Authors: Ergun Bi\c{c}ici
Abstract: We identify the similarity between two words in English by casting the task as machine translation performance prediction (MTPP) between the words given the context and the distance between their similarities. We use referential translation machines (RTMs), which allows a common representation for training and test sets and stacked machine learning models. RTMs can achieve the top results in Graded Word Similarity in Context (GWSC) task.
Authors: Zeyu Leo Liu, Shrey Pandit, Xi Ye, Eunsol Choi, Greg Durrett
Abstract: Large language models (LLMs) are increasingly being used to synthesize and reason about source code. However, the static nature of these models' knowledge does not reflect the fact that libraries and API functions they invoke are continuously evolving, with functionality being added or changing. While numerous benchmarks evaluate how LLMs can generate code, no prior work has studied how an LLMs' knowledge about code API functions can be updated. To fill this gap, we present CodeUpdateArena, a benchmark for knowledge editing in the code domain. An instance in our benchmark consists of a synthetic API function update paired with a program synthesis example that uses the updated functionality; our goal is to update an LLM to be able to solve this program synthesis example without providing documentation of the update at inference time. Compared to knowledge editing for facts encoded in text, success here is more challenging: a code LLM must correctly reason about the semantics of the modified function rather than just reproduce its syntax. Our dataset is constructed by first prompting GPT-4 to generate atomic and executable function updates. Then, for each update, we generate program synthesis examples whose code solutions are prone to use the update. Our benchmark covers updates of various types to 54 functions from seven diverse Python packages, with a total of 670 program synthesis examples. Our experiments show that prepending documentation of the update to open-source code LLMs (i.e., DeepSeek, CodeLlama) does not allow them to incorporate changes for problem solving, and existing knowledge editing techniques also have substantial room for improvement. We hope our benchmark will inspire new methods for knowledge updating in code LLMs.
Authors: Pedro Ruas, Fernando Gallego, Francisco J. Veredas, Francisco M. Couto
Abstract: State-of-the-art deep learning entity linking methods rely on extensive human-labelled data, which is costly to acquire. Current datasets are limited in size, leading to inadequate coverage of biomedical concepts and diminished performance when applied to new data. In this work, we propose to automatically generate data to create large-scale training datasets, which allows the exploration of approaches originally developed for the task of extreme multi-label ranking in the biomedical entity linking task. We propose the hybrid X-Linker pipeline that includes different modules to link disease and chemical entity mentions to concepts in the MEDIC and the CTD-Chemical vocabularies, respectively. X-Linker was evaluated on several biomedical datasets: BC5CDR-Disease, BioRED-Disease, NCBI-Disease, BC5CDR-Chemical, BioRED-Chemical, and NLM-Chem, achieving top-1 accuracies of 0.8307, 0.7969, 0.8271, 0.9511, 0.9248, and 0.7895, respectively. X-Linker demonstrated superior performance in three datasets: BC5CDR-Disease, NCBI-Disease, and BioRED-Chemical. In contrast, SapBERT outperformed X-Linker in the remaining three datasets. Both models rely only on the mention string for their operations. The source code of X-Linker and its associated data are publicly available for performing biomedical entity linking without requiring pre-labelled entities with identifiers from specific knowledge organization systems.
Authors: Mokhaiber Dandash, Masoud Asadpour
Abstract: Social media is heading toward personalization more and more, where individuals reveal their beliefs, interests, habits, and activities, simply offering glimpses into their personality traits. This study, explores the correlation between the use of Arabic language on twitter, personality traits and its impact on sentiment analysis. We indicated the personality traits of users based on the information extracted from their profile activities, and the content of their tweets. Our analysis incorporated linguistic features, profile statistics (including gender, age, bio, etc.), as well as additional features like emoticons. To obtain personality data, we crawled the timelines and profiles of users who took the 16personalities test in Arabic on 16personalities.com. Our dataset comprised 3,250 users who shared their personality results on twitter. We implemented various machine learning techniques, to reveal personality traits and developed a dedicated model for this purpose, achieving a 74.86% accuracy rate with BERT, analysis of this dataset proved that linguistic features, profile features and derived model can be used to differentiate between different personality traits. Furthermore, our findings demonstrated that personality affect sentiment in social media. This research contributes to the ongoing efforts in developing robust understanding of the relation between human behaviour on social media and personality features for real-world applications, such as political discourse analysis, and public opinion tracking.
Authors: Manish Nagireddy, Inkit Padhi, Soumya Ghosh, Prasanna Sattigeri
Abstract: Large language models (LLMs) have convincing performance in a variety of downstream tasks. However, these systems are prone to generating undesirable outputs such as harmful and biased text. In order to remedy such generations, the development of guardrail (or detector) models has gained traction. Motivated by findings from developing a detector for social bias, we adopt the notion of a use-mention distinction - which we identified as the primary source of under-performance in the preliminary versions of our social bias detector. Armed with this information, we describe a fully extensible and reproducible synthetic data generation pipeline which leverages taxonomy-driven instructions to create targeted and labeled data. Using this pipeline, we generate over 300K unique contrastive samples and provide extensive experiments to systematically evaluate performance on a suite of open source datasets. We show that our method achieves competitive performance with a fraction of the cost in compute and offers insight into iteratively developing efficient and capable guardrail models. Warning: This paper contains examples of text which are toxic, biased, and potentially harmful.
Authors: Krishnakant Bhatt, Karthika N J, Ganesh Ramakrishnan, Preethi Jyothi
Abstract: Subword tokens in Indian languages inherently carry meaning, and isolating them can enhance NLP tasks, making sub-word segmentation a crucial process. Segmenting Sanskrit and other Indian languages into subtokens is not straightforward, as it may include sandhi, which may lead to changes in the word boundaries. We propose a new approach of utilizing a Character-level Transformer model for Sanskrit Word Segmentation (CharSS). We perform experiments on three benchmark datasets to compare the performance of our method against existing methods. On the UoH+SandhiKosh dataset, our method outperforms the current state-of-the-art system by an absolute gain of 6.72 points in split prediction accuracy. On the hackathon dataset, our method achieves a gain of 2.27 points over the current SOTA system in terms of perfect match metric. We also propose a use-case of Sanskrit-based segments for a linguistically informed translation of technical terms to lexically similar low-resource Indian languages. In two separate experimental settings for this task, we achieve an average improvement of 8.46 and 6.79 chrF++ scores, respectively.
Authors: Jesse Roberts, Kyle Moore, Thao Pham, Oseremhen Ewaleifoh, Doug Fisher
Abstract: This paper evaluates whether large language models (LLMs) exhibit cognitive fan effects, similar to those discovered by Anderson in humans, after being pre-trained on human textual data. We conduct two sets of in-context recall experiments designed to elicit fan effects. Consistent with human results, we find that LLM recall uncertainty, measured via token probability, is influenced by the fan effect. Our results show that removing uncertainty disrupts the observed effect. The experiments suggest the fan effect is consistent whether the fan value is induced in-context or in the pre-training data. Finally, these findings provide in-silico evidence that fan effects and typicality are expressions of the same phenomena.
Authors: Jupinder Parmar, Shrimai Prabhumoye, Joseph Jennings, Bo Liu, Aastha Jhunjhunwala, Zhilin Wang, Mostofa Patwary, Mohammad Shoeybi, Bryan Catanzaro
Abstract: The impressive capabilities of recent language models can be largely attributed to the multi-trillion token pretraining datasets that they are trained on. However, model developers fail to disclose their construction methodology which has lead to a lack of open information on how to develop effective pretraining sets. To address this issue, we perform the first systematic study across the entire pipeline of pretraining set construction. First, we run ablations on existing techniques for pretraining set development to identify which methods translate to the largest gains in model accuracy on downstream evaluations. Then, we categorize the most widely used data source, web crawl snapshots, across the attributes of toxicity, quality, type of speech, and domain. Finally, we show how such attribute information can be used to further refine and improve the quality of a pretraining set. These findings constitute an actionable set of steps that practitioners can use to develop high quality pretraining sets.
Authors: Luke Yoffe, Alfonso Amayuelas, William Yang Wang
Abstract: To enhance Large Language Model (LLM) capabilities, multi-agent debates have been introduced, where multiple LLMs discuss solutions to a problem over several rounds of debate. However, LLMs often produce incorrect responses that appear deceptively confident, which can mislead other agents. This is partly because agents do not express their confidence levels during standard debates. To address this, we introduce DebUnc, a multi-agent debate framework that uses uncertainty metrics to assess agent confidence levels. We adapted the LLM attention mechanism to adjust token weights based on confidence levels and also explored using textual prompts to convey confidence. Our evaluations across various benchmarks show that attention-based methods are particularly effective, and that as uncertainty metrics evolve, performance will continue to increase. The code is available at https://github.com/lukeyoffe/debunc
Authors: Jayanta Sadhu, Maneesha Rani Saha, Rifat Shahriyar
Abstract: The influence of Large Language Models (LLMs) is rapidly growing, automating more jobs over time. Assessing the fairness of LLMs is crucial due to their expanding impact. Studies reveal the reflection of societal norms and biases in LLMs, which creates a risk of propagating societal stereotypes in downstream tasks. Many studies on bias in LLMs focus on gender bias in various NLP applications. However, there's a gap in research on bias in emotional attributes, despite the close societal link between emotion and gender. This gap is even larger for low-resource languages like Bangla. Historically, women are associated with emotions like empathy, fear, and guilt, while men are linked to anger, bravado, and authority. This pattern reflects societal norms in Bangla-speaking regions. We offer the first thorough investigation of gendered emotion attribution in Bangla for both closed and open source LLMs in this work. Our aim is to elucidate the intricate societal relationship between gender and emotion specifically within the context of Bangla. We have been successful in showing the existence of gender bias in the context of emotions in Bangla through analytical methods and also show how emotion attribution changes on the basis of gendered role selection in LLMs. All of our resources including code and data are made publicly available to support future research on Bangla NLP. Warning: This paper contains explicit stereotypical statements that many may find offensive.
Authors: Weijia Shi, Jaechan Lee, Yangsibo Huang, Sadhika Malladi, Jieyu Zhao, Ari Holtzman, Daogao Liu, Luke Zettlemoyer, Noah A. Smith, Chiyuan Zhang
Abstract: Language models (LMs) are trained on vast amounts of text data, which may include private and copyrighted content. Data owners may request the removal of their data from a trained model due to privacy or copyright concerns. However, exactly unlearning only these datapoints (i.e., retraining with the data removed) is intractable in modern-day models. This has led to the development of many approximate unlearning algorithms. The evaluation of the efficacy of these algorithms has traditionally been narrow in scope, failing to precisely quantify the success and practicality of the algorithm from the perspectives of both the model deployers and the data owners. We address this issue by proposing MUSE, a comprehensive machine unlearning evaluation benchmark that enumerates six diverse desirable properties for unlearned models: (1) no verbatim memorization, (2) no knowledge memorization, (3) no privacy leakage, (4) utility preservation on data not intended for removal, (5) scalability with respect to the size of removal requests, and (6) sustainability over sequential unlearning requests. Using these criteria, we benchmark how effectively eight popular unlearning algorithms on 7B-parameter LMs can unlearn Harry Potter books and news articles. Our results demonstrate that most algorithms can prevent verbatim memorization and knowledge memorization to varying degrees, but only one algorithm does not lead to severe privacy leakage. Furthermore, existing algorithms fail to meet deployer's expectations because they often degrade general model utility and also cannot sustainably accommodate successive unlearning requests or large-scale content removal. Our findings identify key issues with the practicality of existing unlearning algorithms on language models, and we release our benchmark to facilitate further evaluations: muse-bench.github.io
Authors: Rena Gao, Carsten Roever, Jey Han Lau
Abstract: We present an evaluation framework for interactive dialogue assessment in the context of English as a Second Language (ESL) speakers. Our framework collects dialogue-level interactivity labels (e.g., topic management; 4 labels in total) and micro-level span features (e.g., backchannels; 17 features in total). Given our annotated data, we study how the micro-level features influence the (higher level) interactivity quality of ESL dialogues by constructing various machine learning-based models. Our results demonstrate that certain micro-level features strongly correlate with interactivity quality, like reference word (e.g., she, her, he), revealing new insights about the interaction between higher-level dialogue quality and lower-level linguistic signals. Our framework also provides a means to assess ESL communication, which is useful for language assessment.
Authors: Yongqi Leng, Deyi Xiong
Abstract: While large language models (LLMs) have demonstrated superior multi-task capabilities, understanding the learning mechanisms behind this is still a challenging problem. In this paper, we attempt to understand such mechanisms from the perspective of neurons. Specifically, we detect task-sensitive neurons in LLMs via gradient attribution on task-specific data. Through extensive deactivation and fine-tuning experiments, we demonstrate that the detected neurons are highly correlated with the given task, which we term as task-specific neurons. With these identified task-specific neurons, we delve into two common problems in multi-task learning and continuous learning: Generalization and Catastrophic Forgetting. We find that the overlap of task-specific neurons is strongly associated with generalization and specialization across tasks. Interestingly, at certain layers of LLMs, there is a high similarity in the parameters of different task-specific neurons, and such similarity is highly correlated with the generalization performance. Inspired by these findings, we propose a neuron-level continuous fine-tuning method that only fine-tunes the current task-specific neurons during continuous learning, and extensive experiments demonstrate the effectiveness of the proposed method. Our study provides insights into the interpretability of LLMs in multi-task learning.
Authors: Youngkyung Seo, Yoonseok Heo, Jun-Seok Koh, Du-Seoung Chang
Abstract: In multi-session dialog system, it is essential to continuously update the memory as the session progresses. Simply accumulating memory can make it difficult to focus on the content of the conversation for inference due to the limited input sentence size. Therefore, efficient and accurate conversation model that is capable of managing memory to reflect the conversation history continuously is necessary. This paper presents a conversation model that efficiently manages memory as sessions progress and incorporates this into the model to reflect the conversation history accurately with 3 methodologies: SFT, DPO and DPO with SFT model. Our model using DPO algorithm shows an improvement about 0.0591 of BERTScore in memory accuracy, and the rate of responses reflecting the memory increased as well. Also, response generation performance enhanced about 4.292 in fluency, 3.935 in coherence, and 2.896 in consistency. This paper describes a training method that yields better performance than models with more than twice the parameter size, even when the model size is smaller. Thus, our model demonstrates efficiency not only in terms of accuracy but also in resource utilization.
Authors: Aniruddha Roy, Pretam Ray, Ayush Maheshwari, Sudeshna Sarkar, Pawan Goyal
Abstract: Neural Machine Translation (NMT) remains a formidable challenge, especially when dealing with low-resource languages. Pre-trained sequence-to-sequence (seq2seq) multi-lingual models, such as mBART-50, have demonstrated impressive performance in various low-resource NMT tasks. However, their pre-training has been confined to 50 languages, leaving out support for numerous low-resource languages, particularly those spoken in the Indian subcontinent. Expanding mBART-50's language support requires complex pre-training, risking performance decline due to catastrophic forgetting. Considering these expanding challenges, this paper explores a framework that leverages the benefits of a pre-trained language model along with knowledge distillation in a seq2seq architecture to facilitate translation for low-resource languages, including those not covered by mBART-50. The proposed framework employs a multilingual encoder-based seq2seq model as the foundational architecture and subsequently uses complementary knowledge distillation techniques to mitigate the impact of imbalanced training. Our framework is evaluated on three low-resource Indic languages in four Indic-to-Indic directions, yielding significant BLEU-4 and chrF improvements over baselines. Further, we conduct human evaluation to confirm effectiveness of our approach. Our code is publicly available at https://github.com/raypretam/Two-step-low-res-NMT.
Authors: Xiao Yu, Qingyang Wu, Yu Li, Zhou Yu
Abstract: Alignment is a crucial step to enhance the instruction-following and conversational abilities of language models. Despite many recent work proposing new algorithms, datasets, and training pipelines, there is a lack of comprehensive studies measuring the impact of various design choices throughout the whole training process. We first conduct a rigorous analysis over a three-stage training pipeline consisting of supervised fine-tuning, offline preference learning, and online preference learning. We have found that using techniques like sequence packing, loss masking in SFT, increasing the preference dataset size in DPO, and online DPO training can significantly improve the performance of language models. We then train from Gemma-2b-base and LLama-3-8b-base, and find that our best models exceed the performance of the official instruct models tuned with closed-source data and algorithms. Our code and models can be found at https://github.com/Columbia-NLP-Lab/LionAlignment.
Authors: Sneha Ray Barman, Shakuntala Mahanta, Neeraj Kumar Sharma
Abstract: Traditional approaches for understanding phonological learning have predominantly relied on curated text data. Although insightful, such approaches limit the knowledge captured in textual representations of the spoken language. To overcome this limitation, we investigate the potential of the Featural InfoWaveGAN model to learn iterative long-distance vowel harmony using raw speech data. We focus on Assamese, a language known for its phonologically regressive and word-bound vowel harmony. We demonstrate that the model is adept at grasping the intricacies of Assamese phonotactics, particularly iterative long-distance harmony with regressive directionality. It also produced non-iterative illicit forms resembling speech errors during human language acquisition. Our statistical analysis reveals a preference for a specific [+high,+ATR] vowel as a trigger across novel items, indicative of feature learning. More data and control could improve model proficiency, contrasting the universality of learning.
Authors: Junsoo Park, Seungyeon Jwa, Meiying Ren, Daeyoung Kim, Sanghyuk Choi
Abstract: Employing Large Language Models (LLMs) to assess the quality of generated responses, such as prompting instruct-tuned models or fine-tuning judge models, has become a widely adopted evaluation method. It is also known that such evaluators are vulnerable to biases, such as favoring longer responses. While it is important to overcome this problem, the specifics of these biases remain under-explored. In this work, we qualitatively identify six types of biases inherent in various judge models. We propose EvalBiasBench as a meta-evaluation collection of hand-crafted test cases for each bias type. Additionally, we present de-biasing dataset construction methods and the associated preference dataset OffsetBias. Experimental results demonstrate that fine-tuning on our dataset significantly enhances the robustness of judge models against biases and improves performance across most evaluation scenarios. We release our datasets and the fine-tuned judge model to public.
Authors: Amanda Kau, Xuzeng He, Aishwarya Nambissan, Aland Astudillo, Hui Yin, Amir Aryani
Abstract: In recent years, Natural Language Processing (NLP) has played a significant role in various Artificial Intelligence (AI) applications such as chatbots, text generation, and language translation. The emergence of large language models (LLMs) has greatly improved the performance of these applications, showing astonishing results in language understanding and generation. However, they still show some disadvantages, such as hallucinations and lack of domain-specific knowledge, that affect their performance in real-world tasks. These issues can be effectively mitigated by incorporating knowledge graphs (KGs), which organise information in structured formats that capture relationships between entities in a versatile and interpretable fashion. Likewise, the construction and validation of KGs present challenges that LLMs can help resolve. The complementary relationship between LLMs and KGs has led to a trend that combines these technologies to achieve trustworthy results. This work collected 28 papers outlining methods for KG-powered LLMs, LLM-based KGs, and LLM-KG hybrid approaches. We systematically analysed and compared these approaches to provide a comprehensive overview highlighting key trends, innovative techniques, and common challenges. This synthesis will benefit researchers new to the field and those seeking to deepen their understanding of how KGs and LLMs can be effectively combined to enhance AI applications capabilities.
Authors: Yangyang Yu, Zhiyuan Yao, Haohang Li, Zhiyang Deng, Yupeng Cao, Zhi Chen, Jordan W. Suchow, Rong Liu, Zhenyu Cui, Denghui Zhang, Zhaozhuo Xu, Koduvayur Subbalakshmi, Guojun Xiong, Yueru He, Jimin Huang, Dong Li, Qianqian Xie
Abstract: Large language models (LLMs) have demonstrated notable potential in conducting complex tasks and are increasingly utilized in various financial applications. However, high-quality sequential financial investment decision-making remains challenging. These tasks require multiple interactions with a volatile environment for every decision, demanding sufficient intelligence to maximize returns and manage risks. Although LLMs have been used to develop agent systems that surpass human teams and yield impressive investment returns, opportunities to enhance multi-sourced information synthesis and optimize decision-making outcomes through timely experience refinement remain unexplored. Here, we introduce the FinCon, an LLM-based multi-agent framework with CONceptual verbal reinforcement tailored for diverse FINancial tasks. Inspired by effective real-world investment firm organizational structures, FinCon utilizes a manager-analyst communication hierarchy. This structure allows for synchronized cross-functional agent collaboration towards unified goals through natural language interactions and equips each agent with greater memory capacity than humans. Additionally, a risk-control component in FinCon enhances decision quality by episodically initiating a self-critiquing mechanism to update systematic investment beliefs. The conceptualized beliefs serve as verbal reinforcement for the future agent's behavior and can be selectively propagated to the appropriate node that requires knowledge updates. This feature significantly improves performance while reducing unnecessary peer-to-peer communication costs. Moreover, FinCon demonstrates strong generalization capabilities in various financial tasks, including single stock trading and portfolio management.
Authors: Suhong Moon, Marwa Abdulhai, Minwoo Kang, Joseph Suh, Widyadewi Soedarmadji, Eran Kohen Behar, David M. Chan
Abstract: Large language models (LLMs) are trained from vast repositories of text authored by millions of distinct authors, reflecting an enormous diversity of human traits. While these models bear the potential to be used as approximations of human subjects in behavioral studies, prior efforts have been limited in steering model responses to match individual human users. In this work, we introduce "Anthology", a method for conditioning LLMs to particular virtual personas by harnessing open-ended life narratives, which we refer to as "backstories." We show that our methodology enhances the consistency and reliability of experimental outcomes while ensuring better representation of diverse sub-populations. Across three nationally representative human surveys conducted as part of Pew Research Center's American Trends Panel (ATP), we demonstrate that Anthology achieves up to 18% improvement in matching the response distributions of human respondents and 27% improvement in consistency metrics. Our code and generated backstories are available at https://github.com/CannyLab/anthology.
Authors: Hongfei Huang, Tingting Liang, Xixi Sun, Zikang Jin, Yuyu Yin
Abstract: Existing research on learning with noisy labels predominantly focuses on synthetic label noise. Although synthetic noise possesses well-defined structural properties, it often fails to accurately replicate real-world noise patterns. In recent years, there has been a concerted effort to construct generalizable and controllable instance-dependent noise datasets for image classification, significantly advancing the development of noise-robust learning in this area. However, studies on noisy label learning for text classification remain scarce. To better understand label noise in real-world text classification settings, we constructed the benchmark dataset NoisyAG-News through manual annotation. Initially, we analyzed the annotated data to gather observations about real-world noise. We qualitatively and quantitatively demonstrated that real-world noisy labels adhere to instance-dependent patterns. Subsequently, we conducted comprehensive learning experiments on NoisyAG-News and its corresponding synthetic noise datasets using pre-trained language models and noise-handling techniques. Our findings reveal that while pre-trained models are resilient to synthetic noise, they struggle against instance-dependent noise, with samples of varying confusion levels showing inconsistent performance during training and testing. These real-world noise patterns pose new, significant challenges, prompting a reevaluation of noisy label handling methods. We hope that NoisyAG-News will facilitate the development and evaluation of future solutions for learning with noisy labels.
Authors: Mana Makinae, Katsuhito Sudoh, Mararu Yamada, Satoshi Nakamura
Abstract: Simultaneous interpretation (SI), the translation of one language to another in real time, starts translation before the original speech has finished. Its evaluation needs to consider both latency and quality. This trade-off is challenging especially for distant word order language pairs such as English and Japanese. To handle this word order gap, interpreters maintain the word order of the source language as much as possible to keep up with original language to minimize its latency while maintaining its quality, whereas in translation reordering happens to keep fluency in the target language. This means outputs synchronized with the source language are desirable based on the real SI situation, and it's a key for further progress in computational SI and simultaneous machine translation (SiMT). In this work, we propose an automatic evaluation metric for SI and SiMT focusing on word order synchronization. Our evaluation metric is based on rank correlation coefficients, leveraging cross-lingual pre-trained language models. Our experimental results on NAIST-SIC-Aligned and JNPC showed our metrics' effectiveness to measure word order synchronization between source and target language.
Authors: Nan He, Weichen Xiong, Hanwen Liu, Yi Liao, Lei Ding, Kai Zhang, Guohua Tang, Xiao Han, Wei Yang
Abstract: The effectiveness of large language models (LLMs) is often hindered by duplicated data in their extensive pre-training datasets. Current approaches primarily focus on detecting and removing duplicates, which risks the loss of valuable information and neglects the varying degrees of duplication. To address this, we propose a soft deduplication method that maintains dataset integrity while selectively reducing the sampling weight of data with high commonness. Central to our approach is the concept of "data commonness", a metric we introduce to quantify the degree of duplication by measuring the occurrence probabilities of samples using an n-gram model. Empirical analysis shows that this method significantly improves training efficiency, achieving comparable perplexity scores with at least a 26% reduction in required training steps. Additionally, it enhances average few-shot downstream accuracy by 1.77% when trained for an equivalent duration. Importantly, this approach consistently improves performance, even on rigorously deduplicated datasets, indicating its potential to complement existing methods and become a standard pre-training process for LLMs.
Authors: Zhuocheng Gong, Ang Lv, Jian Guan, Junxi Yan, Wei Wu, Huishuai Zhang, Minlie Huang, Dongyan Zhao, Rui Yan
Abstract: Is it always necessary to compute tokens from shallow to deep layers in Transformers? The continued success of vanilla Transformers and their variants suggests an undoubted "yes". In this work, however, we attempt to break the depth-ordered convention by proposing a novel architecture dubbed mixture-of-modules (MoM), which is motivated by an intuition that any layer, regardless of its position, can be used to compute a token as long as it possesses the needed processing capabilities. The construction of MoM starts from a finite set of modules defined by multi-head attention and feed-forward networks, each distinguished by its unique parameterization. Two routers then iteratively select attention modules and feed-forward modules from the set to process a token. The selection dynamically expands the computation graph in the forward pass of the token, culminating in an assembly of modules. We show that MoM provides not only a unified framework for Transformers and their numerous variants but also a flexible and learnable approach for reducing redundancy in Transformer parameterization. We pre-train various MoMs using OpenWebText. Empirical results demonstrate that MoMs, of different parameter counts, consistently outperform vanilla transformers on both GLUE and XSUM benchmarks. More interestingly, with a fixed parameter budget, MoM-large enables an over 38% increase in depth for computation graphs compared to GPT-2-large, resulting in absolute gains of 1.4 on GLUE and 1 on XSUM. On the other hand, MoM-large also enables an over 60% reduction in depth while involving more modules per layer, yielding a 16% reduction in TFLOPs and a 43% decrease in memory usage compared to GPT-2-large, while maintaining comparable performance.
Authors: Ali Modarressi, Abdullatif K\"oksal, Hinrich Sch\"utze
Abstract: Many datasets have been developed to train and evaluate document-level relation extraction (RE) models. Most of these are constructed using real-world data. It has been shown that RE models trained on real-world data suffer from factual biases. To evaluate and address this issue, we present CovEReD, a counterfactual data generation approach for document-level relation extraction datasets using entity replacement. We first demonstrate that models trained on factual data exhibit inconsistent behavior: while they accurately extract triples from factual data, they fail to extract the same triples after counterfactual modification. This inconsistency suggests that models trained on factual data rely on spurious signals such as specific entities and external knowledge $\unicode{x2013}$ rather than on the input context $\unicode{x2013}$ to extract triples. We show that by generating document-level counterfactual data with CovEReD and training models on them, consistency is maintained with minimal impact on RE performance. We release our CovEReD pipeline as well as Re-DocRED-CF, a dataset of counterfactual RE documents, to assist in evaluating and addressing inconsistency in document-level RE.
Authors: Wenxin Zhou, Thuy Hang Ngo
Abstract: Our team participated in the BioASQ 2024 Task12b and Synergy tasks to build a system that can answer biomedical questions by retrieving relevant articles and snippets from the PubMed database and generating exact and ideal answers. We propose a two-level information retrieval and question-answering system based on pre-trained large language models (LLM), focused on LLM prompt engineering and response post-processing. We construct prompts with in-context few-shot examples and utilize post-processing techniques like resampling and malformed response detection. We compare the performance of various pre-trained LLM models on this challenge, including Mixtral, OpenAI GPT and Llama2. Our best-performing system achieved 0.14 MAP score on document retrieval, 0.05 MAP score on snippet retrieval, 0.96 F1 score for yes/no questions, 0.38 MRR score for factoid questions and 0.50 F1 score for list questions in Task 12b.
Authors: Jinseok Kim, Jaewon Jung, Sangyeop Kim, Sohyung Park, Sungzoon Cho
Abstract: Despite the impressive capabilities of Large Language Models (LLMs) in various tasks, their vulnerability to unsafe prompts remains a critical issue. These prompts can lead LLMs to generate responses on illegal or sensitive topics, posing a significant threat to their safe and ethical use. Existing approaches attempt to address this issue using classification models, but they have several drawbacks. With the increasing complexity of unsafe prompts, similarity search-based techniques that identify specific features of unsafe prompts provide a more robust and effective solution to this evolving problem. This paper investigates the potential of sentence encoders to distinguish safe from unsafe prompts, and the ability to classify various unsafe prompts according to a safety taxonomy. We introduce new pairwise datasets and the Categorical Purity (CP) metric to measure this capability. Our findings reveal both the effectiveness and limitations of existing sentence encoders, proposing directions to improve sentence encoders to operate as more robust safety detectors. Our code is available at https://github.com/JwdanielJung/Safe-Embed.
Authors: Victoria R. Li, Yida Chen, Naomi Saphra
Abstract: While the biases of language models in production are extensively documented, the biases of their guardrails have been neglected. This paper studies how contextual information about the user influences the likelihood of an LLM to refuse to execute a request. By generating user biographies that offer ideological and demographic information, we find a number of biases in guardrail sensitivity on GPT-3.5. Younger, female, and Asian-American personas are more likely to trigger a refusal guardrail when requesting censored or illegal information. Guardrails are also sycophantic, refusing to comply with requests for a political position the user is likely to disagree with. We find that certain identity groups and seemingly innocuous information, e.g., sports fandom, can elicit changes in guardrail sensitivity similar to direct statements of political ideology. For each demographic category and even for American football team fandom, we find that ChatGPT appears to infer a likely political ideology and modify guardrail behavior accordingly.
Authors: Mayank Singh, Nazia Nafis, Abhijeet Kumar, Mridul Mishra
Abstract: Global sustainable fund universe encompasses open-end funds and exchange-traded funds (ETF) that, by prospectus or other regulatory filings, claim to focus on Environment, Social and Governance (ESG). Challengingly, the claims can only be confirmed by examining the textual disclosures to check if there is presence of intentionality and ESG focus on its investment strategy. Currently, there is no regulation to enforce sustainability in ESG products space. This paper proposes a unique method and system to classify and score the fund prospectuses in the sustainable universe regarding specificity and transparency of language. We aim to employ few-shot learners to identify specific, ambiguous, and generic sustainable investment-related language. Additionally, we construct a ratio metric to determine language score and rating to rank products and quantify sustainability claims for US sustainable universe. As a by-product, we publish manually annotated quality training dataset on Hugging Face (ESG-Prospectus-Clarity-Category under cc-by-nc-sa-4.0) of more than 1K ESG textual statements. The performance of the few-shot finetuning approach is compared with zero-shot models e.g., Llama-13B, GPT 3.5 Turbo etc. We found that prompting large language models are not accurate for domain specific tasks due to misalignment issues. The few-shot finetuning techniques outperform zero-shot models by large margins of more than absolute ~30% in precision, recall and F1 metrics on completely unseen ESG languages (test set). Overall, the paper attempts to establish a systematic and scalable approach to measure and rate sustainability intention quantitatively for sustainable funds using texts in prospectus. Regulatory bodies, investors, and advisors may utilize the findings of this research to reduce cognitive load in investigating or screening of ESG funds which accurately reflects the ESG intention.
Authors: Flor Miriam Plaza-del-Arco, Amanda Cercas Curry, Susanna Paoli, Alba Curry, Dirk Hovy
Abstract: Emotions play important epistemological and cognitive roles in our lives, revealing our values and guiding our actions. Previous work has shown that LLMs display biases in emotion attribution along gender lines. However, unlike gender, which says little about our values, religion, as a socio-cultural system, prescribes a set of beliefs and values for its followers. Religions, therefore, cultivate certain emotions. Moreover, these rules are explicitly laid out and interpreted by religious leaders. Using emotion attribution, we explore how different religions are represented in LLMs. We find that: Major religions in the US and European countries are represented with more nuance, displaying a more shaded model of their beliefs. Eastern religions like Hinduism and Buddhism are strongly stereotyped. Judaism and Islam are stigmatized -- the models' refusal skyrocket. We ascribe these to cultural bias in LLMs and the scarcity of NLP literature on religion. In the rare instances where religion is discussed, it is often in the context of toxic language, perpetuating the perception of these religions as inherently toxic. This finding underscores the urgent need to address and rectify these biases. Our research underscores the crucial role emotions play in our lives and how our values influence them.
Authors: Zara Siddique, Liam D. Turner, Luis Espinosa-Anke
Abstract: Large language models (LLMs) have been shown to propagate and amplify harmful stereotypes, particularly those that disproportionately affect marginalised communities. To understand the effect of these stereotypes more comprehensively, we introduce GlobalBias, a dataset of 876k sentences incorporating 40 distinct gender-by-ethnicity groups alongside descriptors typically used in bias literature, which enables us to study a broad set of stereotypes from around the world. We use GlobalBias to directly probe a suite of LMs via perplexity, which we use as a proxy to determine how certain stereotypes are represented in the model's internal representations. Following this, we generate character profiles based on given names and evaluate the prevalence of stereotypes in model outputs. We find that the demographic groups associated with various stereotypes remain consistent across model likelihoods and model outputs. Furthermore, larger models consistently display higher levels of stereotypical outputs, even when explicitly instructed not to.
Authors: Omar Manil Bendali, Samir Ferroum, Ekaterina Kozachenko, Youssef Parviz, Hanna Shcharbakova, Anna Tokareva, Shemair Williams
Abstract: The task of writing rap is challenging and involves producing complex rhyming schemes, yet meaningful lyrics. In this work, we propose Raply, a fine-tuned GPT-2 model capable of producing meaningful rhyming text in the style of rap. In addition to its rhyming capabilities, the model is able to generate less offensive content. It was achieved through the fine-tuning the model on a new dataset Mitislurs, a profanity-mitigated corpus. We evaluate the output of the model on two criteria: 1) rhyming based on the rhyme density metric; 2) profanity content, using the list of profanities for the English language. To our knowledge, this is the first attempt at profanity mitigation for rap lyrics generation.
Authors: Tim R. Davidson, Viacheslav Surkov, Veniamin Veselovsky, Giuseppe Russo, Robert West, Caglar Gulcehre
Abstract: A rapidly growing number of applications rely on a small set of closed-source language models (LMs). This dependency might introduce novel security risks if LMs develop self-recognition capabilities. Inspired by human identity verification methods, we propose a novel approach for assessing self-recognition in LMs using model-generated "security questions". Our test can be externally administered to keep track of frontier models as it does not require access to internal model parameters or output probabilities. We use our test to examine self-recognition in ten of the most capable open- and closed-source LMs currently publicly available. Our extensive experiments found no empirical evidence of general or consistent self-recognition in any examined LM. Instead, our results suggest that given a set of alternatives, LMs seek to pick the "best" answer, regardless of its origin. Moreover, we find indications that preferences about which models produce the best answers are consistent across LMs. We additionally uncover novel insights on position bias considerations for LMs in multiple-choice settings.
Authors: Angel Navarro, Francisco Casacuberta
Abstract: Pre-trained large language models (LLM) are starting to be widely used in many applications. In this work, we explore the use of these models in interactive machine translation (IMT) environments. In particular, we have chosen mBART (multilingual Bidirectional and Auto-Regressive Transformer) and mT5 (multilingual Text-to-Text Transfer Transformer) as the LLMs to perform our experiments. The system generates perfect translations interactively using the feedback provided by the user at each iteration. The Neural Machine Translation (NMT) model generates a preliminary hypothesis with the feedback, and the user validates new correct segments and performs a word correction--repeating the process until the sentence is correctly translated. We compared the performance of mBART, mT5, and a state-of-the-art (SoTA) machine translation model on a benchmark dataset regarding user effort, Word Stroke Ratio (WSR), Key Stroke Ratio (KSR), and Mouse Action Ratio (MAR). The experimental results indicate that mBART performed comparably with SoTA models, suggesting that it is a viable option for this field of IMT. The implications of this finding extend to the development of new machine translation models for interactive environments, as it indicates that some novel pre-trained models exhibit SoTA performance in this domain, highlighting the potential benefits of adapting these models to specific needs.
Authors: Rapha\"el Tinarrage, Henrique Ennes, Lucas E. Resck, Lucas T. Gomes, Jean R. Ponciano, Jorge Poco
Abstract: Binding precedents (S\'umulas Vinculantes) constitute a juridical instrument unique to the Brazilian legal system and whose objectives include the protection of the Federal Supreme Court against repetitive demands. Studies of the effectiveness of these instruments in decreasing the Court's exposure to similar cases, however, indicate that they tend to fail in such a direction, with some of the binding precedents seemingly creating new demands. We empirically assess the legal impact of five binding precedents, 11, 14, 17, 26 and 37, at the highest court level through their effects on the legal subjects they address. This analysis is only possible through the comparison of the Court's ruling about the precedents' themes before they are created, which means that these decisions should be detected through techniques of Similar Case Retrieval. The contributions of this article are therefore twofold: on the mathematical side, we compare the uses of different methods of Natural Language Processing -- TF-IDF, LSTM, BERT, and regex -- for Similar Case Retrieval, whereas on the legal side, we contrast the inefficiency of these binding precedents with a set of hypotheses that may justify their repeated usage. We observe that the deep learning models performed significantly worse in the specific Similar Case Retrieval task and that the reasons for binding precedents to fail in responding to repetitive demand are heterogeneous and case-dependent, making it impossible to single out a specific cause.
Authors: J. Crosbie, E. Shutova
Abstract: Large language models (LLMs) have shown a remarkable ability to learn and perform complex tasks through in-context learning (ICL). However, a comprehensive understanding of its internal mechanisms is still lacking. This paper explores the role of induction heads in a few-shot ICL setting. We analyse two state-of-the-art models, Llama-3-8B and InternLM2-20B on abstract pattern recognition and NLP tasks. Our results show that even a minimal ablation of induction heads leads to ICL performance decreases of up to ~32% for abstract pattern recognition tasks, bringing the performance close to random. For NLP tasks, this ablation substantially decreases the model's ability to benefit from examples, bringing few-shot ICL performance close to that of zero-shot prompts. We further use attention knockout to disable specific induction patterns, and present fine-grained evidence for the role that the induction mechanism plays in ICL.
Authors: Inwon Kang, William Van Woensel, Oshani Seneviratne
Abstract: We explore using Large Language Models (LLMs) to generate application code that automates health insurance processes from text-based policies. We target blockchain-based smart contracts as they offer immutability, verifiability, scalability, and a trustless setting: any number of parties can use the smart contracts, and they need not have previously established trust relationships with each other. Our methodology generates outputs at increasing levels of technical detail: (1) textual summaries, (2) declarative decision logic, and (3) smart contract code with unit tests. We ascertain LLMs are good at the task (1), and the structured output is useful to validate tasks (2) and (3). Declarative languages (task 2) are often used to formalize healthcare policies, but their execution on blockchain is non-trivial. Hence, task (3) attempts to directly automate the process using smart contracts. To assess the LLM output, we propose completeness, soundness, clarity, syntax, and functioning code as metrics. Our evaluation employs three health insurance policies (scenarios) with increasing difficulty from Medicare's official booklet. Our evaluation uses GPT-3.5 Turbo, GPT-3.5 Turbo 16K, GPT-4, GPT-4 Turbo and CodeLLaMA. Our findings confirm that LLMs perform quite well in generating textual summaries. Although outputs from tasks (2)-(3) are useful starting points, they require human oversight: in multiple cases, even "runnable" code will not yield sound results; the popularity of the target language affects the output quality; and more complex scenarios still seem a bridge too far. Nevertheless, our experiments demonstrate the promise of LLMs for translating textual process descriptions into smart contracts.
Authors: Yue Zhang, Ziqiao Ma, Jialu Li, Yanyuan Qiao, Zun Wang, Joyce Chai, Qi Wu, Mohit Bansal, Parisa Kordjamshidi
Abstract: Vision-and-Language Navigation (VLN) has gained increasing attention over recent years and many approaches have emerged to advance their development. The remarkable achievements of foundation models have shaped the challenges and proposed methods for VLN research. In this survey, we provide a top-down review that adopts a principled framework for embodied planning and reasoning, and emphasizes the current methods and future opportunities leveraging foundation models to address VLN challenges. We hope our in-depth discussions could provide valuable resources and insights: on one hand, to milestone the progress and explore opportunities and potential roles for foundation models in this field, and on the other, to organize different challenges and solutions in VLN to foundation model researchers.
Authors: Ruiran Su, Janet B. Pierrehumbert
Abstract: This work introduces the ClimateSent-GAT Model, an innovative method that integrates Graph Attention Networks (GATs) with techniques from natural language processing to accurately identify and predict disagreements within Reddit comment-reply pairs. Our model classifies disagreements into three categories: agree, disagree, and neutral. Leveraging the inherent graph structure of Reddit comment-reply pairs, the model significantly outperforms existing benchmarks by capturing complex interaction patterns and sentiment dynamics. This research advances graph-based NLP methodologies and provides actionable insights for policymakers and educators in climate science communication.
Authors: Weize Chen, Ziming You, Ran Li, Yitong Guan, Chen Qian, Chenyang Zhao, Cheng Yang, Ruobing Xie, Zhiyuan Liu, Maosong Sun
Abstract: The rapid advancement of large language models (LLMs) has paved the way for the development of highly capable autonomous agents. However, existing multi-agent frameworks often struggle with integrating diverse capable third-party agents due to reliance on agents defined within their own ecosystems. They also face challenges in simulating distributed environments, as most frameworks are limited to single-device setups. Furthermore, these frameworks often rely on hard-coded communication pipelines, limiting their adaptability to dynamic task requirements. Inspired by the concept of the Internet, we propose the Internet of Agents (IoA), a novel framework that addresses these limitations by providing a flexible and scalable platform for LLM-based multi-agent collaboration. IoA introduces an agent integration protocol, an instant-messaging-like architecture design, and dynamic mechanisms for agent teaming and conversation flow control. Through extensive experiments on general assistant tasks, embodied AI tasks, and retrieval-augmented generation benchmarks, we demonstrate that IoA consistently outperforms state-of-the-art baselines, showcasing its ability to facilitate effective collaboration among heterogeneous agents. IoA represents a step towards linking diverse agents in an Internet-like environment, where agents can seamlessly collaborate to achieve greater intelligence and capabilities. Our codebase has been released at \url{https://github.com/OpenBMB/IoA}.
Authors: Yung-Sung Chuang, Linlu Qiu, Cheng-Yu Hsieh, Ranjay Krishna, Yoon Kim, James Glass
Abstract: When asked to summarize articles or answer questions given a passage, large language models (LLMs) can hallucinate details and respond with unsubstantiated answers that are inaccurate with respect to the input context. This paper describes a simple approach for detecting such contextual hallucinations. We hypothesize that contextual hallucinations are related to the extent to which an LLM attends to information in the provided context versus its own generations. Based on this intuition, we propose a simple hallucination detection model whose input features are given by the ratio of attention weights on the context versus newly generated tokens (for each attention head). We find that a linear classifier based on these lookback ratio features is as effective as a richer detector that utilizes the entire hidden states of an LLM or a text-based entailment model. The lookback ratio-based detector -- Lookback Lens -- is found to transfer across tasks and even models, allowing a detector that is trained on a 7B model to be applied (without retraining) to a larger 13B model. We further apply this detector to mitigate contextual hallucinations, and find that a simple classifier-guided decoding approach is able to reduce the amount of hallucination, for example by 9.6% in the XSum summarization task.
Authors: Shaltiel Shmidman, Avi Shmidman, Amir DN Cohen, Moshe Koppel
Abstract: Training large language models (LLMs) in low-resource languages such as Hebrew poses unique challenges. In this paper, we introduce DictaLM2.0 and DictaLM2.0-Instruct, two LLMs derived from the Mistral model, trained on a substantial corpus of approximately 200 billion tokens in both Hebrew and English. Adapting a pre-trained model to a new language involves specialized techniques that differ significantly from training a model from scratch or further training existing models on well-resourced languages such as English. We outline these novel training methodologies, which facilitate effective learning and adaptation to the linguistic properties of Hebrew. Additionally, we fine-tuned DictaLM2.0-Instruct on a comprehensive instruct dataset to enhance its performance on task-specific instructions. To rigorously evaluate our models, we introduce a new benchmark suite for Hebrew LLM evaluation, covering a diverse set of tasks including Question Answering, Sentiment Analysis, Winograd Schema Challenge, Translation, and Summarization. Our work not only addresses the intricacies of training LLMs in low-resource languages but also proposes a framework that can be leveraged for adapting other LLMs to various non-English languages, contributing to the broader field of multilingual NLP.
Authors: Tong Chen, Akari Asai, Niloofar Mireshghallah, Sewon Min, James Grimmelmann, Yejin Choi, Hannaneh Hajishirzi, Luke Zettlemoyer, Pang Wei Koh
Abstract: Evaluating the degree of reproduction of copyright-protected content by language models (LMs) is of significant interest to the AI and legal communities. Although both literal and non-literal similarities are considered by courts when assessing the degree of reproduction, prior research has focused only on literal similarities. To bridge this gap, we introduce CopyBench, a benchmark designed to measure both literal and non-literal copying in LM generations. Using copyrighted fiction books as text sources, we provide automatic evaluation protocols to assess literal and non-literal copying, balanced against the model utility in terms of the ability to recall facts from the copyrighted works and generate fluent completions. We find that, although literal copying is relatively rare, two types of non-literal copying -- event copying and character copying -- occur even in models as small as 7B parameters. Larger models demonstrate significantly more copying, with literal copying rates increasing from 0.2% to 10.5% and non-literal copying from 2.3% to 6.9% when comparing Llama3-8B and 70B models, respectively. We further evaluate the effectiveness of current strategies for mitigating copying and show that (1) training-time alignment can reduce literal copying but may increase non-literal copying, and (2) current inference-time mitigation methods primarily reduce literal but not non-literal copying.
Authors: Liqun Ma, Mingjie Sun, Zhiqiang Shen
Abstract: This work presents a Fully BInarized Large Language Model (FBI-LLM), demonstrating for the first time how to train a large-scale binary language model from scratch (not the partial binary or ternary LLM like BitNet b1.58) to match the performance of its full-precision counterparts (e.g., FP16 or BF16) in transformer-based LLMs. It achieves this by employing an autoregressive distillation (AD) loss with maintaining equivalent model dimensions (130M, 1.3B, 7B) and training data volume as regular LLM pretraining, while delivering competitive results in terms of perplexity and task-specific effectiveness. Intriguingly, by analyzing the training trajectory, we find that the pretrained weight is not necessary for training binarized LLMs from scratch. This research encourages a new computational framework and may facilitate the future design of specialized hardware tailored for fully 1-bit LLMs. We make all models, code, and training dataset fully accessible and transparent to support further research (Code: https://github.com/LiqunMa/FBI-LLM. Model: https://huggingface.co/LiqunMa/).
URLs: https://github.com/LiqunMa/FBI-LLM., https://huggingface.co/LiqunMa/).
Authors: Jiaxi Cui, Wentao Zhang, Jing Tang, Xudong Tong, Zhenwei Zhang, Amie, Jing Wen, Rongsheng Wang, Pengfei Wu
Abstract: The pervasive deployment of Large Language Models-LLMs in various sectors often neglects the nuanced requirements of individuals and small organizations, who benefit more from models precisely tailored to their specific business contexts rather than those with broadly superior general capabilities. This work introduces \textbf{AnyTaskTune}, a novel fine-tuning methodology coined as \textbf{Task-Fine-Tune}, specifically developed to elevate model performance on a diverse array of domain-specific tasks. This method involves a meticulous process to identify and define targeted sub-tasks within a domain, followed by the creation of specialized enhancement datasets for fine-tuning, thereby optimizing task-specific model performance. We conducted comprehensive fine-tuning experiments not only in the legal domain for tasks such as keyword extraction and sentence prediction but across over twenty different sub-tasks derived from the domains of finance, healthcare, law, psychology, consumer services, and human resources. To substantiate our approach and facilitate community engagement, we will open-source these bilingual task datasets. Our findings demonstrate that models fine-tuned using the \textbf{Task-Fine-Tune} methodology not only achieve superior performance on these specific tasks but also significantly outperform models with higher general capabilities in their respective domains. Our work is publicly available at \url{https://github.com/PandaVT/DataTager}.
Authors: Messi H. J. Lee, Jacob M. Montgomery, Calvin K. Lai
Abstract: Vision Language Models (VLMs), exemplified by GPT-4V, adeptly integrate text and vision modalities. This integration enhances Large Language Models' ability to mimic human perception, allowing them to process image inputs. Despite VLMs' advanced capabilities, however, there is a concern that VLMs inherit biases of both modalities in ways that make biases more pervasive and difficult to mitigate. Our study explores how VLMs perpetuate homogeneity bias and trait associations with regards to race and gender. When prompted to write stories based on images of human faces, GPT-4V describes subordinate racial and gender groups with greater homogeneity than dominant groups and relies on distinct, yet generally positive, stereotypes. Importantly, VLM stereotyping is driven by visual cues rather than group membership alone such that faces that are rated as more prototypically Black and feminine are subject to greater stereotyping. These findings suggest that VLMs may associate subtle visual cues related to racial and gender groups with stereotypes in ways that could be challenging to mitigate. We explore the underlying reasons behind this behavior and discuss its implications and emphasize the importance of addressing these biases as VLMs come to mirror human perception.
Authors: Weilin Cai, Juyong Jiang, Fan Wang, Jing Tang, Sunghun Kim, Jiayi Huang
Abstract: Large language models (LLMs) have garnered unprecedented advancements across diverse fields, ranging from natural language processing to computer vision and beyond. The prowess of LLMs is underpinned by their substantial model size, extensive and diverse datasets, and the vast computational power harnessed during training, all of which contribute to the emergent abilities of LLMs (e.g., in-context learning) that are not present in small models. Within this context, the mixture of experts (MoE) has emerged as an effective method for substantially scaling up model capacity with minimal computation overhead, gaining significant attention from academia and industry. Despite its growing prevalence, there lacks a systematic and comprehensive review of the literature on MoE. This survey seeks to bridge that gap, serving as an essential resource for researchers delving into the intricacies of MoE. We first briefly introduce the structure of the MoE layer, followed by proposing a new taxonomy of MoE. Next, we overview the core designs for various MoE models including both algorithmic and systemic aspects, alongside collections of available open-source implementations, hyperparameter configurations and empirical evaluations. Furthermore, we delineate the multifaceted applications of MoE in practice, and outline some potential directions for future research. To facilitate ongoing updates and the sharing of cutting-edge developments in MoE research, we have established a resource repository accessible at https://github.com/withinmiaov/A-Survey-on-Mixture-of-Experts.
URLs: https://github.com/withinmiaov/A-Survey-on-Mixture-of-Experts.
Authors: Shreekanth M Prabhu, Gopalpillai Radhakrishnan
Abstract: Establishing the chronology of the Vedas has interested scholars for the last two centuries. The oldest among them is Rig-Veda which has ten Mandalas, each composed separately. In this paper, we look at deciphering plausible pointers to the internal chronology of the Mandalas, by focusing on Gods and Goddesses worshiped in different Mandalas. We apply text analysis to the Mandalas using Clustering Techniques based on Cosine Similarity. Then we represent the association of deities with Mandalas using a grid-based Social Network that is amenable to chronological analysis and demonstrates the benefits of using Social Network Analysis for the problem at hand. Further, we analyze references to rivers to arrive at additional correlations. The approach used can be deployed generically to analyze other kinds of references and mentions and arrive at more substantive inferences.
Authors: Pranshav Gajjar, Vijay K. Shah
Abstract: Large Language Models (LLMs) can revolutionize how we deploy and operate Open Radio Access Networks (O-RAN) by enhancing network analytics, anomaly detection, and code generation and significantly increasing the efficiency and reliability of a plethora of O-RAN tasks. In this paper, we present ORAN-Bench-13K, the first comprehensive benchmark designed to evaluate the performance of Large Language Models (LLMs) within the context of O-RAN. Our benchmark consists of 13,952 meticulously curated multiple-choice questions generated from 116 O-RAN specification documents. We leverage a novel three-stage LLM framework, and the questions are categorized into three distinct difficulties to cover a wide spectrum of ORAN-related knowledge. We thoroughly evaluate the performance of several state-of-the-art LLMs, including Gemini, Chat-GPT, and Mistral. Additionally, we propose ORANSight, a Retrieval-Augmented Generation (RAG)-based pipeline that demonstrates superior performance on ORAN-Bench-13K compared to other tested closed-source models. Our findings indicate that current popular LLM models are not proficient in O-RAN, highlighting the need for specialized models. We observed a noticeable performance improvement when incorporating the RAG-based ORANSight pipeline, with a Macro Accuracy of 0.784 and a Weighted Accuracy of 0.776, which was on average 21.55% and 22.59% better than the other tested LLMs.
Authors: Yuwei Fang, Willi Menapace, Aliaksandr Siarohin, Tsai-Shien Chen, Kuan-Chien Wang, Ivan Skorokhodov, Graham Neubig, Sergey Tulyakov
Abstract: Existing text-to-video diffusion models rely solely on text-only encoders for their pretraining. This limitation stems from the absence of large-scale multimodal prompt video datasets, resulting in a lack of visual grounding and restricting their versatility and application in multimodal integration. To address this, we construct a large-scale multimodal prompt dataset by employing retrieval methods to pair in-context examples with the given text prompts and then utilize a two-stage training strategy to enable diverse video generation tasks within the same model. In the first stage, we propose a multimodal conditional video generation framework for pretraining on these augmented datasets, establishing a foundational model for grounded video generation. Secondly, we finetune the model from the first stage on three video generation tasks, incorporating multi-modal instructions. This process further refines the model's ability to handle diverse inputs and tasks, ensuring seamless integration of multi-modal information. After this two-stage train-ing process, VIMI demonstrates multimodal understanding capabilities, producing contextually rich and personalized videos grounded in the provided inputs, as shown in Figure 1. Compared to previous visual grounded video generation methods, VIMI can synthesize consistent and temporally coherent videos with large motion while retaining the semantic control. Lastly, VIMI also achieves state-of-the-art text-to-video generation results on UCF101 benchmark.
Authors: Luca Zancato, Arjun Seshadri, Yonatan Dukler, Aditya Golatkar, Yantao Shen, Benjamin Bowman, Matthew Trager, Alessandro Achille, Stefano Soatto
Abstract: We describe a family of architectures to support transductive inference by allowing memory to grow to a finite but a-priori unknown bound while making efficient use of finite resources for inference. Current architectures use such resources to represent data either eidetically over a finite span ("context" in Transformers), or fading over an infinite span (in State Space Models, or SSMs). Recent hybrid architectures have combined eidetic and fading memory, but with limitations that do not allow the designer or the learning process to seamlessly modulate the two, nor to extend the eidetic memory span. We leverage ideas from Stochastic Realization Theory to develop a class of models called B'MOJO to seamlessly combine eidetic and fading memory within an elementary composable module. The overall architecture can be used to implement models that can access short-term eidetic memory "in-context," permanent structural memory "in-weights," fading memory "in-state," and long-term eidetic memory "in-storage" by natively incorporating retrieval from an asynchronously updated memory. We show that Transformers, existing SSMs such as Mamba, and hybrid architectures such as Jamba are special cases of B'MOJO and describe a basic implementation, to be open sourced, that can be stacked and scaled efficiently in hardware. We test B'MOJO on transductive inference tasks, such as associative recall, where it outperforms existing SSMs and Hybrid models; as a baseline, we test ordinary language modeling where B'MOJO achieves perplexity comparable to similarly-sized Transformers and SSMs up to 1.4B parameters, while being up to 10% faster to train. Finally, we show that B'MOJO's ability to modulate eidetic and fading memory results in better inference on longer sequences tested up to 32K tokens, four-fold the length of the longest sequences seen during training.
Authors: Adriano Hernandez
Abstract: Large language models (LLMs) sometimes exhibit dangerous unintended behaviors. Finding and fixing these is challenging because the attack surface is massive -- it is not tractable to exhaustively search for all possible inputs that may elicit such behavior. One specific and particularly challenging case is that if data-poisoning-injected trojans, since there is no way to know what they are to search for them. To our knowledge, there is no generally applicable method to unlearn unknown trojans injected during pre-training. This work seeks to provide a general purpose recipe (filters) and a specific implementation (LoRA) filters that work in practice on small to medium sized models. The focus is primarily empirical, though some perplexing behavior opens the door to the fundamental question of how LLMs store and process information. Not unexpectedly, we find that our filters work best on the residual stream and the latest layers.
Authors: Yiming Zhu, Peixian Zhang, Ehsan-Ul Haq, Pan Hui, Gareth Tyson
Abstract: Harnessing the potential of large language models (LLMs) like ChatGPT can help address social challenges through inclusive, ethical, and sustainable means. In this paper, we investigate the extent to which ChatGPT can annotate data for social computing tasks, aiming to reduce the complexity and cost of undertaking web research. To evaluate ChatGPT's potential, we re-annotate seven datasets using ChatGPT, covering topics related to pressing social issues like COVID-19 misinformation, social bot deception, cyberbully, clickbait news, and the Russo-Ukrainian War. Our findings demonstrate that ChatGPT exhibits promise in handling these data annotation tasks, albeit with some challenges. Across the seven datasets, ChatGPT achieves an average annotation F1-score of 72.00%. Its performance excels in clickbait news annotation, correctly labeling 89.66% of the data. However, we also observe significant variations in performance across individual labels. Our study reveals predictable patterns in ChatGPT's annotation performance. Thus, we propose GPT-Rater, a tool to predict if ChatGPT can correctly label data for a given annotation task. Researchers can use this to identify where ChatGPT might be suitable for their annotation requirements. We show that GPT-Rater effectively predicts ChatGPT's performance. It performs best on a clickbait headlines dataset by achieving an average F1-score of 95.00%. We believe that this research opens new avenues for analysis and can reduce barriers to engaging in social computing research.
Authors: Yangyi Chen, Xingyao Wang, Hao Peng, Heng Ji
Abstract: We present SOLO, a single transformer for Scalable visiOn-Language mOdeling. Current large vision-language models (LVLMs) such as LLaVA mostly employ heterogeneous architectures that connect pre-trained visual encoders with large language models (LLMs) to facilitate visual recognition and complex reasoning. Although achieving remarkable performance with relatively lightweight training, we identify four primary scalability limitations: (1) The visual capacity is constrained by pre-trained visual encoders, which are typically an order of magnitude smaller than LLMs. (2) The heterogeneous architecture complicates the use of established hardware and software infrastructure. (3) Study of scaling laws on such architecture must consider three separate components - visual encoder, connector, and LLMs, which complicates the analysis. (4) The use of existing visual encoders typically requires following a pre-defined specification of image inputs pre-processing, for example, by reshaping inputs to fixed-resolution square images, which presents difficulties in processing and training on high-resolution images or those with unusual aspect ratio. A unified single Transformer architecture, like SOLO, effectively addresses these scalability concerns in LVLMs; however, its limited adoption in the modern context likely stems from the absence of reliable training recipes that balance both modalities and ensure stable training for billion-scale models. In this paper, we introduce the first open-source training recipe for developing SOLO, an open-source 7B LVLM using moderate academic resources. The training recipe involves initializing from LLMs, sequential pre-training on ImageNet and web-scale data, and instruction fine-tuning on our curated high-quality datasets. On extensive evaluation, SOLO demonstrates performance comparable to LLaVA-v1.5-7B, particularly excelling in visual mathematical reasoning.
Authors: Arinbjorn Kolbeinsson, Kyle O'Brien, Tianjin Huang, Shanghua Gao, Shiwei Liu, Jonathan Richard Schwarz, Anurag Vaidya, Faisal Mahmood, Marinka Zitnik, Tianlong Chen, Thomas Hartvigsen
Abstract: Test-time interventions for language models can enhance factual accuracy, mitigate harmful outputs, and improve model efficiency without costly retraining. But despite a flood of new methods, different types of interventions are largely developing independently. In practice, multiple interventions must be applied sequentially to the same model, yet we lack standardized ways to study how interventions interact. We fill this gap by introducing composable interventions, a framework to study the effects of using multiple interventions on the same language models, featuring new metrics and a unified codebase. Using our framework, we conduct extensive experiments and compose popular methods from three emerging intervention categories -- Knowledge Editing, Model Compression, and Machine Unlearning. Our results from 310 different compositions uncover meaningful interactions: compression hinders editing and unlearning, composing interventions hinges on their order of application, and popular general-purpose metrics are inadequate for assessing composability. Taken together, our findings showcase clear gaps in composability, suggesting a need for new multi-objective interventions. All of our code is public: https://github.com/hartvigsen-group/composable-interventions.
URLs: https://github.com/hartvigsen-group/composable-interventions.
Authors: Melanie Subbiah, Faisal Ladhak, Akankshya Mishra, Griffin Adams, Lydia B. Chilton, Kathleen McKeown
Abstract: Human evaluation has been the gold standard for checking faithfulness in abstractive summarization. However, with a challenging source domain like narrative, multiple annotators can agree a summary is faithful, while missing details that are obvious errors only once pointed out. We therefore introduce a new dataset, STORYSUMM, comprising LLM summaries of short stories with localized faithfulness labels and error explanations. This benchmark is for evaluation methods, testing whether a given method can detect challenging inconsistencies. Using this dataset, we first show that any one human annotation protocol is likely to miss inconsistencies, and we advocate for pursuing a range of methods when establishing ground truth for a summarization dataset. We finally test recent automatic metrics and find that none of them achieve more than 70% balanced accuracy on this task, demonstrating that it is a challenging benchmark for future work in faithfulness evaluation.
Authors: Rachneet Kaur, Zhen Zeng, Tucker Balch, Manuela Veloso
Abstract: Recent advancements in language modeling have shown promising results when applied to time series data. In particular, fine-tuning pre-trained large language models (LLMs) for time series classification tasks has achieved state-of-the-art (SOTA) performance on standard benchmarks. However, these LLM-based models have a significant drawback due to the large model size, with the number of trainable parameters in the millions. In this paper, we propose an alternative approach to leveraging the success of language modeling in the time series domain. Instead of fine-tuning LLMs, we utilize a language embedding model to embed time series and then pair the embeddings with a simple classification head composed of convolutional neural networks (CNN) and multilayer perceptron (MLP). We conducted extensive experiments on well-established time series classification benchmark datasets. We demonstrated LETS-C not only outperforms the current SOTA in classification accuracy but also offers a lightweight solution, using only 14.5% of the trainable parameters on average compared to the SOTA model. Our findings suggest that leveraging language encoders to embed time series data, combined with a simple yet effective classification head, offers a promising direction for achieving high-performance time series classification while maintaining a lightweight model architecture.
Authors: Shouchang Guo, Sonam Damani, Keng-hao Chang
Abstract: Ads relevance models are crucial in determining the relevance between user search queries and ad offers, often framed as a classification problem. The complexity of modeling increases significantly with multiple ad types and varying scenarios that exhibit both similarities and differences. In this work, we introduce a novel multi-faceted attention model that performs task aware feature combination and cross task interaction modeling. Our technique formulates the feature combination problem as "language" modeling with auto-regressive attentions across both feature and task dimensions. Specifically, we introduce a new dimension of task ID encoding for task representations, thereby enabling precise relevance modeling across diverse ad scenarios with substantial improvement in generality capability for unseen tasks. We demonstrate that our model not only effectively handles the increased computational and maintenance demands as scenarios proliferate, but also outperforms generalized DNN models and even task-specific models across a spectrum of ad applications using a single unified model.
Authors: David Gimeno-G\'omez, Carlos-D. Mart\'inez-Hinarejos
Abstract: Recent advances in Audio-Visual Speech Recognition (AVSR) have led to unprecedented achievements in the field, improving the robustness of this type of system in adverse, noisy environments. In most cases, this task has been addressed through the design of models composed of two independent encoders, each dedicated to a specific modality. However, while recent works have explored unified audio-visual encoders, determining the optimal cross-modal architecture remains an ongoing challenge. Furthermore, such approaches often rely on models comprising vast amounts of parameters and high computational cost training processes. In this paper, we aim to bridge this research gap by introducing a novel audio-visual framework. Our proposed method constitutes, to the best of our knowledge, the first attempt to harness the flexibility and interpretability offered by encoder architectures, such as the Branchformer, in the design of parameter-efficient AVSR systems. To be more precise, the proposed framework consists of two steps: first, estimating audio- and video-only systems, and then designing a tailored audio-visual unified encoder based on the layer-level branch scores provided by the modality-specific models. Extensive experiments on English and Spanish AVSR benchmarks covering multiple data conditions and scenarios demonstrated the effectiveness of our proposed method. Results reflect how our tailored AVSR system is able to reach state-of-the-art recognition rates while significantly reducing the model complexity w.r.t. the prevalent approach in the field. Code and pre-trained models are available at https://github.com/david-gimeno/tailored-avsr.
Authors: Mingjia Yin, Chuhan Wu, Yufei Wang, Hao Wang, Wei Guo, Yasheng Wang, Yong Liu, Ruiming Tang, Defu Lian, Enhong Chen
Abstract: Data is the cornerstone of large language models (LLMs), but not all data is useful for model learning. Carefully selected data can better elicit the capabilities of LLMs with much less computational overhead. Most methods concentrate on evaluating the quality of individual samples in data selection, while the combinatorial effects among samples are neglected. Even if each sample is of perfect quality, their combinations may be suboptimal in teaching LLMs due to their intrinsic homogeneity or contradiction. In this paper, we aim to uncover the underlying relationships between LLM performance and data selection. Inspired by the information compression nature of LLMs, we uncover an ``entropy law'' that connects LLM performance with data compression ratio and first-epoch training loss, which reflect the information redundancy of a dataset and the mastery of inherent knowledge encoded in this dataset, respectively. Through both theoretical deduction and empirical evaluation, we find that model performance is negatively correlated to the compression ratio of training data, which usually yields a lower training loss. Based on the findings of the entropy law, we propose a quite efficient and universal data selection method named \textbf{ZIP} for training LLMs, which aim to prioritize data subsets exhibiting a low compression ratio. Based on a multi-stage algorithm that selects diverse data in a greedy manner, we can obtain a good data subset with satisfactory diversity. Extensive experiments have been conducted to validate the entropy law and the superiority of ZIP across different LLM backbones and alignment stages. We also present an interesting application of entropy law that can detect potential performance risks at the beginning of model training.
Authors: Mengjie Qian, Siyuan Tang, Rao Ma, Kate M. Knill, Mark J. F. Gales
Abstract: Foundation ASR models often support many languages, e.g. 100 languages in Whisper. However, there has been limited work on integrating an additional, typically low-resource, language, while maintaining performance on the original language set. Fine-tuning, while simple, may degrade the accuracy of the original set. We compare three approaches that exploit adaptation parameters: soft language code tuning, train only the language code; soft prompt tuning, train prepended tokens; and LoRA where a small set of additional parameters are optimised. Elastic Weight Consolidation (EWC) offers an alternative compromise with the potential to maintain performance in specific target languages. Results show that direct fine-tuning yields the best performance for the new language but degrades existing language capabilities. EWC can address this issue for specific languages. If only adaptation parameters are used, the language capabilities are maintained but at the cost of performance in the new language.
Authors: Filipe Lauar, Valentin Laurent
Abstract: This study explores the transfer learning capabilities of the TrOCR architecture to Spanish. TrOCR is a transformer-based Optical Character Recognition (OCR) model renowned for its state-of-the-art performance in English benchmarks. Inspired by Li et al. assertion regarding its adaptability to multilingual text recognition, we investigate two distinct approaches to adapt the model to a new language: integrating an English TrOCR encoder with a language specific decoder and train the model on this specific language, and fine-tuning the English base TrOCR model on a new language data. Due to the scarcity of publicly available datasets, we present a resource-efficient pipeline for creating OCR datasets in any language, along with a comprehensive benchmark of the different image generation methods employed with a focus on Visual Rich Documents (VRDs). Additionally, we offer a comparative analysis of the two approaches for the Spanish language, demonstrating that fine-tuning the English TrOCR on Spanish yields superior recognition than the language specific decoder for a fixed dataset size. We evaluate our model employing character and word error rate metrics on a public available printed dataset, comparing the performance against other open-source and cloud OCR spanish models. As far as we know, these resources represent the best open-source model for OCR in Spanish. The Spanish TrOCR models are publicly available on HuggingFace [20] and the code to generate the dataset is available on Github [25].
Authors: Wai Man Si, Michael Backes, Yang Zhang
Abstract: In-context learning (ICL) is a recent advancement in the capabilities of large language models (LLMs). This feature allows users to perform a new task without updating the model. Concretely, users can address tasks during the inference time by conditioning on a few input-label pair demonstrations along with the test input. It is different than the conventional fine-tuning paradigm and offers more flexibility. However, this capability also introduces potential issues. For example, users may use the model on any data without restriction, such as performing tasks with improper or sensitive content, which might violate the model policy or conflict with the model owner's interests. As a model owner, it is crucial to establish a mechanism to control the model's behavior under ICL, depending on the model owner's requirements for various content. To this end, we introduce the concept of "applicability authorization" tailored for LLMs, particularly for ICL behavior, and propose a simple approach, ICLGuard. It is a fine-tuning framework designed to allow the model owner to regulate ICL behavior on different data. ICLGuard preserves the original LLM and fine-tunes only a minimal set of additional trainable parameters to "guard" the LLM. Empirical results show that the guarded LLM can deactivate its ICL ability on target data without affecting its ICL ability on other data and its general functionality across all data.
Authors: Yi-Cheng Lin, Tzu-Quan Lin, Chih-Kai Yang, Ke-Han Lu, Wei-Chih Chen, Chun-Yi Kuan, Hung-yi Lee
Abstract: Speech Integrated Large Language Models (SILLMs) combine large language models with speech perception to perform diverse tasks, such as emotion recognition to speaker verification, demonstrating universal audio understanding capability. However, these models may amplify biases present in training data, potentially leading to biased access to information for marginalized groups. This work introduces a curated spoken bias evaluation toolkit and corresponding dataset. We evaluate gender bias in SILLMs across four semantic-related tasks: speech-to-text translation (STT), spoken coreference resolution (SCR), spoken sentence continuation (SSC), and spoken question answering (SQA). Our analysis reveals that bias levels are language-dependent and vary with different evaluation methods. Our findings emphasize the necessity of employing multiple approaches to comprehensively assess biases in SILLMs, providing insights for developing fairer SILLM systems.
Authors: Yu-An Liu, Ruqing Zhang, Jiafeng Guo, Maarten de Rijke, Yixing Fan, Xueqi Cheng
Abstract: Recent advances in neural information retrieval (IR) models have significantly enhanced their effectiveness over various IR tasks. The robustness of these models, essential for ensuring their reliability in practice, has also garnered significant attention. With a wide array of research on robust IR being proposed, we believe it is the opportune moment to consolidate the current status, glean insights from existing methodologies, and lay the groundwork for future development. We view the robustness of IR to be a multifaceted concept, emphasizing its necessity against adversarial attacks, out-of-distribution (OOD) scenarios and performance variance. With a focus on adversarial and OOD robustness, we dissect robustness solutions for dense retrieval models (DRMs) and neural ranking models (NRMs), respectively, recognizing them as pivotal components of the neural IR pipeline. We provide an in-depth discussion of existing methods, datasets, and evaluation metrics, shedding light on challenges and future directions in the era of large language models. To the best of our knowledge, this is the first comprehensive survey on the robustness of neural IR models, and we will also be giving our first tutorial presentation at SIGIR 2024 \url{https://sigir2024-robust-information-retrieval.github.io}. Along with the organization of existing work, we introduce a Benchmark for robust IR (BestIR), a heterogeneous evaluation benchmark for robust neural information retrieval, which is publicly available at \url{https://github.com/Davion-Liu/BestIR}. We hope that this study provides useful clues for future research on the robustness of IR models and helps to develop trustworthy search engines \url{https://github.com/Davion-Liu/Awesome-Robustness-in-Information-Retrieval}.
URLs: https://sigir2024-robust-information-retrieval.github.io, https://github.com/Davion-Liu/BestIR, https://github.com/Davion-Liu/Awesome-Robustness-in-Information-Retrieval
Authors: Amey Agrawal, Anmol Agarwal, Nitin Kedia, Jayashree Mohan, Souvik Kundu, Nipun Kwatra, Ramachandran Ramjee, Alexey Tumanov
Abstract: Serving large language models (LLMs) in production can incur substantial costs, which has prompted recent advances in inference system optimizations. Today, these systems are evaluated against conventional latency and throughput metrics (eg. TTFT, TBT, Normalised Latency and TPOT). However, these metrics fail to fully capture the nuances of LLM inference, leading to an incomplete assessment of user-facing performance crucial for real-time applications such as chat and translation. In this paper, we first identify the pitfalls of current performance metrics in evaluating LLM inference systems. We then propose Metron, a comprehensive performance evaluation framework that includes fluidity-index -- a novel metric designed to reflect the intricacies of the LLM inference process and its impact on real-time user experience. Finally, we evaluate various existing open-source platforms and model-as-a-service offerings using Metron, discussing their strengths and weaknesses. Metron is available at https://github.com/project-metron/metron.
Authors: Nikita Dhawan, Leonardo Cotta, Karen Ullrich, Rahul G. Krishnan, Chris J. Maddison
Abstract: Knowing the effect of an intervention is critical for human decision-making, but current approaches for causal effect estimation rely on manual data collection and structuring, regardless of the causal assumptions. This increases both the cost and time-to-completion for studies. We show how large, diverse observational text data can be mined with large language models (LLMs) to produce inexpensive causal effect estimates under appropriate causal assumptions. We introduce NATURAL, a novel family of causal effect estimators built with LLMs that operate over datasets of unstructured text. Our estimators use LLM conditional distributions (over variables of interest, given the text data) to assist in the computation of classical estimators of causal effect. We overcome a number of technical challenges to realize this idea, such as automating data curation and using LLMs to impute missing information. We prepare six (two synthetic and four real) observational datasets, paired with corresponding ground truth in the form of randomized trials, which we used to systematically evaluate each step of our pipeline. NATURAL estimators demonstrate remarkable performance, yielding causal effect estimates that fall within 3 percentage points of their ground truth counterparts, including on real-world Phase 3/4 clinical trials. Our results suggest that unstructured text data is a rich source of causal effect information, and NATURAL is a first step towards an automated pipeline to tap this resource.
Authors: Daiqing Wu, Dongbao Yang, Huawen Shen, Can Ma, Yu Zhou
Abstract: With the proliferation of social media posts in recent years, the need to detect sentiments in multimodal (image-text) content has grown rapidly. Since posts are user-generated, the image and text from the same post can express different or even contradictory sentiments, leading to potential \textbf{sentiment discrepancy}. However, existing works mainly adopt a single-branch fusion structure that primarily captures the consistent sentiment between image and text. The ignorance or implicit modeling of discrepant sentiment results in compromised unimodal encoding and limited performances. In this paper, we propose a semantics Completion and Decomposition (CoDe) network to resolve the above issue. In the semantics completion module, we complement image and text representations with the semantics of the OCR text embedded in the image, helping bridge the sentiment gap. In the semantics decomposition module, we decompose image and text representations with exclusive projection and contrastive learning, thereby explicitly capturing the discrepant sentiment between modalities. Finally, we fuse image and text representations by cross-attention and combine them with the learned discrepant sentiment for final classification. Extensive experiments conducted on four multimodal sentiment datasets demonstrate the superiority of CoDe against SOTA methods.
Authors: Tao Ni, Qing Wang, Gabriela Ferraro
Abstract: Extracting multiple relations from text sentences is still a challenge for current Open Relation Extraction (Open RE) tasks. In this paper, we develop several Open RE models based on the bidirectional LSTM-CRF (BiLSTM-CRF) neural network and different contextualized word embedding methods. We also propose a new tagging scheme to solve overlapping problems and enhance models' performance. From the evaluation results and comparisons between models, we select the best combination of tagging scheme, word embedder, and BiLSTM-CRF network to achieve an Open RE model with a remarkable extracting ability on multiple-relation sentences.
Authors: Jiajia Li, Lu Yang, Letian Peng, Shitou Zhang, Ping Wang, Zuchao Li, Hai Zhao
Abstract: In recent years, machine learning - particularly deep learning - has significantly impacted the field of information management. While several strategies have been proposed to restrict models from learning and memorizing sensitive information from raw texts, this paper suggests a more linguistically-grounded approach to distort texts while maintaining semantic integrity. To this end, we leverage Neighboring Distribution Divergence, a novel metric to assess the preservation of semantic meaning during distortion. Building on this metric, we present two distinct frameworks for semantic-preserving distortion: a generative approach and a substitutive approach. Our evaluations across various tasks, including named entity recognition, constituency parsing, and machine reading comprehension, affirm the plausibility and efficacy of our distortion technique in personal privacy protection. We also test our method against attribute attacks in three privacy-focused assignments within the NLP domain, and the findings underscore the simplicity and efficacy of our data-based improvement approach over structural improvement approaches. Moreover, we explore privacy protection in a specific medical information management scenario, showing our method effectively limits sensitive data memorization, underscoring its practicality.
Authors: Vittoria Dentella, Fritz Guenther, Elliot Murphy, Gary Marcus, Evelina Leivada
Abstract: Large Language Models (LLMs) are recruited in applications that span from clinical assistance and legal support to question answering and education. Their success in specialized tasks has led to the claim that they possess human-like linguistic capabilities related to compositional understanding and reasoning. Yet, reverse-engineering is bound by Moravec's Paradox, according to which easy skills are hard. We systematically assess 7 state-of-the-art models on a novel benchmark. Models answered a series of comprehension questions, each prompted multiple times in two settings, permitting one-word or open-length replies. Each question targets a short text featuring high-frequency linguistic constructions. To establish a baseline for achieving human-like performance, we tested 400 humans on the same prompts. Based on a dataset of n=26,680 datapoints, we discovered that LLMs perform at chance accuracy and waver considerably in their answers. Quantitatively, the tested models are outperformed by humans, and qualitatively their answers showcase distinctly non-human errors in language understanding. We interpret this evidence as suggesting that, despite their usefulness in various tasks, current AI models fall short of understanding language in a way that matches humans, and we argue that this may be due to their lack of a compositional operator for regulating grammatical and semantic information.
Authors: Thilo Hagendorff
Abstract: Large language models (LLMs) are currently at the forefront of intertwining AI systems with human communication and everyday life. Due to rapid technological advances and their extreme versatility, LLMs nowadays have millions of users and are at the cusp of being the main go-to technology for information retrieval, content generation, problem-solving, etc. Therefore, it is of great importance to thoroughly assess and scrutinize their capabilities. Due to increasingly complex and novel behavioral patterns in current LLMs, this can be done by treating them as participants in psychology experiments that were originally designed to test humans. For this purpose, the paper introduces a new field of research called "machine psychology". The paper outlines how different subfields of psychology can inform behavioral tests for LLMs. It defines methodological standards for machine psychology research, especially by focusing on policies for prompt designs. Additionally, it describes how behavioral patterns discovered in LLMs are to be interpreted. In sum, machine psychology aims to discover emergent abilities in LLMs that cannot be detected by most traditional natural language processing benchmarks.
Authors: Hongyuan Lu, Haoyang Huang, Dongdong Zhang, Haoran Yang, Wai Lam, Furu Wei
Abstract: Large language models (LLMs) have shown surprisingly good performance in multilingual neural machine translation (MNMT) even when trained without parallel data. Yet, despite the fact that the amount of training data is gigantic, they still struggle with translating rare words, particularly for low-resource languages. Even worse, it is usually unrealistic to retrieve relevant demonstrations for in-context learning with low-resource languages on LLMs, which restricts the practical use of LLMs for translation -- how should we mitigate this problem? To this end, we present a novel method, CoD, which augments LLMs with prior knowledge with the chains of multilingual dictionaries for a subset of input words to elicit translation abilities for LLMs. Extensive experiments indicate that augmenting ChatGPT with CoD elicits large gains by up to 13x chrF++ points for MNMT (3.08 to 42.63 for English to Serbian written in Cyrillic script) on FLORES-200 full devtest set. We further demonstrate the importance of chaining the multilingual dictionaries, as well as the superiority of CoD to few-shot demonstration for low-resource languages.
Authors: Helan Hu, Shuzheng Si, Haozhe Zhao, Shuang Zeng, Kaikai An, Zefan Cai, Baobao Chang
Abstract: Distantly-Supervised Named Entity Recognition (DS-NER) is widely used in real-world scenarios. It can effectively alleviate the burden of annotation by matching entities in existing knowledge bases with snippets in the text but suffer from the label noise. Recent works attempt to adopt the teacher-student framework to gradually refine the training labels and improve the overall robustness. However, these teacher-student methods achieve limited performance because the poor calibration of the teacher network produces incorrectly pseudo-labeled samples, leading to error propagation. Therefore, we propose: (1) Uncertainty-Aware Teacher Learning that leverages the prediction uncertainty to reduce the number of incorrect pseudo labels in the self-training stage; (2) Student-Student Collaborative Learning that allows the transfer of reliable labels between two student networks instead of indiscriminately relying on all pseudo labels from its teacher, and further enables a full exploration of mislabeled samples rather than simply filtering unreliable pseudo-labeled samples. We evaluate our proposed method on five DS-NER datasets, demonstrating that our method is superior to the state-of-the-art DS-NER methods.
Authors: Chenyang Lyu, Minghao Wu, Alham Fikri Aji
Abstract: Large Language Models (LLMs) have demonstrated remarkable capabilities across various applications, fundamentally reshaping the landscape of natural language processing (NLP) research. However, recent evaluation frameworks often rely on the output probabilities of LLMs for predictions, primarily due to computational constraints, diverging from real-world LLM usage scenarios. While widely employed, the efficacy of these probability-based evaluation strategies remains an open research question. This study aims to scrutinize the validity of such probability-based evaluation methods within the context of using LLMs for Multiple Choice Questions (MCQs), highlighting their inherent limitations. Our empirical investigation reveals that the prevalent probability-based evaluation method inadequately aligns with generation-based prediction. Furthermore, current evaluation frameworks typically assess LLMs through predictive tasks based on output probabilities rather than directly generating responses, owing to computational limitations. We illustrate that these probability-based approaches do not effectively correspond with generative predictions. The outcomes of our study can enhance the understanding of LLM evaluation methodologies and provide insights for future research in this domain.
Authors: Melanie Subbiah, Sean Zhang, Lydia B. Chilton, Kathleen McKeown
Abstract: We evaluate recent Large Language Models (LLMs) on the challenging task of summarizing short stories, which can be lengthy, and include nuanced subtext or scrambled timelines. Importantly, we work directly with authors to ensure that the stories have not been shared online (and therefore are unseen by the models), and to obtain informed evaluations of summary quality using judgments from the authors themselves. Through quantitative and qualitative analysis grounded in narrative theory, we compare GPT-4, Claude-2.1, and LLama-2-70B. We find that all three models make faithfulness mistakes in over 50% of summaries and struggle with specificity and interpretation of difficult subtext. We additionally demonstrate that LLM ratings and other automatic metrics for summary quality do not correlate well with the quality ratings from the writers.
Authors: Siqi Fan, Xin Jiang, Xiang Li, Xuying Meng, Peng Han, Shuo Shang, Aixin Sun, Yequan Wang, Zhongyuan Wang
Abstract: Due to the large number of parameters, the inference phase of Large Language Models (LLMs) is resource-intensive. However, not all requests posed to LLMs are equally difficult to handle. Through analysis, we show that for some tasks, LLMs can achieve results comparable to the final output at some intermediate layers. That is, not all layers of LLMs are necessary during inference. If we can predict at which layer the inferred results match the final results (produced by evaluating all layers), we could significantly reduce the inference cost. To this end, we propose a simple yet effective algorithm named AdaInfer to adaptively terminate the inference process for an input instance. AdaInfer relies on easily obtainable statistical features and classic classifiers like SVM. Experiments on well-known LLMs like the Llama2 series and OPT, show that AdaInfer can achieve an average of 17.8% pruning ratio, and up to 43% on sentiment tasks, with nearly no performance drop (<1%). Because AdaInfer does not alter LLM parameters, the LLMs incorporated with AdaInfer maintain generalizability across tasks.
Authors: Yusheng Liao, Yutong Meng, Yuhao Wang, Hongcheng Liu, Yanfeng Wang, Yu Wang
Abstract: Large Language Models (LLMs) have demonstrated remarkable proficiency in human interactions, yet their application within the medical field remains insufficiently explored. Previous works mainly focus on the performance of medical knowledge with examinations, which is far from the realistic scenarios, falling short in assessing the abilities of LLMs on clinical tasks. In the quest to enhance the application of Large Language Models (LLMs) in healthcare, this paper introduces the Automated Interactive Evaluation (AIE) framework and the State-Aware Patient Simulator (SAPS), targeting the gap between traditional LLM evaluations and the nuanced demands of clinical practice. Unlike prior methods that rely on static medical knowledge assessments, AIE and SAPS provide a dynamic, realistic platform for assessing LLMs through multi-turn doctor-patient simulations. This approach offers a closer approximation to real clinical scenarios and allows for a detailed analysis of LLM behaviors in response to complex patient interactions. Our extensive experimental validation demonstrates the effectiveness of the AIE framework, with outcomes that align well with human evaluations, underscoring its potential to revolutionize medical LLM testing for improved healthcare delivery.
Authors: Hunter McNichols, Jaewook Lee, Stephen Fancsali, Steve Ritter, Andrew Lan
Abstract: Intelligent Tutoring Systems (ITSs) often contain an automated feedback component, which provides a predefined feedback message to students when they detect a predefined error. To such a feedback component, we often resort to template-based approaches. These approaches require significant effort from human experts to detect a limited number of possible student errors and provide corresponding feedback. This limitation is exemplified in open-ended math questions, where there can be a large number of different incorrect errors. In our work, we examine the capabilities of large language models (LLMs) to generate feedback for open-ended math questions, similar to that of an established ITS that uses a template-based approach. We fine-tune both open-source and proprietary LLMs on real student responses and corresponding ITS-provided feedback. We measure the quality of the generated feedback using text similarity metrics. We find that open-source and proprietary models both show promise in replicating the feedback they see during training, but do not generalize well to previously unseen student errors. These results suggest that despite being able to learn the formatting of feedback, LLMs are not able to fully understand mathematical errors made by students.
Authors: Aditya Joshi, Jake Renzella, Pushpak Bhattacharyya, Saurav Jha, Xiangyu Zhang
Abstract: While deep learning approaches represent the state-of-the-art of natural language processing (NLP) today, classical algorithms and approaches still find a place in NLP textbooks and courses of recent years. This paper discusses the perspectives of conveners of two introductory NLP courses taught in Australia and India, and examines how classical and deep learning approaches can be balanced within the lecture plan and assessments of the courses. We also draw parallels with the objects-first and objects-later debate in CS1 education. We observe that teaching classical approaches adds value to student learning by building an intuitive understanding of NLP problems, potential solutions, and even deep learning models themselves. Despite classical approaches not being state-of-the-art, the paper makes a case for their inclusion in NLP courses today.
Authors: Tingyu Xia, Bowen Yu, Yuan Wu, Yi Chang, Chang Zhou
Abstract: In this paper, we initiate our discussion by demonstrating how Large Language Models (LLMs), when tasked with responding to queries, display a more even probability distribution in their answers if they are more adept, as opposed to their less skilled counterparts. Expanding on this foundational insight, we propose a new self-evaluation method ProbDiff for assessing the efficacy of various LLMs. This approach obviates the necessity for an additional evaluation model or the dependence on external, proprietary models like GPT-4 for judgment. It uniquely utilizes the LLMs being tested to compute the probability discrepancy between the initial response and its revised versions. A higher discrepancy for a given query between two LLMs indicates a relatively weaker capability. Our findings reveal that ProbDiff achieves results on par with those obtained from evaluations based on GPT-4, spanning a range of scenarios that include natural language generation (NLG) tasks such as translation, summarization, and our proposed Xiaohongshu blog writing task, and benchmarks for LLM evaluation like AlignBench, MT-Bench, and AlpacaEval, across LLMs of varying magnitudes.
Authors: Ziyun Cui, Chang Lei, Wen Wu, Yinan Duan, Diyang Qu, Ji Wu, Runsen Chen, Chao Zhang
Abstract: The early detection of suicide risk is important since it enables the intervention to prevent potential suicide attempts. This paper studies the automatic detection of suicide risk based on spontaneous speech from adolescents, and collects a Mandarin dataset with 15 hours of suicide speech from more than a thousand adolescents aged from ten to eighteen for our experiments. To leverage the diverse acoustic and linguistic features embedded in spontaneous speech, both the Whisper speech model and textual large language models (LLMs) are used for suicide risk detection. Both all-parameter finetuning and parameter-efficient finetuning approaches are used to adapt the pre-trained models for suicide risk detection, and multiple audio-text fusion approaches are evaluated to combine the representations of Whisper and the LLM. The proposed system achieves a detection accuracy of 0.807 and an F1-score of 0.846 on the test set with 119 subjects, indicating promising potential for real suicide risk detection applications.
Authors: Salomon Kabongo, Jennifer D'Souza, S\"oren Auer
Abstract: The rapid advancements in Large Language Models (LLMs) have opened new avenues for automating complex tasks in AI research. This paper investigates the efficacy of different LLMs-Mistral 7B, Llama-2, GPT-4-Turbo and GPT-4.o in extracting leaderboard information from empirical AI research articles. We explore three types of contextual inputs to the models: DocTAET (Document Title, Abstract, Experimental Setup, and Tabular Information), DocREC (Results, Experiments, and Conclusions), and DocFULL (entire document). Our comprehensive study evaluates the performance of these models in generating (Task, Dataset, Metric, Score) quadruples from research papers. The findings reveal significant insights into the strengths and limitations of each model and context type, providing valuable guidance for future AI research automation efforts.
Authors: Chih-Kai Yang, Kuan-Po Huang, Hung-yi Lee
Abstract: This research explores how the information of prompts interacts with the high-performing speech recognition model, Whisper. We compare its performances when prompted by prompts with correct information and those corrupted with incorrect information. Our results unexpectedly show that Whisper may not understand the textual prompts in a human-expected way. Additionally, we find that performance improvement is not guaranteed even with stronger adherence to the topic information in textual prompts. It is also noted that English prompts generally outperform Mandarin ones on datasets of both languages, likely due to differences in training data distributions for these languages despite the mismatch with pre-training scenarios. Conversely, we discover that Whisper exhibits awareness of misleading information in language tokens by ignoring incorrect language tokens and focusing on the correct ones. In sum, We raise insightful questions about Whisper's prompt understanding and reveal its counter-intuitive behaviors. We encourage further studies.
Authors: Zhewen Shen, Aditya Joshi, Ruey-Cheng Chen
Abstract: Children from bilingual backgrounds benefit from interactions with parents and teachers to re-acquire their heritage language. In this paper, we investigate how this insight from behavioral study can be incorporated into the learning of small-scale language models. We introduce BAMBINO-LM, a continual pre-training strategy for BabyLM that uses a novel combination of alternation and PPO-based perplexity reward induced from a parent Italian model. Upon evaluation on zero-shot classification tasks for English and Italian, BAMBINO-LM improves the Italian language capability of a BabyLM baseline. Our ablation analysis demonstrates that employing both the alternation strategy and PPO-based modeling is key to this effectiveness gain. We also show that, as a side effect, the proposed method leads to a similar degradation in L1 effectiveness as human children would have had in an equivalent learning scenario. Through its modeling and findings, BAMBINO-LM makes a focused contribution to the pre-training of small-scale language models by first developing a human-inspired strategy for pre-training and then showing that it results in behaviours similar to that of humans.
Authors: Steffi Chern, Zhulin Hu, Yuqing Yang, Ethan Chern, Yuan Guo, Jiahe Jin, Binjie Wang, Pengfei Liu
Abstract: Previous works on Large Language Models (LLMs) have mainly focused on evaluating their helpfulness or harmlessness. However, honesty, another crucial alignment criterion, has received relatively less attention. Dishonest behaviors in LLMs, such as spreading misinformation and defrauding users, present severe risks that intensify as these models approach superintelligent levels. Enhancing honesty in LLMs addresses critical limitations and helps uncover latent capabilities that are not readily expressed. This underscores the urgent need for reliable methods and benchmarks to effectively ensure and evaluate the honesty of LLMs. In this paper, we introduce BeHonest, a pioneering benchmark specifically designed to assess honesty in LLMs comprehensively. BeHonest evaluates three essential aspects of honesty: awareness of knowledge boundaries, avoidance of deceit, and consistency in responses. Building on this foundation, we designed 10 scenarios to evaluate and analyze 9 popular LLMs on the market, including both closed-source and open-source models from different model families with varied model sizes. Our findings indicate that there is still significant room for improvement in the honesty of LLMs. We encourage the AI community to prioritize honesty alignment in these models, which can harness their full potential to benefit society while preventing them from causing harm through deception or inconsistency. Our benchmark and code can be found at: \url{https://github.com/GAIR-NLP/BeHonest}.
Authors: Jing Yang, Yu Zhao, Linyao Yang, 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 design temporal event reasoning in the form of masked language modeling as auxiliary tasks 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 on designing and manufacturing printed circuit boards is provided to validate its effectiveness in crowdsourcing scenarios.
Authors: Zhifeng Kong, Sang-gil Lee, Deepanway Ghosal, Navonil Majumder, Ambuj Mehrish, Rafael Valle, Soujanya Poria, Bryan Catanzaro
Abstract: It is an open challenge to obtain high quality training data, especially captions, for text-to-audio models. Although prior methods have leveraged \textit{text-only language models} to augment and improve captions, such methods have limitations related to scale and coherence between audio and captions. In this work, we propose an audio captioning pipeline that uses an \textit{audio language model} to synthesize accurate and diverse captions for audio at scale. We leverage this pipeline to produce a dataset of synthetic captions for AudioSet, named \texttt{AF-AudioSet}, and then evaluate the benefit of pre-training text-to-audio models on these synthetic captions. Through systematic evaluations on AudioCaps and MusicCaps, we find leveraging our pipeline and synthetic captions leads to significant improvements on audio generation quality, achieving a new \textit{state-of-the-art}.
Authors: Linqing Chen, Weilei Wang, Zilong Bai, Peng Xu, Yan Fang, Jie Fang, Wentao Wu, Lizhi Zhou, Ruiji Zhang, Yubin Xia, Chaobo Xu, Ran Hu, Licong Xu, Qijun Cai, Haoran Hua, Jing Sun, Jin Liu, Tian Qiu, Haowen Liu, Meng Hu, Xiuwen Li, Fei Gao, Yufu Wang, Lin Tie, Chaochao Wang, Jianping Lu, Cheng Sun, Yixin Wang, Shengjie Yang, Yuancheng Li, Lu Jin, Lisha Zhang, Fu Bian, Zhongkai Ye, Lidong Pei, Changyang Tu
Abstract: Large language models (LLMs) have revolutionized Natural Language Processing (NLP) by minimizing the need for complex feature engineering. However, the application of LLMs in specialized domains like biopharmaceuticals and chemistry remains largely unexplored. These fields are characterized by intricate terminologies, specialized knowledge, and a high demand for precision areas where general purpose LLMs often fall short. In this study, we introduce PharmaGPT, a suite of domain specilized LLMs with 13 billion and 70 billion parameters, specifically trained on a comprehensive corpus tailored to the Bio-Pharmaceutical and Chemical domains. Our evaluation shows that PharmaGPT surpasses existing general models on specific-domain benchmarks such as NAPLEX, demonstrating its exceptional capability in domain-specific tasks. Remarkably, this performance is achieved with a model that has only a fraction, sometimes just one-tenth-of the parameters of general-purpose large models. This advancement establishes a new benchmark for LLMs in the bio-pharmaceutical and chemical fields, addressing the existing gap in specialized language modeling. It also suggests a promising path for enhanced research and development, paving the way for more precise and effective NLP applications in these areas.
Authors: Seungju Han, Kavel Rao, Allyson Ettinger, Liwei Jiang, Bill Yuchen Lin, Nathan Lambert, Yejin Choi, Nouha Dziri
Abstract: We introduce WildGuard -- an open, light-weight moderation tool for LLM safety that achieves three goals: (1) identifying malicious intent in user prompts, (2) detecting safety risks of model responses, and (3) determining model refusal rate. Together, WildGuard serves the increasing needs for automatic safety moderation and evaluation of LLM interactions, providing a one-stop tool with enhanced accuracy and broad coverage across 13 risk categories. While existing open moderation tools such as Llama-Guard2 score reasonably well in classifying straightforward model interactions, they lag far behind a prompted GPT-4, especially in identifying adversarial jailbreaks and in evaluating models' refusals, a key measure for evaluating safety behaviors in model responses. To address these challenges, we construct WildGuardMix, a large-scale and carefully balanced multi-task safety moderation dataset with 92K labeled examples that cover vanilla (direct) prompts and adversarial jailbreaks, paired with various refusal and compliance responses. WildGuardMix is a combination of WildGuardTrain, the training data of WildGuard, and WildGuardTest, a high-quality human-annotated moderation test set with 5K labeled items covering broad risk scenarios. Through extensive evaluations on WildGuardTest and ten existing public benchmarks, we show that WildGuard establishes state-of-the-art performance in open-source safety moderation across all the three tasks compared to ten strong existing open-source moderation models (e.g., up to 26.4% improvement on refusal detection). Importantly, WildGuard matches and sometimes exceeds GPT-4 performance (e.g., up to 3.9% improvement on prompt harmfulness identification). WildGuard serves as a highly effective safety moderator in an LLM interface, reducing the success rate of jailbreak attacks from 79.8% to 2.4%.
Authors: Hwanjun Song, Hang Su, Igor Shalyminov, Jason Cai, Saab Mansour
Abstract: Automated evaluation is crucial for streamlining text summarization benchmarking and model development, given the costly and time-consuming nature of human evaluation. Traditional methods like ROUGE do not correlate well with human judgment, while recently proposed LLM-based metrics provide only summary-level assessment using Likert-scale scores. This limits deeper model analysis, e.g., we can only assign one hallucination score at the summary level, while at the sentence level, we can count sentences containing hallucinations. To remedy those limitations, we propose FineSurE, a fine-grained evaluator specifically tailored for the summarization task using large language models (LLMs). It also employs completeness and conciseness criteria, in addition to faithfulness, enabling multi-dimensional assessment. We compare various open-source and proprietary LLMs as backbones for FineSurE. In addition, we conduct extensive benchmarking of FineSurE against SOTA methods including NLI-, QA-, and LLM-based methods, showing improved performance especially on the completeness and conciseness dimensions. The code is available at https://github.com/DISL-Lab/FineSurE-ACL24.
Authors: Sayan Ghosh, Tejas Srinivasan, Swabha Swayamdipta
Abstract: Human evaluation of generated language through pairwise preference judgments is pervasive. However, under common scenarios, such as when generations from a model pair are very similar, or when stochastic decoding results in large variations in generations, it results in inconsistent preference ratings. We address these challenges by introducing a meta-evaluation measure, separability, which estimates how suitable a test instance is for pairwise preference evaluation. For a candidate test instance, separability samples multiple generations from a pair of models, and measures how distinguishable the two sets of generations are. Our experiments show that instances with high separability values yield more consistent preference ratings from both human- and auto-raters. Further, the distribution of separability allows insights into which test benchmarks are more valuable for comparing models. Finally, we incorporate separability into ELO ratings, accounting for how suitable each test instance might be for reliably ranking LLMs. Overall, separability has implications for consistent, efficient and robust preference evaluation of LLMs with both human- and auto-raters.
Authors: Linzhuang Sun, Hao Liang, Jingxuan Wei, Linkun Sun, Bihui Yu, Bin Cui, Wentao Zhang
Abstract: In recent years, with the rapid advancements in large language models (LLMs), achieving excellent empathetic response capability has become a crucial prerequisite. Consequently, managing and understanding large-scale video datasets has gained increasing importance. However, empathetic data are typically trained without any quality selection, leading to inefficient data usage and wasted computational resources. Additionally, using raw data can result in low performance in empathetic dialogues. In this work, we present Efficient-Empathy, a sensibility and rationality score-based data selection algorithm that automatically selects sensibility and rationality data while discarding low-quality data. With only the sensibility data (59% of the full dataset), our trained sensibility model efficiently achieves state-of-the-art (SoTA) performance. Furthermore, with multiple data selection hyperparameters, the sensibility model demonstrates SoTA performance, showcasing the robustness of our method. By integrating sensibility and rationality data with a MoE structure, we achieve even higher performance, demonstrating the effectiveness of our Efficient-Empathy algorithm.
Authors: Hasna Chouikhi, Manel Aloui, Cyrine Ben Hammou, Ghaith Chaabane, Haithem Kchaou, Chehir Dhaouadi
Abstract: Large language models (LLMs) have greatly impacted the natural language processing (NLP) field, particularly for the English language. These models have demonstrated capabilities in understanding and generating human-like text. The success of language models largely depends on the availability of high-quality instruction datasets, which consist of detailed task descriptions and corresponding responses that are essential for training the models to address a variety of prompts accurately. However, the availability and quality of these resources vary by language. While models perform well in English, they often need help with languages like Arabic, due to the lack of datasets for fine-tuning Arabic-specific tasks. To address this issue, we introduce InstAr-500k, a new Arabic instruction dataset created by generating and collecting content that covers several domains and instruction types. We assess this dataset by fine-tuning an open-source Gemma-7B model on several downstream tasks to improve its functionality. Based on multiple evaluations, our fine-tuned model achieves excellent performance on several Arabic NLP benchmarks. These outcomes emphasize the effectiveness of our dataset in elevating the capabilities of language models for Arabic. Our instruction dataset bridges the performance gap between English and Arabic language models by providing resources that amplify Arabic NLP development. Building on this foundation, we developed a model, GemmAr-7B-V1, specifically tuned to excel at a wide range of Arabic NLP tasks.
Authors: Mingxu Tao, Chen Zhang, Quzhe Huang, Tianyao Ma, Songfang Huang, Dongyan Zhao, Yansong Feng
Abstract: Adapting large language models (LLMs) to new languages typically involves continual pre-training (CT) followed by supervised fine-tuning (SFT). However, this CT-then-SFT approach struggles with limited data in the context of low-resource languages, failing to balance language modeling and task-solving capabilities. We thus propose model merging as an alternative for low-resource languages, combining models with distinct capabilities into a single model without additional training. We use model merging to develop task-solving LLMs for low-resource languages without SFT data in the target languages. Our experiments based on Llama-2-7B demonstrate that model merging effectively endows LLMs for low-resource languages with task-solving abilities, outperforming CT-then-SFT in scenarios with extremely scarce data. Observing performance saturation in model merging with more training tokens, we further analyze the merging process and introduce a slack variable to the model merging algorithm to mitigate the loss of important parameters, thereby enhancing performance. We hope that model merging can benefit more human languages suffering from data scarcity with its higher data efficiency.
Authors: Chengpeng Li, Guanting Dong, Mingfeng Xue, Ru Peng, Xiang Wang, Dayiheng Liu
Abstract: Large language models (LLMs) have made impressive progress in handling simple math problems, yet they still struggle with more challenging and complex mathematical tasks. In this paper, we introduce a series of LLMs that employs the Decomposition of thought with code assistance and self-correction for mathematical reasoning, dubbed as DotaMath. DotaMath models tackle complex mathematical tasks by decomposing them into simpler logical subtasks, leveraging code to solve these subtasks, obtaining fine-grained feedback from the code interpreter, and engaging in self-reflection and correction. By annotating diverse interactive tool-use trajectories and employing query evolution on GSM8K and MATH datasets, we generate an instruction fine-tuning dataset called DotaMathQA with 574K query-response pairs. We train a series of base LLMs using imitation learning on DotaMathQA, resulting in DotaMath models that achieve remarkable performance compared to open-source LLMs across various in-domain and out-of-domain benchmarks. Notably, DotaMath-deepseek-7B showcases an outstanding performance of 64.8% on the competitive MATH dataset and 86.7% on GSM8K. Besides, DotaMath-deepseek-7B maintains strong competitiveness on a series of in-domain and out-of-domain benchmarks (Avg. 80.1%). Looking forward, we anticipate that the DotaMath paradigm will open new pathways for addressing intricate mathematical problems. Our code is publicly available at https://github.com/ChengpengLi1003/DotaMath.
Authors: Salima Mdhaffar, Haroun Elleuch, Fethi Bougares, Yannick Est\`eve
Abstract: Speech encoders pretrained through self-supervised learning (SSL) have demonstrated remarkable performance in various downstream tasks, including Spoken Language Understanding (SLU) and Automatic Speech Recognition (ASR). For instance, fine-tuning SSL models for such tasks has shown significant potential, leading to improvements in the SOTA performance across challenging datasets. In contrast to existing research, this paper contributes by comparing the effectiveness of SSL approaches in the context of (i) the low-resource spoken Tunisian Arabic dialect and (ii) its combination with a low-resource SLU and ASR scenario, where only a few semantic annotations are available for fine-tuning. We conduct experiments using many SSL speech encoders on the TARIC-SLU dataset. We use speech encoders that were pre-trained on either monolingual or multilingual speech data. Some of them have also been refined without in-domain nor Tunisian data through multimodal supervised teacher-student paradigm. This study yields numerous significant findings that we are discussing in this paper.
Authors: Shangrui Nie, Michael Fromm, Charles Welch, Rebekka G\"orge, Akbar Karimi, Joan Plepi, Nazia Afsan Mowmita, Nicolas Flores-Herr, Mehdi Ali, Lucie Flek
Abstract: While preliminary findings indicate that multilingual LLMs exhibit reduced bias compared to monolingual ones, a comprehensive understanding of the effect of multilingual training on bias mitigation, is lacking. This study addresses this gap by systematically training six LLMs of identical size (2.6B parameters) and architecture: five monolingual models (English, German, French, Italian, and Spanish) and one multilingual model trained on an equal distribution of data across these languages, all using publicly available data. To ensure robust evaluation, standard bias benchmarks were automatically translated into the five target languages and verified for both translation quality and bias preservation by human annotators. Our results consistently demonstrate that multilingual training effectively mitigates bias. Moreover, we observe that multilingual models achieve not only lower bias but also superior prediction accuracy when compared to monolingual models with the same amount of training data, model architecture, and size.
Authors: Chani Jung, Dongkwan Kim, Jiho Jin, Jiseon Kim, Yeon Seonwoo, Yejin Choi, Alice Oh, Hyunwoo Kim
Abstract: While humans naturally develop theory of mind (ToM), the capability to understand other people's mental states and beliefs, state-of-the-art large language models (LLMs) underperform on simple ToM benchmarks. We posit that we can extend our understanding of LLMs' ToM abilities by evaluating key human ToM precursors -- perception inference and perception-to-belief inference -- in LLMs. We introduce two datasets, Percept-ToMi and Percept-FANToM, to evaluate these precursory inferences for ToM in LLMs by annotating characters' perceptions on ToMi and FANToM, respectively. Our evaluation of eight state-of-the-art LLMs reveals that the models generally perform well in perception inference while exhibiting limited capability in perception-to-belief inference (e.g., lack of inhibitory control). Based on these results, we present PercepToM, a novel ToM method leveraging LLMs' strong perception inference capability while supplementing their limited perception-to-belief inference. Experimental results demonstrate that PercepToM significantly enhances LLM's performance, especially in false belief scenarios.
Authors: Ping Yu, Jing Xu, Jason Weston, Ilia Kulikov
Abstract: Large language models (LLMs) can spend extra compute during inference to generate intermediate thoughts, which helps to produce better final responses. Since Chain-of-Thought (Wei et al., 2022), many such System 2 techniques have been proposed such as Rephrase and Respond (Deng et al., 2023a), System 2 Attention (Weston and Sukhbaatar, 2023) and Branch-Solve-Merge (Saha et al., 2023). In this work we investigate self-supervised methods to ``compile'' (distill) higher quality outputs from System 2 techniques back into LLM generations without intermediate reasoning token sequences, as this reasoning has been distilled into System 1. We show that several such techniques can be successfully distilled, resulting in improved results compared to the original System 1 performance, and with less inference cost than System 2. We posit that such System 2 distillation will be an important feature of future continually learning AI systems, enabling them to focus System 2 capabilities on the reasoning tasks that they cannot yet do well.
Authors: Lukas Netz, Jan Reimer, Bernhard Rumpe
Abstract: We present and evaluate a method called grammar masking, which is used to guide large language models (LLMs) toward producing syntactically correct models for a given context-free grammar. Prompt engineering methods such as few-shot learning or priming can be used to improve the chances of an LLM producing correct syntax, but the more complex the grammar, the more time-consuming and less promising these methods become. Previous work is focused primarily on the usage of either language model training or prompt engineering. In this work, a method is presented that restricts the output to a given grammar using constrained decoding to ensure the output adheres to a valid syntax. We use several DSLs built with MontiCore and task multiple LLMs to produce models with and without constrained decoding. A corresponding parser is used to confirm the syntactic correctness of each model. We show that grammar masking can dramatically improve the modeling capabilities of several LLMs, reducing the need for well-refined prompting while increasing the chance of producing correct models.
Authors: Yinheng Li, Shaofei Wang, Han Ding, Hang Chen
Abstract: Recent advances in large language models (LLMs) have opened new possibilities for artificial intelligence applications in finance. In this paper, we provide a practical survey focused on two key aspects of utilizing LLMs for financial tasks: existing solutions and guidance for adoption. First, we review current approaches employing LLMs in finance, including leveraging pretrained models via zero-shot or few-shot learning, fine-tuning on domain-specific data, and training custom LLMs from scratch. We summarize key models and evaluate their performance improvements on financial natural language processing tasks. Second, we propose a decision framework to guide financial professionals in selecting the appropriate LLM solution based on their use case constraints around data, compute, and performance needs. The framework provides a pathway from lightweight experimentation to heavy investment in customized LLMs. Lastly, we discuss limitations and challenges around leveraging LLMs in financial applications. Overall, this survey aims to synthesize the state-of-the-art and provide a roadmap for responsibly applying LLMs to advance financial AI.
Authors: Haoyu He, Zizheng Pan, Jing Liu, Jianfei Cai, Bohan Zhuang
Abstract: The paradigm of pre-training and fine-tuning has laid the foundation for deploying deep learning models. However, most fine-tuning methods are designed to meet a specific resource budget. Recently, considering diverse deployment scenarios with various resource budgets, SN-Net is introduced to quickly obtain numerous new networks (stitches) from the pre-trained models (anchors) in a model family via model stitching. Although promising, SN-Net confronts new challenges when adapting it to new target domains, including huge memory and storage requirements and a long and sub-optimal multistage adaptation process. In this work, we present a novel framework, Efficient Stitchable Task Adaptation (ESTA), to efficiently produce a palette of fine-tuned models that adhere to diverse resource constraints. Specifically, we first tailor parameter-efficient fine-tuning to share low-rank updates among the stitches while maintaining independent bias terms. In this way, we largely reduce fine-tuning memory burdens and mitigate the interference among stitches that arises in task adaptation. Furthermore, we streamline a simple yet effective one-stage deployment pipeline, which estimates the important stitches to deploy with training-time gradient statistics. By assigning higher sampling probabilities to important stitches, we also get a boosted Pareto frontier. Extensive experiments on 25 downstream visual recognition tasks demonstrate that our ESTA is capable of generating stitches with smooth accuracy-efficiency trade-offs and surpasses the direct SN-Net adaptation by remarkable margins with significantly lower training time and fewer trainable parameters. Furthermore, we demonstrate the flexibility and scalability of our ESTA framework by stitching LLMs from LLaMA family, obtaining chatbot stitches of assorted sizes. Source code is available at https://github.com/ziplab/Stitched_LLaMA
Authors: Andy Zhou, Bo Li, Haohan Wang
Abstract: Despite advances in AI alignment, large language models (LLMs) remain vulnerable to adversarial attacks or jailbreaking, in which adversaries can modify prompts to induce unwanted behavior. While some defenses have been proposed, they have not been adapted to newly proposed attacks and more challenging threat models. To address this, we propose an optimization-based objective for defending LLMs against jailbreaking attacks and an algorithm, Robust Prompt Optimization (RPO) to create robust system-level defenses. Our approach directly incorporates the adversary into the defensive objective and optimizes a lightweight and transferable suffix, enabling RPO to adapt to worst-case adaptive attacks. Our theoretical and experimental results show improved robustness to both jailbreaks seen during optimization and unknown jailbreaks, reducing the attack success rate (ASR) on GPT-4 to 6% and Llama-2 to 0% on JailbreakBench, setting the state-of-the-art. Code can be found at https://github.com/lapisrocks/rpo
Authors: Lance Ying, Tan Zhi-Xuan, Lionel Wong, Vikash Mansinghka, Joshua Tenenbaum
Abstract: Despite the fact that beliefs are mental states that cannot be directly observed, humans talk about each others' beliefs on a regular basis, often using rich compositional language to describe what others think and know. What explains this capacity to interpret the hidden epistemic content of other minds? In this paper, we take a step towards an answer by grounding the semantics of belief statements in a Bayesian theory-of-mind: By modeling how humans jointly infer coherent sets of goals, beliefs, and plans that explain an agent's actions, then evaluating statements about the agent's beliefs against these inferences via epistemic logic, our framework provides a conceptual role semantics for belief, explaining the gradedness and compositionality of human belief attributions, as well as their intimate connection with goals and plans. We evaluate this framework by studying how humans attribute goals and beliefs while watching an agent solve a doors-and-keys gridworld puzzle that requires instrumental reasoning about hidden objects. In contrast to pure logical deduction, non-mentalizing baselines, and mentalizing that ignores the role of instrumental plans, our model provides a much better fit to human goal and belief attributions, demonstrating the importance of theory-of-mind for a semantics of belief.
Authors: Creston Brooks, Robert Calef, Charlie Cowen-Breen, Anna Sappington
Abstract: Beam search with masked language models (MLMs) is challenging in part because joint probability distributions over sequences are not readily available, unlike for autoregressive models. However, estimating such distributions has important domain-specific applications such as ancient text restoration and protein engineering. Here we present probabilistically-sound methods for beam search with MLMs. First, we clarify the conditions under which it is theoretically sound to perform text infilling with MLMs using standard beam search. When these conditions fail, we provide a probabilistically-sound modification with no additional computational complexity and demonstrate that it is superior to the aforementioned beam search in the expected conditions. We then present empirical results comparing several infilling approaches with MLMs across several domains.
Authors: Yifang Chen, Shuohang Wang, Ziyi Yang, Hiteshi Sharma, Nikos Karampatziakis, Donghan Yu, Kevin Jamieson, Simon Shaolei Du, Yelong Shen
Abstract: Reinforcement learning with human feedback (RLHF), as a widely adopted approach in current large language model pipelines, is \textit{bottlenecked by the size of human preference data}. While traditional methods rely on offline preference dataset constructions, recent approaches have shifted towards online settings, where a learner uses a small amount of labeled seed data and a large pool of unlabeled prompts to iteratively construct new preference data through self-generated responses and high-quality reward/preference feedback. However, most current online algorithms still focus on preference labeling during policy model updating with given feedback oracles, which incurs significant expert query costs. \textit{We are the first to explore cost-effective proxy reward oracles construction strategies for further labeling preferences or rewards with extremely limited labeled data and expert query budgets}. Our approach introduces two key innovations: (1) on-policy query to avoid OOD and imbalance issues in seed data, and (2) active learning to select the most informative data for preference queries. Using these methods, we train a evaluation model with minimal expert-labeled data, which then effectively labels nine times more preference pairs for further RLHF training. For instance, our model using Direct Preference Optimization (DPO) gains around over 1% average improvement on AlpacaEval2, MMLU-5shot and MMLU-0shot, with only 1.7K query cost. Our methodology is orthogonal to other direct expert query-based strategies and therefore might be integrated with them to further reduce query costs.
Authors: Shmuel Berman, Kathleen McKeown, Baishakhi Ray
Abstract: Prior research has enhanced the ability of Large Language Models (LLMs) to solve logic puzzles using techniques such as chain-of-thought prompting or introducing a symbolic representation. These frameworks are still usually insufficient to solve complicated logical problems, such as Zebra puzzles, due to the inherent complexity of translating natural language clues into logical statements. We introduce a multi-agent system, ZPS, that integrates LLMs with an off the shelf theorem prover. This system tackles the complex puzzle-solving task by breaking down the problem into smaller, manageable parts, generating SMT (Satisfiability Modulo Theories) code to solve them with a theorem prover, and using feedback between the agents to repeatedly improve their answers. We also introduce an automated grid puzzle grader to assess the correctness of our puzzle solutions and show that the automated grader is reliable by evaluating it in a user-study. Our approach shows improvement in all three LLMs we tested, with GPT-4 showing 166% improvement in the number of fully correct solutions.
Authors: Hannes Kunstmann, Joseph Ollier, Joel Persson, Florian von Wangenheim
Abstract: Large language models (LLMs) present an enormous evolution in the strategic potential of conversational recommender systems (CRS). Yet to date, research has predominantly focused upon technical frameworks to implement LLM-driven CRS, rather than end-user evaluations or strategic implications for firms, particularly from the perspective of a small to medium enterprises (SME) that makeup the bedrock of the global economy. In the current paper, we detail the design of an LLM-driven CRS in an SME setting, and its subsequent performance in the field using both objective system metrics and subjective user evaluations. While doing so, we additionally outline a short-form revised ResQue model for evaluating LLM-driven CRS, enabling replicability in a rapidly evolving field. Our results reveal good system performance from a user experience perspective (85.5% recommendation accuracy) but underscore latency, cost, and quality issues challenging business viability. Notably, with a median cost of $0.04 per interaction and a latency of 5.7s, cost-effectiveness and response time emerge as crucial areas for achieving a more user-friendly and economically viable LLM-driven CRS for SME settings. One major driver of these costs is the use of an advanced LLM as a ranker within the retrieval-augmented generation (RAG) technique. Our results additionally indicate that relying solely on approaches such as Prompt-based learning with ChatGPT as the underlying LLM makes it challenging to achieve satisfying quality in a production environment. Strategic considerations for SMEs deploying an LLM-driven CRS are outlined, particularly considering trade-offs in the current technical landscape.