Authors: Sarmad N. Mohammed, Semra G\"und\"u\c{c}
Abstract: One of the prime problems of computer science and machine learning is to extract information efficiently from large-scale, heterogeneous data. Text data, with its syntax, semantics, and even hidden information content, possesses an exceptional place among the data types in concern. The processing of the text data requires embedding, a method of translating the content of the text to numeric vectors. A correct embedding algorithm is the starting point for obtaining the full information content of the text data. In this work, a new embedding method based on the graph structure of the meaningful sentences is proposed. The design of the algorithm aims to construct an embedding vector that constitutes syntactic and semantic elements as well as the hidden content of the text data. The success of the proposed embedding method is tested in classification problems. Among the wide range of application areas, text classification is the best laboratory for embedding methods; the classification power of the method can be tested using dimensional reduction without any further processing. Furthermore, the method can be compared with different embedding algorithms and machine learning methods. The proposed method is tested with real-world data sets and eight well-known and successful embedding algorithms. The proposed embedding method shows significantly better classification for binary and multiclass datasets compared to well-known algorithms.
Authors: Zacharias Chrysidis, Stefanos-Iordanis Papadopoulos, Symeon Papadopoulos, Panagiotis C. Petrantonakis
Abstract: Automated fact-checking (AFC) is garnering increasing attention by researchers aiming to help fact-checkers combat the increasing spread of misinformation online. While many existing AFC methods incorporate external information from the Web to help examine the veracity of claims, they often overlook the importance of verifying the source and quality of collected "evidence". One overlooked challenge involves the reliance on "leaked evidence", information gathered directly from fact-checking websites and used to train AFC systems, resulting in an unrealistic setting for early misinformation detection. Similarly, the inclusion of information from unreliable sources can undermine the effectiveness of AFC systems. To address these challenges, we present a comprehensive approach to evidence verification and filtering. We create the "CREDible, Unreliable or LEaked" (CREDULE) dataset, which consists of 91,632 articles classified as Credible, Unreliable and Fact checked (Leaked). Additionally, we introduce the EVidence VERification Network (EVVER-Net), trained on CREDULE to detect leaked and unreliable evidence in both short and long texts. EVVER-Net can be used to filter evidence collected from the Web, thus enhancing the robustness of end-to-end AFC systems. We experiment with various language models and show that EVVER-Net can demonstrate impressive performance of up to 91.5% and 94.4% accuracy, while leveraging domain credibility scores along with short or long texts, respectively. Finally, we assess the evidence provided by widely-used fact-checking datasets including LIAR-PLUS, MOCHEG, FACTIFY, NewsCLIPpings+ and VERITE, some of which exhibit concerning rates of leaked and unreliable evidence.
Authors: Mike Zhang
Abstract: [Abridged Abstract] Recent technological advances underscore labor market dynamics, yielding significant consequences for employment prospects and increasing job vacancy data across platforms and languages. Aggregating such data holds potential for valuable insights into labor market demands, new skills emergence, and facilitating job matching for various stakeholders. However, despite prevalent insights in the private sector, transparent language technology systems and data for this domain are lacking. This thesis investigates Natural Language Processing (NLP) technology for extracting relevant information from job descriptions, identifying challenges including scarcity of training data, lack of standardized annotation guidelines, and shortage of effective extraction methods from job ads. We frame the problem, obtaining annotated data, and introducing extraction methodologies. Our contributions include job description datasets, a de-identification dataset, and a novel active learning algorithm for efficient model training. We propose skill extraction using weak supervision, a taxonomy-aware pre-training methodology adapting multilingual language models to the job market domain, and a retrieval-augmented model leveraging multiple skill extraction datasets to enhance overall performance. Finally, we ground extracted information within a designated taxonomy.
Authors: Scott Viteri, Max Lamparth, Peter Chatain, Clark Barrett
Abstract: Chain-of-Thought (CoT) reasoning could in principle enable a deeper understanding of a language model's (LM) internal reasoning. However, prior work suggests that some LMs answer questions similarly despite changes in their CoT, suggesting that those models are not truly using the CoT. We propose a training method to produce CoTs that are sufficient alone for predicting future text, independent of other context. This methodology gives a guarantee that if the LM can predict future tokens, then it must have used the CoT to understand its context. We formalize the idea that the truthfulness of a sender to a receiver LM is the degree to which the sender helps the receiver predict their future observations. Then we define a "Markovian" LM as one which predicts future text given only a CoT as context. We derive a "Markovian training" procedure by applying our definition of truthfulness to a Markovian LM and optimizing via policy gradient and Proximal Policy Optimization (PPO). We demonstrate the effectiveness of our training algorithm on long-context arithmetic problems, show that the model utilizes the CoT, and validate that the generated CoT is meaningful and usable by other models.
Authors: Yilun Hua, Nicholas Chernogor, Yuzhe Gu, Seoyeon Julie Jeong, Miranda Luo, Cristian Danescu-Niculescu-Mizil
Abstract: Throughout a conversation, the way participants interact with each other is in constant flux: their tones may change, they may resort to different strategies to convey their points, or they might alter their interaction patterns. An understanding of these dynamics can complement that of the actual facts and opinions discussed, offering a more holistic view of the trajectory of the conversation: how it arrived at its current state and where it is likely heading. In this work, we introduce the task of summarizing the dynamics of conversations, by constructing a dataset of human-written summaries, and exploring several automated baselines. We evaluate whether such summaries can capture the trajectory of conversations via an established downstream task: forecasting whether an ongoing conversation will eventually derail into toxic behavior. We show that they help both humans and automated systems with this forecasting task. Humans make predictions three times faster, and with greater confidence, when reading the summaries than when reading the transcripts. Furthermore, automated forecasting systems are more accurate when constructing, and then predicting based on, summaries of conversation dynamics, compared to directly predicting on the transcripts.
Authors: Ximing Dong, Dayi Lin, Shaowei Wang, Ahmed E. Hassan
Abstract: Large Language Models (LLMs) have significantly advanced natural language processing (NLP) tasks but also pose ethical and societal risks due to their propensity to generate harmful content. To address this, various approaches have been developed to safeguard LLMs from producing unsafe content. However, existing methods have limitations, including the need for training specific control models and proactive intervention during text generation, that lead to quality degradation and increased computational overhead. To mitigate those limitations, we propose LLMSafeGuard, a lightweight framework to safeguard LLM text generation in real-time. LLMSafeGuard integrates an external validator into the beam search algorithm during decoding, rejecting candidates that violate safety constraints while allowing valid ones to proceed. We introduce a similarity based validation approach, simplifying constraint introduction and eliminating the need for control model training. Additionally, LLMSafeGuard employs a context-wise timing selection strategy, intervening LLMs only when necessary. We evaluate LLMSafe-Guard on two tasks, detoxification and copyright safeguarding, and demonstrate its superior performance over SOTA baselines. For instance, LLMSafeGuard reduces the average toxic score of. LLM output by 29.7% compared to the best baseline meanwhile preserving similar linguistic quality as natural output in detoxification task. Similarly, in the copyright task, LLMSafeGuard decreases the Longest Common Subsequence (LCS) by 56.2% compared to baselines. Moreover, our context-wise timing selection strategy reduces inference time by at least 24% meanwhile maintaining comparable effectiveness as validating each time step. LLMSafeGuard also offers tunable parameters to balance its effectiveness and efficiency.
Authors: Houjun Liu
Abstract: While language models (LMs) offer significant capability in zero-shot reasoning tasks across a wide range of domains, they do not perform satisfactorily in problems which requires multi-step reasoning. Previous approaches to mitigate this involves breaking a larger, multi-step task into sub-tasks and asking the language model to generate proposals ("thoughts") for each sub-task and using exhaustive planning approaches such as DFS to compose a solution. In this work, we leverage this idea to introduce two new contributions: first, we formalize a planning-based approach to perform multi-step problem solving with LMs via Partially Observable Markov Decision Processes (POMDPs), with the LM's own reflections about the value of a state used as a search heuristic; second, leveraging the online POMDP solver POMCP, we demonstrate a superior success rate of 89.4% on the Game of 24 task as compared to existing approaches while also offering better anytime performance characteristics than fixed tree-search which is used previously. Taken together, these contributions allow modern LMs to decompose and solve larger-scale reasoning tasks more effectively.
Authors: Liang Xu, Lei Zhu, Yaotong Wu, Hang Xue
Abstract: The SuperCLUE-Fin (SC-Fin) benchmark is a pioneering evaluation framework tailored for Chinese-native financial large language models (FLMs). It assesses FLMs across six financial application domains and twenty-five specialized tasks, encompassing theoretical knowledge and practical applications such as compliance, risk management, and investment analysis. Using multi-turn, open-ended conversations that mimic real-life scenarios, SC-Fin measures models on a range of criteria, including accurate financial understanding, logical reasoning, clarity, computational efficiency, business acumen, risk perception, and compliance with Chinese regulations. In a rigorous evaluation involving over a thousand questions, SC-Fin identifies a performance hierarchy where domestic models like GLM-4 and MoonShot-v1-128k outperform others with an A-grade, highlighting the potential for further development in transforming theoretical knowledge into pragmatic financial solutions. This benchmark serves as a critical tool for refining FLMs in the Chinese context, directing improvements in financial knowledge databases, standardizing financial interpretations, and promoting models that prioritize compliance, risk management, and secure practices. We create a contextually relevant and comprehensive benchmark that drives the development of AI in the Chinese financial sector. SC-Fin facilitates the advancement and responsible deployment of FLMs, offering valuable insights for enhancing model performance and usability for both individual and institutional users in the Chinese market..~\footnote{Our benchmark can be found at \url{https://www.CLUEbenchmarks.com}}.
Authors: Matteo Merler, Nicola Dainese, Katsiaryna Haitsiukevich
Abstract: Symbolic Regression (SR) is a task which aims to extract the mathematical expression underlying a set of empirical observations. Transformer-based methods trained on SR datasets detain the current state-of-the-art in this task, while the application of Large Language Models (LLMs) to SR remains unexplored. This work investigates the integration of pre-trained LLMs into the SR pipeline, utilizing an approach that iteratively refines a functional form based on the prediction error it achieves on the observation set, until it reaches convergence. Our method leverages LLMs to propose an initial set of possible functions based on the observations, exploiting their strong pre-training prior. These functions are then iteratively refined by the model itself and by an external optimizer for their coefficients. The process is repeated until the results are satisfactory. We then analyze Vision-Language Models in this context, exploring the inclusion of plots as visual inputs to aid the optimization process. Our findings reveal that LLMs are able to successfully recover good symbolic equations that fit the given data, outperforming SR baselines based on Genetic Programming, with the addition of images in the input showing promising results for the most complex benchmarks.
Authors: Arif Ahmed (The University of Arizona), Gondy Leroy (The University of Arizona), Stephen A. Rains (The University of Arizona), Philip Harber (The University of Arizona), David Kauchak (Pomona College), Prosanta Barai (The University of Arizona)
Abstract: Health literacy is crucial to supporting good health and is a major national goal. Audio delivery of information is becoming more popular for informing oneself. In this study, we evaluate the effect of audio enhancements in the form of information emphasis and pauses with health texts of varying difficulty and we measure health information comprehension and retention. We produced audio snippets from difficult and easy text and conducted the study on Amazon Mechanical Turk (AMT). Our findings suggest that emphasis matters for both information comprehension and retention. When there is no added pause, emphasizing significant information can lower the perceived difficulty for difficult and easy texts. Comprehension is higher (54%) with correctly placed emphasis for the difficult texts compared to not adding emphasis (50%). Adding a pause lowers perceived difficulty and can improve retention but adversely affects information comprehension.
Authors: Davis Wertheimer, Joshua Rosenkranz, Thomas Parnell, Sahil Suneja, Pavithra Ranganathan, Raghu Ganti, Mudhakar Srivatsa
Abstract: This technical report describes the design and training of novel speculative decoding draft models, for accelerating the inference speeds of large language models in a production environment. By conditioning draft predictions on both context vectors and sampled tokens, we can train our speculators to efficiently predict high-quality n-grams, which the base model then accepts or rejects. This allows us to effectively predict multiple tokens per inference forward pass, accelerating wall-clock inference speeds of highly optimized base model implementations by a factor of 2-3x. We explore these initial results and describe next steps for further improvements.
Authors: Ning An, Lei Hei, Yong Jiang, Weiping Meng, Jingjing Hu, Boran Huang, Feiliang Ren
Abstract: Relational triple extraction is crucial work for the automatic construction of knowledge graphs. Existing methods only construct shallow representations from a token or token pair-level. However, previous works ignore local spatial dependencies of relational triples, resulting in a weakness of entity pair boundary detection. To tackle this problem, we propose a novel Region-based Table Filling method (RTF). We devise a novel region-based tagging scheme and bi-directional decoding strategy, which regard each relational triple as a region on the relation-specific table, and identifies triples by determining two endpoints of each region. We also introduce convolution to construct region-level table representations from a spatial perspective which makes triples easier to be captured. In addition, we share partial tagging scores among different relations to improve learning efficiency of relation classifier. Experimental results show that our method achieves state-of-the-art with better generalization capability on three variants of two widely used benchmark datasets.
Authors: Sina Bagheri Nezhad, Ameeta Agrawal
Abstract: This study investigates the factors influencing the performance of multilingual large language models (MLLMs) across diverse languages. We study 6 MLLMs, including masked language models, autoregressive models, and instruction-tuned LLMs, on the SIB-200 dataset, a topic classification dataset encompassing 204 languages. Our analysis considers three scenarios: ALL languages, SEEN languages (present in the model's pretraining data), and UNSEEN languages (not present or documented in the model's pretraining data in any meaningful way). We examine the impact of factors such as pretraining data size, general resource availability, language family, and script type on model performance. Decision tree analysis reveals that pretraining data size is the most influential factor for SEEN languages. However, interestingly, script type and language family are crucial for UNSEEN languages, highlighting the importance of cross-lingual transfer learning. Notably, model size and architecture do not significantly alter the most important features identified. Our findings provide valuable insights into the strengths and limitations of current MLLMs and hope to guide the development of more effective and equitable multilingual NLP systems.
Authors: Zhihao Zhang, Feiqi Cao, Yingbin Mo, Yiran Zhang, Josiah Poon, Caren Han
Abstract: The dynamic nature of esports makes the situation relatively complicated for average viewers. Esports broadcasting involves game expert casters, but the caster-dependent game commentary is not enough to fully understand the game situation. It will be richer by including diverse multimodal esports information, including audiences' talks/emotions, game audio, and game match event information. This paper introduces GAME-MUG, a new multimodal game situation understanding and audience-engaged commentary generation dataset and its strong baseline. Our dataset is collected from 2020-2022 LOL game live streams from YouTube and Twitch, and includes multimodal esports game information, including text, audio, and time-series event logs, for detecting the game situation. In addition, we also propose a new audience conversation augmented commentary dataset by covering the game situation and audience conversation understanding, and introducing a robust joint multimodal dual learning model as a baseline. We examine the model's game situation/event understanding ability and commentary generation capability to show the effectiveness of the multimodal aspects coverage and the joint integration learning approach.
Authors: James A. Michaelov, Catherine Arnett, Benjamin K. Bergen
Abstract: Transformers have supplanted Recurrent Neural Networks as the dominant architecture for both natural language processing tasks and, despite criticisms of cognitive implausibility, for modelling the effect of predictability on online human language comprehension. However, two recently developed recurrent neural network architectures, RWKV and Mamba, appear to perform natural language tasks comparably to or better than transformers of equivalent scale. In this paper, we show that contemporary recurrent models are now also able to match - and in some cases, exceed - performance of comparably sized transformers at modeling online human language comprehension. This suggests that transformer language models are not uniquely suited to this task, and opens up new directions for debates about the extent to which architectural features of language models make them better or worse models of human language comprehension.
Authors: Xinwei Chen, Kun Li, Tianyou Song, Jiangjian Guo
Abstract: Named Entity Recognition (NER) is an essential steppingstone in the field of natural language processing. Although promising performance has been achieved by various distantly supervised models, we argue that distant supervision inevitably introduces incomplete and noisy annotations, which may mislead the model training process. To address this issue, we propose a robust NER model named BOND-MoE based on Mixture of Experts (MoE). Instead of relying on a single model for NER prediction, multiple models are trained and ensembled under the Expectation-Maximization (EM) framework, so that noisy supervision can be dramatically alleviated. In addition, we introduce a fair assignment module to balance the document-model assignment process. Extensive experiments on real-world datasets show that the proposed method achieves state-of-the-art performance compared with other distantly supervised NER.
Authors: Xinzhe Li, Ming Liu, Shang Gao
Abstract: Retrieval-augmented Generation (RAG) systems have been actively studied and deployed across various industries to query on domain-specific knowledge base. However, evaluating these systems presents unique challenges due to the scarcity of domain-specific queries and corresponding ground truths, as well as a lack of systematic approaches to diagnosing the cause of failure cases -- whether they stem from knowledge deficits or issues related to system robustness. To address these challenges, we introduce GRAMMAR (GRounded And Modular Methodology for Assessment of RAG), an evaluation framework comprising two key elements: 1) a data generation process that leverages relational databases and LLMs to efficiently produce scalable query-answer pairs. This method facilitates the separation of query logic from linguistic variations for enhanced debugging capabilities; and 2) an evaluation framework that differentiates knowledge gaps from robustness and enables the identification of defective modules. Our empirical results underscore the limitations of current reference-free evaluation approaches and the reliability of GRAMMAR to accurately identify model vulnerabilities.
Authors: Chunlin Tian, Zhan Shi, Zhijiang Guo, Li Li, Chengzhong Xu
Abstract: Adapting Large Language Models (LLMs) to new tasks through fine-tuning has been made more efficient by the introduction of Parameter-Efficient Fine-Tuning (PEFT) techniques, such as LoRA. However, these methods often underperform compared to full fine-tuning, particularly in scenarios involving complex datasets. This issue becomes even more pronounced in complex domains, highlighting the need for improved PEFT approaches that can achieve better performance. Through a series of experiments, we have uncovered two critical insights that shed light on the training and parameter inefficiency of LoRA. Building on these insights, we have developed HydraLoRA, a LoRA framework with an asymmetric structure that eliminates the need for domain expertise. Our experiments demonstrate that HydraLoRA outperforms other PEFT approaches, even those that rely on domain knowledge during the training and inference phases. \href{https://github.com/Clin0212/HydraLoRA}{Code}.
Authors: Cuong Nhat Vo, Khanh Bao Huynh, Son T. Luu, Trong-Hop Do
Abstract: The growth of social networks makes toxic content spread rapidly. Hate speech detection is a task to help decrease the number of harmful comments. With the diversity in the hate speech created by users, it is necessary to interpret the hate speech besides detecting it. Hence, we propose a methodology to construct a system for targeted hate speech detection from online streaming texts from social media. We first introduce the ViTHSD - a targeted hate speech detection dataset for Vietnamese Social Media Texts. The dataset contains 10K comments, each comment is labeled to specific targets with three levels: clean, offensive, and hate. There are 5 targets in the dataset, and each target is labeled with the corresponding level manually by humans with strict annotation guidelines. The inter-annotator agreement obtained from the dataset is 0.45 by Cohen's Kappa index, which is indicated as a moderate level. Then, we construct a baseline for this task by combining the Bi-GRU-LSTM-CNN with the pre-trained language model to leverage the power of text representation of BERTology. Finally, we suggest a methodology to integrate the baseline model for targeted hate speech detection into the online streaming system for practical application in preventing hateful and offensive content on social media.
Authors: Vaishak Narayanan, Prabin Raj KP, Saifudheen Nouphal
Abstract: Current evaluation benchmarks for question answering (QA) in Indic languages often rely on machine translation of existing English datasets. This approach suffers from bias and inaccuracies inherent in machine translation, leading to datasets that may not reflect the true capabilities of EQA models for Indic languages. This paper proposes a new benchmark specifically designed for evaluating Hindi EQA models and discusses the methodology to do the same for any task. This method leverages large language models (LLMs) to generate a high-quality dataset in an extractive setting, ensuring its relevance for the target language. We believe this new resource will foster advancements in Hindi NLP research by providing a more accurate and reliable evaluation tool.
Authors: Abir Chakraborty
Abstract: In this work we investigate the capability of Graph Attention Network for extracting aspect and opinion terms. Aspect and opinion term extraction is posed as a token-level classification task akin to named entity recognition. We use the dependency tree of the input query as additional feature in a Graph Attention Network along with the token and part-of-speech features. We show that the dependency structure is a powerful feature that in the presence of a CRF layer substantially improves the performance and generates the best result on the commonly used datasets from SemEval 2014, 2015 and 2016. We experiment with additional layers like BiLSTM and Transformer in addition to the CRF layer. We also show that our approach works well in the presence of multiple aspects or sentiments in the same query and it is not necessary to modify the dependency tree based on a single aspect as was the original application for sentiment classification.
Authors: Wei Chen, Zhiyuan Li
Abstract: Language models have been effective in a wide range of applications, yet the most sophisticated models are often proprietary. For example, GPT-4 by OpenAI and various models by Anthropic are expensive and consume substantial energy. In contrast, the open-source community has produced competitive models, like Llama3. Furthermore, niche-specific smaller language models, such as those tailored for legal, medical or financial tasks, have outperformed their proprietary counterparts. This paper introduces a novel approach that employs \textit{functional tokens} to integrate \textbf{multiple open-source models}, each optimized for particular tasks. Our newly developed Octopus v4 model leverages \textit{functional tokens} to intelligently direct user queries to the most appropriate vertical model and reformat the query to achieve the best performance. Octopus v4, an evolution of the Octopus v1, v2, and v3 models, excels in selection and parameter understanding and reformatting. Additionally, we explore the use of graph as a versatile data structure that effectively coordinates multiple open-source models by harnessing the capabilities of the Octopus model and \textit{functional tokens}. Use our open-sourced GitHub (\url{https://www.nexa4ai.com/}) to try Octopus v4 models (\url{https://huggingface.co/NexaAIDev/Octopus-v4}), and contrite to a larger graph of language models. By activating models less than 10B parameters, we achieved SOTA MMLU score of 74.8 among the same level models.
URLs: https://www.nexa4ai.com/, https://huggingface.co/NexaAIDev/Octopus-v4
Authors: Eyal Liron Dolev, Clemens Fidel Lutz, No\"emi Aepli
Abstract: Whisper is a state-of-the-art automatic speech recognition (ASR) model (Radford et al., 2022). Although Swiss German dialects are allegedly not part of Whisper's training data, preliminary experiments showed that Whisper can transcribe Swiss German quite well, with the output being a speech translation into Standard German. To gain a better understanding of Whisper's performance on Swiss German, we systematically evaluate it using automatic, qualitative, and human evaluation. We test its performance on three existing test sets: SwissDial (Dogan-Sch\"onberger et al., 2021), STT4SG-350 (Pl\"uss et al., 2023), and Swiss Parliaments Corpus (Pl\"uss et al., 2021). In addition, we create a new test set for this work, based on short mock clinical interviews. For automatic evaluation, we used word error rate (WER) and BLEU. In the qualitative analysis, we discuss Whisper's strengths and weaknesses and anylyze some output examples. For the human evaluation, we conducted a survey with 28 participants who were asked to evaluate Whisper's performance. All of our evaluations suggest that Whisper is a viable ASR system for Swiss German, so long as the Standard German output is desired.
Authors: Zachary William Hopton (Language,Space Lab, University of Zurich), No\"emi Aepli (Department of Computational Linguistics, University of Zurich)
Abstract: Effectively normalizing textual data poses a considerable challenge, especially for low-resource languages lacking standardized writing systems. In this study, we fine-tuned a multilingual model with data from several Occitan dialects and conducted a series of experiments to assess the model's representations of these dialects. For evaluation purposes, we compiled a parallel lexicon encompassing four Occitan dialects. Intrinsic evaluations of the model's embeddings revealed that surface similarity between the dialects strengthened representations. When the model was further fine-tuned for part-of-speech tagging and Universal Dependency parsing, its performance was robust to dialectical variation, even when trained solely on part-of-speech data from a single dialect. Our findings suggest that large multilingual models minimize the need for spelling normalization during pre-processing.
Authors: Sheng Ouyang, Jianzong Wang, Yong Zhang, Zhitao Li, Ziqi Liang, Xulong Zhang, Ning Cheng, Jing Xiao
Abstract: Extractive Question Answering (EQA) in Machine Reading Comprehension (MRC) often faces the challenge of dealing with semantically identical but format-variant inputs. Our work introduces a novel approach, called the ``Query Latent Semantic Calibrator (QLSC)'', designed as an auxiliary module for existing MRC models. We propose a unique scaling strategy to capture latent semantic center features of queries. These features are then seamlessly integrated into traditional query and passage embeddings using an attention mechanism. By deepening the comprehension of the semantic queries-passage relationship, our approach diminishes sensitivity to variations in text format and boosts the model's capability in pinpointing accurate answers. Experimental results on robust Question-Answer datasets confirm that our approach effectively handles format-variant but semantically identical queries, highlighting the effectiveness and adaptability of our proposed method.
Authors: Minh Duc Bui, Fabian David Schmidt, Goran Glava\v{s}, Katharina von der Wense
Abstract: Compared to standard language model (LM) pretraining (i.e., from scratch), Knowledge Distillation (KD) entails an additional forward pass through a teacher model that is typically substantially larger than the target student model. As such, KD in LM pretraining materially slows down throughput of pretraining instances vis-a-vis pretraining from scratch. Scaling laws of LM pretraining suggest that smaller models can close the gap to larger counterparts if trained on more data (i.e., processing more tokens)-and under a fixed computation budget, smaller models are able be process more data than larger models. We thus hypothesize that KD might, in fact, be suboptimal to pretraining from scratch for obtaining smaller LMs, when appropriately accounting for the compute budget. To test this, we compare pretraining from scratch against several KD strategies for masked language modeling (MLM) in a fair experimental setup, with respect to amount of computation as well as pretraining data. Downstream results on GLUE, however, do not confirm our hypothesis: while pretraining from scratch performs comparably to ordinary KD under a fixed computation budget, more sophisticated KD strategies, namely TinyBERT (Jiao et al., 2020) and MiniLM (Wang et al., 2023), outperform it by a notable margin. We further find that KD yields larger gains over pretraining from scratch when the data must be repeated under the fixed computation budget.
Authors: Luise H\"auser, Gerhard J\"ager, Alexandros Stamatakis
Abstract: Working with cognate data involves handling synonyms, that is, multiple words that describe the same concept in a language. In the early days of language phylogenetics it was recommended to select one synonym only. However, as we show here, binary character matrices, which are used as input for computational methods, do allow for representing the entire dataset including all synonyms. Here we address the question how one can and if one should include all synonyms or whether it is preferable to select synonyms a priori. To this end, we perform maximum likelihood tree inferences with the widely used RAxML-NG tool and show that it yields plausible trees when all synonyms are used as input. Furthermore, we show that a priori synonym selection can yield topologically substantially different trees and we therefore advise against doing so. To represent cognate data including all synonyms, we introduce two types of character matrices beyond the standard binary ones: probabilistic binary and probabilistic multi-valued character matrices. We further show that it is dataset-dependent for which character matrix type the inferred RAxML-NG tree is topologically closest to the gold standard. We also make available a Python interface for generating all of the above character matrix types for cognate data provided in CLDF format.
Authors: Xiaoming Liu, Chen Liu, Zhaohan Zhang, Chengzhengxu Li, Longtian Wang, Yu Lan, Chao Shen
Abstract: Large language models have shown their ability to become effective few-shot learners with prompting, revoluting the paradigm of learning with data scarcity. However, this approach largely depends on the quality of prompt initialization, and always exhibits large variability among different runs. Such property makes prompt tuning highly unreliable and vulnerable to poorly constructed prompts, which limits its extension to more real-world applications. To tackle this issue, we propose to treat the hard prompt and soft prompt as separate inputs to mitigate noise brought by the prompt initialization. Furthermore, we optimize soft prompts with contrastive learning for utilizing class-aware information in the training process to maintain model performance. Experimental results demonstrate that \sysname outperforms state-of-the-art methods by 7.20% in accuracy and reduces the standard deviation by 2.02 on average. Furthermore, extensive experiments underscore its robustness and stability across 7 datasets covering various tasks.
Authors: Kirill Milintsevich, Ga\"el Dias, Kairit Sirts
Abstract: This paper explores the impact of incorporating sentiment, emotion, and domain-specific lexicons into a transformer-based model for depression symptom estimation. Lexicon information is added by marking the words in the input transcripts of patient-therapist conversations as well as in social media posts. Overall results show that the introduction of external knowledge within pre-trained language models can be beneficial for prediction performance, while different lexicons show distinct behaviours depending on the targeted task. Additionally, new state-of-the-art results are obtained for the estimation of depression level over patient-therapist interviews.
Authors: Andreas Triantafyllopoulos, Bj\"orn W. Schuller
Abstract: Imbuing machines with the ability to talk has been a longtime pursuit of artificial intelligence (AI) research. From the very beginning, the community has not only aimed to synthesise high-fidelity speech that accurately conveys the semantic meaning of an utterance, but also to colour it with inflections that cover the same range of affective expressions that humans are capable of. After many years of research, it appears that we are on the cusp of achieving this when it comes to single, isolated utterances. This unveils an abundance of potential avenues to explore when it comes to combining these single utterances with the aim of synthesising more complex, longer-term behaviours. In the present chapter, we outline the methodological advances that brought us so far and sketch out the ongoing efforts to reach that coveted next level of artificial expressivity. We also discuss the societal implications coupled with rapidly advancing expressive speech synthesis (ESS) technology and highlight ways to mitigate those risks and ensure the alignment of ESS capabilities with ethical norms.
Authors: Yunhao Zhang, Shaonan Wang, Xinyi Dong, Jiajun Yu, Chengqing Zong
Abstract: Neural language models, particularly large-scale ones, have been consistently proven to be most effective in predicting brain neural activity across a range of studies. However, previous research overlooked the comparison of these models with psychologically plausible ones. Moreover, evaluations were reliant on limited, single-modality, and English cognitive datasets. To address these questions, we conducted an analysis comparing encoding performance of various neural language models and psychologically plausible models. Our study utilized extensive multi-modal cognitive datasets, examining bilingual word and discourse levels. Surprisingly, our findings revealed that psychologically plausible models outperformed neural language models across diverse contexts, encompassing different modalities such as fMRI and eye-tracking, and spanning languages from English to Chinese. Among psychologically plausible models, the one incorporating embodied information emerged as particularly exceptional. This model demonstrated superior performance at both word and discourse levels, exhibiting robust prediction of brain activation across numerous regions in both English and Chinese.
Authors: Katikela Sreeharsha Kishore, Rahimanuddin Shaik
Abstract: The growing prominence of large language models (LLMs) necessitates the exploration of their capabilities beyond English. This research investigates the Telugu language proficiency of ChatGPT and Gemini, two leading LLMs. Through a designed set of 20 questions encompassing greetings, grammar, vocabulary, common phrases, task completion, and situational reasoning, the study delves into their strengths and weaknesses in handling Telugu. The analysis aims to identify the LLM that demonstrates a deeper understanding of Telugu grammatical structures, possesses a broader vocabulary, and exhibits superior performance in tasks like writing and reasoning. By comparing their ability to comprehend and use everyday Telugu expressions, the research sheds light on their suitability for real-world language interaction. Furthermore, the evaluation of adaptability and reasoning capabilities provides insights into how each LLM leverages Telugu to respond to dynamic situations. This comparative analysis contributes to the ongoing discussion on multilingual capabilities in AI and paves the way for future research in developing LLMs that can seamlessly integrate with Telugu-speaking communities.
Authors: Mathieu Rita, Florian Strub, Rahma Chaabouni, Paul Michel, Emmanuel Dupoux, Olivier Pietquin
Abstract: While Reinforcement Learning (RL) has been proven essential for tuning large language models (LLMs), it can lead to reward over-optimization (ROO). Existing approaches address ROO by adding KL regularization, requiring computationally expensive hyperparameter tuning. Additionally, KL regularization focuses solely on regularizing the language policy, neglecting a potential source of regularization: the reward function itself. Inspired by demonstration-guided RL, we here introduce the Reward Calibration from Demonstration (RCfD), which leverages human demonstrations and a reward model to recalibrate the reward objective. Formally, given a prompt, the RCfD objective minimizes the distance between the demonstrations' and LLM's rewards rather than directly maximizing the reward function. This objective shift avoids incentivizing the LLM to exploit the reward model and promotes more natural and diverse language generation. We show the effectiveness of RCfD on three language tasks, which achieves comparable performance to carefully tuned baselines while mitigating ROO.
Authors: Aleksei Dorkin, Kairit Sirts
Abstract: We present an information retrieval based reverse dictionary system using modern pre-trained language models and approximate nearest neighbors search algorithms. The proposed approach is applied to an existing Estonian language lexicon resource, S\~onaveeb (word web), with the purpose of enhancing and enriching it by introducing cross-lingual reverse dictionary functionality powered by semantic search. The performance of the system is evaluated using both an existing labeled English dataset of words and definitions that is extended to contain also Estonian and Russian translations, and a novel unlabeled evaluation approach that extracts the evaluation data from the lexicon resource itself using synonymy relations. Evaluation results indicate that the information retrieval based semantic search approach without any model training is feasible, producing median rank of 1 in the monolingual setting and median rank of 2 in the cross-lingual setting using the unlabeled evaluation approach, with models trained for cross-lingual retrieval and including Estonian in their training data showing superior performance in our particular task.
Authors: D. Panas, S. Seth, V. Belle
Abstract: Two major areas of interest in the era of Large Language Models regard questions of what do LLMs know, and if and how they may be able to reason (or rather, approximately reason). Since to date these lines of work progressed largely in parallel (with notable exceptions), we are interested in investigating the intersection: probing for reasoning about the implicitly-held knowledge. Suspecting the performance to be lacking in this area, we use a very simple set-up of comparisons between cardinalities associated with elements of various subjects (e.g. the number of legs a bird has versus the number of wheels on a tricycle). We empirically demonstrate that although LLMs make steady progress in knowledge acquisition and (pseudo)reasoning with each new GPT release, their capabilities are limited to statistical inference only. It is difficult to argue that pure statistical learning can cope with the combinatorial explosion inherent in many commonsense reasoning tasks, especially once arithmetical notions are involved. Further, we argue that bigger is not always better and chasing purely statistical improvements is flawed at the core, since it only exacerbates the dangerous conflation of the production of correct answers with genuine reasoning ability.
Authors: David Ifeoluwa Adelani, A. Seza Do\u{g}ru\"oz, Iyanuoluwa Shode, Anuoluwapo Aremu
Abstract: Naija is the Nigerian-Pidgin spoken by approx. 120M speakers in Nigeria and it is a mixed language (e.g., English, Portuguese and Indigenous languages). Although it has mainly been a spoken language until recently, there are currently two written genres (BBC and Wikipedia) in Naija. Through statistical analyses and Machine Translation experiments, we prove that these two genres do not represent each other (i.e., there are linguistic differences in word order and vocabulary) and Generative AI operates only based on Naija written in the BBC genre. In other words, Naija written in Wikipedia genre is not represented in Generative AI.
Authors: Vinay Setty
Abstract: We introduce 'FactCheck Editor', an advanced text editor designed to automate fact-checking and correct factual inaccuracies. Given the widespread issue of misinformation, often a result of unintentional mistakes by content creators, our tool aims to address this challenge. It supports over 90 languages and utilizes transformer models to assist humans in the labor-intensive process of fact verification. This demonstration showcases a complete workflow that detects text claims in need of verification, generates relevant search engine queries, and retrieves appropriate documents from the web. It employs Natural Language Inference (NLI) to predict the veracity of claims and uses LLMs to summarize the evidence and suggest textual revisions to correct any errors in the text. Additionally, the effectiveness of models used in claim detection and veracity assessment is evaluated across multiple languages.
Authors: Mariia Ignashina, Julia Ive
Abstract: Current text generation models are trained using real data which can potentially contain sensitive information, such as confidential patient information and the like. Under certain conditions output of the training data which they have memorised can be triggered, exposing sensitive data. To mitigate against this risk we propose a safer alternative which sees fragmented data in the form of domain-specific short phrases randomly grouped together shared instead of full texts. Thus, text fragments that could re-identify an individual cannot be reproduced by the model in one sequence, giving significant protection against linkage attacks. We fine-tune several state-of-the-art LLMs using meaningful syntactic chunks to explore their utility. In particular, we fine-tune BERT-based models to predict two cardiovascular diagnoses. Our results demonstrate the capacity of LLMs to benefit from the pre-trained knowledge and deliver classification results when fine-tuned with fragmented data comparable to fine-tuning with full training data.
Authors: Huy Hien Vu, Hidetaka Kamigaito, Taro Watanabe
Abstract: Despite significant improvements in enhancing the quality of translation, context-aware machine translation (MT) models underperform in many cases. One of the main reasons is that they fail to utilize the correct features from context when the context is too long or their models are overly complex. This can lead to the explain-away effect, wherein the models only consider features easier to explain predictions, resulting in inaccurate translations. To address this issue, we propose a model that explains the decisions made for translation by predicting coreference features in the input. We construct a model for input coreference by exploiting contextual features from both the input and translation output representations on top of an existing MT model. We evaluate and analyze our method in the WMT document-level translation task of English-German dataset, the English-Russian dataset, and the multilingual TED talk dataset, demonstrating an improvement of over 1.0 BLEU score when compared with other context-aware models.
Authors: Shisen Yue, Siyuan Song, Xinyuan Cheng, Hai Hu
Abstract: Understanding the non-literal meaning of an utterance is critical for large language models (LLMs) to become human-like social communicators. In this work, we introduce SwordsmanImp, the first Chinese multi-turn-dialogue-based dataset aimed at conversational implicature, sourced from dialogues in the Chinese sitcom $\textit{My Own Swordsman}$. It includes 200 carefully handcrafted questions, all annotated on which Gricean maxims have been violated. We test eight close-source and open-source LLMs under two tasks: a multiple-choice question task and an implicature explanation task. Our results show that GPT-4 attains human-level accuracy (94%) on multiple-choice questions. CausalLM demonstrates a 78.5% accuracy following GPT-4. Other models, including GPT-3.5 and several open-source models, demonstrate a lower accuracy ranging from 20% to 60% on multiple-choice questions. Human raters were asked to rate the explanation of the implicatures generated by LLMs on their reasonability, logic and fluency. While all models generate largely fluent and self-consistent text, their explanations score low on reasonability except for GPT-4, suggesting that most LLMs cannot produce satisfactory explanations of the implicatures in the conversation. Moreover, we find LLMs' performance does not vary significantly by Gricean maxims, suggesting that LLMs do not seem to process implicatures derived from different maxims differently. Our data and code are available at https://github.com/sjtu-compling/llm-pragmatics.
Authors: Yucheng Hu, Yuxing Lu
Abstract: Large Language Models (LLMs) have catalyzed significant advancements in Natural Language Processing (NLP), yet they encounter challenges such as hallucination and the need for domain-specific knowledge. To mitigate these, recent methodologies have integrated information retrieved from external resources with LLMs, substantially enhancing their performance across NLP tasks. This survey paper addresses the absence of a comprehensive overview on Retrieval-Augmented Language Models (RALMs), both Retrieval-Augmented Generation (RAG) and Retrieval-Augmented Understanding (RAU), providing an in-depth examination of their paradigm, evolution, taxonomy, and applications. The paper discusses the essential components of RALMs, including Retrievers, Language Models, and Augmentations, and how their interactions lead to diverse model structures and applications. RALMs demonstrate utility in a spectrum of tasks, from translation and dialogue systems to knowledge-intensive applications. The survey includes several evaluation methods of RALMs, emphasizing the importance of robustness, accuracy, and relevance in their assessment. It also acknowledges the limitations of RALMs, particularly in retrieval quality and computational efficiency, offering directions for future research. In conclusion, this survey aims to offer a structured insight into RALMs, their potential, and the avenues for their future development in NLP. The paper is supplemented with a Github Repository containing the surveyed works and resources for further study: https://github.com/2471023025/RALM_Survey.
Authors: Peitian Zhang, Ninglu Shao, Zheng Liu, Shitao Xiao, Hongjin Qian, Qiwei Ye, Zhicheng Dou
Abstract: We extend the context length of Llama-3-8B-Instruct from 8K to 80K via QLoRA fine-tuning. The entire training cycle is super efficient, which takes 8 hours on one 8xA800 (80G) GPU machine. The resulted model exhibits superior performances across a broad range of evaluation tasks, such as NIHS, topic retrieval, and long-context language understanding; meanwhile, it also well preserves the original capability over short contexts. The dramatic context extension is mainly attributed to merely 3.5K synthetic training samples generated by GPT-4 , which indicates the LLMs' inherent (yet largely underestimated) potential to extend its original context length. In fact, the context length could be extended far beyond 80K with more computation resources. Therefore, the team will publicly release the entire resources (including data, model, data generation pipeline, training code) so as to facilitate the future research from the community: \url{https://github.com/FlagOpen/FlagEmbedding}.
Authors: Shuqian Sheng, Yi Xu, Tianhang Zhang, Zanwei Shen, Luoyi Fu, Jiaxin Ding, Lei Zhou, Xinbing Wang, Chenghu Zhou
Abstract: Automatic evaluation metrics for generated texts play an important role in the NLG field, especially with the rapid growth of LLMs. However, existing metrics are often limited to specific scenarios, making it challenging to meet the evaluation requirements of expanding LLM applications. Therefore, there is a demand for new, flexible, and effective metrics. In this study, we introduce RepEval, the first metric leveraging the projection of LLM representations for evaluation. RepEval requires minimal sample pairs for training, and through simple prompt modifications, it can easily transition to various tasks. Results on ten datasets from three tasks demonstrate the high effectiveness of our method, which exhibits stronger correlations with human judgments compared to previous metrics, even outperforming GPT-4. Our work underscores the richness of information regarding text quality embedded within LLM representations, offering insights for the development of new metrics.
Authors: Xuanli He, Jun Wang, Qiongkai Xu, Pasquale Minervini, Pontus Stenetorp, Benjamin I. P. Rubinstein, Trevor Cohn
Abstract: The implications of backdoor attacks on English-centric large language models (LLMs) have been widely examined - such attacks can be achieved by embedding malicious behaviors during training and activated under specific conditions that trigger malicious outputs. However, the impact of backdoor attacks on multilingual models remains under-explored. Our research focuses on cross-lingual backdoor attacks against multilingual LLMs, particularly investigating how poisoning the instruction-tuning data in one or two languages can affect the outputs in languages whose instruction-tuning data was not poisoned. Despite its simplicity, our empirical analysis reveals that our method exhibits remarkable efficacy in models like mT5, BLOOM, and GPT-3.5-turbo, with high attack success rates, surpassing 95% in several languages across various scenarios. Alarmingly, our findings also indicate that larger models show increased susceptibility to transferable cross-lingual backdoor attacks, which also applies to LLMs predominantly pre-trained on English data, such as Llama2, Llama3, and Gemma. Moreover, our experiments show that triggers can still work even after paraphrasing, and the backdoor mechanism proves highly effective in cross-lingual response settings across 25 languages, achieving an average attack success rate of 50%. Our study aims to highlight the vulnerabilities and significant security risks present in current multilingual LLMs, underscoring the emergent need for targeted security measures.
Authors: Tiziano Labruna, Jon Ander Campos, Gorka Azkune
Abstract: In this paper, we demonstrate how Large Language Models (LLMs) can effectively learn to use an off-the-shelf information retrieval (IR) system specifically when additional context is required to answer a given question. Given the performance of IR systems, the optimal strategy for question answering does not always entail external information retrieval; rather, it often involves leveraging the parametric memory of the LLM itself. Prior research has identified this phenomenon in the PopQA dataset, wherein the most popular questions are effectively addressed using the LLM's parametric memory, while less popular ones require IR system usage. Following this, we propose a tailored training approach for LLMs, leveraging existing open-domain question answering datasets. Here, LLMs are trained to generate a special token,
Authors: Scott Sumpter
Abstract: This study introduces a transformative framework for medical education by integrating semi-structured data with Large Language Models (LLMs), primarily OpenAIs ChatGPT3.5, to automate the creation of medical simulation scenarios. Traditionally, developing these scenarios was a time-intensive process with limited flexibility to meet diverse educational needs. The proposed approach utilizes AI to efficiently generate detailed, clinically relevant scenarios that are tailored to specific educational objectives. This innovation has significantly reduced the time and resources required for scenario development, allowing for a broader variety of simulations. Preliminary feedback from educators and learners has shown enhanced engagement and improved knowledge acquisition, confirming the effectiveness of this AI-enhanced methodology in simulation-based learning. The integration of structured data with LLMs not only streamlines the creation process but also offers a scalable, dynamic solution that could revolutionize medical training, highlighting the critical role of AI in advancing educational outcomes and patient care standards.
Authors: Hoang-Thang Ta, Abu Bakar Siddiqur Rahman, Lotfollah Najjar, Alexander Gelbukh
Abstract: This paper describes our participation in Task 3 and Task 5 of the #SMM4H (Social Media Mining for Health) 2024 Workshop, explicitly targeting the classification challenges within tweet data. Task 3 is a multi-class classification task centered on tweets discussing the impact of outdoor environments on symptoms of social anxiety. Task 5 involves a binary classification task focusing on tweets reporting medical disorders in children. We applied transfer learning from pre-trained encoder-decoder models such as BART-base and T5-small to identify the labels of a set of given tweets. We also presented some data augmentation methods to see their impact on the model performance. Finally, the systems obtained the best F1 score of 0.627 in Task 3 and the best F1 score of 0.841 in Task 5.
Authors: Richard Yuanzhe Pang, Weizhe Yuan, Kyunghyun Cho, He He, Sainbayar Sukhbaatar, Jason Weston
Abstract: Iterative preference optimization methods have recently been shown to perform well for general instruction tuning tasks, but typically make little improvement on reasoning tasks (Yuan et al., 2024, Chen et al., 2024). In this work we develop an iterative approach that optimizes the preference between competing generated Chain-of-Thought (CoT) candidates by optimizing for winning vs. losing reasoning steps that lead to the correct answer. We train using a modified DPO loss (Rafailov et al., 2023) with an additional negative log-likelihood term, which we find to be crucial. We show reasoning improves across repeated iterations of this scheme. While only relying on examples in the training set, our approach results in increasing accuracy for Llama-2-70B-Chat from 55.6% to 81.6% on GSM8K (and 88.7% with majority voting out of 32 samples), from 12.5% to 20.8% on MATH, and from 77.8% to 86.7% on ARC-Challenge, which outperforms other Llama-2-based models not relying on additionally sourced datasets.
Authors: Fabian Gloeckle, Badr Youbi Idrissi, Baptiste Rozi\`ere, David Lopez-Paz, Gabriel Synnaeve
Abstract: Large language models such as GPT and Llama are trained with a next-token prediction loss. In this work, we suggest that training language models to predict multiple future tokens at once results in higher sample efficiency. More specifically, at each position in the training corpus, we ask the model to predict the following n tokens using n independent output heads, operating on top of a shared model trunk. Considering multi-token prediction as an auxiliary training task, we measure improved downstream capabilities with no overhead in training time for both code and natural language models. The method is increasingly useful for larger model sizes, and keeps its appeal when training for multiple epochs. Gains are especially pronounced on generative benchmarks like coding, where our models consistently outperform strong baselines by several percentage points. Our 13B parameter models solves 12 % more problems on HumanEval and 17 % more on MBPP than comparable next-token models. Experiments on small algorithmic tasks demonstrate that multi-token prediction is favorable for the development of induction heads and algorithmic reasoning capabilities. As an additional benefit, models trained with 4-token prediction are up to 3 times faster at inference, even with large batch sizes.
Authors: Venkatesh C, Harshit Oberoi, Anurag Kumar Pandey, Anil Goyal, Nikhil Sikka
Abstract: In recent years, digital platform companies have faced increasing challenges in managing customer complaints, driven by widespread consumer adoption. This paper introduces an end-to-end pipeline, named RE-GrievanceAssist, designed specifically for real estate customer complaint management. The pipeline consists of three key components: i) response/no-response ML model using TF-IDF vectorization and XGBoost classifier ; ii) user type classifier using fasttext classifier; iii) issue/sub-issue classifier using TF-IDF vectorization and XGBoost classifier. Finally, it has been deployed as a batch job in Databricks, resulting in a remarkable 40% reduction in overall manual effort with monthly cost reduction of Rs 1,50,000 since August 2023.
Authors: Paul Pu Liang
Abstract: Building multisensory AI systems that learn from multiple sensory inputs such as text, speech, video, real-world sensors, wearable devices, and medical data holds great promise for impact in many scientific areas with practical benefits, such as in supporting human health and well-being, enabling multimedia content processing, and enhancing real-world autonomous agents. By synthesizing a range of theoretical frameworks and application domains, this thesis aims to advance the machine learning foundations of multisensory AI. In the first part, we present a theoretical framework formalizing how modalities interact with each other to give rise to new information for a task. These interactions are the basic building blocks in all multimodal problems, and their quantification enables users to understand their multimodal datasets, design principled approaches to learn these interactions, and analyze whether their model has succeeded in learning. In the second part, we study the design of practical multimodal foundation models that generalize over many modalities and tasks, which presents a step toward grounding large language models to real-world sensory modalities. We introduce MultiBench, a unified large-scale benchmark across a wide range of modalities, tasks, and research areas, followed by the cross-modal attention and multimodal transformer architectures that now underpin many of today's multimodal foundation models. Scaling these architectures on MultiBench enables the creation of general-purpose multisensory AI systems, and we discuss our collaborative efforts in applying these models for real-world impact in affective computing, mental health, cancer prognosis, and robotics. Finally, we conclude this thesis by discussing how future work can leverage these ideas toward more general, interactive, and safe multisensory AI.
Authors: Gabriel Sarch, Sahil Somani, Raghav Kapoor, Michael J. Tarr, Katerina Fragkiadaki
Abstract: Recent research on instructable agents has used memory-augmented Large Language Models (LLMs) as task planners, a technique that retrieves language-program examples relevant to the input instruction and uses them as in-context examples in the LLM prompt to improve the performance of the LLM in inferring the correct action and task plans. In this technical report, we extend the capabilities of HELPER, by expanding its memory with a wider array of examples and prompts, and by integrating additional APIs for asking questions. This simple expansion of HELPER into a shared memory enables the agent to work across the domains of executing plans from dialogue, natural language instruction following, active question asking, and commonsense room reorganization. We evaluate the agent on four diverse interactive visual-language embodied agent benchmarks: ALFRED, TEACh, DialFRED, and the Tidy Task. HELPER-X achieves few-shot, state-of-the-art performance across these benchmarks using a single agent, without requiring in-domain training, and remains competitive with agents that have undergone in-domain training.
Authors: Sanjana Gautam, Mukund Srinath
Abstract: With the rapid proliferation of artificial intelligence, there is growing concern over its potential to exacerbate existing biases and societal disparities and introduce novel ones. This issue has prompted widespread attention from academia, policymakers, industry, and civil society. While evidence suggests that integrating human perspectives can mitigate bias-related issues in AI systems, it also introduces challenges associated with cognitive biases inherent in human decision-making. Our research focuses on reviewing existing methodologies and ongoing investigations aimed at understanding annotation attributes that contribute to bias.
Authors: Navid Rajabi, Jana Kosecka
Abstract: Vision and Language Models (VLMs) continue to demonstrate remarkable zero-shot (ZS) performance across various tasks. However, many probing studies have revealed that even the best-performing VLMs struggle to capture aspects of compositional scene understanding, lacking the ability to properly ground and localize linguistic phrases in images. Recent VLM advancements include scaling up both model and dataset sizes, additional training objectives and levels of supervision, and variations in the model architectures. To characterize the grounding ability of VLMs, such as phrase grounding, referring expressions comprehension, and relationship understanding, Pointing Game has been used as an evaluation metric for datasets with bounding box annotations. In this paper, we introduce a novel suite of quantitative metrics that utilize GradCAM activations to rigorously evaluate the grounding capabilities of pre-trained VLMs like CLIP, BLIP, and ALBEF. These metrics offer an explainable and quantifiable approach for a more detailed comparison of the zero-shot capabilities of VLMs and enable measuring models' grounding uncertainty. This characterization reveals interesting tradeoffs between the size of the model, the dataset size, and their performance.
Authors: Jiading Fang, Xiangshan Tan, Shengjie Lin, Igor Vasiljevic, Vitor Guizilini, Hongyuan Mei, Rares Ambrus, Gregory Shakhnarovich, Matthew R Walter
Abstract: If robots are to work effectively alongside people, they must be able to interpret natural language references to objects in their 3D environment. Understanding 3D referring expressions is challenging -- it requires the ability to both parse the 3D structure of the scene and correctly ground free-form language in the presence of distraction and clutter. We introduce Transcrib3D, an approach that brings together 3D detection methods and the emergent reasoning capabilities of large language models (LLMs). Transcrib3D uses text as the unifying medium, which allows us to sidestep the need to learn shared representations connecting multi-modal inputs, which would require massive amounts of annotated 3D data. As a demonstration of its effectiveness, Transcrib3D achieves state-of-the-art results on 3D reference resolution benchmarks, with a great leap in performance from previous multi-modality baselines. To improve upon zero-shot performance and facilitate local deployment on edge computers and robots, we propose self-correction for fine-tuning that trains smaller models, resulting in performance close to that of large models. We show that our method enables a real robot to perform pick-and-place tasks given queries that contain challenging referring expressions. Project site is at https://ripl.github.io/Transcrib3D.
Authors: Abir Chakraborty
Abstract: Knowledge graphs (KGs) are large datasets with specific structures representing large knowledge bases (KB) where each node represents a key entity and relations amongst them are typed edges. Natural language queries formed to extract information from a KB entail starting from specific nodes and reasoning over multiple edges of the corresponding KG to arrive at the correct set of answer nodes. Traditional approaches of question answering on KG are based on (a) semantic parsing (SP), where a logical form (e.g., S-expression, SPARQL query, etc.) is generated using node and edge embeddings and then reasoning over these representations or tuning language models to generate the final answer directly, or (b) information-retrieval based that works by extracting entities and relations sequentially. In this work, we evaluate the capability of (LLMs) to answer questions over KG that involve multiple hops. We show that depending upon the size and nature of the KG we need different approaches to extract and feed the relevant information to an LLM since every LLM comes with a fixed context window. We evaluate our approach on six KGs with and without the availability of example-specific sub-graphs and show that both the IR and SP-based methods can be adopted by LLMs resulting in an extremely competitive performance.
Authors: Sol\`ene Tarride, Christopher Kermorvant
Abstract: In recent advances in automatic text recognition (ATR), deep neural networks have demonstrated the ability to implicitly capture language statistics, potentially reducing the need for traditional language models. This study directly addresses whether explicit language models, specifically n-gram models, still contribute to the performance of state-of-the-art deep learning architectures in the field of handwriting recognition. We evaluate two prominent neural network architectures, PyLaia and DAN, with and without the integration of explicit n-gram language models. Our experiments on three datasets - IAM, RIMES, and NorHand v2 - at both line and page level, investigate optimal parameters for n-gram models, including their order, weight, smoothing methods and tokenization level. The results show that incorporating character or subword n-gram models significantly improves the performance of ATR models on all datasets, challenging the notion that deep learning models alone are sufficient for optimal performance. In particular, the combination of DAN with a character language model outperforms current benchmarks, confirming the value of hybrid approaches in modern document analysis systems.
Authors: Yuvraj Virk, Premkumar Devanbu, Toufique Ahmed
Abstract: A good summary can often be very useful during program comprehension. While a brief, fluent, and relevant summary can be helpful, it does require significant human effort to produce. Often, good summaries are unavailable in software projects, thus making maintenance more difficult. There has been a considerable body of research into automated AI-based methods, using Large Language models (LLMs), to generate summaries of code; there also has been quite a bit work on ways to measure the performance of such summarization methods, with special attention paid to how closely these AI-generated summaries resemble a summary a human might have produced. Measures such as BERTScore and BLEU have been suggested and evaluated with human-subject studies. However, LLMs often err and generate something quite unlike what a human might say. Given an LLM-produced code summary, is there a way to gauge whether it's likely to be sufficiently similar to a human produced summary, or not? In this paper, we study this question, as a calibration problem: given a summary from an LLM, can we compute a confidence measure, which is a good indication of whether the summary is sufficiently similar to what a human would have produced in this situation? We examine this question using several LLMs, for several languages, and in several different settings. We suggest an approach which provides well-calibrated predictions of likelihood of similarity to human summaries.
Authors: Hao-Cheng Lo, Jung-Mei Chu, Jieh Hsiang, Chun-Chieh Cho
Abstract: In patent prosecution, image-based retrieval systems for identifying similarities between current patent images and prior art are pivotal to ensure the novelty and non-obviousness of patent applications. Despite their growing popularity in recent years, existing attempts, while effective at recognizing images within the same patent, fail to deliver practical value due to their limited generalizability in retrieving relevant prior art. Moreover, this task inherently involves the challenges posed by the abstract visual features of patent images, the skewed distribution of image classifications, and the semantic information of image descriptions. Therefore, we propose a language-informed, distribution-aware multimodal approach to patent image feature learning, which enriches the semantic understanding of patent image by integrating Large Language Models and improves the performance of underrepresented classes with our proposed distribution-aware contrastive losses. Extensive experiments on DeepPatent2 dataset show that our proposed method achieves state-of-the-art or comparable performance in image-based patent retrieval with mAP +53.3%, Recall@10 +41.8%, and MRR@10 +51.9%. Furthermore, through an in-depth user analysis, we explore our model in aiding patent professionals in their image retrieval efforts, highlighting the model's real-world applicability and effectiveness.
Authors: Zhen Guo
Abstract: Large language model pre-training has become increasingly expensive, with most practitioners relying on scaling laws to allocate compute budgets for model size and training tokens, commonly referred to as Compute-Optimal or Chinchilla Optimal. In this paper, we hypothesize a new scaling law that suggests model performance depends mostly on the amount of compute spent for transformer-based models, independent of the specific allocation to model size and dataset size. Using this unified scaling law, we predict that (a) for inference efficiency, training should prioritize smaller model sizes and larger training datasets, and (b) assuming the exhaustion of available web datasets, scaling the model size might be the only way to further improve model performance.
Authors: Chun Feng, Joy Hsu, Weiyu Liu, Jiajun Wu
Abstract: 3D visual grounding is a challenging task that often requires direct and dense supervision, notably the semantic label for each object in the scene. In this paper, we instead study the naturally supervised setting that learns from only 3D scene and QA pairs, where prior works underperform. We propose the Language-Regularized Concept Learner (LARC), which uses constraints from language as regularization to significantly improve the accuracy of neuro-symbolic concept learners in the naturally supervised setting. Our approach is based on two core insights: the first is that language constraints (e.g., a word's relation to another) can serve as effective regularization for structured representations in neuro-symbolic models; the second is that we can query large language models to distill such constraints from language properties. We show that LARC improves performance of prior works in naturally supervised 3D visual grounding, and demonstrates a wide range of 3D visual reasoning capabilities-from zero-shot composition, to data efficiency and transferability. Our method represents a promising step towards regularizing structured visual reasoning frameworks with language-based priors, for learning in settings without dense supervision.
Authors: Nicholas S. Kersting, Mohammad Rahman, Suchismitha Vedala, Yang Wang
Abstract: We introduce an intuitive method to test the robustness (stability and explainability) of any black-box LLM in real-time, based upon the local deviation from harmoniticity, denoted as $\gamma$. To the best of our knowledge this is the first completely model-agnostic and unsupervised method of measuring the robustness of any given response from an LLM, based upon the model itself conforming to a purely mathematical standard. We conduct human annotation experiments to show the positive correlation of $\gamma$ with false or misleading answers, and demonstrate that following the gradient of $\gamma$ in stochastic gradient ascent efficiently exposes adversarial prompts. Measuring $\gamma$ across thousands of queries in popular LLMs (GPT-4, ChatGPT, Claude-2.1, Mixtral-8x7B, Smaug-72B, Llama2-7B, and MPT-7B) allows us to estimate the liklihood of wrong or hallucinatory answers automatically and quantitatively rank the reliability of these models in various objective domains (Web QA, TruthfulQA, and Programming QA). Across all models and domains tested, human ratings confirm that $\gamma \to 0$ indicates trustworthiness, and the low-$\gamma$ leaders among these models are GPT-4, ChatGPT, and Smaug-72B.
Authors: Steph Buongiorno, Lawrence Jake Klinkert, Tanishq Chawla, Zixin Zhuang, Corey Clark
Abstract: This research introduces Procedural Artificial Narrative using Generative AI (PANGeA), a structured approach for leveraging large language models (LLMs), guided by a game designer's high-level criteria, to generate narrative content for turn-based role-playing video games (RPGs). Distinct from prior applications of LLMs used for video game design, PANGeA innovates by not only generating game level data (which includes, but is not limited to, setting, key items, and non-playable characters (NPCs)), but by also fostering dynamic, free-form interactions between the player and the environment that align with the procedural game narrative. The NPCs generated by PANGeA are personality-biased and express traits from the Big 5 Personality Model in their generated responses. PANGeA addresses challenges behind ingesting free-form text input, which can prompt LLM responses beyond the scope of the game narrative. A novel validation system that uses the LLM's intelligence evaluates text input and aligns generated responses with the unfolding narrative. Making these interactions possible, PANGeA is supported by a server that hosts a custom memory system that supplies context for augmenting generated responses thus aligning them with the procedural narrative. For its broad application, the server has a REST interface enabling any game engine to integrate directly with PANGeA, as well as an LLM interface adaptable with local or private LLMs. PANGeA's ability to foster dynamic narrative generation by aligning responses with the procedural narrative is demonstrated through an empirical study and ablation test of two versions of a demo game. These are, a custom, browser-based GPT and a Unity demo. As the results show, PANGeA holds potential to assist game designers in using LLMs to generate narrative-consistent content even when provided varied and unpredictable, free-form text input.
Authors: Yasumasa Onoe, Sunayana Rane, Zachary Berger, Yonatan Bitton, Jaemin Cho, Roopal Garg, Alexander Ku, Zarana Parekh, Jordi Pont-Tuset, Garrett Tanzer, Su Wang, Jason Baldridge
Abstract: Vision-language datasets are vital for both text-to-image (T2I) and image-to-text (I2T) research. However, current datasets lack descriptions with fine-grained detail that would allow for richer associations to be learned by models. To fill the gap, we introduce Descriptions of Connected and Contrasting Images (DOCCI), a dataset with long, human-annotated English descriptions for 15k images that were taken, curated and donated by a single researcher intent on capturing key challenges such as spatial relations, counting, text rendering, world knowledge, and more. We instruct human annotators to create comprehensive descriptions for each image; these average 136 words in length and are crafted to clearly distinguish each image from those that are related or similar. Each description is highly compositional and typically encompasses multiple challenges. Through both quantitative and qualitative analyses, we demonstrate that DOCCI serves as an effective training resource for image-to-text generation -- a PaLI 5B model finetuned on DOCCI shows equal or superior results compared to highly-performant larger models like LLaVA-1.5 7B and InstructBLIP 7B. Furthermore, we show that DOCCI is a useful testbed for text-to-image generation, highlighting the limitations of current text-to-image models in capturing long descriptions and fine details.
Authors: Emil Carlsson, Devdatt Dubhashi, Fredrik D. Johansson
Abstract: Recent work (Xu et al., 2020) has suggested that numeral systems in different languages are shaped by a functional need for efficient communication in an information-theoretic sense. Here we take a learning-theoretic approach and show how efficient communication emerges via reinforcement learning. In our framework, two artificial agents play a Lewis signaling game where the goal is to convey a numeral concept. The agents gradually learn to communicate using reinforcement learning and the resulting numeral systems are shown to be efficient in the information-theoretic framework of Regier et al. (2015); Gibson et al. (2017). They are also shown to be similar to human numeral systems of same type. Our results thus provide a mechanistic explanation via reinforcement learning of the recent results in Xu et al. (2020) and can potentially be generalized to other semantic domains.
Authors: Pai Liu, Wenyang Gao, Wenjie Dong, Songfang Huang, Yue Zhang
Abstract: Open information extraction is an important NLP task that targets extracting structured information from unstructured text without limitations on the relation type or the domain of the text. This survey paper covers open information extraction technologies from 2007 to 2022 with a focus on new models not covered by previous surveys. We propose a new categorization method from the source of information perspective to accommodate the development of recent OIE technologies. In addition, we summarize three major approaches based on task settings as well as current popular datasets and model evaluation metrics. Given the comprehensive review, several future directions are shown from datasets, source of information, output form, method, and evaluation metric aspects.
Authors: Soroush Abbasi Koohpayegani, KL Navaneet, Parsa Nooralinejad, Soheil Kolouri, Hamed Pirsiavash
Abstract: Fine-tuning Large Language Models (LLMs) and storing them for each downstream task or domain is impractical because of the massive model size (e.g., 350GB in GPT-3). Current literature, such as LoRA, showcases the potential of low-rank modifications to the original weights of an LLM, enabling efficient adaptation and storage for task-specific models. These methods can reduce the number of parameters needed to fine-tune an LLM by several orders of magnitude. Yet, these methods face two primary limitations: (1) the parameter count is lower-bounded by the rank one decomposition, and (2) the extent of reduction is heavily influenced by both the model architecture and the chosen rank. We introduce NOLA, which overcomes the rank one lower bound present in LoRA. It achieves this by re-parameterizing the low-rank matrices in LoRA using linear combinations of randomly generated matrices (basis) and optimizing the linear mixture coefficients only. This approach allows us to decouple the number of trainable parameters from both the choice of rank and the network architecture. We present adaptation results using GPT-2, LLaMA-2, and ViT in natural language and computer vision tasks. NOLA performs as well as LoRA models with much fewer number of parameters compared to LoRA with rank one, the best compression LoRA can archive. Particularly, on LLaMA-2 70B, our method is almost 20 times more compact than the most compressed LoRA without degradation in accuracy. Our code is available here: https://github.com/UCDvision/NOLA
Authors: Jessica Ojo, Kelechi Ogueji, Pontus Stenetorp, David Ifeoluwa Adelani
Abstract: Recent advancements in natural language processing have led to the proliferation of large language models (LLMs). These models have been shown to yield good performance, using in-context learning, even on tasks and languages they are not trained on. However, their performance on African languages is largely understudied relative to high-resource languages. We present an analysis of four popular large language models (mT0, Aya, LLaMa 2, and GPT-4) on six tasks (topic classification, sentiment classification, machine translation, summarization, question answering, and named entity recognition) across 60 African languages, spanning different language families and geographical regions. Our results suggest that all LLMs produce lower performance for African languages, and there is a large gap in performance compared to high-resource languages (such as English) for most tasks. We find that GPT-4 has an average to good performance on classification tasks, yet its performance on generative tasks such as machine translation and summarization is significantly lacking. Surprisingly, we find that mT0 had the best overall performance for cross-lingual QA, better than the state-of-the-art supervised model (i.e. fine-tuned mT5) and GPT-4 on African languages. Similarly, we find the recent Aya model to have comparable result to mT0 in almost all tasks except for topic classification where it outperform mT0. Overall, LLaMa 2 showed the worst performance, which we believe is due to its English and code-centric~(around 98%) pre-training corpus. Our findings confirms that performance on African languages continues to remain a hurdle for the current LLMs, underscoring the need for additional efforts to close this gap.
Authors: Micha Livne, Zulfat Miftahutdinov, Elena Tutubalina, Maksim Kuznetsov, Daniil Polykovskiy, Annika Brundyn, Aastha Jhunjhunwala, Anthony Costa, Alex Aliper, Al\'an Aspuru-Guzik, Alex Zhavoronkov
Abstract: Large Language Models (LLMs) have substantially driven scientific progress in various domains, and many papers have demonstrated their ability to tackle complex problems with creative solutions. Our paper introduces a new foundation model, nach0, capable of solving various chemical and biological tasks: biomedical question answering, named entity recognition, molecular generation, molecular synthesis, attributes prediction, and others. nach0 is a multi-domain and multi-task encoder-decoder LLM pre-trained on unlabeled text from scientific literature, patents, and molecule strings to incorporate a range of chemical and linguistic knowledge. We employed instruction tuning, where specific task-related instructions are utilized to fine-tune nach0 for the final set of tasks. To train nach0 effectively, we leverage the NeMo framework, enabling efficient parallel optimization of both base and large model versions. Extensive experiments demonstrate that our model outperforms state-of-the-art baselines on single-domain and cross-domain tasks. Furthermore, it can generate high-quality outputs in molecular and textual formats, showcasing its effectiveness in multi-domain setups.
Authors: Simon A. Lee, Sujay Jain, Alex Chen, Kyoka Ono, Jennifer Fang, Akos Rudas, Jeffrey N. Chiang
Abstract: In this work, we introduce the Multiple Embedding Model for EHR (MEME), an approach that serializes multimodal EHR tabular data into text using pseudo-notes, mimicking clinical text generation. This conversion not only preserves better representations of categorical data and learns contexts but also enables the effective employment of pretrained foundation models for rich feature representation. To address potential issues with context length, our framework encodes embeddings for each EHR modality separately. We demonstrate the effectiveness of MEME by applying it to several decision support tasks within the Emergency Department across multiple hospital systems. Our findings indicate that MEME outperforms traditional machine learning, EHR-specific foundation models, and general LLMs, highlighting its potential as a general and extendible EHR representation strategy.
Authors: Vinay Setty
Abstract: In this paper, we explore the challenges associated with establishing an end-to-end fact-checking pipeline in a real-world context, covering over 90 languages. Our real-world experimental benchmarks demonstrate that fine-tuning Transformer models specifically for fact-checking tasks, such as claim detection and veracity prediction, provide superior performance over large language models (LLMs) like GPT-4, GPT-3.5-Turbo, and Mistral-7b. However, we illustrate that LLMs excel in generative tasks such as question decomposition for evidence retrieval. Through extensive evaluation, we show the efficacy of fine-tuned models for fact-checking in a multilingual setting and complex claims that include numerical quantities.
Authors: Grgur Kova\v{c}, R\'emy Portelas, Masataka Sawayama, Peter Ford Dominey, Pierre-Yves Oudeyer
Abstract: The standard way to study Large Language Models (LLMs) with benchmarks or psychology questionnaires is to provide many different queries from similar minimal contexts (e.g. multiple choice questions). However, due to LLMs' highly context-dependent nature, conclusions from such minimal-context evaluations may be little informative about the model's behavior in deployment (where it will be exposed to many new contexts). We argue that context-dependence (specifically, value stability) should be studied a specific property of LLMs and used as another dimension of LLM comparison (alongside others such as cognitive abilities, knowledge, or model size). We present a case-study on the stability of value expression over different contexts (simulated conversations on different topics) as measured using a standard psychology questionnaire (PVQ) and on behavioral downstream tasks. Reusing methods from psychology, we study Rank-order stability on the population (interpersonal) level, and Ipsative stability on the individual (intrapersonal) level. We consider two settings (with and without instructing LLMs to simulate particular personas), two simulated populations, and three downstream tasks. We observe consistent trends in the stability of models and model families - Mixtral, Mistral, GPT-3.5 and Qwen families are more stable than LLaMa-2 and Phi. The consistency of these trends implies that some models exhibit higher value-stability than others, and that value stability can be estimated with the set of introduced methodological tools. When instructed to simulate particular personas, LLMs exhibit low Rank-Order stability, which further diminishes with conversation length. This highlights the need for future research on LLMs that coherently simulate different personas. This paper provides a foundational step in that direction, and, to our knowledge, it is the first study of value stability in LLMs.
Authors: Shivchander Sudalairaj, Abhishek Bhandwaldar, Aldo Pareja, Kai Xu, David D. Cox, Akash Srivastava
Abstract: This work introduces LAB (Large-scale Alignment for chatBots), a novel methodology designed to overcome the scalability challenges in the instruction-tuning phase of large language model (LLM) training. Leveraging a taxonomy-guided synthetic data generation process and a multi-phase tuning framework, LAB significantly reduces reliance on expensive human annotations and proprietary models like GPT-4. We demonstrate that LAB-trained models can achieve competitive performance across several benchmarks compared to models trained with traditional human-annotated or GPT-4 generated synthetic data. Thus offering a scalable, cost-effective solution for enhancing LLM capabilities and instruction-following behaviors without the drawbacks of catastrophic forgetting, marking a step forward in the efficient training of LLMs for a wide range of applications.
Authors: Kartik Kartik, Sanjana Soni, Anoop Kunchukuttan, Tanmoy Chakraborty, Md Shad Akhtar
Abstract: The widespread online communication in a modern multilingual world has provided opportunities to blend more than one language (aka code-mixed language) in a single utterance. This has resulted a formidable challenge for the computational models due to the scarcity of annotated data and presence of noise. A potential solution to mitigate the data scarcity problem in low-resource setup is to leverage existing data in resource-rich language through translation. In this paper, we tackle the problem of code-mixed (Hinglish and Bengalish) to English machine translation. First, we synthetically develop HINMIX, a parallel corpus of Hinglish to English, with ~4.2M sentence pairs. Subsequently, we propose RCMT, a robust perturbation based joint-training model that learns to handle noise in the real-world code-mixed text by parameter sharing across clean and noisy words. Further, we show the adaptability of RCMT in a zero-shot setup for Bengalish to English translation. Our evaluation and comprehensive analyses qualitatively and quantitatively demonstrate the superiority of RCMT over state-of-the-art code-mixed and robust translation methods.
Authors: Venktesh V, Abhijit Anand, Avishek Anand, Vinay Setty
Abstract: Automated fact checking has gained immense interest to tackle the growing misinformation in the digital era. Existing systems primarily focus on synthetic claims on Wikipedia, and noteworthy progress has also been made on real-world claims. In this work, we release Numtemp, a diverse, multi-domain dataset focused exclusively on numerical claims, encompassing temporal, statistical and diverse aspects with fine-grained metadata and an evidence collection without leakage. This addresses the challenge of verifying real-world numerical claims, which are complex and often lack precise information, not addressed by existing works that mainly focus on synthetic claims. We evaluate and quantify the limitations of existing solutions for the task of verifying numerical claims. We also evaluate claim decomposition based methods, numerical understanding based models and our best baselines achieves a macro-F1 of 58.32. This demonstrates that Numtemp serves as a challenging evaluation set for numerical claim verification.
Authors: Yunlong Feng, Yang Xu, Libo Qin, Yasheng Wang, Wanxiang Che
Abstract: Large-scale high-quality training data is important for improving the performance of models. After trained with data that has rationales (reasoning steps), models gain reasoning capability. However, the dataset with high-quality rationales is relatively scarce due to the high annotation cost. To address this issue, we propose \textit{Self-motivated Learning} framework. The framework motivates the model itself to automatically generate rationales on existing datasets. Based on the inherent rank from correctness across multiple rationales, the model learns to generate better rationales, leading to higher reasoning capability. Specifically, we train a reward model with the rank to evaluate the quality of rationales, and improve the performance of reasoning through reinforcement learning. Experiment results of Llama2 7B on multiple reasoning datasets show that our method significantly improves the reasoning ability of models, even outperforming text-davinci-002 in some datasets.
Authors: Robert Vacareanu, Vlad-Andrei Negru, Vasile Suciu, Mihai Surdeanu
Abstract: We analyze how well pre-trained large language models (e.g., Llama2, GPT-4, Claude 3, etc) can do linear and non-linear regression when given in-context examples, without any additional training or gradient updates. Our findings reveal that several large language models (e.g., GPT-4, Claude 3) are able to perform regression tasks with a performance rivaling (or even outperforming) that of traditional supervised methods such as Random Forest, Bagging, or Gradient Boosting. For example, on the challenging Friedman #2 regression dataset, Claude 3 outperforms many supervised methods such as AdaBoost, SVM, Random Forest, KNN, or Gradient Boosting. We then investigate how well the performance of large language models scales with the number of in-context exemplars. We borrow from the notion of regret from online learning and empirically show that LLMs are capable of obtaining a sub-linear regret.
Authors: Michael Dalvean
Abstract: Word complexity is defined in a number of different ways. Psycholinguistic, morphological and lexical proxies are often used. Human ratings are also used. The problem here is that these proxies do not measure complexity directly, and human ratings are susceptible to subjective bias. In this study we contend that some form of 'latent complexity' can be approximated by using samples of simple and complex words. We use a sample of 'simple' words from primary school picture books and a sample of 'complex' words from high school and academic settings. In order to analyse the differences between these classes, we look at the letter positional probabilities (LPPs). We find strong statistical associations between several LPPs and complexity. For example, simple words are significantly (p<.001) more likely to start with w, b, s, h, g, k, j, t, y or f, while complex words are significantly (p<.001) more likely to start with i, a, e, r, v, u or d. We find similar strong associations for subsequent letter positions, with 84 letter-position variables in the first 6 positions being significant at the p<.001 level. We then use LPPs as variables in creating a classifier which can classify the two classes with an 83% accuracy. We test these findings using a second data set, with 66 LPPs significant (p<.001) in the first 6 positions common to both datasets. We use these 66 variables to create a classifier that is able to classify a third dataset with an accuracy of 70%. Finally, we create a fourth sample by combining the extreme high and low scoring words generated by three classifiers built on the first three separate datasets and use this sample to build a classifier which has an accuracy of 97%. We use this to score the four levels of English word groups from an ESL program.
Authors: Taylor Webb, Keith J. Holyoak, Hongjing Lu
Abstract: We recently reported evidence that large language models are capable of solving a wide range of text-based analogy problems in a zero-shot manner, indicating the presence of an emergent capacity for analogical reasoning. Two recent commentaries have challenged these results, citing evidence from so-called `counterfactual' tasks in which the standard sequence of the alphabet is arbitrarily permuted so as to decrease similarity with materials that may have been present in the language model's training data. Here, we reply to these critiques, clarifying some misunderstandings about the test materials used in our original work, and presenting evidence that language models are also capable of generalizing to these new counterfactual task variants.
Authors: Constanza Fierro, Jiaang Li, Anders S{\o}gaard
Abstract: The purpose of instruction tuning is enabling zero-shot performance, but instruction tuning has also been shown to improve chain-of-thought reasoning and value alignment (Si et al., 2023). Here we consider the impact on $\textit{consistency}$, i.e., the sensitivity of language models to small perturbations in the input. We compare 10 instruction-tuned LLaMA models to the original LLaMA-7b model and show that almost across-the-board they become more consistent, both in terms of their representations and their predictions in zero-shot and downstream tasks. We explain these improvements through mechanistic analyses of factual recall.
Authors: Timothee Mickus, Ra\'ul V\'azquez, Joseph Attieh
Abstract: Modularity is a paradigm of machine translation with the potential of bringing forth models that are large at training time and small during inference. Within this field of study, modular approaches, and in particular attention bridges, have been argued to improve the generalization capabilities of models by fostering language-independent representations. In the present paper, we study whether modularity affects translation quality; as well as how well modular architectures generalize across different evaluation scenarios. For a given computational budget, we find non-modular architectures to be always comparable or preferable to all modular designs we study.
Authors: Zilong Bai, Ruiji Zhang, Linqing Chen, Qijun Cai, Yuan Zhong, Cong Wang, Yan Fang, Jie Fang, Jing Sun, Weikuan Wang, Lizhi Zhou, Haoran Hua, Tian Qiu, Chaochao Wang, Cheng Sun, Jianping Lu, Yixin Wang, Yubin Xia, Meng Hu, Haowen Liu, Peng Xu, Licong Xu, Fu Bian, Xiaolong Gu, Lisha Zhang, Weilei Wang, Changyang Tu
Abstract: In recent years, large language models have attracted significant attention due to their exceptional performance across a multitude of natural language process tasks, and have been widely applied in various fields. However, the application of large language models in the Intellectual Property (IP) space is challenging due to the strong need for specialized knowledge, privacy protection, processing of extremely long text in this field. In this technical report, we present for the first time a low-cost, standardized procedure for training IP-oriented LLMs, meeting the unique requirements of the IP domain. Using this standard process, we have trained the PatentGPT series models based on open-source pretrained models. By evaluating them on the open-source IP-oriented benchmark MOZIP, our domain-specific LLMs outperforms GPT-4, indicating the effectiveness of the proposed training procedure and the expertise of the PatentGPT models in the IP demain. What is impressive is that our model significantly outperformed GPT-4 on the 2019 China Patent Agent Qualification Examination by achieving a score of 65, reaching the level of human experts. Additionally, the PatentGPT model, which utilizes the SMoE architecture, achieves performance comparable to that of GPT-4 in the IP domain and demonstrates a better cost-performance ratio on long-text tasks, potentially serving as an alternative to GPT-4 within the IP domain.
Authors: David Ifeoluwa Adelani, A. Seza Do\u{g}ru\"oz, Andr\'e Coneglian, Atul Kr. Ojha
Abstract: Large Language Models are transforming NLP for a variety of tasks. However, how LLMs perform NLP tasks for low-resource languages (LRLs) is less explored. In line with the goals of the AmericasNLP workshop, we focus on 12 LRLs from Brazil, 2 LRLs from Africa and 2 high-resource languages (HRLs) (e.g., English and Brazilian Portuguese). Our results indicate that the LLMs perform worse for the part of speech (POS) labeling of LRLs in comparison to HRLs. We explain the reasons behind this failure and provide an error analysis through examples observed in our data set.
Authors: Sebastian Schmidt, Ines Zelch, Janek Bevendorff, Benno Stein, Matthias Hagen, Martin Potthast
Abstract: Conversational search engines such as YouChat and Microsoft Copilot use large language models (LLMs) to generate responses to queries. It is only a small step to also let the same technology insert ads within the generated responses - instead of separately placing ads next to a response. Inserted ads would be reminiscent of native advertising and product placement, both of which are very effective forms of subtle and manipulative advertising. Considering the high computational costs associated with LLMs, for which providers need to develop sustainable business models, users of conversational search engines may very well be confronted with generated native ads in the near future. In this paper, we thus take a first step to investigate whether LLMs can also be used as a countermeasure, i.e., to block generated native ads. We compile the Webis Generated Native Ads 2024 dataset of queries and generated responses with automatically inserted ads, and evaluate whether LLMs or fine-tuned sentence transformers can detect the ads. In our experiments, the investigated LLMs struggle with the task but sentence transformers achieve precision and recall values above 0.9.
Authors: Tarun Gupta, Wenbo Gong, Chao Ma, Nick Pawlowski, Agrin Hilmkil, Meyer Scetbon, Marc Rigter, Ade Famoti, Ashley Juan Llorens, Jianfeng Gao, Stefan Bauer, Danica Kragic, Bernhard Sch\"olkopf, Cheng Zhang
Abstract: Recent advances in foundation models, especially in large multi-modal models and conversational agents, have ignited interest in the potential of generally capable embodied agents. Such agents will require the ability to perform new tasks in many different real-world environments. However, current foundation models fail to accurately model physical interactions and are therefore insufficient for Embodied AI. The study of causality lends itself to the construction of veridical world models, which are crucial for accurately predicting the outcomes of possible interactions. This paper focuses on the prospects of building foundation world models for the upcoming generation of embodied agents and presents a novel viewpoint on the significance of causality within these. We posit that integrating causal considerations is vital to facilitating meaningful physical interactions with the world. Finally, we demystify misconceptions about causality in this context and present our outlook for future research.
Authors: Junzhe Zhang, Huixuan Zhang, Xunjian Yin, Xiaojun Wan
Abstract: News image captioning requires model to generate an informative caption rich in entities, with the news image and the associated news article. Though Multimodal Large Language Models (MLLMs) have demonstrated remarkable capabilities in addressing various vision-language tasks, our research finds that current MLLMs still bear limitations in handling entity information on news image captioning task. Besides, while MLLMs have the ability to process long inputs, generating high-quality news image captions still requires a trade-off between sufficiency and conciseness of textual input information. To explore the potential of MLLMs and address problems we discovered, we propose : an Entity-Aware Multimodal Alignment based approach for news image captioning. Our approach first aligns the MLLM through Balance Training Strategy with two extra alignment tasks: Entity-Aware Sentence Selection task and Entity Selection task, together with News Image Captioning task, to enhance its capability in handling multimodal entity information. The aligned MLLM will utilizes the additional entity-related information it explicitly extracts to supplement its textual input while generating news image captions. Our approach achieves better results than all previous models in CIDEr score on GoodNews dataset (72.33 -> 88.39) and NYTimes800k dataset (70.83 -> 85.61).
Authors: Ekaterina Trofimova, Emil Sataev, Andrey E. Ustyuzhanin
Abstract: In the ever-evolving landscape of machine learning, seamless translation of natural language descriptions into executable code remains a formidable challenge. This paper introduces Linguacodus, an innovative framework designed to tackle this challenge by deploying a dynamic pipeline that iteratively transforms natural language task descriptions into code through high-level data-shaping instructions. The core of Linguacodus is a fine-tuned large language model (LLM), empowered to evaluate diverse solutions for various problems and select the most fitting one for a given task. This paper details the fine-tuning process, and sheds light on how natural language descriptions can be translated into functional code. Linguacodus represents a substantial leap towards automated code generation, effectively bridging the gap between task descriptions and executable code. It holds great promise for advancing machine learning applications across diverse domains. Additionally, we propose an algorithm capable of transforming a natural description of an ML task into code with minimal human interaction. In extensive experiments on a vast machine learning code dataset originating from Kaggle, we showcase the effectiveness of Linguacodus. The investigations highlight its potential applications across diverse domains, emphasizing its impact on applied machine learning in various scientific fields.
Authors: Matthias Gerstgrasser, Rylan Schaeffer, Apratim Dey, Rafael Rafailov, Henry Sleight, John Hughes, Tomasz Korbak, Rajashree Agrawal, Dhruv Pai, Andrey Gromov, Daniel A. Roberts, Diyi Yang, David L. Donoho, Sanmi Koyejo
Abstract: The proliferation of generative models, combined with pretraining on web-scale data, raises a timely question: what happens when these models are trained on their own generated outputs? Recent investigations into model-data feedback loops proposed that such loops would lead to a phenomenon termed model collapse, under which performance progressively degrades with each model-data feedback iteration until fitted models become useless. However, those studies largely assumed that new data replace old data over time, where an arguably more realistic assumption is that data accumulate over time. In this paper, we ask: what effect does accumulating data have on model collapse? We empirically study this question by pretraining sequences of language models on text corpora. We confirm that replacing the original real data by each generation's synthetic data does indeed tend towards model collapse, then demonstrate that accumulating the successive generations of synthetic data alongside the original real data avoids model collapse; these results hold across a range of model sizes, architectures, and hyperparameters. We obtain similar results for deep generative models on other types of real data: diffusion models for molecule conformation generation and variational autoencoders for image generation. To understand why accumulating data can avoid model collapse, we use an analytically tractable framework introduced by prior work in which a sequence of linear models are fit to the previous models' outputs. Previous work used this framework to show that if data are replaced, the test error increases with the number of model-fitting iterations; we extend this argument to prove that if data instead accumulate, the test error has a finite upper bound independent of the number of iterations, meaning model collapse no longer occurs.
Authors: Clemencia Siro, Mohammad Aliannejadi, Maarten de Rijke
Abstract: In ad-hoc retrieval, evaluation relies heavily on user actions, including implicit feedback. In a conversational setting such signals are usually unavailable due to the nature of the interactions, and, instead, the evaluation often relies on crowdsourced evaluation labels. The role of user feedback in annotators' assessment of turns in a conversational perception has been little studied. We focus on how the evaluation of task-oriented dialogue systems (TDSs), is affected by considering user feedback, explicit or implicit, as provided through the follow-up utterance of a turn being evaluated. We explore and compare two methodologies for assessing TDSs: one includes the user's follow-up utterance and one without. We use both crowdworkers and large language models (LLMs) as annotators to assess system responses across four aspects: relevance, usefulness, interestingness, and explanation quality. Our findings indicate that there is a distinct difference in ratings assigned by both annotator groups in the two setups, indicating user feedback does influence system evaluation. Workers are more susceptible to user feedback on usefulness and interestingness compared to LLMs on interestingness and relevance. User feedback leads to a more personalized assessment of usefulness by workers, aligning closely with the user's explicit feedback. Additionally, in cases of ambiguous or complex user requests, user feedback improves agreement among crowdworkers. These findings emphasize the significance of user feedback in refining system evaluations and suggest the potential for automated feedback integration in future research. We publicly release the annotated data to foster research in this area.
Authors: Qixin Deng, Qikai Yang, Ruibin Yuan, Yipeng Huang, Yi Wang, Xubo Liu, Zeyue Tian, Jiahao Pan, Ge Zhang, Hanfeng Lin, Yizhi Li, Yinghao Ma, Jie Fu, Chenghua Lin, Emmanouil Benetos, Wenwu Wang, Guangyu Xia, Wei Xue, Yike Guo
Abstract: Music composition represents the creative side of humanity, and itself is a complex task that requires abilities to understand and generate information with long dependency and harmony constraints. While demonstrating impressive capabilities in STEM subjects, current LLMs easily fail in this task, generating ill-written music even when equipped with modern techniques like In-Context-Learning and Chain-of-Thoughts. To further explore and enhance LLMs' potential in music composition by leveraging their reasoning ability and the large knowledge base in music history and theory, we propose ComposerX, an agent-based symbolic music generation framework. We find that applying a multi-agent approach significantly improves the music composition quality of GPT-4. The results demonstrate that ComposerX is capable of producing coherent polyphonic music compositions with captivating melodies, while adhering to user instructions.
Authors: Mladjan Jovanovic, Peter Voss
Abstract: Real-time learning concerns the ability of learning systems to acquire knowledge over time, enabling their adaptation and generalization to novel tasks. It is a critical ability for intelligent, real-world systems, especially when data may be insufficient or difficult to obtain. This review provides a comprehensive analysis of real-time learning in Large Language Models. It synthesizes the state-of-the-art real-time learning paradigms, including continual learning, meta-learning, parameter-efficient learning, and mixture-of-experts learning. We demonstrate their utility for real-time learning by describing specific achievements from these related topics and their critical factors. Finally, the paper highlights current problems and challenges for future research in the field. By consolidating the latest relevant research developments, this review offers a comprehensive understanding of real-time learning and its implications for designing and developing LLM-based learning systems addressing real-world problems.
Authors: Yuxuan Yan, Shunpu Tang, Zhiguo Shi, Qianqian Yang
Abstract: Pre-trained Language Models (PLMs) have shown excellent performance on various downstream tasks after fine-tuning. Nevertheless, the escalating concerns surrounding user privacy have posed significant challenges to centralized training reliant on extensive data collection. Federated learning, which only requires training on the clients and aggregates weights on the server without sharing data, has emerged as a solution. However, the substantial parameter size of PLMs places a significant burden on the computational resources of client devices, while also leading to costly communication expenses. Introducing Parameter-Efficient Fine-Tuning(PEFT) into federated learning can effectively address this problem. However, we observe that the non-IID data in federated learning leads to a gap in performance between the PEFT method and full parameter fine-tuning(FFT). To overcome this, we propose FeDeRA, an improvement over the Low-Rank Adaption(LoRA) method in federated learning. FeDeRA uses the same adapter module as LoRA. However, the difference lies in FeDeRA's initialization of the adapter module by performing Singular Value Decomposition (SVD) on the pre-trained matrix and selecting its principal components. We conducted extensive experiments, using RoBERTa and DeBERTaV3, on six datasets, comparing the methods including FFT and the other three different PEFT methods. FeDeRA outperforms all other PEFT methods and is comparable to or even surpasses the performance of FFT method. We also deployed federated learning on Jetson AGX Orin and compared the time required by different methods to achieve the target accuracy on specific tasks. Compared to FFT, FeDeRA reduces the training time by 95.9\%, 97.9\%, 96.9\% and 97.3\%, 96.5\%, 96.5\% respectively on three tasks using RoBERTa and DeBERTaV3. The overall experiments indicate that FeDeRA achieves good performance while also maintaining efficiency.