthermore, integrating anomaly maps results in an 18% accuracy increase in detecting open-set anomalies, thereby enhancing the language model's generalizability to previously unseen medical conditions. The code and dataset are available at https://github.com/compai-lab/miccai-2024-junli?tab=readme-ov-file
URLs: https://github.com/compai-lab/miccai-2024-junli?tab=readme-ov-file
Authors: Selvan Sunitha Ravi, Bartosz Mielczarek, Anand Kannappan, Douwe Kiela, Rebecca Qian
Abstract: Retrieval Augmented Generation (RAG) techniques aim to mitigate hallucinations in Large Language Models (LLMs). However, LLMs can still produce information that is unsupported or contradictory to the retrieved contexts. We introduce LYNX, a SOTA hallucination detection LLM that is capable of advanced reasoning on challenging real-world hallucination scenarios. To evaluate LYNX, we present HaluBench, a comprehensive hallucination evaluation benchmark, consisting of 15k samples sourced from various real-world domains. Our experiment results show that LYNX outperforms GPT-4o, Claude-3-Sonnet, and closed and open-source LLM-as-a-judge models on HaluBench. We release LYNX, HaluBench and our evaluation code for public access.
Authors: Nuo Chen, Yang Deng, Jia Li
Abstract: This survey explores the burgeoning field of role-playing with language models, focusing on their development from early persona-based models to advanced character-driven simulations facilitated by Large Language Models (LLMs). Initially confined to simple persona consistency due to limited model capabilities, role-playing tasks have now expanded to embrace complex character portrayals involving character consistency, behavioral alignment, and overall attractiveness. We provide a comprehensive taxonomy of the critical components in designing these systems, including data, models and alignment, agent architecture and evaluation. This survey not only outlines the current methodologies and challenges, such as managing dynamic personal profiles and achieving high-level persona consistency but also suggests avenues for future research in improving the depth and realism of role-playing applications. The goal is to guide future research by offering a structured overview of current methodologies and identifying potential areas for improvement. Related resources and papers are available at https://github.com/nuochenpku/Awesome-Role-Play-Papers.
URLs: https://github.com/nuochenpku/Awesome-Role-Play-Papers.
Authors: Alessio Buscemi, Daniele Proverbio
Abstract: The ethical implications and potentials for misuse of Generative Artificial Intelligence are increasingly worrying topics. This paper explores how easily the default ethical guardrails of ChatGPT, using its latest customization features, can be bypassed by simple prompts and fine-tuning, that can be effortlessly accessed by the broad public. This malevolently altered version of ChatGPT, nicknamed "RogueGPT", responded with worrying behaviours, beyond those triggered by jailbreak prompts. We conduct an empirical study of RogueGPT responses, assessing its flexibility in answering questions pertaining to what should be disallowed usage. Our findings raise significant concerns about the model's knowledge about topics like illegal drug production, torture methods and terrorism. The ease of driving ChatGPT astray, coupled with its global accessibility, highlights severe issues regarding the data quality used for training the foundational model and the implementation of ethical safeguards. We thus underline the responsibilities and dangers of user-driven modifications, and the broader effects that these may have on the design of safeguarding and ethical modules implemented by AI programmers.
Authors: Jianxin Liang, Xiaojun Meng, Yueqian Wang, Chang Liu, Qun Liu, Dongyan Zhao
Abstract: Video Question Answering (VideoQA) has emerged as a challenging frontier in the field of multimedia processing, requiring intricate interactions between visual and textual modalities. Simply uniformly sampling frames or indiscriminately aggregating frame-level visual features often falls short in capturing the nuanced and relevant contexts of videos to well perform VideoQA. To mitigate these issues, we propose VidF4, a novel VideoQA framework equipped with tailored frame selection strategy for effective and efficient VideoQA. We propose three frame-scoring mechanisms that consider both question relevance and inter-frame similarity to evaluate the importance of each frame for a given question on the video. Furthermore, we design a differentiable adaptive frame sampling mechanism to facilitate end-to-end training for the frame selector and answer generator. The experimental results across three widely adopted benchmarks demonstrate that our model consistently outperforms existing VideoQA methods, establishing a new SOTA across NExT-QA (+0.3%), STAR (+0.9%), and TVQA (+1.0%). Furthermore, through both quantitative and qualitative analyses, we validate the effectiveness of each design choice.
Authors: Jiaxing Zhang, Jiayi Liu, Dongsheng Luo, Jennifer Neville, Hua Wei
Abstract: Recent studies seek to provide Graph Neural Network (GNN) interpretability via multiple unsupervised learning models. Due to the scarcity of datasets, current methods easily suffer from learning bias. To solve this problem, we embed a Large Language Model (LLM) as knowledge into the GNN explanation network to avoid the learning bias problem. We inject LLM as a Bayesian Inference (BI) module to mitigate learning bias. The efficacy of the BI module has been proven both theoretically and experimentally. We conduct experiments on both synthetic and real-world datasets. The innovation of our work lies in two parts: 1. We provide a novel view of the possibility of an LLM functioning as a Bayesian inference to improve the performance of existing algorithms; 2. We are the first to discuss the learning bias issues in the GNN explanation problem.