Authors: Shravan Cheekati, Mridul Gupta, Vibha Raghu, Pranav Raj
Abstract: Age-related language patterns play a crucial role in understanding linguistic differences and developing age-appropriate communication strategies. However, the lack of comprehensive and diverse datasets has hindered the progress of research in this area. To address this issue, we present TextAge, a curated text dataset that maps sentences to the age and age group of the producer, as well as an underage (under 13) label. TextAge covers a wide range of ages and includes both spoken and written data from various sources such as CHILDES, Meta, Poki Poems-by-kids, JUSThink, and the TV show "Survivor." The dataset undergoes extensive cleaning and preprocessing to ensure data quality and consistency. We demonstrate the utility of TextAge through two applications: Underage Detection and Generational Classification. For Underage Detection, we train a Naive Bayes classifier, fine-tuned RoBERTa, and XLNet models to differentiate between language patterns of minors and young-adults and over. For Generational Classification, the models classify language patterns into different age groups (kids, teens, twenties, etc.). The models excel at classifying the "kids" group but struggle with older age groups, particularly "fifties," "sixties," and "seventies," likely due to limited data samples and less pronounced linguistic differences. TextAge offers a valuable resource for studying age-related language patterns and developing age-sensitive language models. The dataset's diverse composition and the promising results of the classification tasks highlight its potential for various applications, such as content moderation, targeted advertising, and age-appropriate communication. Future work aims to expand the dataset further and explore advanced modeling techniques to improve performance on older age groups.
Authors: Michael Ackerman
Abstract: This research paper discusses the advances made in the past decade in biomedicine and Large Language Models. To understand how the advances have been made hand-in-hand with one another, the paper also discusses the integration of Natural Language Processing techniques and tools into biomedicine. Finally, the goal of this paper is to expand on a survey conducted last year (2023) by introducing a new list of questions and prompts for the top two language models. Through this survey, this paper seeks to quantify the improvement made in the reasoning abilities in LLMs and to what extent those improvements are felt by the average user. Additionally, this paper seeks to extend research on retrieval of biological literature by prompting the LLM to answer open-ended questions in great depth.
Authors: Dominik Farhan
Abstract: Entity linking (EL) is the computational process of connecting textual mentions to corresponding entities. Like many areas of natural language processing, the EL field has greatly benefited from deep learning, leading to significant performance improvements. However, present-day approaches are expensive to train and rely on diverse data sources, complicating their reproducibility. In this thesis, we develop multiple systems that are fast to train, demonstrating that competitive entity linking can be achieved without a large GPU cluster. Moreover, we train on a publicly available dataset, ensuring reproducibility and accessibility. Our models are evaluated for 9 languages giving an accurate overview of their strengths. Furthermore, we offer a~detailed analysis of bi-encoder training hyperparameters, a popular approach in EL, to guide their informed selection. Overall, our work shows that building competitive neural network based EL systems that operate in multiple languages is possible even with limited resources, thus making EL more approachable.
Authors: Wazib Ansar, Saptarsi Goswami, Amlan Chakrabarti
Abstract: The advent of transformers with attention mechanisms and associated pre-trained models have revolutionized the field of Natural Language Processing (NLP). However, such models are resource-intensive due to highly complex architecture. This limits their application to resource-constrained environments. While choosing an appropriate NLP model, a major trade-off exists over choosing accuracy over efficiency and vice versa. This paper presents a commentary on the evolution of NLP and its applications with emphasis on their accuracy as-well-as efficiency. Following this, a survey of research contributions towards enhancing the efficiency of transformer-based models at various stages of model development along with hardware considerations has been conducted. The goal of this survey is to determine how current NLP techniques contribute towards a sustainable society and to establish a foundation for future research.
Authors: Jieh-Sheng Lee
Abstract: In this research, patent prosecution is conceptualized as a system of reinforcement learning from human feedback. The objective of the system is to increase the likelihood for a language model to generate patent claims that have a higher chance of being granted. To showcase the controllability of the language model, the system learns from granted patents and pre-grant applications with different rewards. The status of "granted" and "pre-grant" are perceived as labeled human feedback implicitly. In addition, specific to patent drafting, the experiments in this research demonstrate the model's capability to learn from adjusting claim length and inclusion of limiting terms for narrowing claim scope. As proof of concept, the experiments focus on claim ones only and the training data originates from a patent dataset tailored specifically for artificial intelligence. Although the available human feedback in patent prosecution are limited and the quality of generated patent text requires improvement, the experiments following the 3-stage reinforcement learning from human feedback have demonstrated that generative language models are capable of reflecting the human feedback or intent in patent prosecution. To enhance the usability of language models, the implementation in this research utilizes modern techniques that enable execution on a single consumer-grade GPU. The demonstrated proof of concept, which reduces hardware requirements, will prove valuable in the future as more human feedback in patent prosecution become available for broader use, either within patent offices or in the public domain.
Authors: Yuni Susanti, Nina Holsmoelle
Abstract: This work aims toward an application of natural language processing (NLP) technology for automatic verification of causal graphs using text sources. A causal graph is often derived from unsupervised causal discovery methods and requires manual evaluation from human experts. NLP technologies, i.e., Large Language Models (LLMs) such as BERT and ChatGPT, can potentially be used to verify the resulted causal graph by predicting if causal relation can be observed between node pairs based on the textual context. In this work, we compare the performance of two types of NLP models: (1) Pre-trained language models fine-tuned for causal relation classification task and, (2) prompt-based LLMs. Contrasted to previous studies where prompt-based LLMs work relatively well over a set of diverse tasks, preliminary experiments on biomedical and open-domain datasets suggest that the fine-tuned models far outperform the prompt-based LLMs, up to 20.5 points improvement of F1 score. We shared the code and the pre-processed datasets in our repository.
Authors: Jingwei Wang
Abstract: This project investigates the behavior of multi-head attention in Transformer models, specifically focusing on the differences between benign and trojan models in the context of sentiment analysis. Trojan attacks cause models to perform normally on clean inputs but exhibit misclassifications when presented with inputs containing predefined triggers. We characterize attention head functions in trojan and benign models, identifying specific 'trojan' heads and analyzing their behavior.
Authors: Sumit Kumar Dam, Choong Seon Hong, Yu Qiao, Chaoning Zhang
Abstract: The past few decades have witnessed an upsurge in data, forming the foundation for data-hungry, learning-based AI technology. Conversational agents, often referred to as AI chatbots, rely heavily on such data to train large language models (LLMs) and generate new content (knowledge) in response to user prompts. With the advent of OpenAI's ChatGPT, LLM-based chatbots have set new standards in the AI community. This paper presents a complete survey of the evolution and deployment of LLM-based chatbots in various sectors. We first summarize the development of foundational chatbots, followed by the evolution of LLMs, and then provide an overview of LLM-based chatbots currently in use and those in the development phase. Recognizing AI chatbots as tools for generating new knowledge, we explore their diverse applications across various industries. We then discuss the open challenges, considering how the data used to train the LLMs and the misuse of the generated knowledge can cause several issues. Finally, we explore the future outlook to augment their efficiency and reliability in numerous applications. By addressing key milestones and the present-day context of LLM-based chatbots, our survey invites readers to delve deeper into this realm, reflecting on how their next generation will reshape conversational AI.
Authors: Nathan A. Chi, Teodor Malchev, Riley Kong, Ryan A. Chi, Lucas Huang, Ethan A. Chi, R. Thomas McCoy, Dragomir Radev
Abstract: We introduce modeLing, a novel benchmark of Linguistics Olympiad-style puzzles which tests few-shot reasoning in AI systems. Solving these puzzles necessitates inferring aspects of a language's grammatical structure from a small number of examples. Such puzzles provide a natural testbed for language models, as they require compositional generalization and few-shot inductive reasoning. Consisting solely of new puzzles written specifically for this work, modeLing has no risk of appearing in the training data of existing AI systems: this ameliorates the risk of data leakage, a potential confounder for many prior evaluations of reasoning. Evaluating several large open source language models and GPT on our benchmark, we observe non-negligible accuracy, demonstrating few-shot emergent reasoning ability which cannot merely be attributed to shallow memorization. However, imperfect model performance suggests that modeLing can be used to measure further progress in linguistic reasoning.
Authors: Ryan Liu, Jiayi Geng, Joshua C. Peterson, Ilia Sucholutsky, Thomas L. Griffiths
Abstract: In order for AI systems to communicate effectively with people, they must understand how we make decisions. However, people's decisions are not always rational, so the implicit internal models of human decision-making in Large Language Models (LLMs) must account for this. Previous empirical evidence seems to suggest that these implicit models are accurate -- LLMs offer believable proxies of human behavior, acting how we expect humans would in everyday interactions. However, by comparing LLM behavior and predictions to a large dataset of human decisions, we find that this is actually not the case: when both simulating and predicting people's choices, a suite of cutting-edge LLMs (GPT-4o & 4-Turbo, Llama-3-8B & 70B, Claude 3 Opus) assume that people are more rational than we really are. Specifically, these models deviate from human behavior and align more closely with a classic model of rational choice -- expected value theory. Interestingly, people also tend to assume that other people are rational when interpreting their behavior. As a consequence, when we compare the inferences that LLMs and people draw from the decisions of others using another psychological dataset, we find that these inferences are highly correlated. Thus, the implicit decision-making models of LLMs appear to be aligned with the human expectation that other people will act rationally, rather than with how people actually act.
Authors: Meiru Zhang, Zaiqiao Meng, Nigel Collier
Abstract: The context window of large language models has been extended to 128k tokens or more. However, language models still suffer from position bias and have difficulty in accessing and using the middle part of the context due to the lack of attention. We examine the relative position awareness of LLMs and the feasibility of mitigating disproportional attention through prompting. We augment the original task instruction with $\texttt{attention instructions}$ that direct language models to allocate more attention towards a selected segment of the context. We conduct a comprehensive investigation on multi-document question answering task with both position-based and index-based instructions. We find that language models do not have relative position awareness of the context. Nevertheless, they demonstrate the capacity to adapt attention to a specific segment using matching indexes. Our analysis contributes to a deeper understanding of position bias in LLMs and provides a pathway to mitigate this bias by instruction, thus benefiting LLMs in locating and utilizing relevant information from retrieved documents in RAG applications.
Authors: Yash Kumar Lal, Preethi Lahoti, Aradhana Sinha, Yao Qin, Ananth Balashankar
Abstract: Safety classifiers are critical in mitigating toxicity on online forums such as social media and in chatbots. Still, they continue to be vulnerable to emergent, and often innumerable, adversarial attacks. Traditional automated adversarial data generation methods, however, tend to produce attacks that are not diverse, but variations of previously observed harm types. We formalize the task of automated adversarial discovery for safety classifiers - to find new attacks along previously unseen harm dimensions that expose new weaknesses in the classifier. We measure progress on this task along two key axes (1) adversarial success: does the attack fool the classifier? and (2) dimensional diversity: does the attack represent a previously unseen harm type? Our evaluation of existing attack generation methods on the CivilComments toxicity task reveals their limitations: Word perturbation attacks fail to fool classifiers, while prompt-based LLM attacks have more adversarial success, but lack dimensional diversity. Even our best-performing prompt-based method finds new successful attacks on unseen harm dimensions of attacks only 5\% of the time. Automatically finding new harmful dimensions of attack is crucial and there is substantial headroom for future research on our new task.
Authors: Venktesh V. Deepali Prabhu, Avishek Anand
Abstract: Open-domain complex Question Answering (QA) is a difficult task with challenges in evidence retrieval and reasoning. The complexity of such questions could stem from questions being compositional, hybrid evidence, or ambiguity in questions. While retrieval performance for classical QA tasks is well explored, their capabilities for heterogeneous complex retrieval tasks, especially in an open-domain setting, and the impact on downstream QA performance, are relatively unexplored. To address this, in this work, we propose a benchmark composing diverse complex QA tasks and provide a toolkit to evaluate state-of-the-art pre-trained dense and sparse retrieval models in an open-domain setting. We observe that late interaction models and surprisingly lexical models like BM25 perform well compared to other pre-trained dense retrieval models. In addition, since context-based reasoning is critical for solving complex QA tasks, we also evaluate the reasoning capabilities of LLMs and the impact of retrieval performance on their reasoning capabilities. Through experiments, we observe that much progress is to be made in retrieval for complex QA to improve downstream QA performance. Our software and related data can be accessed at https://github.com/VenkteshV/DEXTER
Authors: Vikas Yadav, Zheng Tang, Vijay Srinivasan
Abstract: Large language models (LLM) have achieved remarkable success in natural language generation but lesser focus has been given to their applicability in decision making tasks such as classification. We show that LLMs like LLaMa can achieve high performance on large multi-class classification tasks but still make classification errors and worse, generate out-of-vocabulary class labels. To address these critical issues, we introduce Paraphrase and AGgregate (PAG)-LLM approach wherein an LLM generates multiple paraphrases of the input query (parallel queries), performs multi-class classification for the original query and each paraphrase, and at the end aggregate all the classification labels based on their confidence scores. We evaluate PAG-LLM on two large multi-class classication datasets: CLINC, and Banking and show 22.7% and 15.1% error reduction. We show that PAG-LLM is especially effective for hard examples where LLM is uncertain, and reduces the critical misclassification and hallucinated label generation errors
Authors: Nisarg Patel, Mohith Kulkarni, Mihir Parmar, Aashna Budhiraja, Mutsumi Nakamura, Neeraj Varshney, Chitta Baral
Abstract: As Large Language Models (LLMs) continue to exhibit remarkable performance in natural language understanding tasks, there is a crucial need to measure their ability for human-like multi-step logical reasoning. Existing logical reasoning evaluation benchmarks often focus primarily on simplistic single-step or multi-step reasoning with a limited set of inference rules. Furthermore, the lack of datasets for evaluating non-monotonic reasoning represents a crucial gap since it aligns more closely with human-like reasoning. To address these limitations, we propose Multi-LogiEval, a comprehensive evaluation dataset encompassing multi-step logical reasoning with various inference rules and depths. Multi-LogiEval covers three logic types--propositional, first-order, and non-monotonic--consisting of more than 30 inference rules and more than 60 of their combinations with various depths. Leveraging this dataset, we conduct evaluations on a range of LLMs including GPT-4, ChatGPT, Gemini-Pro, Yi, Orca, and Mistral, employing a zero-shot chain-of-thought. Experimental results show that there is a significant drop in the performance of LLMs as the reasoning steps/depth increases (average accuracy of ~68% at depth-1 to ~43% at depth-5). We further conduct a thorough investigation of reasoning chains generated by LLMs which reveals several important findings. We believe that Multi-LogiEval facilitates future research for evaluating and enhancing the logical reasoning ability of LLMs. Data is available at https://github.com/Mihir3009/Multi-LogiEval.
Authors: Koichi Akabe, Shunsuke Kanda, Yusuke Oda, Shinsuke Mori
Abstract: This paper proposes an approach to improve the runtime efficiency of Japanese tokenization based on the pointwise linear classification (PLC) framework, which formulates the whole tokenization process as a sequence of linear classification problems. Our approach optimizes tokenization by leveraging the characteristics of the PLC framework and the task definition. Our approach involves (1) composing multiple classifications into array-based operations, (2) efficient feature lookup with memory-optimized automata, and (3) three orthogonal pre-processing methods for reducing actual score calculation. Thus, our approach makes the tokenization speed 5.7 times faster than the current approach based on the same model without decreasing tokenization accuracy. Our implementation is available at https://github.com/daac-tools/vaporetto under the MIT or Apache-2.0 license.
Authors: Abe Bohan Hou, Orion Weller, Guanghui Qin, Eugene Yang, Dawn Lawrie, Nils Holzenberger, Andrew Blair-Stanek, Benjamin Van Durme
Abstract: Legal professionals need to write analyses that rely on citations to relevant precedents, i.e., previous case decisions. Intelligent systems assisting legal professionals in writing such documents provide great benefits but are challenging to design. Such systems need to help locate, summarize, and reason over salient precedents in order to be useful. To enable systems for such tasks, we work with legal professionals to transform a large open-source legal corpus into a dataset supporting two important backbone tasks: information retrieval (IR) and retrieval-augmented generation (RAG). This dataset CLERC (Case Law Evaluation Retrieval Corpus), is constructed for training and evaluating models on their ability to (1) find corresponding citations for a given piece of legal analysis and to (2) compile the text of these citations (as well as previous context) into a cogent analysis that supports a reasoning goal. We benchmark state-of-the-art models on CLERC, showing that current approaches still struggle: GPT-4o generates analyses with the highest ROUGE F-scores but hallucinates the most, while zero-shot IR models only achieve 48.3% recall@1000.
Authors: Isidora Chara Tourni, Lei Guo, Hengchang Hu, Edward Halim, Prakash Ishwar, Taufiq Daryanto, Mona Jalal, Boqi Chen, Margrit Betke, Fabian Zhafransyah, Sha Lai, Derry Tanti Wijaya
Abstract: News media structure their reporting of events or issues using certain perspectives. When describing an incident involving gun violence, for example, some journalists may focus on mental health or gun regulation, while others may emphasize the discussion of gun rights. Such perspectives are called \say{frames} in communication research. We study, for the first time, the value of combining lead images and their contextual information with text to identify the frame of a given news article. We observe that using multiple modes of information(article- and image-derived features) improves prediction of news frames over any single mode of information when the images are relevant to the frames of the headlines. We also observe that frame image relevance is related to the ease of conveying frames via images, which we call frame concreteness. Additionally, we release the first multimodal news framing dataset related to gun violence in the U.S., curated and annotated by communication researchers. The dataset will allow researchers to further examine the use of multiple information modalities for studying media framing.
Authors: Tong Zhou, Yubo Chen, Kang Liu, Jun Zhao
Abstract: Large language models have become integral to question-answering applications despite their propensity for generating hallucinations and factually inaccurate content. Querying knowledge graphs to reduce hallucinations in LLM meets the challenge of incomplete knowledge coverage in knowledge graphs. On the other hand, updating knowledge graphs by information extraction and knowledge graph completion faces the knowledge update misalignment issue. In this work, we introduce a collaborative augmentation framework, CogMG, leveraging knowledge graphs to address the limitations of LLMs in QA scenarios, explicitly targeting the problems of incomplete knowledge coverage and knowledge update misalignment. The LLMs identify and decompose required knowledge triples that are not present in the KG, enriching them and aligning updates with real-world demands. We demonstrate the efficacy of this approach through a supervised fine-tuned LLM within an agent framework, showing significant improvements in reducing hallucinations and enhancing factual accuracy in QA responses. Our code and video are publicly available.
Authors: Yun-Shiuan Chuang, Zach Studdiford, Krirk Nirunwiroj, Agam Goyal, Vincent V. Frigo, Sijia Yang, Dhavan Shah, Junjie Hu, Timothy T. Rogers
Abstract: Creating human-like large language model (LLM) agents is crucial for faithful social simulation. Having LLMs role-play based on demographic information sometimes improves human likeness but often does not. This study assessed whether LLM alignment with human behavior can be improved by integrating information from empirically-derived human belief networks. Using data from a human survey, we estimated a belief network encompassing 18 topics loading on two non-overlapping latent factors. We then seeded LLM-based agents with an opinion on one topic, and assessed the alignment of its expressed opinions on remaining test topics with corresponding human data. Role-playing based on demographic information alone did not align LLM and human opinions, but seeding the agent with a single belief greatly improved alignment for topics related in the belief network, and not for topics outside the network. These results suggest a novel path for human-LLM belief alignment in work seeking to simulate and understand patterns of belief distributions in society.
Authors: Huaizhi Ge, Frank Rudzicz, Zining Zhu
Abstract: In the field of language model interpretability, circuit discovery is gaining popularity. Despite this, the true meaning of these circuits remain largely unanswered. We introduce a novel method to learn their meanings as a holistic object through the lens of knowledge editing. We extract circuits in the GPT2-XL model using diverse text classification datasets, and use hierarchical relations datasets to explore knowledge editing in the circuits. Our findings indicate that these circuits contain entity knowledge but resist new knowledge more than complementary circuits during knowledge editing. Additionally, we examine the impact of circuit size, discovering that an ideal "theoretical circuit" where essential knowledge is concentrated likely incorporates more than 5% but less than 50% of the model's parameters. We also assess the overlap between circuits from different datasets, finding moderate similarities. What constitutes these circuits, then? We find that up to 60% of the circuits consist of layer normalization modules rather than attention or MLP modules, adding evidence to the ongoing debates regarding knowledge localization. In summary, our findings offer new insights into the functions of the circuits, and introduce research directions for further interpretability and safety research of language models.
Authors: Huaizhi Ge, Frank Rudzicz, Zining Zhu
Abstract: As large language models (LLMs) are widely deployed, targeted editing of their knowledge has become a critical challenge. Recently, advancements in model editing techniques, such as Rank-One Model Editing (ROME), have paved the way for updating LLMs with new knowledge. However, the efficacy of these methods varies across different types of knowledge. This study investigates the capability of knowledge editing methods to incorporate new knowledge with varying degrees of "perplexingness", a term we use to describe the initial difficulty LLMs have in understanding new concepts. We begin by quantifying the "perplexingness" of target knowledge using pre-edit conditional probabilities, and assess the efficacy of edits through post-edit conditional probabilities. Utilizing the widely-used CounterFact dataset, we find significant negative correlations between the "perplexingness" of the new knowledge and the edit efficacy across all 12 scenarios. To dive deeper into this phenomenon, we introduce a novel dataset, HierarchyData, consisting of 99 hyponym-hypernym pairs across diverse categories. Our analysis reveal that more abstract concepts (hypernyms) tend to be more perplexing than their specific counterparts (hyponyms). Further exploration into the influence of knowledge hierarchy on editing outcomes indicates that knowledge positioned at higher hierarchical levels is more challenging to modify in some scenarios. Our research highlights a previously overlooked aspect of LLM editing: the variable efficacy of editing methods in handling perplexing knowledge. By revealing how hierarchical relationships can influence editing outcomes, our findings offer new insights into the challenges of updating LLMs and pave the way for more nuanced approaches to model editing in the future.
Authors: Zhenlong Dai, Chang Yao, WenKang Han, Ying Yuan, Zhipeng Gao, Jingyuan Chen
Abstract: Large Language Models (LLMs) have demonstrated great potential for assisting developers in their daily development. However, most research focuses on generating correct code, how to use LLMs to generate personalized code has seldom been investigated. To bridge this gap, we proposed MPCoder (Multi-user Personalized Code Generator) to generate personalized code for multiple users. To better learn coding style features, we utilize explicit coding style residual learning to capture the syntax code style standards and implicit style learning to capture the semantic code style conventions. We train a multi-user style adapter to better differentiate the implicit feature representations of different users through contrastive learning, ultimately enabling personalized code generation for multiple users. We further propose a novel evaluation metric for estimating similarities between codes of different coding styles. The experimental results show the effectiveness of our approach for this novel task.
Authors: Yingting Li, Ambuj Mehrish, Bryan Chew, Bo Cheng, Soujanya Poria
Abstract: Different languages have distinct phonetic systems and vary in their prosodic features making it challenging to develop a Text-to-Speech (TTS) model that can effectively synthesise speech in multilingual settings. Furthermore, TTS architecture needs to be both efficient enough to capture nuances in multiple languages and efficient enough to be practical for deployment. The standard approach is to build transformer based model such as SpeechT5 and train it on large multilingual dataset. As the size of these models grow the conventional fine-tuning for adapting these model becomes impractical due to heavy computational cost. In this paper, we proposes to integrate parameter-efficient transfer learning (PETL) methods such as adapters and hypernetwork with TTS architecture for multilingual speech synthesis. Notably, in our experiments PETL methods able to achieve comparable or even better performance compared to full fine-tuning with only $\sim$2.5\% tunable parameters.The code and samples are available at: https://anonymous.4open.science/r/multilingualTTS-BA4C.
URLs: https://anonymous.4open.science/r/multilingualTTS-BA4C.
Authors: Nafis Sadeq, Zhouhang Xie, Byungkyu Kang, Prarit Lamba, Xiang Gao, Julian McAuley
Abstract: Role-playing has wide-ranging applications in customer support, embodied agents, computational social science, etc. The influence of parametric world knowledge of large language models (LLMs) often causes role-playing characters to act out of character and hallucinate about things outside the scope of their knowledge. In this work, we focus on the evaluation and mitigation of hallucination in fictional character role-play. We introduce a dataset with more than 2,000 characters and 72,000 interviews, including 18,000 adversarial questions. We propose RoleFact, a role-playing method that mitigates hallucination by modulating the influence of parametric knowledge using a pre-calibrated confidence threshold. Experiments show that the proposed method improves the factual precision of generated responses by 18% for adversarial questions with a 44% reduction in temporal hallucination for time-sensitive interviews. The code and the dataset will be available at https://github.com/NafisSadeq/rolefact.git.
Authors: Yiran Luo, Het Patel, Yu Fu, Dawon Ahn, Jia Chen, Yue Dong, Evangelos E. Papalexakis
Abstract: Large language models (LLMs) have fundamentally transformed artificial intelligence, catalyzing recent advancements while imposing substantial environmental and computational burdens. We introduce TRAWL (Tensor Reduced and Approximated Weights for Large Language Models), a novel methodology for optimizing LLMs through tensor decomposition. TRAWL leverages diverse strategies to exploit matrices within transformer-based architectures, realizing notable performance enhancements without necessitating retraining. The most significant improvements were observed through a layer-by-layer intervention strategy, particularly when applied to fully connected weights of the final layers, yielding up to 16% enhancement in accuracy without the need for additional data or fine-tuning. These results underscore the importance of targeted and adaptive techniques in increasing the efficiency and effectiveness of large language model optimization, thereby promoting the development of more sustainable and accessible AI systems.
Authors: Zihan Liao, Hang Yu, Jianguo Li, Jun Wang, Wei Zhang
Abstract: The key challenge in semantic search is to create models that are both accurate and efficient in pinpointing relevant sentences for queries. While BERT-style bi-encoders excel in efficiency with pre-computed embeddings, they often miss subtle nuances in search tasks. Conversely, GPT-style LLMs with cross-encoder designs capture these nuances but are computationally intensive, hindering real-time applications. In this paper, we present D2LLMs-Decomposed and Distilled LLMs for semantic search-that combines the best of both worlds. We decompose a cross-encoder into an efficient bi-encoder integrated with Pooling by Multihead Attention and an Interaction Emulation Module, achieving nuanced understanding and pre-computability. Knowledge from the LLM is distilled into this model using contrastive, rank, and feature imitation techniques. Our experiments show that D2LLM surpasses five leading baselines in terms of all metrics across three tasks, particularly improving NLI task performance by at least 6.45%. The source code is available at https://github.com/codefuse-ai/D2LLM.
Authors: Zhehao Zhang, Jiaao Chen, Diyi Yang
Abstract: The current paradigm of evaluating Large Language Models (LLMs) through static benchmarks comes with significant limitations, such as vulnerability to data contamination and a lack of adaptability to the evolving capabilities of LLMs. Therefore, evaluation methods that can adapt and generate evaluation data with controlled complexity are urgently needed. In this work, we introduce Dynamic Evaluation of LLMs via Adaptive Reasoning Graph Evolvement (DARG) to dynamically extend current benchmarks with controlled complexity and diversity. Specifically, we first extract the reasoning graphs of data points in current benchmarks and then perturb the reasoning graphs to generate novel testing data. Such newly generated test samples can have different levels of complexity while maintaining linguistic diversity similar to the original benchmarks. We further use a code-augmented LLM to ensure the label correctness of newly generated data. We apply our DARG framework to diverse reasoning tasks in four domains with 15 state-of-the-art LLMs. Experimental results show that almost all LLMs experience a performance decrease with increased complexity and certain LLMs exhibit significant drops. Additionally, we find that LLMs exhibit more biases when being evaluated via the data generated by DARG with higher complexity levels. These observations provide useful insights into how to dynamically and adaptively evaluate LLMs. The code is available at https://github.com/SALT-NLP/DARG.
Authors: Jianfeng He, Runing Yang, Linlin Yu, Changbin Li, Ruoxi Jia, Feng Chen, Ming Jin, Chang-Tien Lu
Abstract: Text summarization, a key natural language generation (NLG) task, is vital in various domains. However, the high cost of inaccurate summaries in risk-critical applications, particularly those involving human-in-the-loop decision-making, raises concerns about the reliability of uncertainty estimation on text summarization (UE-TS) evaluation methods. This concern stems from the dependency of uncertainty model metrics on diverse and potentially conflicting NLG metrics. To address this issue, we introduce a comprehensive UE-TS benchmark incorporating 31 NLG metrics across four dimensions. The benchmark evaluates the uncertainty estimation capabilities of two large language models and one pre-trained language model on three datasets, with human-annotation analysis incorporated where applicable. We also assess the performance of 14 common uncertainty estimation methods within this benchmark. Our findings emphasize the importance of considering multiple uncorrelated NLG metrics and diverse uncertainty estimation methods to ensure reliable and efficient evaluation of UE-TS techniques.
Authors: Jikai Wang, Yi Su, Juntao Li, Qinrong Xia, Zi Ye, Xinyu Duan, Zhefeng Wang, Min Zhang
Abstract: Autoregressive language models demonstrate excellent performance in various scenarios. However, the inference efficiency is limited by its one-step-one-word generation mode, which has become a pressing problem recently as the models become increasingly larger. Speculative decoding employs a "draft and then verify" mechanism to allow multiple tokens to be generated in one step, realizing lossless acceleration. Existing methods mainly adopt fixed heuristic draft structures, which fail to adapt to different situations to maximize the acceptance length during verification. To alleviate this dilemma, we proposed OPT-Tree, an algorithm to construct adaptive and scalable draft trees. It searches the optimal tree structure that maximizes the mathematical expectation of the acceptance length in each decoding step. Experimental results reveal that OPT-Tree outperforms the existing draft structures and achieves a speed-up ratio of up to 3.2 compared with autoregressive decoding. If the draft model is powerful enough and the node budget is sufficient, it can generate more than ten tokens in a single step. Our code is available at https://github.com/Jikai0Wang/OPT-Tree.
Authors: Quan Mai, Susan Gauch, Douglas Adams
Abstract: We introduce SetBERT, a fine-tuned BERT-based model designed to enhance query embeddings for set operations and Boolean logic queries, such as Intersection (AND), Difference (NOT), and Union (OR). SetBERT significantly improves retrieval performance for logic-structured queries, an area where both traditional and neural retrieval methods typically underperform. We propose an innovative use of inversed-contrastive loss, focusing on identifying the negative sentence, and fine-tuning BERT with a dataset generated via prompt GPT. Furthermore, we demonstrate that, unlike other BERT-based models, fine-tuning with triplet loss actually degrades performance for this specific task. Our experiments reveal that SetBERT-base not only significantly outperforms BERT-base (up to a 63% improvement in Recall) but also achieves performance comparable to the much larger BERT-large model, despite being only one-third the size.
Authors: Daniel M. Stelzer (University of Illinois at Urbana-Champaign)
Abstract: One of the most significant problems in cuneiform pedagogy is the process of looking up unknown signs, which often involves a tedious page-by-page search through a sign list. This paper proposes a new "recursive encoding" for signs, which represents the arrangement of strokes in a way a computer can process. A series of new algorithms then offers students a new way to look up signs by any distinctive component, as well as providing new ways to render signs and tablets electronically.
Authors: Yang Yan, Lizhi Ma, Anqi Li, Jingsong Ma, Zhenzhong Lan
Abstract: Accurate assessment of personality traits is crucial for effective psycho-counseling, yet traditional methods like self-report questionnaires are time-consuming and biased. This study exams whether Large Language Models (LLMs) can predict the Big Five personality traits directly from counseling dialogues and introduces an innovative framework to perform the task. Our framework applies role-play and questionnaire-based prompting to condition LLMs on counseling sessions, simulating client responses to the Big Five Inventory. We evaluated our framework on 853 real-world counseling sessions, finding a significant correlation between LLM-predicted and actual Big Five traits, proving the validity of framework. Moreover, ablation studies highlight the importance of role-play simulations and task simplification via questionnaires in enhancing prediction accuracy. Meanwhile, our fine-tuned Llama3-8B model, utilizing Direct Preference Optimization with Supervised Fine-Tuning, achieves a 130.95\% improvement, surpassing the state-of-the-art Qwen1.5-110B by 36.94\% in personality prediction validity. In conclusion, LLMs can predict personality based on counseling dialogues. Our code and model are publicly available at \url{https://github.com/kuri-leo/BigFive-LLM-Predictor}, providing a valuable tool for future research in computational psychometrics.
Authors: Wenhao Shi, Zhiqiang Hu, Yi Bin, Junhua Liu, Yang Yang, See-Kiong Ng, Lidong Bing, Roy Ka-Wei Lee
Abstract: Large language models (LLMs) have demonstrated impressive reasoning capabilities, particularly in textual mathematical problem-solving. However, existing open-source image instruction fine-tuning datasets, containing limited question-answer pairs per image, do not fully exploit visual information to enhance the multimodal mathematical reasoning capabilities of Multimodal LLMs (MLLMs). To bridge this gap, we address the lack of high-quality, diverse multimodal mathematical datasets by collecting 40K high-quality images with question-answer pairs from 24 existing datasets and synthesizing 320K new pairs, creating the MathV360K dataset, which enhances both the breadth and depth of multimodal mathematical questions. We introduce Math-LLaVA, a LLaVA-1.5-based model fine-tuned with MathV360K. This novel approach significantly improves the multimodal mathematical reasoning capabilities of LLaVA-1.5, achieving a 19-point increase and comparable performance to GPT-4V on MathVista's minitest split. Furthermore, Math-LLaVA demonstrates enhanced generalizability, showing substantial improvements on the MMMU benchmark. Our research highlights the importance of dataset diversity and synthesis in advancing MLLMs' mathematical reasoning abilities. The code and data are available at: \url{https://github.com/HZQ950419/Math-LLaVA}.
Authors: Tao Feng, Lizhen Qu, Xiaoxi Kang, Gholamreza Haffari
Abstract: Automatically evaluating the quality of responses in open-domain dialogue systems is a challenging but crucial task. Current evaluation metrics often fail to align with human judgments, especially when assessing responses that are grammatically correct. To address this issue, we propose a novel metric, called CausalScore, which assesses the relevance of responses by measuring the causal strength between dialogue histories and responses. The causal strength is estimated by utilizing both unconditional dependence and conditional dependencies from the dialogue history to responses. We compare our metric with the existing competitive metrics in terms of their alignment with human judgements. Our experimental results demonstrate that CausalScore significantly surpasses existing state-of-the-art metrics by aligning better with human judgements. Additionally, we collect a new dialogue dataset CGDIALOG+ with human-annotated causal relations and a set of pairwise human judgements to facilitate the development of future automatic metrics.
Authors: Jinghan Jia, Abi Komma, Timothy Leffel, Xujun Peng, Ajay Nagesh, Tamer Soliman, Aram Galstyan, Anoop Kumar
Abstract: In task-oriented conversational AI evaluation, unsupervised methods poorly correlate with human judgments, and supervised approaches lack generalization. Recent advances in large language models (LLMs) show robust zeroshot and few-shot capabilities across NLP tasks. This paper explores using LLMs for automated dialogue quality evaluation, experimenting with various configurations on public and proprietary datasets. Manipulating factors such as model size, in-context examples, and selection techniques, we examine "chain-of-thought" (CoT) reasoning and label extraction procedures. Our results show that (1) larger models yield more accurate dialogue labels; (2) algorithmic selection of in-context examples outperforms random selection; (3) CoT reasoning where an LLM is asked to provide justifications before outputting final labels improves performance; and (4) fine-tuned LLMs outperform out-of-the-box ones. Our results indicate that LLMs that are suitably fine-tuned and have sufficient reasoning capabilities can be leveraged for automated dialogue evaluation.
Authors: Tingyu Xie, Jian Zhang, Yan Zhang, Yuanyuan Liang, Qi Li, Hongwei Wang
Abstract: The strong capability of large language models (LLMs) has been applied to information extraction (IE) through either retrieval augmented prompting or instruction tuning (IT). However, the best way to incorporate information with LLMs for IE remains an open question. In this paper, we explore Retrieval Augmented Instruction Tuning (RA-IT) for IE, focusing on the task of open named entity recognition (NER). Specifically, for each training sample, we retrieve semantically similar examples from the training dataset as the context and prepend them to the input of the original instruction. To evaluate our RA-IT approach more thoroughly, we construct a Chinese IT dataset for open NER and evaluate RA-IT in both English and Chinese scenarios. Experimental results verify the effectiveness of RA-IT across various data sizes and in both English and Chinese scenarios. We also conduct thorough studies to explore the impacts of various retrieval strategies in the proposed RA-IT framework. Code and data are available at: https://github.com/Emma1066/Retrieval-Augmented-IT-OpenNER
URLs: https://github.com/Emma1066/Retrieval-Augmented-IT-OpenNER
Authors: Sen Yang, Leyang Cui, Deng Cai, Xinting Huang, Shuming Shi, Wai Lam
Abstract: Iterative preference learning, though yielding superior performances, requires online annotated preference labels. In this work, we study strategies to select worth-annotating response pairs for cost-efficient annotation while achieving competitive or even better performances compared with the random selection baseline for iterative preference learning. Built on assumptions regarding uncertainty and distribution shifts, we propose a comparative view to rank the implicit reward margins as predicted by DPO to select the response pairs that yield more benefits. Through extensive experiments, we show that annotating those response pairs with small margins is generally better than large or random, under both single- and multi-iteration scenarios. Besides, our empirical results suggest allocating more annotation budgets in the earlier iterations rather than later across multiple iterations.
Authors: Simone Astarita, Sandor Kruk, Jan Reerink, Pablo G\'omez
Abstract: Rapid progress in the capabilities of machine learning approaches in natural language processing has culminated in the rise of large language models over the last two years. Recent works have shown unprecedented adoption of these for academic writing, especially in some fields, but their pervasiveness in astronomy has not been studied sufficiently. To remedy this, we extract words that ChatGPT uses more often than humans when generating academic text and search a total of 1 million articles for them. This way, we assess the frequency of word occurrence in published works in astronomy tracked by the NASA Astrophysics Data System since 2000. We then perform a statistical analysis of the occurrences. We identify a list of words favoured by ChatGPT and find a statistically significant increase for these words against a control group in 2024, which matches the trend in other disciplines. These results suggest a widespread adoption of these models in the writing of astronomy papers. We encourage organisations, publishers, and researchers to work together to identify ethical and pragmatic guidelines to maximise the benefits of these systems while maintaining scientific rigour.
Authors: Songming Zhang, Xue Zhang, Zengkui Sun, Yufeng Chen, Jinan Xu
Abstract: Knowledge distillation (KD) is known as a promising solution to compress large language models (LLMs) via transferring their knowledge to smaller models. During this process, white-box KD methods usually minimize the distance between the output distributions of the two models so that more knowledge can be transferred. However, in the current white-box KD framework, the output distributions are from the respective output spaces of the two models, using their own prediction heads. We argue that the space discrepancy will lead to low similarity between the teacher model and the student model on both representation and distribution levels. Furthermore, this discrepancy also hinders the KD process between models with different vocabularies, which is common for current LLMs. To address these issues, we propose a dual-space knowledge distillation (DSKD) framework that unifies the output spaces of the two models for KD. On the basis of DSKD, we further develop a cross-model attention mechanism, which can automatically align the representations of the two models with different vocabularies. Thus, our framework is not only compatible with various distance functions for KD (e.g., KL divergence) like the current framework, but also supports KD between any two LLMs regardless of their vocabularies. Experiments on task-agnostic instruction-following benchmarks show that DSKD significantly outperforms the current white-box KD framework with various distance functions, and also surpasses existing KD methods for LLMs with different vocabularies.
Authors: Yasmin Moslem
Abstract: This paper describes our system submission to the International Conference on Spoken Language Translation (IWSLT 2024) for Irish-to-English speech translation. We built end-to-end systems based on Whisper, and employed a number of data augmentation techniques, such as speech back-translation and noise augmentation. We investigate the effect of using synthetic audio data and discuss several methods for enriching signal diversity.
Authors: Jayanta Sadhu, Ayan Antik Khan, Abhik Bhattacharjee, Rifat Shahriyar
Abstract: Pretrained language models inherently exhibit various social biases, prompting a crucial examination of their social impact across various linguistic contexts due to their widespread usage. Previous studies have provided numerous methods for intrinsic bias measurements, predominantly focused on high-resource languages. In this work, we aim to extend these investigations to Bangla, a low-resource language. Specifically, in this study, we (1) create a dataset for intrinsic gender bias measurement in Bangla, (2) discuss necessary adaptations to apply existing bias measurement methods for Bangla, and (3) examine the impact of context length variation on bias measurement, a factor that has been overlooked in previous studies. Through our experiments, we demonstrate a clear dependency of bias metrics on context length, highlighting the need for nuanced considerations in Bangla bias analysis. We consider our work as a stepping stone for bias measurement in the Bangla Language and make all of our resources publicly available to support future research.
Authors: Vaibhav Singh, Amrith Krishna, Karthika NJ, Ganesh Ramakrishnan
Abstract: Low-resource languages, by its very definition, tend to be under represented in the pre-training corpora of Large Language Models. In this work, we investigate three low-resource cross-lingual approaches that enable an LLM adapt to tasks in previously unseen languages. Llama-2 is an LLM where Indic languages, among many other language families, contribute to less than $0.005\%$ of the total $2$ trillion token pre-training corpora. In this work, we experiment with the English-dominated Llama-2 for cross-lingual transfer to three Indic languages, Bengali, Hindi, and Tamil as target languages. We study three approaches for cross-lingual transfer, under ICL and fine-tuning. One, we find that adding additional supervisory signals via a dominant language in the LLM, leads to improvements, both under in-context learning and fine-tuning. Two, adapting the target languages to word reordering may be beneficial under ICL, but its impact diminishes with fine tuning. Finally, continued pre-training in one low-resource language can improve model performance for other related low-resource languages.
Authors: Zhijie Nie, Richong Zhang, Zhanyu Wu
Abstract: Text embeddings from large language models (LLMs) have achieved excellent results in tasks such as information retrieval, semantic textual similarity, etc. In this work, we show an interesting finding: when feeding a text into the embedding LLMs, the obtained text embedding will be able to be aligned with the key tokens in the input text. We first fully analyze this phenomenon on eight embedding LLMs and show that this phenomenon is universal and is not affected by model architecture, training strategy, and embedding method. With a deeper analysis, we then find that the main change in embedding space between the embedding LLMs and their original generative LLMs is in the first principal component. By adjusting the first principal component, we can align text embedding with the key tokens. Finally, we give several examples to demonstrate the vast application potential of this finding: (1) we propose a simple and practical sparse retrieval method based on the aligned tokens, which can achieve 80\% of the dense retrieval effect of the same model while reducing the computation significantly; (2) we show that our findings provide a fresh perspective to help understand fuzzy concepts (e.g., semantic relatedness vs. semantic similarity) and emerging technologies (e.g., instruction-following embedding) in this field.
Authors: Manon Reusens, Philipp Borchert, Jochen De Weerdt, Bart Baesens
Abstract: Large Language Models (LLMs) excel at providing information acquired during pretraining on large-scale corpora and following instructions through user prompts. This study investigates whether the quality of LLM responses varies depending on the demographic profile of users. Considering English as the global lingua franca, along with the diversity of its dialects among speakers of different native languages, we explore whether non-native English speakers receive lower-quality or even factually incorrect responses from LLMs more frequently. Our results show that performance discrepancies occur when LLMs are prompted by native versus non-native English speakers and persist when comparing native speakers from Western countries with others. Additionally, we find a strong anchoring effect when the model recognizes or is made aware of the user's nativeness, which further degrades the response quality when interacting with non-native speakers. Our analysis is based on a newly collected dataset with over 12,000 unique annotations from 124 annotators, including information on their native language and English proficiency.
Authors: Yixuan Wang, Xianzhen Luo, Fuxuan Wei, Yijun Liu, Qingfu Zhu, Xuanyu Zhang, Qing Yang, Dongliang Xu, Wanxiang Che
Abstract: Existing speculative decoding methods typically require additional model structure and training processes to assist the model for draft token generation. This makes the migration of acceleration methods to the new model more costly and more demanding on device memory. To address this problem, we propose the Make Some Noise (MSN) training framework as a replacement for the supervised fine-tuning stage of the large language model. The training method simply introduces some noise at the input for the model to learn the denoising task. It significantly enhances the parallel decoding capability of the model without affecting the original task capability. In addition, we propose a tree-based retrieval-augmented Jacobi (TR-Jacobi) decoding strategy to further improve the inference speed of MSN models. Experiments in both the general and code domains have shown that MSN can improve inference speed by 2.3-2.7x times without compromising model performance. The MSN model also achieves comparable acceleration ratios to the SOTA model with additional model structure on Spec-Bench.
Authors: Razvan-Gabriel Dumitru, Vikas Yadav, Rishabh Maheshwary, Paul-Ioan Clotan, Sathwik Tejaswi Madhusudhan, Mihai Surdeanu
Abstract: We present a simple variable quantization approach that quantizes different layers of a large language model (LLM) at different bit levels. Specifically, we quantize the most important layers to higher bit precision and less important layers to lower bits to achieve floating point quantization levels. We propose two effective strategies to measure the importance of layers within LLMs: the first measures the importance of a layer based on how different its output embeddings are from the input embeddings (the higher the better); the second estimates the importance of a layer using the number of layer weights that are much larger than average (the smaller the better). We show that quantizing different layers at varying bits according to our importance scores results in minimal performance drop with a far more compressed model size. Finally, we present several practical key takeaways from our variable layer-wise quantization experiments: (a) LLM performance under variable quantization remains close to the original model until 25-50% of layers are moved in lower quantization using our proposed ordering but only until 5-10% if moved using no specific ordering; (b) Quantizing LLMs to lower bits performs substantially better than pruning unless extreme quantization (2-bit) is used; and (c) Layer-wise quantization to lower bits works better in the case of larger LLMs with more layers compared to smaller LLMs with fewer layers. The code used to run the experiments is available at: https://github.com/RazvanDu/LayerwiseQuant.
Authors: Minzheng Wang, Longze Chen, Cheng Fu, Shengyi Liao, Xinghua Zhang, Bingli Wu, Haiyang Yu, Nan Xu, Lei Zhang, Run Luo, Yunshui Li, Min Yang, Fei Huang, Yongbin Li
Abstract: Long-context modeling capabilities have garnered widespread attention, leading to the emergence of Large Language Models (LLMs) with ultra-context windows. Meanwhile, benchmarks for evaluating long-context LLMs are gradually catching up. However, existing benchmarks employ irrelevant noise texts to artificially extend the length of test cases, diverging from the real-world scenarios of long-context applications. To bridge this gap, we propose a novel long-context benchmark, Loong, aligning with realistic scenarios through extended multi-document question answering (QA). Unlike typical document QA, in Loong's test cases, each document is relevant to the final answer, ignoring any document will lead to the failure of the answer. Furthermore, Loong introduces four types of tasks with a range of context lengths: Spotlight Locating, Comparison, Clustering, and Chain of Reasoning, to facilitate a more realistic and comprehensive evaluation of long-context understanding. Extensive experiments indicate that existing long-context language models still exhibit considerable potential for enhancement. Retrieval augmented generation (RAG) achieves poor performance, demonstrating that Loong can reliably assess the model's long-context modeling capabilities.
Authors: Hao Yang, Lizhen Qu, Ehsan Shareghi, Gholamreza Haffari
Abstract: Large Multimodal Models (LMMs) have achieved great success recently, demonstrating a strong capability to understand multimodal information and to interact with human users. Despite the progress made, the challenge of detecting high-risk interactions in multimodal settings, and in particular in speech modality, remains largely unexplored. Conventional research on risk for speech modality primarily emphasises the content (e.g., what is captured as transcription). However, in speech-based interactions, paralinguistic cues in audio can significantly alter the intended meaning behind utterances. In this work, we propose a speech-specific risk taxonomy, covering 8 risk categories under hostility (malicious sarcasm and threats), malicious imitation (age, gender, ethnicity), and stereotypical biases (age, gender, ethnicity). Based on the taxonomy, we create a small-scale dataset for evaluating current LMMs capability in detecting these categories of risk. We observe even the latest models remain ineffective to detect various paralinguistic-specific risks in speech (e.g., Gemini 1.5 Pro is performing only slightly above random baseline). Warning: this paper contains biased and offensive examples.
Authors: Davide Mazzaccara, Alberto Testoni, Raffaella Bernardi
Abstract: Questions are essential tools for acquiring the necessary information to complete information-seeking tasks. However, large language models (LLMs), especially open-source models, often perform poorly in generating informative questions, as measured by expected information gain (EIG). In this paper, we propose a method to enhance the informativeness of LLM-generated questions in 20-question game dialogues. We sample multiple questions from the same model (LLAMA 2-CHAT 7B) for each game and create pairs of low-EIG and high-EIG questions to apply a Direct Preference Optimization (DPO) algorithm. Our results show that this method produces more effective questions (in terms of EIG), even in domains different from those used to train the DPO model.
Authors: Yixuan Wang, Baoxin Wang, Yijun Liu, Qingfu Zhu, Dayong Wu, Wanxiang Che
Abstract: Nowadays, data augmentation through synthetic data has been widely used in the field of Grammatical Error Correction (GEC) to alleviate the problem of data scarcity. However, these synthetic data are mainly used in the pre-training phase rather than the data-limited fine-tuning phase due to inconsistent error distribution and noisy labels. In this paper, we propose a synthetic data construction method based on contextual augmentation, which can ensure an efficient augmentation of the original data with a more consistent error distribution. Specifically, we combine rule-based substitution with model-based generation, using the generative model to generate a richer context for the extracted error patterns. Besides, we also propose a relabeling-based data cleaning method to mitigate the effects of noisy labels in synthetic data. Experiments on CoNLL14 and BEA19-Test show that our proposed augmentation method consistently and substantially outperforms strong baselines and achieves the state-of-the-art level with only a few synthetic data.
Authors: Qiancheng Xu, Yongqi Li, Heming Xia, Wenjie Li
Abstract: Tool learning aims to enhance and expand large language models' (LLMs) capabilities with external tools, which has gained significant attention recently. Current methods have shown that LLMs can effectively handle a certain amount of tools through in-context learning or fine-tuning. However, in real-world scenarios, the number of tools is typically extensive and irregularly updated, emphasizing the necessity for a dedicated tool retrieval component. Tool retrieval is nontrivial due to the following challenges: 1) complex user instructions and tool descriptions; 2) misalignment between tool retrieval and tool usage models. To address the above issues, we propose to enhance tool retrieval with iterative feedback from the large language model. Specifically, we prompt the tool usage model, i.e., the LLM, to provide feedback for the tool retriever model in multi-round, which could progressively improve the tool retriever's understanding of instructions and tools and reduce the gap between the two standalone components. We build a unified and comprehensive benchmark to evaluate tool retrieval models. The extensive experiments indicate that our proposed approach achieves advanced performance in both in-domain evaluation and out-of-domain evaluation.
Authors: Micha{\l} Marci\'nczuk
Abstract: This study examines transformer-based models and their effectiveness in named entity recognition tasks. The study investigates data representation strategies, including single, merged, and context, which respectively use one sentence, multiple sentences, and sentences joined with attention to context per vector. Analysis shows that training models with a single strategy may lead to poor performance on different data representations. To address this limitation, the study proposes a combined training procedure that utilizes all three strategies to improve model stability and adaptability. The results of this approach are presented and discussed for four languages (English, Polish, Czech, and German) across various datasets, demonstrating the effectiveness of the combined strategy.
Authors: Yusheng Liao, Shuyang Jiang, Yanfeng Wang, Yu Wang
Abstract: Large language models (LLMs) have shown substantial progress in natural language understanding and generation, proving valuable especially in the medical field. Despite advancements, challenges persist due to the complexity and diversity inherent in medical tasks, which can be categorized as knowledge-intensive tasks and alignment-required tasks. Previous approaches either ignore the latter task or focus on a minority of tasks and hence lose generalization. To address these drawbacks, we propose a progressive fine-tuning pipeline. This pipeline employs a Knowledge Aggregator and a Noise aggregator to encode diverse knowledge in the first stage and filter out detrimental information. In the second stage, we drop the Noise Aggregator to avoid the interference of suboptimal representation and leverage an additional alignment module optimized towards an orthogonal direction to the knowledge space to mitigate knowledge forgetting. Based on this two-stage paradigm, we proposed a Medical LLM through decoupling Clinical Alignment and Knowledge Aggregation (MedCare), which is designed to achieve state-of-the-art (SOTA) performance on over 20 medical tasks, as well as SOTA results on specific medical alignment tasks. Various model sizes of MedCare (1.8B, 7B, 14B) all demonstrate significant improvements over existing models with similar model sizes.
Authors: Matteo Bortoletto, Constantin Ruhdorfer, Lei Shi, Andreas Bulling
Abstract: While numerous works have assessed the generative performance of language models (LMs) on tasks requiring Theory of Mind reasoning, research into the models' internal representation of mental states remains limited. Recent work has used probing to demonstrate that LMs can represent beliefs of themselves and others. However, these claims are accompanied by limited evaluation, making it difficult to assess how mental state representations are affected by model design and training choices. We report an extensive benchmark with various LM types with different model sizes, fine-tuning approaches, and prompt designs to study the robustness of mental state representations and memorisation issues within the probes. Our results show that the quality of models' internal representations of the beliefs of others increases with model size and, more crucially, with fine-tuning. We are the first to study how prompt variations impact probing performance on theory of mind tasks. We demonstrate that models' representations are sensitive to prompt variations, even when such variations should be beneficial. Finally, we complement previous activation editing experiments on Theory of Mind tasks and show that it is possible to improve models' reasoning performance by steering their activations without the need to train any probe.
Authors: Zexuan Qiu, Zijing Ou, Bin Wu, Jingjing Li, Aiwei Liu, Irwin King
Abstract: Augmenting Large Language Models (LLMs) with retrieved external knowledge has proven effective for improving the factual accuracy of generated responses. Despite their success, retrieval-augmented LLMs still face the distractibility issue, where the generated responses are negatively influenced by noise from both external and internal knowledge sources. In this paper, we introduce a novel, training-free decoding method guided by entropy considerations to mitigate this issue. Our approach utilizes entropy-based document-parallel ensemble decoding to prioritize low-entropy distributions from retrieved documents, thereby enhancing the extraction of relevant information of context. Additionally, it incorporates a contrastive decoding mechanism that contrasts the obtained low-entropy ensemble distribution with the high-entropy distribution derived from the model's internal knowledge across layers, which ensures a greater emphasis on reliable external information. Extensive experiments on open-domain question answering datasets demonstrate the superiority of our method.
Authors: Andr\'e V. Duarte, Jo\~ao Marques, Miguel Gra\c{c}a, Miguel Freire, Lei Li, Arlindo L. Oliveira
Abstract: Modern NLP tasks increasingly rely on dense retrieval methods to access up-to-date and relevant contextual information. We are motivated by the premise that retrieval benefits from segments that can vary in size such that a content's semantic independence is better captured. We propose LumberChunker, a method leveraging an LLM to dynamically segment documents, which iteratively prompts the LLM to identify the point within a group of sequential passages where the content begins to shift. To evaluate our method, we introduce GutenQA, a benchmark with 3000 "needle in a haystack" type of question-answer pairs derived from 100 public domain narrative books available on Project Gutenberg. Our experiments show that LumberChunker not only outperforms the most competitive baseline by 7.37% in retrieval performance (DCG@20) but also that, when integrated into a RAG pipeline, LumberChunker proves to be more effective than other chunking methods and competitive baselines, such as the Gemini 1.5M Pro. Our Code and Data are available at https://github.com/joaodsmarques/LumberChunker
Authors: Huiyao Chen, Yu Zhao, Zulong Chen, Mengjia Wang, Liangyue Li, Meishan Zhang, Min Zhang
Abstract: Hierarchical text classification (HTC) is an important task with broad applications, while few-shot HTC has gained increasing interest recently. While in-context learning (ICL) with large language models (LLMs) has achieved significant success in few-shot learning, it is not as effective for HTC because of the expansive hierarchical label sets and extremely-ambiguous labels. In this work, we introduce the first ICL-based framework with LLM for few-shot HTC. We exploit a retrieval database to identify relevant demonstrations, and an iterative policy to manage multi-layer hierarchical labels. Particularly, we equip the retrieval database with HTC label-aware representations for the input texts, which is achieved by continual training on a pretrained language model with masked language modeling (MLM), layer-wise classification (CLS, specifically for HTC), and a novel divergent contrastive learning (DCL, mainly for adjacent semantically-similar labels) objective. Experimental results on three benchmark datasets demonstrate superior performance of our method, and we can achieve state-of-the-art results in few-shot HTC.
Authors: Fabio Mercorio, Mario Mezzanzanica, Daniele Potert\`i, Antonio Serino, Andrea Seveso
Abstract: Recent advancements in Large Language Models (LLMs) have significantly enhanced their ability to generate and manipulate human language, highlighting their potential across various applications. Evaluating LLMs in languages other than English is crucial for ensuring their linguistic versatility, cultural relevance, and applicability in diverse global contexts, thus broadening their usability and effectiveness. We tackle this challenge by introducing a structured benchmark using the INVALSI tests, a set of well-established assessments designed to measure educational competencies across Italy. Our study makes three primary contributions: Firstly, we adapt the INVALSI benchmark for automated LLM evaluation, which involves rigorous adaptation of the test format to suit automated processing while retaining the essence of the original tests. Secondly, we provide a detailed assessment of current LLMs, offering a crucial reference point for the academic community. Finally, we visually compare the performance of these models against human results. Additionally, researchers are invited to submit their models for ongoing evaluation, ensuring the benchmark remains a current and valuable resource.
Authors: Chalamalasetti Kranti, Sherzod Hakimov, David Schlangen
Abstract: In the Minecraft Collaborative Building Task, two players collaborate: an Architect (A) provides instructions to a Builder (B) to assemble a specified structure using 3D blocks. In this work, we investigate the use of large language models (LLMs) to predict the sequence of actions taken by the Builder. Leveraging LLMs' in-context learning abilities, we use few-shot prompting techniques, that significantly improve performance over baseline methods. Additionally, we present a detailed analysis of the gaps in performance for future work
Authors: Guilherme Penedo, Hynek Kydl\'i\v{c}ek, Loubna Ben allal, Anton Lozhkov, Margaret Mitchell, Colin Raffel, Leandro Von Werra, Thomas Wolf
Abstract: The performance of a large language model (LLM) depends heavily on the quality and size of its pretraining dataset. However, the pretraining datasets for state-of-the-art open LLMs like Llama 3 and Mixtral are not publicly available and very little is known about how they were created. In this work, we introduce FineWeb, a 15-trillion token dataset derived from 96 Common Crawl snapshots that produces better-performing LLMs than other open pretraining datasets. To advance the understanding of how best to curate high-quality pretraining datasets, we carefully document and ablate all of the design choices used in FineWeb, including in-depth investigations of deduplication and filtering strategies. In addition, we introduce FineWeb-Edu, a 1.3-trillion token collection of educational text filtered from FineWeb. LLMs pretrained on FineWeb-Edu exhibit dramatically better performance on knowledge- and reasoning-intensive benchmarks like MMLU and ARC. Along with our datasets, we publicly release our data curation codebase and all of the models trained during our ablation experiments.
Authors: Daniel Scalena, Gabriele Sarti, Malvina Nissim
Abstract: Activation steering methods were shown to be effective in conditioning language model generation by additively intervening over models' intermediate representations. However, the evaluation of these techniques has so far been limited to single conditioning properties and synthetic settings. In this work, we conduct a comprehensive evaluation of various activation steering strategies, highlighting the property-dependent nature of optimal parameters to ensure a robust effect throughout generation. To address this issue, we propose Dynamic Activation Composition, an information-theoretic approach to modulate the steering intensity of one or more properties throughout generation. Our experiments on multi-property steering show that our method successfully maintains high conditioning while minimizing the impact of conditioning on generation fluency.
Authors: Caroline Brun, Vassilina Nikoulina
Abstract: Large language models (LLMs) are increasingly popular but are also prone to generating bias, toxic or harmful language, which can have detrimental effects on individuals and communities. Although most efforts is put to assess and mitigate toxicity in generated content, it is primarily concentrated on English, while it's essential to consider other languages as well. For addressing this issue, we create and release FrenchToxicityPrompts, a dataset of 50K naturally occurring French prompts and their continuations, annotated with toxicity scores from a widely used toxicity classifier. We evaluate 14 different models from four prevalent open-sourced families of LLMs against our dataset to assess their potential toxicity across various dimensions. We hope that our contribution will foster future research on toxicity detection and mitigation beyond Englis
Authors: Ryan Pavlich, Nima Ebadi, Richard Tarbell, Billy Linares, Adrian Tan, Rachael Humphreys, Jayanta Kumar Das, Rambod Ghandiparsi, Hannah Haley, Jerris George, Rocky Slavin, Kim-Kwang Raymond Choo, Glenn Dietrich, Anthony Rios
Abstract: Recognizing the promise of natural language interfaces to databases, prior studies have emphasized the development of text-to-SQL systems. While substantial progress has been made in this field, existing research has concentrated on generating SQL statements from text queries. The broader challenge, however, lies in inferring new information about the returned data. Our research makes two major contributions to address this gap. First, we introduce a novel Internet-of-Things (IoT) text-to-SQL dataset comprising 10,985 text-SQL pairs and 239,398 rows of network traffic activity. The dataset contains additional query types limited in prior text-to-SQL datasets, notably temporal-related queries. Our dataset is sourced from a smart building's IoT ecosystem exploring sensor read and network traffic data. Second, our dataset allows two-stage processing, where the returned data (network traffic) from a generated SQL can be categorized as malicious or not. Our results show that joint training to query and infer information about the data can improve overall text-to-SQL performance, nearly matching substantially larger models. We also show that current large language models (e.g., GPT3.5) struggle to infer new information about returned data, thus our dataset provides a novel test bed for integrating complex domain-specific reasoning into LLMs.
Authors: Shawn Gavin, Tuney Zheng, Jiaheng Liu, Quehry Que, Noah Wang, Jian Yang, Chenchen Zhang, Wenhao Huang, Wenhu Chen, Ge Zhang
Abstract: The long-context capabilities of large language models (LLMs) have been a hot topic in recent years. To evaluate the performance of LLMs in different scenarios, various assessment benchmarks have emerged. However, as most of these benchmarks focus on identifying key information to answer questions, which mainly requires the retrieval ability of LLMs, these benchmarks can partially represent the reasoning performance of LLMs from large amounts of information. Meanwhile, although LLMs often claim to have context windows of 32k, 128k, 200k, or even longer, these benchmarks fail to reveal the actual supported length of these LLMs. To address these issues, we propose the LongIns benchmark dataset, a challenging long-context instruction-based exam for LLMs, which is built based on the existing instruction datasets. Specifically, in our LongIns, we introduce three evaluation settings: Global Instruction & Single Task (GIST), Local Instruction & Single Task (LIST), and Local Instruction & Multiple Tasks (LIMT). Based on LongIns, we perform comprehensive evaluations on existing LLMs and have the following important findings: (1). The top-performing GPT-4 with 128k context length performs poorly on the evaluation context window of 16k in our LongIns. (2). For the multi-hop reasoning ability of many existing LLMs, significant efforts are still needed under short context windows (less than 4k).
Authors: Beiduo Chen, Xinpeng Wang, Siyao Peng, Robert Litschko, Anna Korhonen, Barbara Plank
Abstract: Human label variation (HLV) is a valuable source of information that arises when multiple human annotators provide different labels for valid reasons. In Natural Language Inference (NLI) earlier approaches to capturing HLV involve either collecting annotations from many crowd workers to represent human judgment distribution (HJD) or use expert linguists to provide detailed explanations for their chosen labels. While the former method provides denser HJD information, obtaining it is resource-intensive. In contrast, the latter offers richer textual information but it is challenging to scale up to many human judges. Besides, large language models (LLMs) are increasingly used as evaluators (``LLM judges'') but with mixed results, and few works aim to study HJDs. This study proposes to exploit LLMs to approximate HJDs using a small number of expert labels and explanations. Our experiments show that a few explanations significantly improve LLMs' ability to approximate HJDs with and without explicit labels, thereby providing a solution to scale up annotations for HJD. However, fine-tuning smaller soft-label aware models with the LLM-generated model judgment distributions (MJDs) presents partially inconsistent results: while similar in distance, their resulting fine-tuned models and visualized distributions differ substantially. We show the importance of complementing instance-level distance measures with a global-level shape metric and visualization to more effectively evaluate MJDs against human judgment distributions.
Authors: Zhiyuan Wen, Yu Yang, Jiannong Cao, Haoming Sun, Ruosong Yang, Shuaiqi Liu
Abstract: As large language models (LLMs) appear to behave increasingly human-like in text-based interactions, more and more researchers become interested in investigating personality in LLMs. However, the diversity of psychological personality research and the rapid development of LLMs have led to a broad yet fragmented landscape of studies in this interdisciplinary field. Extensive studies across different research focuses, different personality psychometrics, and different LLMs make it challenging to have a holistic overview and further pose difficulties in applying findings to real-world applications. In this paper, we present a comprehensive review by categorizing current studies into three research problems: self-assessment, exhibition, and recognition, based on the intrinsic characteristics and external manifestations of personality in LLMs. For each problem, we provide a thorough analysis and conduct in-depth comparisons of their corresponding solutions. Besides, we summarize research findings and open challenges from current studies and further discuss their underlying causes. We also collect extensive publicly available resources to facilitate interested researchers and developers. Lastly, we discuss the potential future research directions and application scenarios. Our paper is the first comprehensive survey of up-to-date literature on personality in LLMs. By presenting a clear taxonomy, in-depth analysis, promising future directions, and extensive resource collections, we aim to provide a better understanding and facilitate further advancements in this emerging field.
Authors: Erxin Yu, Jing Li, Ming Liao, Siqi Wang, Zuchen Gao, Fei Mi, Lanqing Hong
Abstract: As large language models (LLMs) constantly evolve, ensuring their safety remains a critical research problem. Previous red-teaming approaches for LLM safety have primarily focused on single prompt attacks or goal hijacking. To the best of our knowledge, we are the first to study LLM safety in multi-turn dialogue coreference. We created a dataset of 1,400 questions across 14 categories, each featuring multi-turn coreference safety attacks. We then conducted detailed evaluations on five widely used open-source LLMs. The results indicated that under multi-turn coreference safety attacks, the highest attack success rate was 56% with the LLaMA2-Chat-7b model, while the lowest was 13.9% with the Mistral-7B-Instruct model. These findings highlight the safety vulnerabilities in LLMs during dialogue coreference interactions.
Authors: Nicholas Pangakis, Samuel Wolken
Abstract: Computational social science (CSS) practitioners often rely on human-labeled data to fine-tune supervised text classifiers. We assess the potential for researchers to augment or replace human-generated training data with surrogate training labels from generative large language models (LLMs). We introduce a recommended workflow and test this LLM application by replicating 14 classification tasks and measuring performance. We employ a novel corpus of English-language text classification data sets from recent CSS articles in high-impact journals. Because these data sets are stored in password-protected archives, our analyses are less prone to issues of contamination. For each task, we compare supervised classifiers fine-tuned using GPT-4 labels against classifiers fine-tuned with human annotations and against labels from GPT-4 and Mistral-7B with few-shot in-context learning. Our findings indicate that supervised classification models fine-tuned on LLM-generated labels perform comparably to models fine-tuned with labels from human annotators. Fine-tuning models using LLM-generated labels can be a fast, efficient and cost-effective method of building supervised text classifiers.
Authors: Johnny Li, Saksham Consul, Eda Zhou, James Wong, Naila Farooqui, Yuxin Ye, Nithyashree Manohar, Zhuxiaona Wei, Tian Wu, Ben Echols, Sharon Zhou, Gregory Diamos
Abstract: Despite their powerful chat, coding, and reasoning abilities, Large Language Models (LLMs) frequently hallucinate. Conventional wisdom suggests that hallucinations are a consequence of a balance between creativity and factuality, which can be mitigated, but not eliminated, by grounding the LLM in external knowledge sources. Through extensive systematic experiments, we show that these traditional approaches fail to explain why LLMs hallucinate in practice. Specifically, we show that LLMs augmented with a massive Mixture of Memory Experts (MoME) can easily memorize large datasets of random numbers. We corroborate these experimental findings with a theoretical construction showing that simple neural networks trained to predict the next token hallucinate when the training loss is above a threshold as it usually does in practice when training on internet scale data. We interpret our findings by comparing against traditional retrieval methods for mitigating hallucinations. We use our findings to design a first generation model for removing hallucinations -- Lamini-1 -- that stores facts in a massive mixture of millions of memory experts that are retrieved dynamically.
Authors: Alan Ramponi, Camilla Casula, Stefano Menini
Abstract: Exploring and understanding language data is a fundamental stage in all areas dealing with human language. It allows NLP practitioners to uncover quality concerns and harmful biases in data before training, and helps linguists and social scientists to gain insight into language use and human behavior. Yet, there is currently a lack of a unified, customizable tool to seamlessly inspect and visualize language variation and bias across multiple variables, language units, and diverse metrics that go beyond descriptive statistics. In this paper, we introduce Variationist, a highly-modular, extensible, and task-agnostic tool that fills this gap. Variationist handles at once a potentially unlimited combination of variable types and semantics across diversity and association metrics with regards to the language unit of choice, and orchestrates the creation of up to five-dimensional interactive charts for over 30 variable type-semantics combinations. Through our case studies on computational dialectology, human label variation, and text generation, we show how Variationist enables researchers from different disciplines to effortlessly answer specific research questions or unveil undesired associations in language data. A Python library, code, documentation, and tutorials are made publicly available to the research community.
Authors: Aditya Kalyanpur, Kailash Saravanakumar, Victor Barres, Jennifer Chu-Carroll, David Melville, David Ferrucci
Abstract: We introduce LLM-ARC, a neuro-symbolic framework designed to enhance the logical reasoning capabilities of Large Language Models (LLMs), by combining them with an Automated Reasoning Critic (ARC). LLM-ARC employs an Actor-Critic method where the LLM Actor generates declarative logic programs along with tests for semantic correctness, while the Automated Reasoning Critic evaluates the code, runs the tests and provides feedback on test failures for iterative refinement. Implemented using Answer Set Programming (ASP), LLM-ARC achieves a new state-of-the-art accuracy of 88.32% on the FOLIO benchmark which tests complex logical reasoning capabilities. Our experiments demonstrate significant improvements over LLM-only baselines, highlighting the importance of logic test generation and iterative self-refinement. We achieve our best result using a fully automated self-supervised training loop where the Actor is trained on end-to-end dialog traces with Critic feedback. We discuss potential enhancements and provide a detailed error analysis, showcasing the robustness and efficacy of LLM-ARC for complex natural language reasoning tasks.
Authors: Yuan Li, Yue Huang, Hongyi Wang, Xiangliang Zhang, James Zou, Lichao Sun
Abstract: Large Language Models (LLMs) have demonstrated exceptional task-solving capabilities, increasingly adopting roles akin to human-like assistants. The broader integration of LLMs into society has sparked interest in whether they manifest psychological attributes, and whether these attributes are stable-inquiries that could deepen the understanding of their behaviors. Inspired by psychometrics, this paper presents a framework for investigating psychology in LLMs, including psychological dimension identification, assessment dataset curation, and assessment with results validation. Following this framework, we introduce a comprehensive psychometrics benchmark for LLMs that covers six psychological dimensions: personality, values, emotion, theory of mind, motivation, and intelligence. This benchmark includes thirteen datasets featuring diverse scenarios and item types. Our findings indicate that LLMs manifest a broad spectrum of psychological attributes. We also uncover discrepancies between LLMs' self-reported traits and their behaviors in real-world scenarios. This paper demonstrates a thorough psychometric assessment of LLMs, providing insights into reliable evaluation and potential applications in AI and social sciences.
Authors: Kun Qian, Shunji Wan, Claudia Tang, Youzhi Wang, Xuanming Zhang, Maximillian Chen, Zhou Yu
Abstract: As large language models achieve impressive scores on traditional benchmarks, an increasing number of researchers are becoming concerned about benchmark data leakage during pre-training, commonly known as the data contamination problem. To ensure fair evaluation, recent benchmarks release only the training and validation sets, keeping the test set labels closed-source. They require anyone wishing to evaluate his language model to submit the model's predictions for centralized processing and then publish the model's result on their leaderboard. However, this submission process is inefficient and prevents effective error analysis. To address this issue, we propose to variabilize benchmarks and evaluate language models dynamically. Specifically, we extract variables from each test case and define a value range for each variable. For each evaluation, we sample new values from these value ranges to create unique test cases, thus ensuring a fresh evaluation each time. We applied this variable perturbation method to four datasets: GSM8K, ARC, CommonsenseQA, and TruthfulQA, which cover mathematical generation and multiple-choice tasks. Our experimental results demonstrate that this approach provides a more accurate assessment of the true capabilities of language models, effectively mitigating the contamination problem.
Authors: Thom Lake, Eunsol Choi, Greg Durrett
Abstract: The alignment process changes several properties of a large language model's (LLM's) output distribution. We analyze two aspects of post-alignment distributional shift of LLM responses. First, we re-examine previously reported reductions in response diversity post-alignment. Our analysis suggests that an apparent drop in the diversity of responses is largely explained by quality control and information aggregation. Alignment suppresses irrelevant and unhelpful content while shifting the output distribution toward longer responses that cover information spanning several responses from the base LLM, essentially presenting diverse information in a single response. Finding little evidence that alignment suppresses useful information, it is natural to ask the opposite question: do aligned models surface information that cannot be recovered from base models? Our second investigation shows this is not the case and the behavior of aligned models is recoverable from base models without fine-tuning. A combination of in-context examples and lower-resolution semantic hints about response content can elicit responses from base LLMs that are as similar to alignment-tuned LLM responses as alignment-tuned LLM responses are to each other. Taken together, these results indicate that current alignment techniques capture but do not extend the useful subset of assistant-like base LLM behavior, providing further evidence for the Superficial Alignment Hypothesis. They also show that in-context alignment can go surprisingly far as a strategy for imitating aligned LLMs without fine-tuning. Our code and data is available at https://github.com/thomlake/investigating-alignment.
Authors: Tin Van Huynh, Kiet Van Nguyen, Ngan Luu-Thuy Nguyen
Abstract: The development of Natural Language Processing (NLI) datasets and models has been inspired by innovations in annotation design. With the rapid development of machine learning models today, the performance of existing machine learning models has quickly reached state-of-the-art results on a variety of tasks related to natural language processing, including natural language inference tasks. By using a pre-trained model during the annotation process, it is possible to challenge current NLI models by having humans produce premise-hypothesis combinations that the machine model cannot correctly predict. To remain attractive and challenging in the research of natural language inference for Vietnamese, in this paper, we introduce the adversarial NLI dataset to the NLP research community with the name ViANLI. This data set contains more than 10K premise-hypothesis pairs and is built by a continuously adjusting process to obtain the most out of the patterns generated by the annotators. ViANLI dataset has brought many difficulties to many current SOTA models when the accuracy of the most powerful model on the test set only reached 48.4%. Additionally, the experimental results show that the models trained on our dataset have significantly improved the results on other Vietnamese NLI datasets.
Authors: Elinor Poole-Dayan, Deb Roy, Jad Kabbara
Abstract: While state-of-the-art Large Language Models (LLMs) have shown impressive performance on many tasks, there has been extensive research on undesirable model behavior such as hallucinations and bias. In this work, we investigate how the quality of LLM responses changes in terms of information accuracy, truthfulness, and refusals depending on three user traits: English proficiency, education level, and country of origin. We present extensive experimentation on three state-of-the-art LLMs and two different datasets targeting truthfulness and factuality. Our findings suggest that undesirable behaviors in state-of-the-art LLMs occur disproportionately more for users with lower English proficiency, of lower education status, and originating from outside the US, rendering these models unreliable sources of information towards their most vulnerable users.
Authors: Fei Xia, Yixuan Weng, Shizhu He, Kang Liu, Jun Zhao
Abstract: Taxonomies, which organize domain concepts into hierarchical structures, are crucial for building knowledge systems and downstream applications. As domain knowledge evolves, taxonomies need to be continuously updated to include new concepts. Previous approaches have mainly focused on adding concepts to the leaf nodes of the existing hierarchical tree, which does not fully utilize the taxonomy's knowledge and is unable to update the original taxonomy structure (usually involving non-leaf nodes). In this paper, we propose a two-stage method called ATTEMPT for taxonomy completion. Our method inserts new concepts into the correct position by finding a parent node and labeling child nodes. Specifically, by combining local nodes with prompts to generate natural sentences, we take advantage of pre-trained language models for hypernym/hyponymy recognition. Experimental results on two public datasets (including six domains) show that ATTEMPT performs best on both taxonomy completion and extension tasks, surpassing existing methods.
Authors: Weizhe Yuan, Ilia Kulikov, Ping Yu, Kyunghyun Cho, Sainbayar Sukhbaatar, Jason Weston, Jing Xu
Abstract: Aligned instruction following models can better fulfill user requests than their unaligned counterparts. However, it has been shown that there is a length bias in evaluation of such models, and that training algorithms tend to exploit this bias by learning longer responses. In this work we show how to train models that can be controlled at inference time with instructions containing desired length constraints. Such models are superior in length instructed evaluations, outperforming standard instruction following models such as GPT4, Llama 3 and Mixtral.
Authors: USVSN Sai Prashanth, Alvin Deng, Kyle O'Brien, Jyothir S V, Mohammad Aflah Khan, Jaydeep Borkar, Christopher A. Choquette-Choo, Jacob Ray Fuehne, Stella Biderman, Tracy Ke, Katherine Lee, Naomi Saphra
Abstract: Memorization in language models is typically treated as a homogenous phenomenon, neglecting the specifics of the memorized data. We instead model memorization as the effect of a set of complex factors that describe each sample and relate it to the model and corpus. To build intuition around these factors, we break memorization down into a taxonomy: recitation of highly duplicated sequences, reconstruction of inherently predictable sequences, and recollection of sequences that are neither. We demonstrate the usefulness of our taxonomy by using it to construct a predictive model for memorization. By analyzing dependencies and inspecting the weights of the predictive model, we find that different factors influence the likelihood of memorization differently depending on the taxonomic category.
Authors: Amalie Brogaard Pauli, Isabelle Augenstein, Ira Assent
Abstract: We are exposed to much information trying to influence us, such as teaser messages, debates, politically framed news, and propaganda - all of which use persuasive language. With the recent interest in Large Language Models (LLMs), we study the ability of LLMs to produce persuasive text. As opposed to prior work which focuses on particular domains or types of persuasion, we conduct a general study across various domains to measure and benchmark to what degree LLMs produce persuasive text - both when explicitly instructed to rewrite text to be more or less persuasive and when only instructed to paraphrase. To this end, we construct a new dataset, Persuasive-Pairs, of pairs each consisting of a short text and of a text rewritten by an LLM to amplify or diminish persuasive language. We multi-annotate the pairs on a relative scale for persuasive language. This data is not only a valuable resource in itself, but we also show that it can be used to train a regression model to predict a score of persuasive language between text pairs. This model can score and benchmark new LLMs across domains, thereby facilitating the comparison of different LLMs. Finally, we discuss effects observed for different system prompts. Notably, we find that different 'personas' in the system prompt of LLaMA3 change the persuasive language in the text substantially, even when only instructed to paraphrase. These findings underscore the importance of investigating persuasive language in LLM generated text.
Authors: Zifeng Wang, Lang Cao, Benjamin Danek, Yichi Zhang, Qiao Jin, Zhiyong Lu, Jimeng Sun
Abstract: Automatic medical discovery by AI is a dream of many. One step toward that goal is to create an AI model to understand clinical studies and synthesize clinical evidence from the literature. Clinical evidence synthesis currently relies on systematic reviews of clinical trials and retrospective analyses from medical literature. However, the rapid expansion of publications presents challenges in efficiently identifying, summarizing, and updating evidence. We introduce TrialMind, a generative AI-based pipeline for conducting medical systematic reviews, encompassing study search, screening, and data extraction phases. We utilize large language models (LLMs) to drive each pipeline component while incorporating human expert oversight to minimize errors. To facilitate evaluation, we also create a benchmark dataset TrialReviewBench, a custom dataset with 870 annotated clinical studies from 25 meta-analysis papers across various medical treatments. Our results demonstrate that TrialMind significantly improves the literature review process, achieving high recall rates (0.897-1.000) in study searching from over 20 million PubMed studies and outperforming traditional language model embeddings-based methods in screening (Recall@20 of 0.227-0.246 vs. 0.000-0.102). Furthermore, our approach surpasses direct GPT-4 performance in result extraction, with accuracy ranging from 0.65 to 0.84. We also support clinical evidence synthesis in forest plots, as validated by eight human annotators who preferred TrialMind over the GPT-4 baseline with a winning rate of 62.5%-100% across the involved reviews. Our findings suggest that an LLM-based clinical evidence synthesis approach, such as TrialMind, can enable reliable and high-quality clinical evidence synthesis to improve clinical research efficiency.
Authors: Shane Arora, Marzena Karpinska, Hung-Ting Chen, Ipsita Bhattacharjee, Mohit Iyyer, Eunsol Choi
Abstract: Large language models (LLMs) are commonly used for long-form question answering, which requires them to generate paragraph-length answers to complex questions. While long-form QA has been well-studied in English via many different datasets and evaluation metrics, this research has not been extended to cover most other languages. To bridge this gap, we introduce CaLMQA, a collection of 2.6K complex questions spanning 23 languages, including under-resourced, rarely-studied languages such as Fijian and Kirundi. Our dataset includes both naturally-occurring questions collected from community web forums as well as questions written by native speakers, whom we hire for this purpose. Our process yields diverse, complex questions that reflect cultural topics (e.g. traditions, laws, news) and the language usage of native speakers. We conduct automatic evaluation across a suite of open- and closed-source models using our novel metric CaLMScore, which detects incorrect language and token repetitions in answers, and observe that the quality of LLM-generated answers degrades significantly for some low-resource languages. We perform human evaluation on a subset of models and see that model performance is significantly worse for culturally specific questions than for culturally agnostic questions. Our findings highlight the need for further research in LLM multilingual capabilities and non-English LFQA evaluation.
Authors: Ercong Nie, Bo Shao, Zifeng Ding, Mingyang Wang, Helmut Schmid, Hinrich Sch\"utze
Abstract: Large language models (LLMs) possess extensive parametric knowledge, but this knowledge is difficult to update with new information because retraining is very expensive and infeasible for closed-source models. Knowledge editing (KE) has emerged as a viable solution for updating the knowledge of LLMs without compromising their overall performance. On-the-fly KE methods, inspired by in-context learning (ICL), have shown great promise and allow LLMs to be treated as black boxes. In the past, KE was primarily employed in English contexts, whereas the potential for cross-lingual KE in current English-centric LLMs has not been fully explored. To foster more research in this direction, we introduce the BMIKE-53 benchmark for evaluating cross-lingual KE on 53 diverse languages across three KE task types. We also propose a gradient-free KE method called Multilingual In-context Knowledge Editing (MIKE) and evaluate it on BMIKE-53. Our evaluation focuses on cross-lingual knowledge transfer in terms of reliability, generality, locality, and portability, offering valuable insights and a framework for future research in cross-lingual KE. Our code and data are publicly accessible via the anonymous repository at https://anonymous.4open.science/r/MIKE.
Authors: Changyu Du, Stavros Nousias, Andr\'e Borrmann
Abstract: Facing increasingly complex BIM authoring software and the accompanying expensive learning costs, designers often seek to interact with the software in a more intelligent and lightweight manner. They aim to automate modeling workflows, avoiding obstacles and difficulties caused by software usage, thereby focusing on the design process itself. To address this issue, we proposed an LLM-based autonomous agent framework that can function as a copilot in the BIM authoring tool, answering software usage questions, understanding the user's design intentions from natural language, and autonomously executing modeling tasks by invoking the appropriate tools. In a case study based on the BIM authoring software Vectorworks, we implemented a software prototype to integrate the proposed framework seamlessly into the BIM authoring scenario. We evaluated the planning and reasoning capabilities of different LLMs within this framework when faced with complex instructions. Our work demonstrates the significant potential of LLM-based agents in design automation and intelligent interaction.
Authors: Justine Frija, Juliette Millet, Emilie Bequignon, Ala Covali, Guillaume Cathelain, Josselin Houenou, Helene Benzaquen, Pierre Alexis Geoffroy, Emmanuel Bacry, Mathieu Grajoszex, Marie-Pia d Ortho
Abstract: Obstructive sleep apnea (OSA) is frequent and responsible for cardiovascular complications and excessive daytime sleepiness. It is underdiagnosed due to the difficulty to access the gold standard for diagnosis, polysomnography (PSG). Alternative methods using smartphone sensors could be useful to increase diagnosis. The objective is to assess the performances of Apneal, an application that records the sound using a smartphone's microphone and movements thanks to a smartphone's accelerometer and gyroscope, to estimate patients' AHI. In this article, we perform a monocentric proof-of-concept study with a first manual scoring step, and then an automatic detection of respiratory events from the recorded signals using a sequential deep-learning model which was released internally at Apneal at the end of 2022 (version 0.1 of Apneal automatic scoring of respiratory events), in adult patients during in-hospital polysomnography.46 patients (women 34 per cent, mean BMI 28.7 kg per m2) were included. For AHI superior to 15, sensitivity of manual scoring was 0.91, and positive predictive value (PPV) 0.89. For AHI superior to 30, sensitivity was 0.85, PPV 0.94. We obtained an AUC-ROC of 0.85 and an AUC-PR of 0.94 for the identification of AHI superior to 15, and AUC-ROC of 0.95 and AUC-PR of 0.93 for AHI superior to 30. Promising results are obtained for the automatic annotations of events.This article shows that manual scoring of smartphone-based signals is possible and accurate compared to PSG-based scorings. Automatic scoring method based on a deep learning model provides promising results. A larger multicentric validation study, involving subjects with different SAHS severity is required to confirm these results.
Authors: Shubhabrata Mukherjee, Cory Beard, Sejun Song
Abstract: Semantic Communication can transform the way we transmit information, prioritizing meaningful and effective content over individual symbols or bits. This evolution promises significant benefits, including reduced latency, lower bandwidth usage, and higher throughput compared to traditional communication. However, the development of Semantic Communication faces a crucial challenge: the need for universal metrics to benchmark the joint effects of semantic information loss and energy consumption. This research introduces an innovative solution: the ``Energy-Optimized Semantic Loss'' (EOSL) function, a novel multi-objective loss function that effectively balances semantic information loss and energy consumption. Through comprehensive experiments on transformer models, including energy benchmarking, we demonstrate the remarkable effectiveness of EOSL-based model selection. We have established that EOSL-based transformer model selection achieves up to 83\% better similarity-to-power ratio (SPR) compared to BLEU score-based selection and 67\% better SPR compared to solely lowest power usage-based selection. Furthermore, we extend the applicability of EOSL to diverse and varying contexts, inspired by the principles of Meta-Learning. By cumulatively applying EOSL, we enable the model selection system to adapt to this change, leveraging historical EOSL values to guide the learning process. This work lays the foundation for energy-efficient model selection and the development of green semantic communication.
Authors: Yukun Zhang
Abstract: This study presents a novel approach that leverages Neural Ordinary Differential Equations (Neural ODEs) to unravel the intricate relationships between inputs and outputs in Large Language Models (LLMs), and employs robust control to fine-tune outputs to meet predefined standards. Central to our methodology is the transformation of LLM inputs and outputs into a lower-dimensional latent space, facilitating a detailed examination of the information processing pathways within LLMs. Neural ODEs play a pivotal role in this investigation by providing a dynamic model that captures the continuous evolution of data within the LLMs. Additionally, robust control mechanisms are applied to strategically adjust the model's outputs, ensuring they not only maintain high quality and reliability but also adhere to specific performance criteria. This fusion of Neural ODEs and robust control represents a significant advancement in LLM interpretability, offering a comprehensive framework that elucidates the previously opaque mechanisms of these complex models. Our empirical results validate the effectiveness of this integrated approach, making a substantial contribution to the field of explainable AI by merging advanced machine learning techniques with the critical need for transparency and control in AI outputs.
Authors: Ixandra Achitouv, David Chavalarias, Bruno Gaume
Abstract: The advent of online social networks has led to the development of an abundant literature on the study of online social groups and their relationship to individuals' personalities as revealed by their textual productions. Social structures are inferred from a wide range of social interactions. Those interactions form complex -- sometimes multi-layered -- networks, on which community detection algorithms are applied to extract higher order structures. The choice of the community detection algorithm is however hardily questioned in relation with the cultural production of the individual they classify. In this work, we assume the entangled nature of social networks and their cultural production to propose a definition of cultural based online social groups as sets of individuals whose online production can be categorized as social group-related. We take advantage of this apparently self-referential description of online social groups with a hybrid methodology that combines a community detection algorithm and a natural language processing classification algorithm. A key result of this analysis is the possibility to score community detection algorithms using their agreement with the natural language processing classification. A second result is that we can assign the opinion of a random user at >85% accuracy.
Authors: Ruochen Wang, Si Si, Felix Yu, Dorothea Wiesmann, Cho-Jui Hsieh, Inderjit Dhillon
Abstract: The trade-off between expressiveness and interpretability remains a core challenge when building human-centric predictive models for classification and decision-making. While symbolic rules offer interpretability, they often lack expressiveness, whereas neural networks excel in performance but are known for being black boxes. In this paper, we show a combination of Large Language Models (LLMs) and symbolic programs can bridge this gap. In the proposed LLM-based Symbolic Programs (LSPs), the pretrained LLM with natural language prompts provides a massive set of interpretable modules that can transform raw input into natural language concepts. Symbolic programs then integrate these modules into an interpretable decision rule. To train LSPs, we develop a divide-and-conquer approach to incrementally build the program from scratch, where the learning process of each step is guided by LLMs. To evaluate the effectiveness of LSPs in extracting interpretable and accurate knowledge from data, we introduce IL-Bench, a collection of diverse tasks, including both synthetic and real-world scenarios across different modalities. Empirical results demonstrate LSP's superior performance compared to traditional neurosymbolic programs and vanilla automatic prompt tuning methods. Moreover, as the knowledge learned by LSP is a combination of natural language descriptions and symbolic rules, it is easily transferable to humans (interpretable), and other LLMs, and generalizes well to out-of-distribution samples.
Authors: Yunlong Feng, Yang Xu, Dechuan Teng, Honglin Mu, Xiao Xu, Libo Qin, Wanxiang Che, Qingfu Zhu
Abstract: Decompilation transforms compiled code back into a high-level programming language for analysis when source code is unavailable. Previous work has primarily focused on enhancing decompilation performance by increasing the scale of model parameters or training data for pre-training. Based on the characteristics of the decompilation task, we propose two methods: (1) Without fine-tuning, the Self-Constructed Context Decompilation (sc$^2$dec) method recompiles the LLM's decompilation results to construct pairs for in-context learning, helping the model improve decompilation performance. (2) Fine-grained Alignment Enhancement (FAE), which meticulously aligns assembly code with source code at the statement level by leveraging debugging information, is employed during the fine-tuning phase to achieve further improvements in decompilation. By integrating these two methods, we achieved a Re-Executability performance improvement of approximately 7.35\% on the Decompile-Eval benchmark, establishing a new state-of-the-art performance of 55.03\%.
Authors: Wenyu Du, Shuang Cheng, Tongxu Luo, Zihan Qiu, Zeyu Huang, Ka Chun Cheung, Reynold Cheng, Jie Fu
Abstract: Language models (LMs) exhibit impressive performance and generalization capabilities. However, LMs struggle with the persistent challenge of catastrophic forgetting, which undermines their long-term sustainability in continual learning (CL). Existing approaches usually address the issue by incorporating old task data or task-wise inductive bias into LMs. However, old data and accurate task information are often unavailable or costly to collect, hindering the availability of current CL approaches for LMs. To address this limitation, we introduce $\textbf{MIGU}$ ($\textbf{M}$agn$\textbf{I}$tude-based $\textbf{G}$radient $\textbf{U}$pdating for continual learning), a rehearsal-free and task-label-free method that only updates the model parameters with large magnitudes of output in LMs' linear layers. MIGU is based on our observation that the L1-normalized magnitude distribution of the output in LMs' linear layers is different when the LM models deal with different task data. By imposing this simple constraint on the gradient update process, we can leverage the inherent behaviors of LMs, thereby unlocking their innate CL abilities. Our experiments demonstrate that MIGU is universally applicable to all three LM architectures (T5, RoBERTa, and Llama2), delivering state-of-the-art or on-par performance across continual finetuning and continual pre-training settings on four CL benchmarks. For example, MIGU brings a 15.2% average accuracy improvement over conventional parameter-efficient finetuning baselines in a 15-task CL benchmark. MIGU can also seamlessly integrate with all three existing CL types to further enhance performance. Code is available at \href{https://github.com/wenyudu/MIGU}{this https URL}.
Authors: Rohit Paturi, Xiang Li, Sundararajan Srinivasan
Abstract: Speaker Diarization (SD) systems are typically audio-based and operate independently of the ASR system in traditional speech transcription pipelines and can have speaker errors due to SD and/or ASR reconciliation, especially around speaker turns and regions of speech overlap. To reduce these errors, a Lexical Speaker Error Correction (LSEC), in which an external language model provides lexical information to correct the speaker errors, was recently proposed. Though the approach achieves good Word Diarization error rate (WDER) improvements, it does not use any additional acoustic information and is prone to miscorrections. In this paper, we propose to enhance and acoustically ground the LSEC system with speaker scores directly derived from the existing SD pipeline. This approach achieves significant relative WDER reductions in the range of 25-40% over the audio-based SD, ASR system and beats the LSEC system by 15-25% relative on RT03-CTS, Callhome American English and Fisher datasets.
Authors: Keyu Wang, Guilin Qi, Jiaqi Li, Songlin Zhai
Abstract: Large language models (LLMs) have shown significant achievements in solving a wide range of tasks. Recently, LLMs' capability to store, retrieve and infer with symbolic knowledge has drawn a great deal of attention, showing their potential to understand structured information. However, it is not yet known whether LLMs can understand Description Logic (DL) ontologies. In this work, we empirically analyze the LLMs' capability of understanding DL-Lite ontologies covering 6 representative tasks from syntactic and semantic aspects. With extensive experiments, we demonstrate both the effectiveness and limitations of LLMs in understanding DL-Lite ontologies. We find that LLMs can understand formal syntax and model-theoretic semantics of concepts and roles. However, LLMs struggle with understanding TBox NI transitivity and handling ontologies with large ABoxes. We hope that our experiments and analyses provide more insights into LLMs and inspire to build more faithful knowledge engineering solutions.
Authors: Pranav Ajit Nair, Arun Sai Suggala
Abstract: Large language models (LLMs) have recently demonstrated remarkable performance across diverse language tasks. But their deployment is often constrained by their substantial computational and storage requirements. Quantization has emerged as a key technique for addressing this challenge, enabling the compression of large models with minimal impact on performance. The recent GPTQ algorithm, a post-training quantization (PTQ) method, has proven highly effective for compressing LLMs, sparking a wave of research that leverages GPTQ as a core component. Recognizing the pivotal role of GPTQ in the PTQ landscape, we introduce CDQuant, a simple and scalable alternative to GPTQ with improved performance. CDQuant uses coordinate descent to minimize the layer-wise reconstruction loss to achieve high-quality quantized weights. Our algorithm is easy to implement and scales efficiently to models with hundreds of billions of parameters. Through extensive evaluation on the PaLM2 model family, we demonstrate that CDQuant consistently outperforms GPTQ across diverse model sizes and quantization levels. In particular, for INT2 quantization of PaLM2-Otter, CDQuant achieves a 10% reduction in perplexity compared to GPTQ.
Authors: Yichi Zhang, Zhuo Chen, Lingbing Guo, Yajing Xu, Binbin Hu, Ziqi Liu, Wen Zhang, Huajun Chen
Abstract: Multi-modal knowledge graph completion (MMKGC) aims to automatically discover the unobserved factual knowledge from a given multi-modal knowledge graph by collaboratively modeling the triple structure and multi-modal information from entities. However, real-world MMKGs present challenges due to their diverse and imbalanced nature, which means that the modality information can span various types (e.g., image, text, numeric, audio, video) but its distribution among entities is uneven, leading to missing modalities for certain entities. Existing works usually focus on common modalities like image and text while neglecting the imbalanced distribution phenomenon of modal information. To address these issues, we propose a comprehensive framework NativE to achieve MMKGC in the wild. NativE proposes a relation-guided dual adaptive fusion module that enables adaptive fusion for any modalities and employs a collaborative modality adversarial training framework to augment the imbalanced modality information. We construct a new benchmark called WildKGC with five datasets to evaluate our method. The empirical results compared with 21 recent baselines confirm the superiority of our method, consistently achieving state-of-the-art performance across different datasets and various scenarios while keeping efficient and generalizable. Our code and data are released at https://github.com/zjukg/NATIVE
Authors: Jiawen Huang, Emmanouil Benetos
Abstract: Multilingual automatic lyrics transcription (ALT) is a challenging task due to the limited availability of labelled data and the challenges introduced by singing, compared to multilingual automatic speech recognition. Although some multilingual singing datasets have been released recently, English continues to dominate these collections. Multilingual ALT remains underexplored due to the scale of data and annotation quality. In this paper, we aim to create a multilingual ALT system with available datasets. Inspired by architectures that have been proven effective for English ALT, we adapt these techniques to the multilingual scenario by expanding the target vocabulary set. We then evaluate the performance of the multilingual model in comparison to its monolingual counterparts. Additionally, we explore various conditioning methods to incorporate language information into the model. We apply analysis by language and combine it with the language classification performance. Our findings reveal that the multilingual model performs consistently better than the monolingual models trained on the language subsets. Furthermore, we demonstrate that incorporating language information significantly enhances performance.
Authors: Sedigheh Eslami, Gerard de Melo
Abstract: Contrastive Language--Image Pre-training (CLIP) has manifested remarkable improvements in zero-shot classification and cross-modal vision-language tasks. Yet, from a geometrical point of view, the CLIP embedding space has been found to have a pronounced modality gap. This gap renders the embedding space overly sparse and disconnected, with different modalities being densely distributed in distinct subregions of the hypersphere. In this work, we aim at answering two main questions: 1. Does sharing the parameter space between the multi-modal encoders reduce the modality gap? 2. Can the gap be mitigated by pushing apart the uni-modal embeddings via intra-modality separation? We design AlignCLIP, in order to answer these questions and show that answers to both questions are positive. Through extensive experiments, we show that AlignCLIP achieves noticeable enhancements in the cross-modal alignment of the embeddings, and thereby, reduces the modality gap, while maintaining the performance across several downstream evaluations, such as zero-shot image classification, zero-shot multi-modal retrieval and zero-shot semantic text similarity.
Authors: Jeff Shrager
Abstract: ELIZA, often considered the world's first chatbot, was written by Joseph Weizenbaum in the early 1960s. Weizenbaum did not intend to invent the chatbot, but rather to build a platform for research into human-machine conversation and the important cognitive processes of interpretation and misinterpretation. His purpose was obscured by ELIZA's fame, resulting in large part from the fortuitous timing of it's creation, and it's escape into the wild. In this paper I provide a rich historical context for ELIZA's creation, demonstrating that ELIZA arose from the intersection of some of the central threads in the technical history of AI. I also briefly discuss how ELIZA escaped into the world, and how its accidental escape, along with several coincidental turns of the programming language screws, led both to the misapprehension that ELIZA was intended as a chatbot, and to the loss of the original ELIZA to history for over 50 years.
Authors: Lukas Christ, Shahin Amiriparian, Friederike Hawighorst, Ann-Kathrin Schill, Angelo Boutalikakis, Lorenz Graf-Vlachy, Andreas K\"onig, Bj\"orn W. Schuller
Abstract: Flattery is an important aspect of human communication that facilitates social bonding, shapes perceptions, and influences behavior through strategic compliments and praise, leveraging the power of speech to build rapport effectively. Its automatic detection can thus enhance the naturalness of human-AI interactions. To meet this need, we present a novel audio textual dataset comprising 20 hours of speech and train machine learning models for automatic flattery detection. In particular, we employ pretrained AST, Wav2Vec2, and Whisper models for the speech modality, and Whisper TTS models combined with a RoBERTa text classifier for the textual modality. Subsequently, we build a multimodal classifier by combining text and audio representations. Evaluation on unseen test data demonstrates promising results, with Unweighted Average Recall scores reaching 82.46% in audio-only experiments, 85.97% in text-only experiments, and 87.16% using a multimodal approach.
Authors: Feijie Wu, Zitao Li, Yaliang Li, Bolin Ding, Jing Gao
Abstract: Large language models (LLMs) show amazing performance on many domain-specific tasks after fine-tuning with some appropriate data. However, many domain-specific data are privately distributed across multiple owners. Thus, this dilemma raises the interest in how to perform LLM fine-tuning in federated learning (FL). However, confronted with limited computation and communication capacities, FL clients struggle to fine-tune an LLM effectively. To this end, we introduce FedBiOT, a resource-efficient LLM fine-tuning approach to FL. Specifically, our method involves the server generating a compressed LLM and aligning its performance with the full model. Subsequently, the clients fine-tune a lightweight yet important part of the compressed model, referred to as an adapter. Notice that as the server has no access to the private data owned by the clients, the data used for alignment by the server has a different distribution from the one used for fine-tuning by clients. We formulate the problem into a bi-level optimization problem to minimize the negative effect of data discrepancy and derive the updating rules for the server and clients. We conduct extensive experiments on LLaMA-2, empirically showing that the adapter has exceptional performance when reintegrated into the global LLM. The results also indicate that the proposed FedBiOT significantly reduces resource consumption compared to existing benchmarks, all while achieving comparable performance levels.
Authors: Aida Kostikova, Benjamin Paassen, Dominik Beese, Ole P\"utz, Gregor Wiedemann, Steffen Eger
Abstract: Solidarity is a crucial concept to understand social relations in societies. In this paper, we explore fine-grained solidarity frames to study solidarity towards women and migrants in German parliamentary debates between 1867 and 2022. Using 2,864 manually annotated text snippets (with a cost exceeding 18k Euro), we evaluate large language models (LLMs) like Llama 3, GPT-3.5, and GPT-4. We find that GPT-4 outperforms other LLMs, approaching human annotation quality. Using GPT-4, we automatically annotate more than 18k further instances (with a cost of around 500 Euro) across 155 years and find that solidarity with migrants outweighs anti-solidarity but that frequencies and solidarity types shift over time. Most importantly, group-based notions of (anti-)solidarity fade in favor of compassionate solidarity, focusing on the vulnerability of migrant groups, and exchange-based anti-solidarity, focusing on the lack of (economic) contribution. Our study highlights the interplay of historical events, socio-economic needs, and political ideologies in shaping migration discourse and social cohesion. We also show that powerful LLMs, if carefully prompted, can be cost-effective alternatives to human annotation for hard social scientific tasks.
Authors: Meiru Zhang, Yixuan Su, Zaiqiao Meng, Zihao Fu, Nigel Collier
Abstract: Event extraction is a complex information extraction task that involves extracting events from unstructured text. Prior classification-based methods require comprehensive entity annotations for joint training, while newer generation-based methods rely on heuristic templates containing oracle information such as event type, which is often unavailable in real-world scenarios. In this study, we consider a more realistic setting of this task, namely the Oracle-Free Event Extraction (OFEE) task, where only the input context is given without any oracle information, including event type, event ontology and trigger word. To solve this task, we propose a new framework, called COFFEE, which extracts the events solely based on the document context without referring to any oracle information. In particular, a contrastive selection model is introduced in COFFEE to rectify the generated triggers and handle multi-event instances. The proposed COFFEE outperforms state-of-the-art approaches under the oracle-free setting of the event extraction task, as evaluated on a public event extraction benchmark ACE05.
Authors: Fangyu Lei, Xiang Li, Yifan Wei, Shizhu He, Yiming Huang, Jun Zhao, Kang Liu
Abstract: Answering multi-hop questions over hybrid factual knowledge from the given text and table (TextTableQA) is a challenging task. Existing models mainly adopt a retriever-reader framework, which have several deficiencies, such as noisy labeling in training retriever, insufficient utilization of heterogeneous information over text and table, and deficient ability for different reasoning operations. In this paper, we propose a three-stage TextTableQA framework S3HQA, which comprises of retriever, selector, and reasoner. We use a retriever with refinement training to solve the noisy labeling problem. Then, a hybrid selector considers the linked relationships between heterogeneous data to select the most relevant factual knowledge. For the final stage, instead of adapting a reading comprehension module like in previous methods, we employ a generation-based reasoner to obtain answers. This includes two approaches: a row-wise generator and an LLM prompting generator~(first time used in this task). The experimental results demonstrate that our method achieves competitive results in the few-shot setting. When trained on the full dataset, our approach outperforms all baseline methods, ranking first on the HybridQA leaderboard.
Authors: Abhishek Arora, Melissa Dell
Abstract: Linking information across sources is fundamental to a variety of analyses in social science, business, and government. While large language models (LLMs) offer enormous promise for improving record linkage in noisy datasets, in many domains approximate string matching packages in popular softwares such as R and Stata remain predominant. These packages have clean, simple interfaces and can be easily extended to a diversity of languages. Our open-source package LinkTransformer aims to extend the familiarity and ease-of-use of popular string matching methods to deep learning. It is a general purpose package for record linkage with transformer LLMs that treats record linkage as a text retrieval problem. At its core is an off-the-shelf toolkit for applying transformer models to record linkage with four lines of code. LinkTransformer contains a rich repository of pre-trained transformer semantic similarity models for multiple languages and supports easy integration of any transformer language model from Hugging Face or OpenAI. It supports standard functionality such as blocking and linking on multiple noisy fields. LinkTransformer APIs also perform other common text data processing tasks, e.g., aggregation, noisy de-duplication, and translation-free cross-lingual linkage. Importantly, LinkTransformer also contains comprehensive tools for efficient model tuning, to facilitate different levels of customization when off-the-shelf models do not provide the required accuracy. Finally, to promote reusability, reproducibility, and extensibility, LinkTransformer makes it easy for users to contribute their custom-trained models to its model hub. By combining transformer language models with intuitive APIs that will be familiar to many users of popular string matching packages, LinkTransformer aims to democratize the benefits of LLMs among those who may be less familiar with deep learning frameworks.
Authors: Wenxuan Ding, Shangbin Feng, Yuhan Liu, Zhaoxuan Tan, Vidhisha Balachandran, Tianxing He, Yulia Tsvetkov
Abstract: We propose Knowledge Crosswords, a geometric knowledge reasoning benchmark consisting of incomplete knowledge networks bounded by structured factual constraints, where LLMs are tasked with inferring the missing facts to meet all constraints. The novel setting of geometric knowledge reasoning necessitates new LM abilities beyond existing atomic/linear multi-hop QA, such as backtracking, verifying facts and constraints, reasoning with uncertainty, and more. Knowledge Crosswords contains 2,101 individual problems, covering diverse knowledge domains, and is further divided into three difficulty levels. We conduct extensive experiments to evaluate existing LLMs and approaches on Knowledge Crosswords. Results demonstrate that baseline approaches struggle with larger knowledge networks and semantically-equivalent entity distractors. In light of their limitations, we propose two new approaches, Staged Prompting and Verify-All, to augment LLMs' abilities for error-aware backtracking and constraint verification. Our Verify-All significantly outperforms prior methods and is more robust towards problems in the hard subset. Further analysis shows that geometric knowledge reasoning poses new challenges to LLMs' knowledge abilities, particularly in robustness towards varying option orders, complex structural constraints in knowledge networks, "none of the above" scenarios, and more.
Authors: Schyan Zafar, Geoff K. Nicholls
Abstract: Word meanings change over time, and word senses evolve, emerge or die out in the process. For ancient languages, where the corpora are often small and sparse, modelling such changes accurately proves challenging, and quantifying uncertainty in sense-change estimates consequently becomes important. GASC (Genre-Aware Semantic Change) and DiSC (Diachronic Sense Change) are existing generative models that have been used to analyse sense change for target words from an ancient Greek text corpus, using unsupervised learning without the help of any pre-training. These models represent the senses of a given target word such as "kosmos" (meaning decoration, order or world) as distributions over context words, and sense prevalence as a distribution over senses. The models are fitted using Markov Chain Monte Carlo (MCMC) methods to measure temporal changes in these representations. This paper introduces EDiSC, an Embedded DiSC model, which combines word embeddings with DiSC to provide superior model performance. It is shown empirically that EDiSC offers improved predictive accuracy, ground-truth recovery and uncertainty quantification, as well as better sampling efficiency and scalability properties with MCMC methods. The challenges of fitting these models are also discussed.
Authors: Wenjie Fu, Huandong Wang, Chen Gao, Guanghua Liu, Yong Li, Tao Jiang
Abstract: Membership Inference Attacks (MIA) aim to infer whether a target data record has been utilized for model training or not. Prior attempts have quantified the privacy risks of language models (LMs) via MIAs, but there is still no consensus on whether existing MIA algorithms can cause remarkable privacy leakage on practical Large Language Models (LLMs). Existing MIAs designed for LMs can be classified into two categories: reference-free and reference-based attacks. They are both based on the hypothesis that training records consistently strike a higher probability of being sampled. Nevertheless, this hypothesis heavily relies on the overfitting of target models, which will be mitigated by multiple regularization methods and the generalization of LLMs. The reference-based attack seems to achieve promising effectiveness in LLMs, which measures a more reliable membership signal by comparing the probability discrepancy between the target model and the reference model. However, the performance of reference-based attack is highly dependent on a reference dataset that closely resembles the training dataset, which is usually inaccessible in the practical scenario. Overall, existing MIAs are unable to effectively unveil privacy leakage over practical fine-tuned LLMs that are overfitting-free and private. We propose a Membership Inference Attack based on Self-calibrated Probabilistic Variation (SPV-MIA). Specifically, since memorization in LLMs is inevitable during the training process and occurs before overfitting, we introduce a more reliable membership signal, probabilistic variation, which is based on memorization rather than overfitting. Furthermore, we introduce a self-prompt approach, which constructs the dataset to fine-tune the reference model by prompting the target LLM itself. In this manner, the adversary can collect a dataset with a similar distribution from public APIs.
Authors: Jiaming Ji, Boyuan Chen, Hantao Lou, Donghai Hong, Borong Zhang, Xuehai Pan, Juntao Dai, Tianyi Qiu, Yaodong Yang
Abstract: With the rapid development of large language models (LLMs) and ever-evolving practical requirements, finding an efficient and effective alignment method has never been more critical. However, the tension between the complexity of current alignment methods and the need for rapid iteration in deployment scenarios necessitates the development of a model-agnostic alignment approach that can operate under these constraints. In this paper, we introduce Aligner, a novel and simple alignment paradigm that learns the correctional residuals between preferred and dispreferred answers using a small model. Designed as a model-agnostic, plug-and-play module, Aligner can be directly applied to various open-source and API-based models with only one-off training, making it suitable for rapid iteration. Notably, Aligner can be applied to any powerful, large-scale upstream models. Moreover, it can even iteratively bootstrap the upstream models using corrected responses as synthetic human preference data, breaking through the model's performance ceiling. Our experiments demonstrate performance improvements by deploying the same Aligner model across 11 different LLMs, evaluated on the 3H dimensions (helpfulness, harmlessness, and honesty). Specifically, Aligner-7B has achieved an average improvement of 68.9% in helpfulness and 23.8% in harmlessness across the tested LLMs while also effectively reducing hallucination. In the Alpaca-Eval leaderboard, stacking Aligner-2B on GPT-4 Turbo improved its LC Win Rate from 55.0% to 58.3%, surpassing GPT-4 Omni's 57.5% Win Rate (community report).
Authors: Andr\'e V. Duarte, Xuandong Zhao, Arlindo L. Oliveira, Lei Li
Abstract: How can we detect if copyrighted content was used in the training process of a language model, considering that the training data is typically undisclosed? We are motivated by the premise that a language model is likely to identify verbatim excerpts from its training text. We propose DE-COP, a method to determine whether a piece of copyrighted content was included in training. DE-COP's core approach is to probe an LLM with multiple-choice questions, whose options include both verbatim text and their paraphrases. We construct BookTection, a benchmark with excerpts from 165 books published prior and subsequent to a model's training cutoff, along with their paraphrases. Our experiments show that DE-COP surpasses the prior best method by 9.6% in detection performance (AUC) on models with logits available. Moreover, DE-COP also achieves an average accuracy of 72% for detecting suspect books on fully black-box models where prior methods give approximately 4% accuracy. The code and datasets are available at https://github.com/LeiLiLab/DE-COP.
Authors: Alex Havrilla, Sharath Raparthy, Christoforus Nalmpantis, Jane Dwivedi-Yu, Maksym Zhuravinskyi, Eric Hambro, Roberta Raileanu
Abstract: State-of-the-art language models can exhibit impressive reasoning refinement capabilities on math, science or coding tasks. However, recent work demonstrates that even the best models struggle to identify \textit{when and where to refine} without access to external feedback. Outcome-based Reward Models (\textbf{ORMs}), trained to predict correctness of the final answer indicating when to refine, offer one convenient solution for deciding when to refine. Process Based Reward Models (\textbf{PRMs}), trained to predict correctness of intermediate steps, can then be used to indicate where to refine. But they are expensive to train, requiring extensive human annotations. In this paper, we propose Stepwise ORMs (\textbf{SORMs}) which are trained, only on synthetic data, to approximate the expected future reward of the optimal policy or $V^{\star}$. More specifically, SORMs are trained to predict the correctness of the final answer when sampling the current policy many times (rather than only once as in the case of ORMs). Our experiments show that SORMs can more accurately detect incorrect reasoning steps compared to ORMs, thus improving downstream accuracy when doing refinements. We then train \textit{global} refinement models, which take only the question and a draft solution as input and predict a corrected solution, and \textit{local} refinement models which also take as input a critique indicating the location of the first reasoning error. We generate training data for both models synthetically by reusing data used to train the SORM. We find combining global and local refinements, using the ORM as a reranker, significantly outperforms either one individually, as well as a best of three sample baseline. With this strategy we can improve the accuracy of a LLaMA-2 13B model (already fine-tuned with RL) on GSM8K from 53\% to 65\% when greedily sampled.
Authors: Shirley Anugrah Hayati, Taehee Jung, Tristan Bodding-Long, Sudipta Kar, Abhinav Sethy, Joo-Kyung Kim, Dongyeop Kang
Abstract: Fine-tuning large language models (LLMs) with a collection of large and diverse instructions has improved the model's generalization to different tasks, even for unseen tasks. However, most existing instruction datasets include only single instructions, and they struggle to follow complex instructions composed of multiple subtasks. In this work, we propose a novel concept of compositional instructions called chain-of-instructions (CoI), where the output of one instruction becomes an input for the next like a chain. Unlike the conventional practice of solving single instruction tasks, our proposed method encourages a model to solve each subtask step by step until the final answer is reached. CoI-tuning (i.e., fine-tuning with CoI instructions) improves the model's ability to handle instructions composed of multiple subtasks as well as unseen composite tasks such as multilingual summarization. Overall, our study find that simple CoI tuning of existing instruction data can provide consistent generalization to solve more complex, unseen, and longer chains of instructions.
Authors: Ge Bai, Jie Liu, Xingyuan Bu, Yancheng He, Jiaheng Liu, Zhanhui Zhou, Zhuoran Lin, Wenbo Su, Tiezheng Ge, Bo Zheng, Wanli Ouyang
Abstract: The advent of Large Language Models (LLMs) has drastically enhanced dialogue systems. However, comprehensively evaluating the dialogue abilities of LLMs remains a challenge. Previous benchmarks have primarily focused on single-turn dialogues or provided coarse-grained and incomplete assessments of multi-turn dialogues, overlooking the complexity and fine-grained nuances of real-life dialogues. To address this issue, we introduce MT-Bench-101, specifically designed to evaluate the fine-grained abilities of LLMs in multi-turn dialogues. By conducting a detailed analysis of real multi-turn dialogue data, we construct a three-tier hierarchical ability taxonomy comprising 4208 turns across 1388 multi-turn dialogues in 13 distinct tasks. We then evaluate 21 popular LLMs based on MT-Bench-101, conducting comprehensive analyses from both ability and task perspectives and observing differing trends in LLMs performance across dialogue turns within various tasks. Further analysis indicates that neither utilizing common alignment techniques nor chat-specific designs has led to obvious enhancements in the multi-turn abilities of LLMs. Extensive case studies suggest that our designed tasks accurately assess the corresponding multi-turn abilities. The data and code are available at \url{https://github.com/mtbench101/mt-bench-101}.
Authors: Stefan Hegselmann, Shannon Zejiang Shen, Florian Gierse, Monica Agrawal, David Sontag, Xiaoyi Jiang
Abstract: Patients often face difficulties in understanding their hospitalizations, while healthcare workers have limited resources to provide explanations. In this work, we investigate the potential of large language models to generate patient summaries based on doctors' notes and study the effect of training data on the faithfulness and quality of the generated summaries. To this end, we release (i) a rigorous labeling protocol for errors in medical texts and (ii) a publicly available dataset of annotated hallucinations in 100 doctor-written and 100 generated summaries. We show that fine-tuning on hallucination-free data effectively reduces hallucinations from 2.60 to 1.55 per summary for Llama 2, while preserving relevant information. We observe a similar effect on GPT-4 (0.70 to 0.40), when the few-shot examples are hallucination-free. We also conduct a qualitative evaluation using hallucination-free and improved training data. We find that common quantitative metrics do not correlate well with faithfulness and quality. Finally, we test GPT-4 for automatic hallucination detection, which clearly outperforms common baselines.
Authors: Guangming Huang, Yunfei Long, Cunjin Luo, Jiaxing Shen, Xia Sun
Abstract: Pre-trained language models (PLMs) leverage chains-of-thought (CoT) to simulate human reasoning and inference processes, achieving proficient performance in multi-hop QA. However, a gap persists between PLMs' reasoning abilities and those of humans when tackling complex problems. Psychological studies suggest a vital connection between explicit information in passages and human prior knowledge during reading. Nevertheless, current research has given insufficient attention to linking input passages and PLMs' pre-training-based knowledge from the perspective of human cognition studies. In this study, we introduce a Prompting Explicit and Implicit knowledge (PEI) framework, which uses prompts to connect explicit and implicit knowledge, aligning with human reading process for multi-hop QA. We consider the input passages as explicit knowledge, employing them to elicit implicit knowledge through unified prompt reasoning. Furthermore, our model incorporates type-specific reasoning via prompts, a form of implicit knowledge. Experimental results show that PEI performs comparably to the state-of-the-art on HotpotQA. Ablation studies confirm the efficacy of our model in bridging and integrating explicit and implicit knowledge.
Authors: Nicola Piovesan, Antonio De Domenico, Fadhel Ayed
Abstract: The increasing interest in Large Language Models (LLMs) within the telecommunications sector underscores their potential to revolutionize operational efficiency. However, the deployment of these sophisticated models is often hampered by their substantial size and computational demands, raising concerns about their viability in resource-constrained environments. Addressing this challenge, recent advancements have seen the emergence of small language models that surprisingly exhibit performance comparable to their larger counterparts in many tasks, such as coding and common-sense reasoning. Phi-2, a compact yet powerful model, exemplifies this new wave of efficient small language models. This paper conducts a comprehensive evaluation of Phi-2's intrinsic understanding of the telecommunications domain. Recognizing the scale-related limitations, we enhance Phi-2's capabilities through a Retrieval-Augmented Generation approach, meticulously integrating an extensive knowledge base specifically curated with telecom standard specifications. The enhanced Phi-2 model demonstrates a profound improvement in accuracy, answering questions about telecom standards with a precision that closely rivals the more resource-intensive GPT-3.5. The paper further explores the refined capabilities of Phi-2 in addressing problem-solving scenarios within the telecom sector, highlighting its potential and limitations.
Authors: Samuel Cahyawijaya, Holy Lovenia, Pascale Fung
Abstract: In-context learning (ICL) empowers large language models (LLMs) to perform diverse tasks in underrepresented languages using only short in-context information, offering a crucial avenue for narrowing the gap between high-resource and low-resource languages. Nonetheless, there is only a handful of works explored ICL for low-resource languages with most of them focusing on relatively high-resource languages, such as French and Spanish. In this work, we extensively study ICL and its cross-lingual variation (X-ICL) on 25 low-resource and 7 relatively higher-resource languages. Our study not only assesses the effectiveness of ICL with LLMs in low-resource languages but also identifies the shortcomings of in-context label alignment, and introduces a more effective alternative: query alignment. Moreover, we provide valuable insights into various facets of ICL for low-resource languages. Our study concludes the significance of few-shot in-context information on enhancing the low-resource understanding quality of LLMs through semantically relevant information by closing the language gap in the target language and aligning the semantics between the targeted low-resource and the high-resource language that the model is proficient in. Our work highlights the importance of advancing ICL research, particularly for low-resource languages. Our code is publicly released at https://github.com/SamuelCahyawijaya/in-context-alignment
URLs: https://github.com/SamuelCahyawijaya/in-context-alignment
Authors: Samuel Cahyawijaya, Delong Chen, Yejin Bang, Leila Khalatbari, Bryan Wilie, Ziwei Ji, Etsuko Ishii, Pascale Fung
Abstract: The widespread application of Large Language Models (LLMs) across various tasks and fields has necessitated the alignment of these models with human values and preferences. Given various approaches of human value alignment, ranging from Reinforcement Learning with Human Feedback (RLHF), to constitutional learning, etc. there is an urgent need to understand the scope and nature of human values injected into these models before their release. There is also a need for model alignment without a costly large scale human annotation effort. We propose UniVaR, a high-dimensional representation of human value distributions in LLMs, orthogonal to model architecture and training data. Trained from the value-relevant output of eight multilingual LLMs and tested on the output from four multilingual LLMs, namely LlaMA2, ChatGPT, JAIS and Yi, we show that UniVaR is a powerful tool to compare the distribution of human values embedded in different LLMs with different langauge sources. Through UniVaR, we explore how different LLMs prioritize various values in different languages and cultures, shedding light on the complex interplay between human values and language modeling.
Authors: Edward Y. Chang
Abstract: This paper explores the integration of human-like emotions and ethical considerations into Large Language Models (LLMs). We first model eight fundamental human emotions, presented as opposing pairs, and employ collaborative LLMs to reinterpret and express these emotions across a spectrum of intensity. Our focus extends to embedding a latent ethical dimension within LLMs, guided by a novel self-supervised learning algorithm with human feedback (SSHF). This approach enables LLMs to perform self-evaluations and adjustments concerning ethical guidelines, enhancing their capability to generate content that is not only emotionally resonant but also ethically aligned. The methodologies and case studies presented herein illustrate the potential of LLMs to transcend mere text and image generation, venturing into the realms of empathetic interaction and principled decision-making, thereby setting a new precedent in the development of emotionally aware and ethically conscious AI systems.
Authors: Ben Hutchinson
Abstract: This position paper concerns the use of religious texts in Natural Language Processing (NLP), which is of special interest to the Ethics of NLP. Religious texts are expressions of culturally important values, and machine learned models have a propensity to reproduce cultural values encoded in their training data. Furthermore, translations of religious texts are frequently used by NLP researchers when language data is scarce. This repurposes the translations from their original uses and motivations, which often involve attracting new followers. This paper argues that NLP's use of such texts raises considerations that go beyond model biases, including data provenance, cultural contexts, and their use in proselytism. We argue for more consideration of researcher positionality, and of the perspectives of marginalized linguistic and religious communities.
Authors: Andrew Liu, Hongjian Zhou, Yining Hua, Omid Rohanian, Anshul Thakur, Lei Clifton, David A. Clifton
Abstract: The adoption of large language models (LLMs) to assist clinicians has attracted remarkable attention. Existing works mainly adopt the close-ended question-answering (QA) task with answer options for evaluation. However, many clinical decisions involve answering open-ended questions without pre-set options. To better understand LLMs in the clinic, we construct a benchmark ClinicBench. We first collect eleven existing datasets covering diverse clinical language generation, understanding, and reasoning tasks. Furthermore, we construct six novel datasets and complex clinical tasks that are close to real-world practice, i.e., referral QA, treatment recommendation, hospitalization (long document) summarization, patient education, pharmacology QA and drug interaction for emerging drugs. We conduct an extensive evaluation of twenty-two LLMs under both zero-shot and few-shot settings. Finally, we invite medical experts to evaluate the clinical usefulness of LLMs.
Authors: Taolin Zhang, Dongyang Li, Qizhou Chen, Chengyu Wang, Longtao Huang, Hui Xue, Xiaofeng He, Jun Huang
Abstract: Retrieval-augmented large language models (LLMs) leverage relevant content retrieved by information retrieval systems to generate correct responses, aiming to alleviate the hallucination problem. However, existing retriever-responder methods typically append relevant documents to the prompt of LLMs to perform text generation tasks without considering the interaction of fine-grained structural semantics between the retrieved documents and the LLMs. This issue is particularly important for accurate response generation as LLMs tend to "lose in the middle" when dealing with input prompts augmented with lengthy documents. In this work, we propose a new pipeline named "Reinforced Retriever-Reorder-Responder" (R$^4$) to learn document orderings for retrieval-augmented LLMs, thereby further enhancing their generation abilities while the large numbers of parameters of LLMs remain frozen. The reordering learning process is divided into two steps according to the quality of the generated responses: document order adjustment and document representation enhancement. Specifically, document order adjustment aims to organize retrieved document orderings into beginning, middle, and end positions based on graph attention learning, which maximizes the reinforced reward of response quality. Document representation enhancement further refines the representations of retrieved documents for responses of poor quality via document-level gradient adversarial learning. Extensive experiments demonstrate that our proposed pipeline achieves better factual question-answering performance on knowledge-intensive tasks compared to strong baselines across various public datasets. The source codes and trained models will be released upon paper acceptance.
Authors: Shan Chen, Jack Gallifant, Mingye Gao, Pedro Moreira, Nikolaj Munch, Ajay Muthukkumar, Arvind Rajan, Jaya Kolluri, Amelia Fiske, Janna Hastings, Hugo Aerts, Brian Anthony, Leo Anthony Celi, William G. La Cava, Danielle S. Bitterman
Abstract: Large language models (LLMs) are increasingly essential in processing natural languages, yet their application is frequently compromised by biases and inaccuracies originating in their training data. In this study, we introduce Cross-Care, the first benchmark framework dedicated to assessing biases and real world knowledge in LLMs, specifically focusing on the representation of disease prevalence across diverse demographic groups. We systematically evaluate how demographic biases embedded in pre-training corpora like $ThePile$ influence the outputs of LLMs. We expose and quantify discrepancies by juxtaposing these biases against actual disease prevalences in various U.S. demographic groups. Our results highlight substantial misalignment between LLM representation of disease prevalence and real disease prevalence rates across demographic subgroups, indicating a pronounced risk of bias propagation and a lack of real-world grounding for medical applications of LLMs. Furthermore, we observe that various alignment methods minimally resolve inconsistencies in the models' representation of disease prevalence across different languages. For further exploration and analysis, we make all data and a data visualization tool available at: www.crosscare.net.
Authors: Yuelyu Ji, Zhuochun Li, Rui Meng, Sonish Sivarajkumar, Yanshan Wang, Zeshui Yu, Hui Ji, Yushui Han, Hanyu Zeng, Daqing He
Abstract: This paper introduces the RAG-RLRC-LaySum framework, designed to make complex biomedical research understandable to laymen through advanced Natural Language Processing (NLP) techniques. Our Retrieval Augmented Generation (RAG) solution, enhanced by a reranking method, utilizes multiple knowledge sources to ensure the precision and pertinence of lay summaries. Additionally, our Reinforcement Learning for Readability Control (RLRC) strategy improves readability, making scientific content comprehensible to non-specialists. Evaluations using the publicly accessible PLOS and eLife datasets show that our methods surpass Plain Gemini model, demonstrating a 20% increase in readability scores, a 15% improvement in ROUGE-2 relevance scores, and a 10% enhancement in factual accuracy. The RAG-RLRC-LaySum framework effectively democratizes scientific knowledge, enhancing public engagement with biomedical discoveries.
Authors: Ethan Shen, Alan Fan, Sarah M. Pratt, Jae Sung Park, Matthew Wallingford, Sham M. Kakade, Ari Holtzman, Ranjay Krishna, Ali Farhadi, Aditya Kusupati
Abstract: Many applications today provide users with multiple auto-complete drafts as they type, including GitHub's code completion, Gmail's smart compose, and Apple's messaging auto-suggestions. Under the hood, language models support this by running an autoregressive inference pass to provide a draft. Consequently, providing $k$ drafts to the user requires running an expensive language model $k$ times. To alleviate the computation cost of running $k$ inference passes, we propose Superposed Decoding, a new decoding algorithm that generates $k$ drafts at the computation cost of one autoregressive inference pass. We achieve this by feeding a superposition of the most recent token embeddings from the $k$ drafts as input to the next decoding step of the language model. At every inference step we combine the $k$ drafts with the top-$k$ tokens to get $k^2$ new drafts and cache the $k$ most likely options, using an n-gram interpolation with minimal compute overhead to filter out incoherent generations. Our experiments show that $k$ drafts from Superposed Decoding are at least as coherent and factual as Nucleus Sampling and Greedy Decoding respectively, while being at least $2.44\times$ faster for $k\ge3$. In a compute-normalized setting, user evaluations demonstrably favor text generated by Superposed Decoding over Nucleus Sampling. Code and more examples open-sourced at https://github.com/RAIVNLab/SuperposedDecoding.
Authors: Ravil Mussabayev
Abstract: The problem of natural language generation, and, more specifically, method name prediction, faces significant difficulties when proposed models need to be evaluated on test data. Such a metric would need to consider the versatility with which a single method can be named, with respect to both semantics and syntax. Measuring the direct overlap between the predicted and reference (true) sequences will not be able to capture these subtleties. Other existing embedding based metrics either do not measure precision and recall or impose strict unrealistic assumptions on both sequences. To address these issues, we propose a new metric that, on the one hand, is very simple and lightweight, and, on the other hand, is able to calculate precision and recall without resorting to any assumptions while obtaining good performance with respect to the human judgement.
Authors: Tomoki Doi, Masaru Isonuma, Hitomi Yanaka
Abstract: Recent work utilizes Large Language Models (LLMs) for topic modeling, generating comprehensible topic labels for given documents. However, their performance has mainly been evaluated qualitatively, and there remains room for quantitative investigation of their capabilities. In this paper, we quantitatively evaluate LLMs from multiple perspectives: the quality of topics, the impact of LLM-specific concerns, such as hallucination and shortcuts for limited documents, and LLMs' controllability of topic categories via prompts. Our findings show that LLMs can identify coherent and diverse topics with few hallucinations but may take shortcuts by focusing only on parts of documents. We also found that their controllability is limited.
Authors: Xinyue Cui, Swabha Swayamdipta
Abstract: Despite the remarkable generative capabilities of language models in producing naturalistic language, their effectiveness on explicit manipulation and generation of linguistic structures remain understudied. In this paper, we investigate the task of generating new sentences preserving a given semantic structure, following the FrameNet formalism. We propose a framework to produce novel frame-semantically annotated sentences following an overgenerate-and-filter approach. Our results show that conditioning on rich, explicit semantic information tends to produce generations with high human acceptance, under both prompting and finetuning. Our generated frame-semantic structured annotations are effective at training data augmentation for frame-semantic role labeling in low-resource settings; however, we do not see benefits under higher resource settings. Our study concludes that while generating high-quality, semantically rich data might be within reach, the downstream utility of such generations remains to be seen, highlighting the outstanding challenges with automating linguistic annotation tasks.
Authors: Rajiv Movva, Pang Wei Koh, Emma Pierson
Abstract: To what extent do LLMs align with human perceptions of safety? We study this question via *annotation alignment*, the extent to which LLMs and humans agree when annotating the safety of user-chatbot conversations. We leverage the recent DICES dataset (Aroyo et al., 2023), in which 350 conversations are each rated for safety by 112 annotators spanning 10 race-gender groups. GPT-4 achieves a Pearson correlation of $r = 0.59$ with the average annotator rating, higher than the median annotator's correlation with the average ($r=0.51$). We show that larger datasets are needed to resolve whether GPT-4 exhibits disparities in how well it correlates with demographic groups. Also, there is substantial idiosyncratic variation in correlation *within* groups, suggesting that race & gender do not fully capture differences in alignment. Finally, we find that GPT-4 cannot predict when one demographic group finds a conversation more unsafe than another.
Authors: Viet Anh Trinh, Rosy Southwell, Yiwen Guan, Xinlu He, Zhiyong Wang, Jacob Whitehill
Abstract: Recent work on discrete speech tokenization has paved the way for models that can seamlessly perform multiple tasks across modalities, e.g., speech recognition, text to speech, speech to speech translation. Moreover, large language models (LLMs) pretrained from vast text corpora contain rich linguistic information that can improve accuracy in a variety of tasks. In this paper, we present a decoder-only Discrete Multimodal Language Model (DMLM), which can be flexibly applied to multiple tasks (ASR, T2S, S2TT, etc.) and modalities (text, speech, vision). We explore several critical aspects of discrete multi-modal models, including the loss function, weight initialization, mixed training supervision, and codebook. Our results show that DMLM benefits significantly, across multiple tasks and datasets, from a combination of supervised and unsupervised training. Moreover, for ASR, it benefits from initializing DMLM from a pretrained LLM, and from a codebook derived from Whisper activations.
Authors: You Zuo (ALMAnaCH), Kim Gerdes (LISN), Eric Villemonte de La Clergerie (ALMAnaCH), Beno\^it Sagot (ALMAnaCH)
Abstract: In this work, we introduce a comprehensive error typology specifically designed for evaluating two distinct tasks in machine-generated patent texts: claims-to-abstract generation, and the generation of the next claim given previous ones. We have also developed a benchmark, PatentEval, for systematically assessing language models in this context. Our study includes a comparative analysis, annotated by humans, of various models. These range from those specifically adapted during training for tasks within the patent domain to the latest general-purpose large language models (LLMs). Furthermore, we explored and evaluated some metrics to approximate human judgments in patent text evaluation, analyzing the extent to which these metrics align with expert assessments. These approaches provide valuable insights into the capabilities and limitations of current language models in the specialized field of patent text generation.
Authors: Cheng Niu, Yang Guan, Yuanhao Wu, Juno Zhu, Juntong Song, Randy Zhong, Kaihua Zhu, Siliang Xu, Shizhe Diao, Tong Zhang
Abstract: The proliferation of fake news poses a significant threat not only by disseminating misleading information but also by undermining the very foundations of democracy. The recent advance of generative artificial intelligence has further exacerbated the challenge of distinguishing genuine news from fabricated stories. In response to this challenge, we introduce VeraCT Scan, a novel retrieval-augmented system for fake news detection. This system operates by extracting the core facts from a given piece of news and subsequently conducting an internet-wide search to identify corroborating or conflicting reports. Then sources' credibility is leveraged for information verification. Besides determining the veracity of news, we also provide transparent evidence and reasoning to support its conclusions, resulting in the interpretability and trust in the results. In addition to GPT-4 Turbo, Llama-2 13B is also fine-tuned for news content understanding, information verification, and reasoning. Both implementations have demonstrated state-of-the-art accuracy in the realm of fake news detection.
Authors: Nikhil Khandekar, Qiao Jin, Guangzhi Xiong, Soren Dunn, Serina S Applebaum, Zain Anwar, Maame Sarfo-Gyamfi, Conrad W Safranek, Abid A Anwar, Andrew Zhang, Aidan Gilson, Maxwell B Singer, Amisha Dave, Andrew Taylor, Aidong Zhang, Qingyu Chen, Zhiyong Lu
Abstract: As opposed to evaluating computation and logic-based reasoning, current benchmarks for evaluating large language models (LLMs) in medicine are primarily focused on question-answering involving domain knowledge and descriptive reasoning. While such qualitative capabilities are vital to medical diagnosis, in real-world scenarios, doctors frequently use clinical calculators that follow quantitative equations and rule-based reasoning paradigms for evidence-based decision support. To this end, we propose MedCalc-Bench, a first-of-its-kind dataset focused on evaluating the medical calculation capability of LLMs. MedCalc-Bench contains an evaluation set of over 1000 manually reviewed instances from 55 different medical calculation tasks. Each instance in MedCalc-Bench consists of a patient note, a question requesting to compute a specific medical value, a ground truth answer, and a step-by-step explanation showing how the answer is obtained. While our evaluation results show the potential of LLMs in this area, none of them are effective enough for clinical settings. Common issues include extracting the incorrect entities, not using the correct equation or rules for a calculation task, or incorrectly performing the arithmetic for the computation. We hope our study highlights the quantitative knowledge and reasoning gaps in LLMs within medical settings, encouraging future improvements of LLMs for various clinical calculation tasks.
Authors: William Merrill, Noah A. Smith, Yanai Elazar
Abstract: How novel are texts generated by language models (LMs) relative to their training corpora? In this work, we investigate the extent to which modern LMs generate $n$-grams from their training data, evaluating both (i) the probability LMs assign to complete training $n$-grams and (ii) $n$-novelty, the proportion of $n$-grams generated by an LM that did not appear in the training data (for arbitrarily large $n$). To enable arbitrary-length $n$-gram search over a corpus in constant time, we develop Rusty-DAWG, a novel search tool inspired by indexing of genomic data. We compare the novelty of LM-generated text to human-written text and explore factors that affect generation novelty, focusing on the Pythia models. We find that, for $n > 4$, LM-generated text is less novel than human-written text, though it is more novel for smaller $n$. Larger LMs and more constrained decoding strategies both decrease novelty. Finally, we show that LMs complete $n$-grams with lower loss if they are more frequent in the training data. Overall, our results reveal factors influencing the novelty of LM-generated text, and we release Rusty-DAWG to facilitate further pretraining data research.
Authors: Roman Koshkin, Katsuhito Sudoh, Satoshi Nakamura
Abstract: The advent of transformers has fueled progress in machine translation. More recently large language models (LLMs) have come to the spotlight thanks to their generality and strong performance in a wide range of language tasks, including translation. Here we show that open-source LLMs perform on par with or better than some state-of-the-art baselines in simultaneous machine translation (SiMT) tasks, zero-shot. We also demonstrate that injection of minimal background information, which is easy with an LLM, brings further performance gains, especially on challenging technical subject-matter. This highlights LLMs' potential for building next generation of massively multilingual, context-aware and terminologically accurate SiMT systems that require no resource-intensive training or fine-tuning.
Authors: Gayane Ghazaryan, Erik Arakelyan, Pasquale Minervini, Isabelle Augenstein
Abstract: Question Answering (QA) datasets have been instrumental in developing and evaluating Large Language Model (LLM) capabilities. However, such datasets are scarce for languages other than English due to the cost and difficulties of collection and manual annotation. This means that producing novel models and measuring the performance of multilingual LLMs in low-resource languages is challenging. To mitigate this, we propose $\textbf{S}$yn$\textbf{DAR}$in, a method for generating and validating QA datasets for low-resource languages. We utilize parallel content mining to obtain $\textit{human-curated}$ paragraphs between English and the target language. We use the English data as context to $\textit{generate}$ synthetic multiple-choice (MC) question-answer pairs, which are automatically translated and further validated for quality. Combining these with their designated non-English $\textit{human-curated}$ paragraphs form the final QA dataset. The method allows to maintain the content quality, reduces the likelihood of factual errors, and circumvents the need for costly annotation. To test the method, we created a QA dataset with $1.2$K samples for the Armenian language. The human evaluation shows that $98\%$ of the generated English data maintains quality and diversity in the question types and topics, while the translation validation pipeline can filter out $\sim70\%$ of data with poor quality. We use the dataset to benchmark state-of-the-art LLMs, showing their inability to achieve human accuracy with some model performances closer to random chance. This shows that the generated dataset is non-trivial and can be used to evaluate reasoning capabilities in low-resource language.
Authors: Jing Yang, Yu Zhao, Yang Linyao, Xiao Wang, Long Chen, Fei-Yue Wang
Abstract: Temporal relation extraction (TRE) aims to grasp the evolution of events or actions, and thus shape the workflow of associated tasks, so it holds promise in helping understand task requests initiated by requesters in crowdsourcing systems. However, existing methods still struggle with limited and unevenly distributed annotated data. Therefore, inspired by the abundant global knowledge stored within pre-trained language models (PLMs), we propose a multi-task prompt learning framework for TRE (TemPrompt), incorporating prompt tuning and contrastive learning to tackle these issues. To elicit more effective prompts for PLMs, we introduce a task-oriented prompt construction approach that thoroughly takes the myriad factors of TRE into consideration for automatic prompt generation. In addition, we present temporal event reasoning as a supplement to bolster the model's focus on events and temporal cues. The experimental results demonstrate that TemPrompt outperforms all compared baselines across the majority of metrics under both standard and few-shot settings. A case study is provided to validate its effectiveness in crowdsourcing scenarios.
Authors: Wentao Shi, Mengqi Yuan, Junkang Wu, Qifan Wang, Fuli Feng
Abstract: Adapting Large Language Models (LLMs) for agent tasks is critical in developing language agents. Direct Preference Optimization (DPO) is a promising technique for this adaptation with the alleviation of compounding errors, offering a means to directly optimize Reinforcement Learning (RL) objectives. However, applying DPO to multi-turn tasks presents challenges due to the inability to cancel the partition function. Overcoming this obstacle involves making the partition function independent of the current state and addressing length disparities between preferred and dis-preferred trajectories. In this light, we replace the policy constraint with the state-action occupancy measure constraint in the RL objective and add length normalization to the Bradley-Terry model, yielding a novel loss function named DMPO for multi-turn agent tasks with theoretical explanations. Extensive experiments on three multi-turn agent task datasets confirm the effectiveness and superiority of the DMPO loss.
Authors: Chia-Yu Hung, Navonil Majumder, Ambuj Mehrish, Soujanya Poria
Abstract: The widespread applicability and increasing omnipresence of LLMs have instigated a need to align LLM responses to user and stakeholder preferences. Many preference optimization approaches have been proposed that fine-tune LLM parameters to achieve good alignment. However, such parameter tuning is known to interfere with model performance on many tasks. Moreover, keeping up with shifting user preferences is tricky in such a situation. Decoding-time alignment with reward model guidance solves these issues at the cost of increased inference time. However, most of such methods fail to strike the right balance between exploration and exploitation of reward -- often due to the conflated formulation of these two aspects - to give well-aligned responses. To remedy this we decouple these two aspects and implement them in an evolutionary fashion: exploration is enforced by decoding from mutated instructions and exploitation is represented as the periodic replacement of poorly-rewarded generations with well-rewarded ones. Empirical evidences indicate that this strategy outperforms many preference optimization and decode-time alignment approaches on two widely accepted alignment benchmarks AlpacaEval 2 and MT-Bench. Our implementation will be available at: https://darwin-alignment.github.io.
Authors: Xingjian Hu, Baole Wei, Liangcai Gao
Abstract: Text line detection is a key task in historical document analysis facing many challenges of arbitrary-shaped text lines, dense texts, and text lines with high aspect ratios, etc. In this paper, we propose a general framework for historical document text detection (SegHist), enabling existing segmentation-based text detection methods to effectively address the challenges, especially text lines with high aspect ratios. Integrating the SegHist framework with the commonly used method DB++, we develop DB-SegHist. This approach achieves SOTA on the CHDAC, MTHv2, and competitive results on HDRC datasets, with a significant improvement of 1.19% on the most challenging CHDAC dataset which features more text lines with high aspect ratios. Moreover, our method attains SOTA on rotated MTHv2 and rotated HDRC, demonstrating its rotational robustness. The code is available at https://github.com/LumionHXJ/SegHist.
Authors: Jingze Shi, Ting Xie, Bingheng Wu, Chunjun Zheng, Kai Wang
Abstract: Recent research has shown that combining Mamba with Transformer architecture, which has selective state space and quadratic self-attention mechanism, outperforms using Mamba or Transformer architecture alone in language modeling tasks. The quadratic self-attention mechanism effectively alleviates the shortcomings of selective state space in handling long-term dependencies of any element in the sequence. We propose a position information injection method that connects the selective state space model with the quadratic attention, and integrates these two architectures with hybrid experts with cross-sharing domains, so that we can enjoy the advantages of both. We design a new architecture with a more biomimetic idea: Observer-Thinker-Conceiver-Expresser (OTCE), which can compete with well-known medium-scale open-source language models on a small scale in language modeling tasks.
Authors: Khai Duy Doan, Abdul Waheed, Muhammad Abdul-Mageed
Abstract: Zero-shot multi-speaker text-to-speech (ZS-TTS) systems have advanced for English, however, it still lags behind due to insufficient resources. We address this gap for Arabic, a language of more than 450 million native speakers, by first adapting a sizeable existing dataset to suit the needs of speech synthesis. Additionally, we employ a set of Arabic dialect identification models to explore the impact of pre-defined dialect labels on improving the ZS-TTS model in a multi-dialect setting. Subsequently, we fine-tune the XTTS\footnote{https://docs.coqui.ai/en/latest/models/xtts.html}\footnote{https://medium.com/machine-learns/xtts-v2-new-version-of-the-open-source-text-to-speech-model-af73914db81f}\footnote{https://medium.com/@erogol/xtts-v1-techincal-notes-eb83ff05bdc} model, an open-source architecture. We then evaluate our models on a dataset comprising 31 unseen speakers and an in-house dialectal dataset. Our automated and human evaluation results show convincing performance while capable of generating dialectal speech. Our study highlights significant potential for improvements in this emerging area of research in Arabic.
URLs: https://docs.coqui.ai/en/latest/models/xtts.html, https://medium.com/machine-learns/xtts-v2-new-version-of-the-open-source-text-to-speech-model-af73914db81f, https://medium.com/@erogol/xtts-v1-techincal-notes-eb83ff05bdc
Authors: Ashwinee Panda, Berivan Isik, Xiangyu Qi, Sanmi Koyejo, Tsachy Weissman, Prateek Mittal
Abstract: Existing methods for adapting large language models (LLMs) to new tasks are not suited to multi-task adaptation because they modify all the model weights -- causing destructive interference between tasks. The resulting effects, such as catastrophic forgetting of earlier tasks, make it challenging to obtain good performance on multiple tasks at the same time. To mitigate this, we propose Lottery Ticket Adaptation (LoTA), a sparse adaptation method that identifies and optimizes only a sparse subnetwork of the model. We evaluate LoTA on a wide range of challenging tasks such as instruction following, reasoning, math, and summarization. LoTA obtains better performance than full fine-tuning and low-rank adaptation (LoRA), and maintains good performance even after training on other tasks -- thus, avoiding catastrophic forgetting. By extracting and fine-tuning over lottery tickets (or sparse task vectors), LoTA also enables model merging over highly dissimilar tasks. Our code is made publicly available at https://github.com/kiddyboots216/lottery-ticket-adaptation.
URLs: https://github.com/kiddyboots216/lottery-ticket-adaptation.
Authors: Shengqiong Wu, Hao Fei, Leigang Qu, Wei Ji, Tat-Seng Chua
Abstract: While recently Multimodal Large Language Models (MM-LLMs) have made exciting strides, they mostly fall prey to the limitation of only input-side multimodal understanding, without the ability to produce content in multiple modalities. As we humans always perceive the world and communicate with people through various modalities, developing any-to-any MM-LLMs capable of accepting and delivering content in any modality becomes essential to human-level AI. To fill the gap, we present an end-to-end general-purpose any-to-any MM-LLM system, NExT-GPT. We connect an LLM with multimodal adaptors and different diffusion decoders, enabling NExT-GPT to perceive inputs and generate outputs in arbitrary combinations of text, images, videos, and audio. By leveraging the existing well-trained highly-performing encoders and decoders, NExT-GPT is tuned with only a small amount of parameter (1%) of certain projection layers, which not only benefits low-cost training and also facilitates convenient expansion to more potential modalities. Moreover, we introduce a modality-switching instruction tuning (MosIT) and manually curate a high-quality dataset for MosIT, based on which NExT-GPT is empowered with complex cross-modal semantic understanding and content generation. Overall, our research showcases the promising possibility of building an AI agent capable of modeling universal modalities, paving the way for more human-like AI research in the community. Project page: https://next-gpt.github.io/
Authors: Yihan Wu, Zhengmian Hu, Junfeng Guo, Hongyang Zhang, Heng Huang
Abstract: Watermarking techniques offer a promising way to identify machine-generated content via embedding covert information into the contents generated from language models. A challenge in the domain lies in preserving the distribution of original generated content after watermarking. Our research extends and improves upon existing watermarking framework, placing emphasis on the importance of a \textbf{Di}stribution-\textbf{P}reserving (DiP) watermark. Contrary to the current strategies, our proposed DiPmark simultaneously preserves the original token distribution during watermarking (distribution-preserving), is detectable without access to the language model API and prompts (accessible), and is provably robust to moderate changes of tokens (resilient). DiPmark operates by selecting a random set of tokens prior to the generation of a word, then modifying the token distribution through a distribution-preserving reweight function to enhance the probability of these selected tokens during the sampling process. Extensive empirical evaluation on various language models and tasks demonstrates our approach's distribution-preserving property, accessibility, and resilience, making it a effective solution for watermarking tasks that demand impeccable quality preservation.
Authors: Keyu Wang, Guilin Qi, Jiaoyan Chen, Yi Huang, Tianxing Wu
Abstract: Ontologies contain rich knowledge within domain, which can be divided into two categories, namely extensional knowledge and intensional knowledge. Extensional knowledge provides information about the concrete instances that belong to specific concepts in the ontology, while intensional knowledge details inherent properties, characteristics, and semantic associations among concepts. However, existing ontology embedding approaches fail to take both extensional knowledge and intensional knowledge into fine consideration simultaneously. In this paper, we propose a novel ontology embedding approach named EIKE (Extensional and Intensional Knowledge Embedding) by representing ontologies in two spaces, called extensional space and intensional space. EIKE presents a unified framework for embedding instances, concepts and their relations in an ontology, applying a geometry-based method to model extensional knowledge and a pretrained language model to model intensional knowledge, which can capture both structure information and textual information. Experimental results show that EIKE significantly outperforms state-of-the-art methods in three datasets for both triple classification and link prediction, indicating that EIKE provides a more comprehensive and representative perspective of the domain.
Authors: Shashwat Singh, Shauli Ravfogel, Jonathan Herzig, Roee Aharoni, Ryan Cotterell, Ponnurangam Kumaraguru
Abstract: Language models often exhibit undesirable behavior, e.g., generating toxic or gender-biased text. In the case of neural language models, an encoding of the undesirable behavior is often present in the model's representations. Thus, one natural (and common) approach to prevent the model from exhibiting undesirable behavior is to steer the model's representations in a manner that reduces the probability of it generating undesirable text. This paper investigates the formal and empirical properties of steering functions, i.e., transformation of the neural language model's representations that alter its behavior. First, we derive two optimal, in the least-squares sense, affine steering functions under different constraints. Our theory provides justification for existing approaches and offers a novel, improved steering approach. Second, we offer a series of experiments that demonstrate the empirical effectiveness of the methods in mitigating bias and reducing toxic generation.
Authors: Zongyu Wu, Hongcheng Gao, Yueze Wang, Xiang Zhang, Suhang Wang
Abstract: Text-to-Image (T2I) models have shown great performance in generating images based on textual prompts. However, these models are vulnerable to unsafe input to generate unsafe content like sexual, harassment and illegal-activity images. Existing studies based on image checker, model fine-tuning and embedding blocking are impractical in real-world applications. Hence, we propose the first universal prompt optimizer for safe T2I (POSI) generation in black-box scenario. We first construct a dataset consisting of toxic-clean prompt pairs by GPT-3.5 Turbo. To guide the optimizer to have the ability of converting toxic prompt to clean prompt while preserving semantic information, we design a novel reward function measuring toxicity and text alignment of generated images and train the optimizer through Proximal Policy Optimization. Experiments show that our approach can effectively reduce the likelihood of various T2I models in generating inappropriate images, with no significant impact on text alignment. It is also flexible to be combined with methods to achieve better performance. Our code is available at https://github.com/wzongyu/POSI.
Authors: Sangkyu Lee, Sungdong Kim, Ashkan Yousefpour, Minjoon Seo, Kang Min Yoo, Youngjae Yu
Abstract: Existing approaches for aligning large language models with human preferences face a trade-off that requires a separate reward model (RM) for on-policy learning. In this paper, we present a novel alignment framework, SELF-JUDGE that (1) does on-policy learning and 2) is parameter efficient, as it does not require an additional RM for evaluating the samples for on-policy learning. To this end, we propose Judge-augmented Supervised Fine-Tuning (JSFT) to train a single model to act as both a policy and a judge. Specifically, we view the pairwise judgment task, choosing the better response from a response pair, as a special case of the instruction-following task. The resulting model can judge preferences of on-the-fly responses from current policy initialized from itself. Experimental results show the efficacy of SELF-JUDGE, outperforming baselines in preference benchmarks. We also show that the rejecting sampling by itself can improve performance further without an additional evaluator.
Authors: Zhiqiang Zhong, Kuangyu Zhou, Davide Mottin
Abstract: As Machine Learning (ML) models grow in size and demand higher-quality training data, the expenses associated with re-training and fine-tuning these models are escalating rapidly. Inspired by recent impressive achievements of Large Language Models (LLMs) in different fields, this paper delves into the question: can LLMs efficiently improve an ML's performance at a minimal cost? We show that, through our proposed training-free framework LlmCorr, an LLM can work as a post-hoc corrector to propose corrections for the predictions of an arbitrary ML model. In particular, we form a contextual knowledge database by incorporating the dataset's label information and the ML model's predictions on the validation dataset. Leveraging the in-context learning capability of LLMs, we ask the LLM to summarise the instances in which the ML model makes mistakes and the correlation between primary predictions and true labels. Following this, the LLM can transfer its acquired knowledge to suggest corrections for the ML model's predictions. Our experimental results on text analysis and the challenging molecular predictions show that \model improves the performance of a number of models by up to 39%.
Authors: Kai Zhang, Yi Luan, Hexiang Hu, Kenton Lee, Siyuan Qiao, Wenhu Chen, Yu Su, Ming-Wei Chang
Abstract: Image retrieval, i.e., finding desired images given a reference image, inherently encompasses rich, multi-faceted search intents that are difficult to capture solely using image-based measures. Recent works leverage text instructions to allow users to more freely express their search intents. However, they primarily focus on image pairs that are visually similar and/or can be characterized by a small set of pre-defined relations. The core thesis of this paper is that text instructions can enable retrieving images with richer relations beyond visual similarity. To show this, we introduce MagicLens, a series of self-supervised image retrieval models that support open-ended instructions. MagicLens is built on a key novel insight: image pairs that naturally occur on the same web pages contain a wide range of implicit relations (e.g., inside view of), and we can bring those implicit relations explicit by synthesizing instructions via foundation models. Trained on 36.7M (query image, instruction, target image) triplets with rich semantic relations mined from the web, MagicLens achieves results comparable with or better than prior best on eight benchmarks of various image retrieval tasks, while maintaining high parameter efficiency with a significantly smaller model size. Additional human analyses on a 1.4M-image unseen corpus further demonstrate the diversity of search intents supported by MagicLens. Code and models are publicly available at https://open-vision-language.github.io/MagicLens/.
Authors: Jinghan Jia, Yihua Zhang, Yimeng Zhang, Jiancheng Liu, Bharat Runwal, James Diffenderfer, Bhavya Kailkhura, Sijia Liu
Abstract: Large Language Models (LLMs) have highlighted the necessity of effective unlearning mechanisms to comply with data regulations and ethical AI practices. LLM unlearning aims at removing undesired data influences and associated model capabilities without compromising utility beyond the scope of unlearning. While interest in studying LLM unlearning is growing, the impact of the optimizer choice for LLM unlearning remains unexplored. In this work, we shed light on the significance of optimizer selection in LLM unlearning for the first time, establishing a clear connection between second-order optimization and influence unlearning (a classical approach using influence functions to update the model for data influence removal). This insight propels us to develop a second-order optimization-based LLM unlearning framework, termed Second-Order UnLearning (SOUL), which extends the static, one-shot model update using influence unlearning to a dynamic, iterative unlearning process. Our extensive experiments show that SOUL consistently outperforms conventional first-order methods across various unlearning tasks, models, and metrics, indicating that second-order optimization offers an effective and broadly applicable solution for LLM unlearning. Codes are available at https://github.com/OPTML-Group/SOUL.
Authors: Danshi Wang, Yidi Wang, Xiaotian Jiang, Yao Zhang, Yue Pang, Min Zhang
Abstract: Since the advent of GPT, large language models (LLMs) have brought about revolutionary advancements in all walks of life. As a superior natural language processing (NLP) technology, LLMs have consistently achieved state-of-the-art performance on numerous areas. However, LLMs are considered to be general-purpose models for NLP tasks, which may encounter challenges when applied to complex tasks in specialized fields such as optical networks. In this study, we propose a framework of LLM-empowered optical networks, facilitating intelligent control of the physical layer and efficient interaction with the application layer through an LLM-driven agent (AI-Agent) deployed in the control layer. The AI-Agent can leverage external tools and extract domain knowledge from a comprehensive resource library specifically established for optical networks. This is achieved through user input and well-crafted prompts, enabling the generation of control instructions and result representations for autonomous operation and maintenance in optical networks. To improve LLM's capability in professional fields and stimulate its potential on complex tasks, the details of performing prompt engineering, establishing domain knowledge library, and implementing complex tasks are illustrated in this study. Moreover, the proposed framework is verified on two typical tasks: network alarm analysis and network performance optimization. The good response accuracies and sematic similarities of 2,400 test situations exhibit the great potential of LLM in optical networks.
Authors: Bhrij Patel, Vishnu Sashank Dorbala, Dinesh Manocha, Amrit Singh Bedi
Abstract: Embodied Question Answering (EQA) is an important problem, which involves an agent exploring the environment to answer user queries. In the existing literature, EQA has exclusively been studied in single-agent scenarios, where exploration can be time-consuming and costly. In this work, we consider EQA in a multi-agent framework involving multiple large language models (LLM) based agents independently answering queries about a household environment. To generate one answer for each query, we use the individual responses to train a Central Answer Model (CAM) that aggregates responses for a robust answer. Using CAM, we observe a $50\%$ higher EQA accuracy when compared against aggregation methods for ensemble LLM, such as voting schemes and debates. CAM does not require any form of agent communication, alleviating it from the associated costs. We ablate CAM with various nonlinear (neural network, random forest, decision tree, XGBoost) and linear (logistic regression classifier, SVM) algorithms. Finally, we present a feature importance analysis for CAM via permutation feature importance (PFI), quantifying CAMs reliance on each independent agent and query context.
Authors: Jizhong Liu, Gang Li, Junbo Zhang, Heinrich Dinkel, Yongqing Wang, Zhiyong Yan, Yujun Wang, Bin Wang
Abstract: Automated audio captioning (AAC) is an audio-to-text task to describe audio contents in natural language. Recently, the advancements in large language models (LLMs), with improvements in training approaches for audio encoders, have opened up possibilities for improving AAC. Thus, we explore enhancing AAC from three aspects: 1) a pre-trained audio encoder via consistent ensemble distillation (CED) is used to improve the effectivity of acoustic tokens, with a querying transformer (Q-Former) bridging the modality gap to LLM and compress acoustic tokens; 2) we investigate the advantages of using a Llama 2 with 7B parameters as the decoder; 3) another pre-trained LLM corrects text errors caused by insufficient training data and annotation ambiguities. Both the audio encoder and text decoder are optimized by low-rank adaptation (LoRA). Experiments show that each of these enhancements is effective. Our method obtains a 33.0 SPIDEr-FL score, outperforming the winner of DCASE 2023 Task 6A.
Authors: Bin Wang, Xunlong Zou, Geyu Lin, Shuo Sun, Zhuohan Liu, Wenyu Zhang, Zhengyuan Liu, AiTi Aw, Nancy F. Chen
Abstract: We introduce AudioBench, a new benchmark designed to evaluate audio large language models (AudioLLMs). AudioBench encompasses 8 distinct tasks and 26 carefully selected or newly curated datasets, focusing on speech understanding, voice interpretation, and audio scene understanding. Despite the rapid advancement of large language models, including multimodal versions, a significant gap exists in comprehensive benchmarks for thoroughly evaluating their capabilities. AudioBench addresses this gap by providing relevant datasets and evaluation metrics. In our study, we evaluated the capabilities of four models across various aspects and found that no single model excels consistently across all tasks. We outline the research outlook for AudioLLMs and anticipate that our open-source code, data, and leaderboard will offer a robust testbed for future model developments.