Authors: Samuel J. Bell, Mariano Coria Meglioli, Megan Richards, Eduardo S\'anchez, Christophe Ropers, Skyler Wang, Adina Williams, Levent Sagun, Marta R. Costa-juss\`a
Abstract: Text toxicity detection systems exhibit significant biases, producing disproportionate rates of false positives on samples mentioning demographic groups. But what about toxicity detection in speech? To investigate the extent to which text-based biases are mitigated by speech-based systems, we produce a set of high-quality group annotations for the multilingual MuTox dataset, and then leverage these annotations to systematically compare speech- and text-based toxicity classifiers. Our findings indicate that access to speech data during inference supports reduced bias against group mentions, particularly for ambiguous and disagreement-inducing samples. Our results also suggest that improving classifiers, rather than transcription pipelines, is more helpful for reducing group bias. We publicly release our annotations and provide recommendations for future toxicity dataset construction.
Authors: Siheng Li, Cheng Yang, Zesen Cheng, Lemao Liu, Mo Yu, Yujiu Yang, Wai Lam
Abstract: Large language models (LLMs) have achieved substantial progress in processing long contexts but still struggle with long-context reasoning. Existing approaches typically involve fine-tuning LLMs with synthetic data, which depends on annotations from human experts or advanced models like GPT-4, thus restricting further advancements. To address this issue, we investigate the potential for LLMs to self-improve in long-context reasoning and propose \ours, an approach specifically designed for this purpose. This approach is straightforward: we sample multiple outputs for each question, score them with Minimum Bayes Risk, and then apply supervised fine-tuning or preference optimization based on these outputs. Extensive experiments on several leading LLMs demonstrate the effectiveness of \ours, with an absolute improvement of $4.2$ points for Llama-3.1-8B-Instruct. Furthermore, \ours achieves superior performance compared to prior approaches that depend on data produced by human experts or advanced models. We anticipate that this work will open new avenues for self-improvement techniques in long-context scenarios, which are essential for the continual advancement of LLMs.
Authors: Khaoula Chehbouni, Jonathan Cola\c{c}o-Carr, Yash More, Jackie CK Cheung, Golnoosh Farnadi
Abstract: In an effort to mitigate the harms of large language models (LLMs), learning from human feedback (LHF) has been used to steer LLMs towards outputs that are intended to be both less harmful and more helpful. Despite the widespread adoption of LHF in practice, the quality of this feedback and its effectiveness as a safety mitigation technique remain unclear. This study addresses these issues by auditing the widely-used Helpful and Harmless (HH) dataset by Anthropic. Our work includes: (1) a thorough investigation of the dataset's content through both manual and automated evaluation; (2) experiments demonstrating the dataset's impact on models' safety; and (3) an analysis of the 100 most influential papers citing this dataset. Through our audit, we showcase how conceptualization failures and quality issues identified in the HH dataset can create additional harms by leading to disparate safety behaviors across demographic groups. Our findings highlight the need for more nuanced, context-sensitive approaches to safety mitigation in LLMs.
Authors: Jiyao Li, Mingze Ni, Yongshun Gong, Wei Liu
Abstract: Deep learning underpins most of the currently advanced natural language processing (NLP) tasks such as textual classification, neural machine translation (NMT), abstractive summarization and question-answering (QA). However, the robustness of the models, particularly QA models, against adversarial attacks is a critical concern that remains insufficiently explored. This paper introduces QA-Attack (Question Answering Attack), a novel word-level adversarial strategy that fools QA models. Our attention-based attack exploits the customized attention mechanism and deletion ranking strategy to identify and target specific words within contextual passages. It creates deceptive inputs by carefully choosing and substituting synonyms, preserving grammatical integrity while misleading the model to produce incorrect responses. Our approach demonstrates versatility across various question types, particularly when dealing with extensive long textual inputs. Extensive experiments on multiple benchmark datasets demonstrate that QA-Attack successfully deceives baseline QA models and surpasses existing adversarial techniques regarding success rate, semantics changes, BLEU score, fluency and grammar error rate.
Authors: Yihe Zhang, Nabin Pakka, Nian-feng Tzeng
Abstract: Large Language Models (LLMs) have received considerable interest in wide applications lately. During pre-training via massive datasets, such a model implicitly memorizes the factual knowledge of trained datasets in its hidden parameters. However, knowledge held implicitly in parameters often makes its use by downstream applications ineffective due to the lack of common-sense reasoning. In this article, we introduce a general framework that permits to build knowledge bases with an aid of LLMs, tailored for processing Web news. The framework applies a rule-based News Information Extractor (NewsIE) to news items for extracting their relational tuples, referred to as knowledge bases, which are then graph-convoluted with the implicit knowledge facts of news items obtained by LLMs, for their classification. It involves two lightweight components: 1) NewsIE: for extracting the structural information of every news item, in the form of relational tuples; 2) BERTGraph: for graph convoluting the implicit knowledge facts with relational tuples extracted by NewsIE. We have evaluated our framework under different news-related datasets for news category classification, with promising experimental results.
Authors: Jiahui Li, Tai-wei Chang, Fengda Zhang, Kun Kuang, Long Chen
Abstract: Reinforcement learning from human feedback (RLHF) provides a paradigm for aligning large language models (LLMs) with human preferences. This involves the initial training of a reward model based on pairwise human feedback. The reward model is subsequently utilized in reinforcement learning to assess the scores of each generated sentence as a whole, further guiding the optimization of LLMs. However, current approaches have a significant shortcoming: \emph{They allocate a single, sparse, and delayed reward to an entire sequence of output}. This may overlook some significant individual contributions of each token towards the desired outcome. To overcome this limitation, our paper proposes a novel reward redistribution method called R3HF, which facilitates a more fine-grained, token-level reward allocation. Specifically, our method treats the reward prediction task of the reward model as a regression problem. As a result, the redistributed rewards are computed by evaluating the specific contribution of each token to the reward model's output. This detailed approach improves the model's understanding of language nuances, leading to more precise enhancements in its performance. Our method is crafted to integrate seamlessly with most current techniques while incurring minimal computational costs. Through comprehensive experiments across diverse datasets and tasks, we have verified the effectiveness and superiority of our approach.
Authors: Hui Dai, Ryan Teehan, Mengye Ren
Abstract: Many existing evaluation benchmarks for Large Language Models (LLMs) quickly become outdated due to the emergence of new models and training data. These benchmarks also fall short in assessing how LLM performance changes over time, as they consist of static questions without a temporal dimension. To address these limitations, we propose using future event prediction as a continuous evaluation method to assess LLMs' temporal generalization and forecasting abilities. Our benchmark, Daily Oracle, automatically generates question-answer (QA) pairs from daily news, challenging LLMs to predict "future" event outcomes. Our findings reveal that as pre-training data becomes outdated, LLM performance degrades over time. While Retrieval Augmented Generation (RAG) has the potential to enhance prediction accuracy, the performance degradation pattern persists, highlighting the need for continuous model updates.
Authors: Shayekh Bin Islam, Ridwanul Hasan Tanvir, Sihat Afnan
Abstract: Bangla is the seventh most spoken language by a total number of speakers in the world, and yet the development of an automated grammar checker in this language is an understudied problem. Bangla grammatical error detection is a task of detecting sub-strings of a Bangla text that contain grammatical, punctuation, or spelling errors, which is crucial for developing an automated Bangla typing assistant. Our approach involves breaking down the task as a token classification problem and utilizing state-of-the-art transformer-based models. Finally, we combine the output of these models and apply rule-based post-processing to generate a more reliable and comprehensive result. Our system is evaluated on a dataset consisting of over 25,000 texts from various sources. Our best model achieves a Levenshtein distance score of 1.04. Finally, we provide a detailed analysis of different components of our system.
Authors: Shangfeng Chen, Xiayang Shi, Pu Li, Yinlin Li, Jingjing Liu
Abstract: Large language models (LLMs) have demonstrated remarkable proficiency in machine translation (MT), even without specific training on the languages in question. However, translating rare words in low-resource or domain-specific contexts remains challenging for LLMs. To address this issue, we propose a multi-step prompt chain that enhances translation faithfulness by prioritizing key terms crucial for semantic accuracy. Our method first identifies these keywords and retrieves their translations from a bilingual dictionary, integrating them into the LLM's context using Retrieval-Augmented Generation (RAG). We further mitigate potential output hallucinations caused by long prompts through an iterative self-checking mechanism, where the LLM refines its translations based on lexical and semantic constraints. Experiments using Llama and Qwen as base models on the FLORES-200 and WMT datasets demonstrate significant improvements over baselines, highlighting the effectiveness of our approach in enhancing translation faithfulness and robustness, particularly in low-resource scenarios.
Authors: Biraj Silwal
Abstract: The distributed representations currently used are dense and uninterpretable, leading to interpretations that themselves are relative, overcomplete, and hard to interpret. We propose a method that transforms these word vectors into reduced syntactic representations. The resulting representations are compact and interpretable allowing better visualization and comparison of the word vectors and we successively demonstrate that the drawn interpretations are in line with human judgment. The syntactic representations are then used to create hierarchical word vectors using an incremental learning approach similar to the hierarchical aspect of human learning. As these representations are drawn from pre-trained vectors, the generation process and learning approach are computationally efficient. Most importantly, we find out that syntactic representations provide a plausible interpretation of the vectors and subsequent hierarchical vectors outperform the original vectors in benchmark tests.
Authors: Aoi Ito, Kota Dohi, Yohei Kawaguchi
Abstract: This paper proposes a foundation model called "CLaSP" that can search time series signals using natural language that describes the characteristics of the signals as queries. Previous efforts to represent time series signal data in natural language have had challenges in designing a conventional class of time series signal characteristics, formulating their quantification, and creating a dictionary of synonyms. To overcome these limitations, the proposed method introduces a neural network based on contrastive learning. This network is first trained using the datasets TRUCE and SUSHI, which consist of time series signals and their corresponding natural language descriptions. Previous studies have proposed vocabularies that data analysts use to describe signal characteristics, and SUSHI was designed to cover these terms. We believe that a neural network trained on these datasets will enable data analysts to search using natural language vocabulary. Furthermore, our method does not require a dictionary of predefined synonyms, and it leverages common sense knowledge embedded in a large-scale language model (LLM). Experimental results demonstrate that CLaSP enables natural language search of time series signal data and can accurately learn the points at which signal data changes.
Authors: Minh Nguyen, Ehsan Shareghi
Abstract: Language agents have shown promising adaptability in dynamic environments to perform complex tasks. However, despite the versatile knowledge embedded in large language models, these agents still fall short when it comes to tasks that require planning. We introduce STEP, a novel framework designed to efficiently learn from previous experiences to enhance the planning capabilities of language agents in future steps. Concretely, STEP functions through four interconnected components. First, the Planner takes on the task, breaks it down into subtasks and provides relevant insights. Then the Executor generates action candidates, while the Evaluator ensures the actions align with learned rules from previous experiences. Lastly, Memory stores experiences to inform future decisions. In the ScienceWorld benchmark, our results show that STEP consistently outperforms state-of-the-art models, achieving an overall score of 67.4 and successfully completing 12 out of 18 tasks. These findings highlight STEP's potential as a framework for enhancing planning capabilities in language agents, paving the way for more sophisticated task-solving in dynamic environments.
Authors: Junhua Liu, Kwan Hui Lim, Roy Ka-Wei Lee
Abstract: How objective and unbiased are we while making decisions? This work investigates cognitive bias identification in high-stake decision making process by human experts, questioning its effectiveness in real-world settings, such as candidates assessments for university admission. We begin with a statistical analysis assessing correlations among different decision points among in the current process, which discovers discrepancies that imply cognitive bias and inconsistency in decisions. This motivates our exploration of bias-aware AI-augmented workflow that surpass human judgment. We propose BGM-HAN, a hierarchical attention network enhanced by byte-pair encoding, multi-head attention and gated residual connection. Using it as backbone model, we further propose a Shortlist-Analyse-Recommend (SAR) agentic workflow, which simulate real-world decision-making. In our experiments, both the proposed model and the agentic workflow significantly improves on both human judgment and alternative models, validated with real-world data.
Authors: Deyi Ji, Lanyun Zhu, Siqi Gao, Peng Xu, Hongtao Lu, Jieping Ye, Feng Zhao
Abstract: The ubiquity and value of tables as semi-structured data across various domains necessitate advanced methods for understanding their complexity and vast amounts of information. Despite the impressive capabilities of large language models (LLMs) in advancing the natural language understanding frontier, their application to large-scale tabular data presents significant challenges, specifically regarding table size and complex intricate relationships. Existing works have shown promise with small-scale tables but often flounder when tasked with the complex reasoning required by larger, interconnected tables found in real-world scenarios. To address this gap, we introduce "Tree-of-Table", a novel approach designed to enhance LLMs' reasoning capabilities over large and complex tables. Our method employs Table Condensation and Decomposition to distill and reorganize relevant data into a manageable format, followed by the construction of a hierarchical Table-Tree that facilitates tree-structured reasoning. Through a meticulous Table-Tree Execution process, we systematically unravel the tree-structured reasoning chain to derive the solutions. Experiments across diverse datasets, including WikiTQ, TableFact, FeTaQA, and BIRD, demonstrate that Tree-of-Table sets a new benchmark with superior performance, showcasing remarkable efficiency and generalization capabilities in large-scale table reasoning.
Authors: Xiaohao Yang, He Zhao, Weijie Xu, Yuanyuan Qi, Jueqing Lu, Dinh Phung, Lan Du
Abstract: Topic modeling is a fundamental task in natural language processing, allowing the discovery of latent thematic structures in text corpora. While Large Language Models (LLMs) have demonstrated promising capabilities in topic discovery, their direct application to topic modeling suffers from issues such as incomplete topic coverage, misalignment of topics, and inefficiency. To address these limitations, we propose LLM-ITL, a novel LLM-in-the-loop framework that integrates LLMs with many existing Neural Topic Models (NTMs). In LLM-ITL, global topics and document representations are learned through the NTM, while an LLM refines the topics via a confidence-weighted Optimal Transport (OT)-based alignment objective. This process enhances the interpretability and coherence of the learned topics, while maintaining the efficiency of NTMs. Extensive experiments demonstrate that LLM-ITL can help NTMs significantly improve their topic interpretability while maintaining the quality of document representation.
Authors: Suhas S Kowshik, Abhishek Divekar, Vijit Malik
Abstract: Large language models (LLMs) have demonstrated remarkable performance in diverse tasks using zero-shot and few-shot prompting. Even though their capabilities of data synthesis have been studied well in recent years, the generated data suffers from a lack of diversity, less adherence to the prompt, and potential biases that creep into the data from the generator model. In this work, we tackle the challenge of generating datasets with high diversity, upon which a student model is trained for downstream tasks. Taking the route of decoding-time guidance-based approaches, we propose CorrSynth, which generates data that is more diverse and faithful to the input prompt using a correlated sampling strategy. Further, our method overcomes the complexity drawbacks of some other guidance-based techniques like classifier-based guidance. With extensive experiments, we show the effectiveness of our approach and substantiate our claims. In particular, we perform intrinsic evaluation to show the improvements in diversity. Our experiments show that CorrSynth improves both student metrics and intrinsic metrics upon competitive baselines across four datasets, showing the innate advantage of our method.
Authors: Felix Stahlberg, Jared Lichtarge, Shankar Kumar
Abstract: We propose a novel parameter-efficient training (PET) method for large language models that adapts models to downstream tasks by optimizing a small subset of the existing model parameters. Unlike prior methods, this subset is not fixed in location but rather which parameters are modified evolves over the course of training. This dynamic parameter selection can yield good performance with many fewer parameters than extant methods. Our method enables a seamless scaling of the subset size across an arbitrary proportion of the total model size, while popular PET approaches like prompt tuning and LoRA cover only a small part of this spectrum. We match or outperform prompt tuning and LoRA in most cases on a variety of NLP tasks (MT, QA, GSM8K, SuperGLUE) for a given parameter budget across different model families and sizes.
Authors: Shaden Shaar, Wayne Chen, Maitreyi Chatterjee, Barry Wang, Wenting Zhao, Claire Cardie
Abstract: Most existing work on event extraction has focused on sentence-level texts and presumes the identification of a trigger-span -- a word or phrase in the input that evokes the occurrence of an event of interest. Event arguments are then extracted with respect to the trigger. Indeed, triggers are treated as integral to, and trigger detection as an essential component of, event extraction. In this paper, we provide the first investigation of the role of triggers for the more difficult and much less studied task of document-level event extraction. We analyze their usefulness in multiple end-to-end and pipelined neural event extraction models for three document-level event extraction datasets, measuring performance using triggers of varying quality (human-annotated, LLM-generated, keyword-based, and random). Our research shows that trigger effectiveness varies based on the extraction task's characteristics and data quality, with basic, automatically-generated triggers serving as a viable alternative to human-annotated ones. Furthermore, providing detailed event descriptions to the extraction model helps maintain robust performance even when trigger quality degrades. Perhaps surprisingly, we also find that the mere existence of trigger input, even random ones, is important for prompt-based LLM approaches to the task.
Authors: Qikai Wei, Mingzhi Yang, Chunlong Han, Jingfu Wei, Minghao Zhang, Feifei Shi, Huansheng Ning
Abstract: Retrieval-Augmented Generation (RAG) mitigates the issue of hallucination in Large Language Models (LLMs) by integrating information retrieval techniques. However, in the tourism domain, since the query is usually brief and the content in the database is diverse, existing RAG may contain a significant amount of irrelevant or contradictory information contents after retrieval. To address this challenge, we propose the QCG-Rerank model. This model first performs an initial retrieval to obtain candidate chunks and then enhances semantics by extracting critical information to expand the original query. Next, we utilize the expanded query and candidate chunks to calculate similarity scores as the initial transition probability and construct the chunks graph. Subsequently, We iteratively compute the transition probabilities based on an initial estimate until convergence. The chunks with the highest score are selected and input into the LLMs to generate responses. We evaluate the model on Cultour, IIRC, StrategyQA, HotpotQA, SQuAD, and MuSiQue datasets. The experimental results demonstrate the effectiveness and superiority of the QCG-Rerank method.
Authors: Rui Liu, Jiayou Liang, Haolong Chen, Yujia Hu
Abstract: This article applies natural language processing (NLP) to extract and quantify textual information to predict stock performance. Using an extensive dataset of Chinese analyst reports and employing a customized BERT deep learning model for Chinese text, this study categorizes the sentiment of the reports as positive, neutral, or negative. The findings underscore the predictive capacity of this sentiment indicator for stock volatility, excess returns, and trading volume. Specifically, analyst reports with strong positive sentiment will increase excess return and intraday volatility, and vice versa, reports with strong negative sentiment also increase volatility and trading volume, but decrease future excess return. The magnitude of this effect is greater for positive sentiment reports than for negative sentiment reports. This article contributes to the empirical literature on sentiment analysis and the response of the stock market to news in the Chinese stock market.
Authors: Somanshu Singla, Zhen Wang, Tianyang Liu, Abdullah Ashfaq, Zhiting Hu, Eric P. Xing
Abstract: Aligning Large Language Models (LLMs) traditionally relies on costly training and human preference annotations. Self-alignment seeks to reduce these expenses by enabling models to align themselves. To further lower costs and achieve alignment without any expensive tuning or annotations, we introduce a new tuning-free approach for self-alignment, Dynamic Rewarding with Prompt Optimization (\ours). Our approach leverages a search-based optimization framework that allows LLMs to iteratively self-improve and craft the optimal alignment instructions, all without additional training or human intervention. The core of \ours is a dynamic rewarding mechanism, which identifies and rectifies model-specific alignment weaknesses, allowing LLMs to adapt efficiently to diverse alignment challenges. Empirical evaluations on eight recent LLMs, both open- and closed-sourced, demonstrate that \ours significantly enhances alignment performance, with base models outperforming their SFT/RLHF-tuned counterparts. Moreover, the prompts automatically optimized by \ours surpass those curated by human experts, further validating the effectiveness of our approach. Our findings highlight the great potential of current LLMs to achieve adaptive self-alignment through inference-time optimization, complementing tuning-based alignment methods.
Authors: Dingdong Wang, Mingyu Cui, Dongchao Yang, Xueyuan Chen, Helen Meng
Abstract: With the rise of Speech Large Language Models (Speech LLMs), there has been growing interest in discrete speech tokens for their ability to integrate with text-based tokens seamlessly. Compared to most studies that focus on continuous speech features, although discrete-token based LLMs have shown promising results on certain tasks, the performance gap between these two paradigms is rarely explored. In this paper, we present a fair and thorough comparison between discrete and continuous features across a variety of semantic-related tasks using a light-weight LLM (Qwen1.5-0.5B). Our findings reveal that continuous features generally outperform discrete tokens, particularly in tasks requiring fine-grained semantic understanding. Moreover, this study goes beyond surface-level comparison by identifying key factors behind the under-performance of discrete tokens, such as limited token granularity and inefficient information retention. To enhance the performance of discrete tokens, we explore potential aspects based on our analysis. We hope our results can offer new insights into the opportunities for advancing discrete speech tokens in Speech LLMs.
Authors: Cl\'ement Dumas, Chris Wendler, Veniamin Veselovsky, Giovanni Monea, Robert West
Abstract: A central question in multilingual language modeling is whether large language models (LLMs) develop a universal concept representation, disentangled from specific languages. In this paper, we address this question by analyzing latent representations (latents) during a word translation task in transformer-based LLMs. We strategically extract latents from a source translation prompt and insert them into the forward pass on a target translation prompt. By doing so, we find that the output language is encoded in the latent at an earlier layer than the concept to be translated. Building on this insight, we conduct two key experiments. First, we demonstrate that we can change the concept without changing the language and vice versa through activation patching alone. Second, we show that patching with the mean over latents across different languages does not impair and instead improves the models' performance in translating the concept. Our results provide evidence for the existence of language-agnostic concept representations within the investigated models.
Authors: Benedetta Muscato, Praveen Bushipaka, Gizem Gezici, Lucia Passaro, Fosca Giannotti
Abstract: Subjective NLP tasks usually rely on human annotations provided by multiple annotators, whose judgments may vary due to their diverse backgrounds and life experiences. Traditional methods often aggregate multiple annotations into a single ground truth, disregarding the diversity in perspectives that arises from annotator disagreement. In this preliminary study, we examine the effect of including multiple annotations on model accuracy in classification. Our methodology investigates the performance of perspective-aware classification models in stance detection task and further inspects if annotator disagreement affects the model confidence. The results show that multi-perspective approach yields better classification performance outperforming the baseline which uses the single label. This entails that designing more inclusive perspective-aware AI models is not only an essential first step in implementing responsible and ethical AI, but it can also achieve superior results than using the traditional approaches.
Authors: Nghia Trung Ngo, Thien Huu Nguyen
Abstract: The majority of previous researches addressing multi-lingual IE are limited to zero-shot cross-lingual single-transfer (one-to-one) setting, with high-resource languages predominantly as source training data. As a result, these works provide little understanding and benefit for the realistic goal of developing a multi-lingual IE system that can generalize to as many languages as possible. Our study aims to fill this gap by providing a detailed analysis on Cross-Lingual Multi-Transferability (many-to-many transfer learning), for the recent IE corpora that cover a diverse set of languages. Specifically, we first determine the correlation between single-transfer performance and a wide range of linguistic-based distances. From the obtained insights, a combined language distance metric can be developed that is not only highly correlated but also robust across different tasks and model scales. Next, we investigate the more general zero-shot multi-lingual transfer settings where multiple languages are involved in the training and evaluation processes. Language clustering based on the newly defined distance can provide directions for achieving the optimal cost-performance trade-off in data (languages) selection problem. Finally, a relational-transfer setting is proposed to further incorporate multi-lingual unlabeled data based on adversarial training using the relation induced from the above linguistic distance.
Authors: Wissam Antoun, Francis Kulumba, Rian Touchent, \'Eric de la Clergerie, Beno\^it Sagot, Djam\'e Seddah
Abstract: French language models, such as CamemBERT, have been widely adopted across industries for natural language processing (NLP) tasks, with models like CamemBERT seeing over 4 million downloads per month. However, these models face challenges due to temporal concept drift, where outdated training data leads to a decline in performance, especially when encountering new topics and terminology. This issue emphasizes the need for updated models that reflect current linguistic trends. In this paper, we introduce two new versions of the CamemBERT base model-CamemBERTav2 and CamemBERTv2-designed to address these challenges. CamemBERTav2 is based on the DeBERTaV3 architecture and makes use of the Replaced Token Detection (RTD) objective for better contextual understanding, while CamemBERTv2 is built on RoBERTa, which uses the Masked Language Modeling (MLM) objective. Both models are trained on a significantly larger and more recent dataset with longer context length and an updated tokenizer that enhances tokenization performance for French. We evaluate the performance of these models on both general-domain NLP tasks and domain-specific applications, such as medical field tasks, demonstrating their versatility and effectiveness across a range of use cases. Our results show that these updated models vastly outperform their predecessors, making them valuable tools for modern NLP systems. All our new models, as well as intermediate checkpoints, are made openly available on Huggingface.
Authors: Daniel P. Jeong, Pranav Mani, Saurabh Garg, Zachary C. Lipton, Michael Oberst
Abstract: Several recent works seek to develop foundation models specifically for medical applications, adapting general-purpose large language models (LLMs) and vision-language models (VLMs) via continued pretraining on publicly available biomedical corpora. These works typically claim that such domain-adaptive pretraining (DAPT) improves performance on downstream medical tasks, such as answering medical licensing exam questions. In this paper, we compare ten public "medical" LLMs and two VLMs against their corresponding base models, arriving at a different conclusion: all medical VLMs and nearly all medical LLMs fail to consistently improve over their base models in the zero-/few-shot prompting and supervised fine-tuning regimes for medical question-answering (QA). For instance, across all tasks and model pairs we consider in the 3-shot setting, medical LLMs only outperform their base models in 22.7% of cases, reach a (statistical) tie in 36.8% of cases, and are significantly worse than their base models in the remaining 40.5% of cases. Our conclusions are based on (i) comparing each medical model head-to-head, directly against the corresponding base model; (ii) optimizing the prompts for each model separately in zero-/few-shot prompting; and (iii) accounting for statistical uncertainty in comparisons. While these basic practices are not consistently adopted in the literature, our ablations show that they substantially impact conclusions. Meanwhile, we find that after fine-tuning on specific QA tasks, medical LLMs can show performance improvements, but the benefits do not carry over to tasks based on clinical notes. Our findings suggest that state-of-the-art general-domain models may already exhibit strong medical knowledge and reasoning capabilities, and offer recommendations to strengthen the conclusions of future studies.
Authors: Muzhi Li, Cehao Yang, Chengjin Xu, Xuhui Jiang, Yiyan Qi, Jian Guo, Ho-fung Leung, Irwin King
Abstract: The Knowledge Graph Completion~(KGC) task aims to infer the missing entity from an incomplete triple. Existing embedding-based methods rely solely on triples in the KG, which is vulnerable to specious relation patterns and long-tail entities. On the other hand, text-based methods struggle with the semantic gap between KG triples and natural language. Apart from triples, entity contexts (e.g., labels, descriptions, aliases) also play a significant role in augmenting KGs. To address these limitations, we propose KGR3, a context-enriched framework for KGC. KGR3 is composed of three modules. Firstly, the Retrieval module gathers supporting triples from the KG, collects plausible candidate answers from a base embedding model, and retrieves context for each related entity. Then, the Reasoning module employs a large language model to generate potential answers for each query triple. Finally, the Re-ranking module combines candidate answers from the two modules mentioned above, and fine-tunes an LLM to provide the best answer. Extensive experiments on widely used datasets demonstrate that KGR3 consistently improves various KGC methods. Specifically, the best variant of KGR3 achieves absolute Hits@1 improvements of 12.3% and 5.6% on the FB15k237 and WN18RR datasets.
Authors: Shivani Upadhyay, Ronak Pradeep, Nandan Thakur, Daniel Campos, Nick Craswell, Ian Soboroff, Hoa Trang Dang, Jimmy Lin
Abstract: The application of large language models to provide relevance assessments presents exciting opportunities to advance information retrieval, natural language processing, and beyond, but to date many unknowns remain. This paper reports on the results of a large-scale evaluation (the TREC 2024 RAG Track) where four different relevance assessment approaches were deployed in situ: the "standard" fully manual process that NIST has implemented for decades and three different alternatives that take advantage of LLMs to different extents using the open-source UMBRELA tool. This setup allows us to correlate system rankings induced by the different approaches to characterize tradeoffs between cost and quality. We find that in terms of nDCG@20, nDCG@100, and Recall@100, system rankings induced by automatically generated relevance assessments from UMBRELA correlate highly with those induced by fully manual assessments across a diverse set of 77 runs from 19 teams. Our results suggest that automatically generated UMBRELA judgments can replace fully manual judgments to accurately capture run-level effectiveness. Surprisingly, we find that LLM assistance does not appear to increase correlation with fully manual assessments, suggesting that costs associated with human-in-the-loop processes do not bring obvious tangible benefits. Overall, human assessors appear to be stricter than UMBRELA in applying relevance criteria. Our work validates the use of LLMs in academic TREC-style evaluations and provides the foundation for future studies.
Authors: Jingyi Zhou, Senlin Luo, Haofan Chen
Abstract: Emotion and personality are central elements in understanding human psychological states. Emotions reflect an individual subjective experiences, while personality reveals relatively stable behavioral and cognitive patterns. Existing affective computing datasets often annotate emotion and personality traits separately, lacking fine-grained labeling of micro-emotions and emotion intensity in both single-label and multi-label classifications. Chinese emotion datasets are extremely scarce, and datasets capturing Chinese user personality traits are even more limited. To address these gaps, this study collected data from the major social media platform Weibo, screening 11,338 valid users from over 50,000 individuals with diverse MBTI personality labels and acquiring 566,900 posts along with the user MBTI personality tags. Using the EQN method, we compiled a multi-label Chinese affective computing dataset that integrates the same user's personality traits with six emotions and micro-emotions, each annotated with intensity levels. Validation results across multiple NLP classification models demonstrate the dataset strong utility. This dataset is designed to advance machine recognition of complex human emotions and provide data support for research in psychology, education, marketing, finance, and politics.
Authors: Siraj Munir, Alessandro Aldini
Abstract: Large Language Models (LLMs) are revolutionizing the landscape of Generative Artificial Intelligence (GenAI), with innovative LLM-backed solutions emerging rapidly. However, when applied to database technologies, specifically query generation for graph databases and Knowledge Graphs (KGs), LLMs still face significant challenges. While research on LLM-driven query generation for Structured Query Language (SQL) exists, similar systems for graph databases remain underdeveloped. This paper presents a comparative study addressing the challenge of generating Cypher queries a powerful language for interacting with graph databases using open-access LLMs. We rigorously evaluate several LLM agents (OpenAI ChatGPT 4o, Claude Sonnet 3.5, Google Gemini Pro 1.5, and a locally deployed Llama 3.1 8B) using a designed few-shot learning prompt and Retrieval Augmented Generation (RAG) backed by Chain-of-Thoughts (CoT) reasoning. Our empirical analysis of query generation accuracy reveals that Claude Sonnet 3.5 outperforms its counterparts in this specific domain. Further, we highlight promising future research directions to address the identified limitations and advance LLM-driven query generation for graph databases.
Authors: Abhinav Java, Simra Shahid, Chirag Agarwal
Abstract: The widespread practice of indiscriminate data scraping to fine-tune language models (LMs) raises significant legal and ethical concerns, particularly regarding compliance with data protection laws such as the General Data Protection Regulation (GDPR). This practice often results in the unauthorized use of personal information, prompting growing debate within the academic and regulatory communities. Recent works have introduced the concept of generating unlearnable datasets (by adding imperceptible noise to the clean data), such that the underlying model achieves lower loss during training but fails to generalize to the unseen test setting. Though somewhat effective, these approaches are predominantly designed for images and are limited by several practical constraints like requiring knowledge of the target model. To this end, we introduce RegText, a framework that injects imperceptible spurious correlations into natural language datasets, effectively rendering them unlearnable without affecting semantic content. We demonstrate RegText's utility through rigorous empirical analysis of small and large LMs. Notably, RegText can restrict newer models like GPT-4o and Llama from learning on our generated data, resulting in a drop in their test accuracy compared to their zero-shot performance and paving the way for generating unlearnable text to protect public data.
Authors: Yingqi Gao, Yifu Liu, Xiaoxia Li, Xiaorong Shi, Yin Zhu, Yiming Wang, Shiqi Li, Wei Li, Yuntao Hong, Zhiling Luo, Jinyang Gao, Liyu Mou, Yu Li
Abstract: To tackle the challenges of large language model performance in natural language to SQL tasks, we introduce XiYan-SQL, an innovative framework that employs a multi-generator ensemble strategy to improve candidate generation. We introduce M-Schema, a semi-structured schema representation method designed to enhance the understanding of database structures. To enhance the quality and diversity of generated candidate SQL queries, XiYan-SQL integrates the significant potential of in-context learning (ICL) with the precise control of supervised fine-tuning. On one hand, we propose a series of training strategies to fine-tune models to generate high-quality candidates with diverse preferences. On the other hand, we implement the ICL approach with an example selection method based on named entity recognition to prevent overemphasis on entities. The refiner optimizes each candidate by correcting logical or syntactical errors. To address the challenge of identifying the best candidate, we fine-tune a selection model to distinguish nuances of candidate SQL queries. The experimental results on multiple dialect datasets demonstrate the robustness of XiYan-SQL in addressing challenges across different scenarios. Overall, our proposed XiYan-SQL achieves the state-of-the-art execution accuracy of 89.65% on the Spider test set, 69.86% on SQL-Eval, 41.20% on NL2GQL, and a competitive score of 72.23% on the Bird development benchmark. The proposed framework not only enhances the quality and diversity of SQL queries but also outperforms previous methods.
Authors: L\'aszl\'o Kozma, Johannes Voderholzer
Abstract: Byte-Pair Encoding (BPE) is a widely used method for subword tokenization, with origins in grammar-based text compression. It is employed in a variety of language processing tasks such as machine translation or large language model (LLM) pretraining, to create a token dictionary of a prescribed size. Most evaluations of BPE to date are empirical, and the reasons for its good practical performance are not well understood. In this paper we focus on the optimization problem underlying BPE: finding a pair encoding that achieves optimal compression utility. We show that this problem is APX-complete, indicating that it is unlikely to admit a polynomial-time approximation scheme. This answers, in a stronger form, a question recently raised by Zouhar et al. On the positive side, we show that BPE approximates the compression utility of the optimal pair encoding to a worst-case factor between $0.333$ and $0.625$. Our results aim to explain the ongoing success of BPE and are, to our knowledge, the first rigorous guarantees on its compression utility that hold for all inputs.
Authors: Harry Mayne, Yushi Yang, Adam Mahdi
Abstract: Steering vectors are a promising approach to control the behaviour of large language models. However, their underlying mechanisms remain poorly understood. While sparse autoencoders (SAEs) may offer a potential method to interpret steering vectors, recent findings show that SAE-reconstructed vectors often lack the steering properties of the original vectors. This paper investigates why directly applying SAEs to steering vectors yields misleading decompositions, identifying two reasons: (1) steering vectors fall outside the input distribution for which SAEs are designed, and (2) steering vectors can have meaningful negative projections in feature directions, which SAEs are not designed to accommodate. These limitations hinder the direct use of SAEs for interpreting steering vectors.
Authors: Yang Trista Cao, Lovely-Frances Domingo, Sarah Ann Gilbert, Michelle Mazurek, Katie Shilton, Hal Daum\'e III
Abstract: Extensive efforts in automated approaches for content moderation have been focused on developing models to identify toxic, offensive, and hateful content with the aim of lightening the load for moderators. Yet, it remains uncertain whether improvements on those tasks have truly addressed moderators' needs in accomplishing their work. In this paper, we surface gaps between past research efforts that have aimed to provide automation for aspects of content moderation and the needs of volunteer content moderators, regarding identifying violations of various moderation rules. To do so, we conduct a model review on Hugging Face to reveal the availability of models to cover various moderation rules and guidelines from three exemplar forums. We further put state-of-the-art LLMs to the test, evaluating how well these models perform in flagging violations of platform rules from one particular forum. Finally, we conduct a user survey study with volunteer moderators to gain insight into their perspectives on useful moderation models. Overall, we observe a non-trivial gap, as missing developed models and LLMs exhibit moderate to low performance on a significant portion of the rules. Moderators' reports provide guides for future work on developing moderation assistant models.
Authors: Hanwen Zhang, Qingyi Si, Peng Fu, Zheng Lin, Weiping Wang
Abstract: Table-based Fact Verification (TFV) aims to extract the entailment relation between statements and structured tables. Existing TFV methods based on small-scaled models suffer from insufficient labeled data and weak zero-shot ability. Recently, the appearance of Large Language Models (LLMs) has gained lots of attraction in research fields. They have shown powerful zero-shot and in-context learning abilities on several NLP tasks, but their potential on TFV is still unknown. In this work, we implement a preliminary study about whether LLMs are table-based fact-checkers. In detail, we design diverse prompts to explore how the in-context learning can help LLMs in TFV, i.e., zero-shot and few-shot TFV capability. Besides, we carefully design and construct TFV instructions to study the performance gain brought by the instruction tuning of LLMs. Experimental results demonstrate that LLMs can achieve acceptable results on zero-shot and few-shot TFV with prompt engineering, while instruction-tuning can stimulate the TFV capability significantly. We also make some valuable findings about the format of zero-shot prompts and the number of in-context examples. Finally, we analyze some possible directions to promote the accuracy of TFV via LLMs, which is beneficial to further research of table reasoning.
Authors: Zhiyuan Hu, Chumin Liu, Xidong Feng, Yilun Zhao, See-Kiong Ng, Anh Tuan Luu, Junxian He, Pang Wei Koh, Bryan Hooi
Abstract: In the face of uncertainty, the ability to *seek information* is of fundamental importance. In many practical applications, such as medical diagnosis and troubleshooting, the information needed to solve the task is not initially given and has to be actively sought by asking follow-up questions (for example, a doctor asking a patient for more details about their symptoms). In this work, we introduce Uncertainty of Thoughts (UoT), an algorithm to augment large language models with the ability to actively seek information by asking effective questions. UoT combines 1) an *uncertainty-aware simulation approach* which enables the model to simulate possible future scenarios and how likely they are to occur, 2) *uncertainty-based rewards* motivated by information gain which incentivizes the model to seek information, and 3) a *reward propagation scheme* to select the optimal question to ask in a way that maximizes the expected reward. In experiments on medical diagnosis, troubleshooting, and the `20 Questions` game, UoT achieves an average performance improvement of 38.1% in the rate of successful task completion across multiple LLMs compared with direct prompting and also improves efficiency (i.e., the number of questions needed to complete the task). Our code has been released [here](https://github.com/zhiyuanhubj/UoT)
Authors: Michael Ginn (University of Colorado), Lindia Tjuatja (Carnegie Mellon University), Taiqi He (Carnegie Mellon University), Enora Rice (University of Colorado), Graham Neubig (Carnegie Mellon University), Alexis Palmer (University of Colorado), Lori Levin (Carnegie Mellon University)
Abstract: Language documentation projects often involve the creation of annotated text in a format such as interlinear glossed text (IGT), which captures fine-grained morphosyntactic analyses in a morpheme-by-morpheme format. However, there are few existing resources providing large amounts of standardized, easily accessible IGT data, limiting their applicability to linguistic research, and making it difficult to use such data in NLP modeling. We compile the largest existing corpus of IGT data from a variety of sources, covering over 450k examples across 1.8k languages, to enable research on crosslingual transfer and IGT generation. We normalize much of our data to follow a standard set of labels across languages. Furthermore, we explore the task of automatically generating IGT in order to aid documentation projects. As many languages lack sufficient monolingual data, we pretrain a large multilingual model on our corpus. We demonstrate the utility of this model by finetuning it on monolingual corpora, outperforming SOTA models by up to 6.6\%. Our pretrained model and dataset are available on Hugging Face.
Authors: Xin Zhang, Tianjie Ju, Huijia Liang, Ying Fu, Qin Zhang
Abstract: The substantial interest in updating Large Language Models (LLMs) without retraining from scratch is accompanied by several challenges. This is particularly true when updating LLMs with datasets that necessitate domain-expert reasoning across extensive texts, despite limited samples. We termed the scenario as the Few-Shot Domain-Expert Reasoning for Updating LLMs (FDoR-UL). Traditional methods such as Low-Rank Adaptation (LoRA) and Retrieval Augmented Generation (RAG) are inadequate for addressing this critical issue, particularly evident in our exploration of a specific medical dataset that epitomizes the distinct needs of FDoR-UL. To tackle this challenge, we introduce a Sequential Fusion method to integrate knowledge from complex contexts into LLMs. This method employs a two-stage framework: initially leveraging general LLMs to perform relation extraction for knowledge acquisition from complex texts, followed by updating domain-specific LLMs through Knowledge Editing (KE). Employing our method, domain-specific LLMs achieved a 71.7% accuracy (an average gain of 39.1%) in question-answering tasks. Furthermore, we expanded our evaluation to a novel economics-management dataset we developed, where our method achieved a 75.0% accuracy (an average gain of 45.0%). These findings underscore the effectiveness and flexibility of our approach in FDoR-UL across various domains.
Authors: Meiqi Chen, Yixin Cao, Yan Zhang, Chaochao Lu
Abstract: Recent advancements in Large Language Models (LLMs) have facilitated the development of Multimodal LLMs (MLLMs). Despite their impressive capabilities, MLLMs often suffer from over-reliance on unimodal biases (e.g., language bias and vision bias), leading to incorrect answers or hallucinations in complex multimodal tasks. To investigate this issue, we propose a causal framework to interpret the biases in Visual Question Answering (VQA) problems. Within this framework, we conduct an in-depth causal analysis to assess the causal effect of these biases on MLLM predictions. Based on the analysis, we introduce 1) a novel MORE dataset with 12,000 challenging VQA instances requiring multi-hop reasoning and overcoming unimodal biases. 2) a causality-enhanced agent framework CAVE that guides models to comprehensively integrate information from different modalities and mitigate biases. Our experiments show that MLLMs perform poorly on MORE, indicating strong unimodal biases and limited semantic understanding. However, when integrated with our CAVE, promising improvements in reasoning and bias mitigation can be seen. These findings provide important insights for the development of more robust MLLMs and contribute to the broader goal of advancing multimodal AI systems capable of deeper understanding and reasoning. Our project page is at https://github.com/OpenCausaLab/MORE.
Authors: Haoran Lian, Yizhe Xiong, Jianwei Niu, Shasha Mo, Zhenpeng Su, Zijia Lin, Hui Chen, Peng Liu, Jungong Han, Guiguang Ding
Abstract: Byte Pair Encoding (BPE) serves as a foundation method for text tokenization in the Natural Language Processing (NLP) field. Despite its wide adoption, the original BPE algorithm harbors an inherent flaw: it inadvertently introduces a frequency imbalance for tokens in the text corpus. Since BPE iteratively merges the most frequent token pair in the text corpus to generate a new token and keeps all generated tokens in the vocabulary, it unavoidably holds tokens that primarily act as components of a longer token and appear infrequently on their own. We term such tokens as Scaffold Tokens. Due to their infrequent occurrences in the text corpus, Scaffold Tokens pose a learning imbalance issue. To address that issue, we propose Scaffold-BPE, which incorporates a dynamic scaffold token removal mechanism by parameter-free, computation-light, and easy-to-implement modifications to the original BPE method. This novel approach ensures the exclusion of low-frequency Scaffold Tokens from the token representations for given texts, thereby mitigating the issue of frequency imbalance and facilitating model training. On extensive experiments across language modeling and even machine translation, Scaffold-BPE consistently outperforms the original BPE, well demonstrating its effectiveness.
Authors: Abhishek Divekar, Greg Durrett
Abstract: It is often desirable to distill the capabilities of large language models (LLMs) into smaller student models due to compute and memory constraints. One way to do this for classification tasks is via dataset synthesis, which can be accomplished by generating examples of each label from the LLM. Prior approaches to synthesis use few-shot prompting, which relies on the LLM's parametric knowledge to generate usable examples. However, this leads to issues of repetition, bias towards popular entities, and stylistic differences from human text. In this work, we propose Synthesize by Retrieval and Refinement (SynthesizRR), which uses retrieval augmentation to introduce variety into the dataset synthesis process: as retrieved passages vary, the LLM is seeded with different content to generate its examples. We empirically study the synthesis of six datasets, covering topic classification, sentiment analysis, tone detection, and humor, requiring complex synthesis strategies. We find that SynthesizRR greatly improves lexical and semantic diversity, similarity to human-written text, and distillation performance, when compared to 32-shot prompting and four prior approaches. We release our code to perform all steps at https://github.com/amazon-science/synthesizrr
Authors: Aleksandr Nikolich, Konstantin Korolev, Sergei Bratchikov, Igor Kiselev, Artem Shelmanov
Abstract: There has been a surge in developing various Large Language Models (LLMs). However, text generation for languages other than English often faces significant challenges, including poor generation quality and reduced computational performance due to the disproportionate representation of tokens in the model's vocabulary. In this work, we address these issues by developing a pipeline for adapting English-oriented pre-trained models to other languages and constructing efficient bilingual LLMs. Using this pipeline, we construct Vikhr, a state-of-the-art bilingual open-source instruction-following LLM designed specifically for the Russian language. "Vikhr" refers to the name of the Mistral LLM series and means a "strong gust of wind." Unlike previous Russian-language models that typically rely on LoRA adapters on top of English-oriented models, sacrificing performance for lower training costs, Vikhr features an adapted tokenizer vocabulary and undergoes continued pre-training and instruction tuning of all weights. This not only enhances the model's performance but also significantly improves its computational and contextual efficiency. The remarkable performance of Vikhr across various Russian-language benchmarks can also be attributed to our efforts in expanding instruction datasets and corpora for continued pre-training. Vikhr not only sets a new state of the art among open-source LLMs for Russian but even outperforms some proprietary closed-source models on certain benchmarks. The model weights, instruction sets, and code are publicly available.
Authors: Pritika Ramu, Koustava Goswami, Apoorv Saxena, Balaji Vasan Srinivasan
Abstract: Accurately attributing answer text to its source document is crucial for developing a reliable question-answering system. However, attribution for long documents remains largely unexplored. Post-hoc attribution systems are designed to map answer text back to the source document, yet the granularity of this mapping has not been addressed. Furthermore, a critical question arises: What exactly should be attributed? This involves identifying the specific information units within an answer that require grounding. In this paper, we propose and investigate a novel approach to the factual decomposition of generated answers for attribution, employing template-based in-context learning. To accomplish this, we utilize the question and integrate negative sampling during few-shot in-context learning for decomposition. This approach enhances the semantic understanding of both abstractive and extractive answers. We examine the impact of answer decomposition by providing a thorough examination of various attribution approaches, ranging from retrieval-based techniques to LLM-based attributors.
Authors: Dipankar Srirag, Jordan Painter, Aditya Joshi, Diptesh Kanojia
Abstract: Existing benchmarks often fail to account for linguistic diversity, like language variants of English. In this paper, we share our experiences from our ongoing project of building a sentiment classification benchmark for three variants of English: Australian (en-AU), Indian (en-IN), and British (en-UK) English. Using Google Places reviews, we explore the effects of various sampling techniques based on label semantics, review length, and sentiment proportion and report performances on three fine-tuned BERT-based models. Our initial evaluation reveals significant performance variations influenced by sample characteristics, label semantics, and language variety, highlighting the need for nuanced benchmark design. We offer actionable insights for researchers to create robust benchmarks, emphasising the importance of diverse sampling, careful label definition, and comprehensive evaluation across linguistic varieties.
Authors: Yun He, Di Jin, Chaoqi Wang, Chloe Bi, Karishma Mandyam, Hejia Zhang, Chen Zhu, Ning Li, Tengyu Xu, Hongjiang Lv, Shruti Bhosale, Chenguang Zhu, Karthik Abinav Sankararaman, Eryk Helenowski, Melanie Kambadur, Aditya Tayade, Hao Ma, Han Fang, Sinong Wang
Abstract: Large Language Models (LLMs) have demonstrated impressive capabilities in various tasks, including instruction following, which is crucial for aligning model outputs with user expectations. However, evaluating LLMs' ability to follow instructions remains challenging due to the complexity and subjectivity of human language. Current benchmarks primarily focus on single-turn, monolingual instructions, which do not adequately reflect the complexities of real-world applications that require handling multi-turn and multilingual interactions. To address this gap, we introduce Multi-IF, a new benchmark designed to assess LLMs' proficiency in following multi-turn and multilingual instructions. Multi-IF, which utilizes a hybrid framework combining LLM and human annotators, expands upon the IFEval by incorporating multi-turn sequences and translating the English prompts into another 7 languages, resulting in a dataset of 4,501 multilingual conversations, where each has three turns. Our evaluation of 14 state-of-the-art LLMs on Multi-IF reveals that it presents a significantly more challenging task than existing benchmarks. All the models tested showed a higher rate of failure in executing instructions correctly with each additional turn. For example, o1-preview drops from 0.877 at the first turn to 0.707 at the third turn in terms of average accuracy over all languages. Moreover, languages with non-Latin scripts (Hindi, Russian, and Chinese) generally exhibit higher error rates, suggesting potential limitations in the models' multilingual capabilities. We release Multi-IF prompts and the evaluation code base to encourage further research in this critical area.
Authors: Yang Zhong, Jiangang Hao, Michael Fauss, Chen Li, Yuan Wang
Abstract: The recent revolutionary advance in generative AI enables the generation of realistic and coherent texts by large language models (LLMs). Despite many existing evaluation metrics on the quality of the generated texts, there is still a lack of rigorous assessment of how well LLMs perform in complex and demanding writing assessments. This study examines essays generated by ten leading LLMs for the analytical writing assessment of the Graduate Record Exam (GRE). We assessed these essays using both human raters and the e-rater automated scoring engine as used in the GRE scoring pipeline. Notably, the top-performing Gemini and GPT-4o received an average score of 4.78 and 4.67, respectively, falling between "generally thoughtful, well-developed analysis of the issue and conveys meaning clearly" and "presents a competent analysis of the issue and conveys meaning with acceptable clarity" according to the GRE scoring guideline. We also evaluated the detection accuracy of these essays, with detectors trained on essays generated by the same and different LLMs.
Authors: Zimo Qi, Guangliang Liu, Kristen Marie Johnson, Lu Cheng
Abstract: Though intensive attentions to the self-correction capability of Large Language Models (LLMs), the underlying mechanism of this capability is still under-explored. In this paper, we aim to answer two fundamental questions for moral self-correction: (1) how different components in self-correction, such as Chain-of-Thought (CoT) reasoning, external feedback, and instructional prompts, interact to enable moral self-correction; and (2) is the self-correction one of LLMs' innate capabilities? To answer the first question, we examine how different self-correction components interact to intervene the embedded morality within hidden states, therefore contributing to different performance. For the second question, we (i) evaluate the robustness of moral self-correction by introducing natural language interventions of weak evidence into prompts; (ii) propose a validation framework, self-distinguish, that requires effective self-correction to enable LLMs to distinguish between desirable and undesirable outputs. Our experimental results indicate that there is no universally optimal self-correction method for the tasks considered, although external feedback and CoT can contribute to additional performance gains. However, our mechanistic analysis reveals negative interactions among instructional prompts, CoT, and external feedback, suggesting a conflict between internal knowledge and external feedback. The self-distinguish experiments demonstrate that while LLMs can self-correct their responses, they are unable to reliably distinguish between desired and undesired outputs. With our empirical evidence, we can conclude that moral self-correction is not an innate capability of LLMs acquired during pretraining.
Authors: Sshubam Verma, Mohammed Safi Ur Rahman Khan, Vishwajeet Kumar, Rudra Murthy, Jaydeep Sen
Abstract: Evaluating Large Language Models (LLMs) in low-resource and linguistically diverse languages remains a significant challenge in NLP, particularly for languages using non-Latin scripts like those spoken in India. Existing benchmarks predominantly focus on English, leaving substantial gaps in assessing LLM capabilities in these languages. We introduce MILU, a Multi task Indic Language Understanding Benchmark, a comprehensive evaluation benchmark designed to address this gap. MILU spans 8 domains and 42 subjects across 11 Indic languages, reflecting both general and culturally specific knowledge. With an India-centric design, incorporates material from regional and state-level examinations, covering topics such as local history, arts, festivals, and laws, alongside standard subjects like science and mathematics. We evaluate over 45 LLMs, and find that current LLMs struggle with MILU, with GPT-4o achieving the highest average accuracy at 72 percent. Open multilingual models outperform language-specific fine-tuned models, which perform only slightly better than random baselines. Models also perform better in high resource languages as compared to low resource ones. Domain-wise analysis indicates that models perform poorly in culturally relevant areas like Arts and Humanities, Law and Governance compared to general fields like STEM. To the best of our knowledge, MILU is the first of its kind benchmark focused on Indic languages, serving as a crucial step towards comprehensive cultural evaluation. All code, benchmarks, and artifacts are publicly available to foster open research.
Authors: Jierui Li, Hung Le, Yingbo Zhou, Caiming Xiong, Silvio Savarese, Doyen Sahoo
Abstract: Pre-trained on massive amounts of code and text data, large language models (LLMs) have demonstrated remarkable achievements in performing code generation tasks. With additional execution-based feedback, these models can act as agents with capabilities to self-refine and improve generated code autonomously. However, on challenging coding tasks with extremely large search space, current agentic approaches still struggle with multi-stage planning, generating, and debugging. To address this problem, we propose CodeTree, a framework for LLM agents to efficiently explore the search space in different stages of the code generation process. Specifically, we adopted a unified tree structure to explicitly explore different coding strategies, generate corresponding coding solutions, and subsequently refine the solutions. In each stage, critical decision-making (ranking, termination, expanding) of the exploration process is guided by both the environmental execution-based feedback and LLM-agent-generated feedback. We comprehensively evaluated CodeTree on 7 code generation benchmarks and demonstrated the significant performance gains of CodeTree against strong baselines. Using GPT-4o as the base model, we consistently achieved top results of 95.1 on HumanEval, 98.7 on MBPP, and 43.0 on CodeContests. On the challenging SWEBench benchmark, our approach led to significant performance gains.
Authors: Sonia K. Murthy, Tomer Ullman, Jennifer Hu
Abstract: Researchers in social science and psychology have recently proposed using large language models (LLMs) as replacements for humans in behavioral research. In addition to arguments about whether LLMs accurately capture population-level patterns, this has raised questions about whether LLMs capture human-like conceptual diversity. Separately, it is debated whether post-training alignment (RLHF or RLAIF) affects models' internal diversity. Inspired by human studies, we use a new way of measuring the conceptual diversity of synthetically-generated LLM "populations" by relating the internal variability of simulated individuals to the population-level variability. We use this approach to evaluate non-aligned and aligned LLMs on two domains with rich human behavioral data. While no model reaches human-like diversity, aligned models generally display less diversity than their instruction fine-tuned counterparts. Our findings highlight potential trade-offs between increasing models' value alignment and decreasing the diversity of their conceptual representations.
Authors: Yancheng He, Shilong Li, Jiaheng Liu, Yingshui Tan, Weixun Wang, Hui Huang, Xingyuan Bu, Hangyu Guo, Chengwei Hu, Boren Zheng, Zhuoran Lin, Xuepeng Liu, Dekai Sun, Shirong Lin, Zhicheng Zheng, Xiaoyong Zhu, Wenbo Su, Bo Zheng
Abstract: New LLM evaluation benchmarks are important to align with the rapid development of Large Language Models (LLMs). In this work, we present Chinese SimpleQA, the first comprehensive Chinese benchmark to evaluate the factuality ability of language models to answer short questions, and Chinese SimpleQA mainly has five properties (i.e., Chinese, Diverse, High-quality, Static, Easy-to-evaluate). Specifically, first, we focus on the Chinese language over 6 major topics with 99 diverse subtopics. Second, we conduct a comprehensive quality control process to achieve high-quality questions and answers, where the reference answers are static and cannot be changed over time. Third, following SimpleQA, the questions and answers are very short, and the grading process is easy-to-evaluate based on OpenAI API. Based on Chinese SimpleQA, we perform a comprehensive evaluation on the factuality abilities of existing LLMs. Finally, we hope that Chinese SimpleQA could guide the developers to better understand the Chinese factuality abilities of their models and facilitate the growth of foundation models.
Authors: Chong Zhang, Mingyu Jin, Dong Shu, Taowen Wang, Dongfang Liu, Xiaobo Jin
Abstract: Current large language models (LLM) provide a strong foundation for large-scale user-oriented natural language tasks. Many users can easily inject adversarial text or instructions through the user interface, thus causing LLM model security challenges like the language model not giving the correct answer. Although there is currently a large amount of research on black-box attacks, most of these black-box attacks use random and heuristic strategies. It is unclear how these strategies relate to the success rate of attacks and thus effectively improve model robustness. To solve this problem, we propose our target-driven black-box attack method to maximize the KL divergence between the conditional probabilities of the clean text and the attack text to redefine the attack's goal. We transform the distance maximization problem into two convex optimization problems based on the attack goal to solve the attack text and estimate the covariance. Furthermore, the projected gradient descent algorithm solves the vector corresponding to the attack text. Our target-driven black-box attack approach includes two attack strategies: token manipulation and misinformation attack. Experimental results on multiple Large Language Models and datasets demonstrate the effectiveness of our attack method.
Authors: Sina Bagheri Nezhad, Sayan Bandyapadhyay, Ameeta Agrawal
Abstract: Fairness in multi-document summarization of user-generated content remains a critical challenge in natural language processing (NLP). Existing summarization methods often fail to ensure equitable representation across different social groups, leading to biased outputs. In this paper, we introduce two novel methods for fair extractive summarization: FairExtract, a clustering-based approach, and FairGPT, which leverages GPT-3.5-turbo with fairness constraints. We evaluate these methods using Divsumm summarization dataset of White-aligned, Hispanic, and African-American dialect tweets and compare them against relevant baselines. The results obtained using a comprehensive set of summarization quality metrics such as SUPERT, BLANC, SummaQA, BARTScore, and UniEval, as well as a fairness metric F, demonstrate that FairExtract and FairGPT achieve superior fairness while maintaining competitive summarization quality. Additionally, we introduce composite metrics (e.g., SUPERT+F, BLANC+F) that integrate quality and fairness into a single evaluation framework, offering a more nuanced understanding of the trade-offs between these objectives. This work highlights the importance of fairness in summarization and sets a benchmark for future research in fairness-aware NLP models.
Authors: Youan Cong, Cheng Wang, Pritom Saha Akash, Kevin Chen-Chuan Chang
Abstract: We introduce the Extract-Refine-Retrieve-Read (ERRR) framework, a novel approach designed to bridge the pre-retrieval information gap in Retrieval-Augmented Generation (RAG) systems through query optimization tailored to meet the specific knowledge requirements of Large Language Models (LLMs). Unlike conventional query optimization techniques used in RAG, the ERRR framework begins by extracting parametric knowledge from LLMs, followed by using a specialized query optimizer for refining these queries. This process ensures the retrieval of only the most pertinent information essential for generating accurate responses. Moreover, to enhance flexibility and reduce computational costs, we propose a trainable scheme for our pipeline that utilizes a smaller, tunable model as the query optimizer, which is refined through knowledge distillation from a larger teacher model. Our evaluations on various question-answering (QA) datasets and with different retrieval systems show that ERRR consistently outperforms existing baselines, proving to be a versatile and cost-effective module for improving the utility and accuracy of RAG systems.
Authors: Qi Pang, Shengyuan Hu, Wenting Zheng, Virginia Smith
Abstract: Advances in generative models have made it possible for AI-generated text, code, and images to mirror human-generated content in many applications. Watermarking, a technique that aims to embed information in the output of a model to verify its source, is useful for mitigating the misuse of such AI-generated content. However, we show that common design choices in LLM watermarking schemes make the resulting systems surprisingly susceptible to attack -- leading to fundamental trade-offs in robustness, utility, and usability. To navigate these trade-offs, we rigorously study a set of simple yet effective attacks on common watermarking systems, and propose guidelines and defenses for LLM watermarking in practice.
Authors: Andreas Madsen, Himabindu Lakkaraju, Siva Reddy, Sarath Chandar
Abstract: Interpretability is the study of explaining models in understandable terms to humans. At present, interpretability is divided into two paradigms: the intrinsic paradigm, which believes that only models designed to be explained can be explained, and the post-hoc paradigm, which believes that black-box models can be explained. At the core of this debate is how each paradigm ensures its explanations are faithful, i.e., true to the model's behavior. This is important, as false but convincing explanations lead to unsupported confidence in artificial intelligence (AI), which can be dangerous. This paper's position is that we should think about new paradigms while staying vigilant regarding faithfulness. First, by examining the history of paradigms in science, we see that paradigms are constantly evolving. Then, by examining the current paradigms, we can understand their underlying beliefs, the value they bring, and their limitations. Finally, this paper presents 3 emerging paradigms for interpretability. The first paradigm designs models such that faithfulness can be easily measured. Another optimizes models such that explanations become faithful. The last paradigm proposes to develop models that produce both a prediction and an explanation.
Authors: Yangyi Chen, Xingyao Wang, Hao Peng, Heng Ji
Abstract: We present SOLO, a single transformer for Scalable visiOn-Language mOdeling. Current large vision-language models (LVLMs) such as LLaVA mostly employ heterogeneous architectures that connect pre-trained visual encoders with large language models (LLMs) to facilitate visual recognition and complex reasoning. Although achieving remarkable performance with relatively lightweight training, we identify four primary scalability limitations: (1) The visual capacity is constrained by pre-trained visual encoders, which are typically an order of magnitude smaller than LLMs. (2) The heterogeneous architecture complicates the use of established hardware and software infrastructure. (3) Study of scaling laws on such architecture must consider three separate components - visual encoder, connector, and LLMs, which complicates the analysis. (4) The use of existing visual encoders typically requires following a pre-defined specification of image inputs pre-processing, for example, by reshaping inputs to fixed-resolution square images, which presents difficulties in processing and training on high-resolution images or those with unusual aspect ratio. A unified single Transformer architecture, like SOLO, effectively addresses these scalability concerns in LVLMs; however, its limited adoption in the modern context likely stems from the absence of reliable training recipes that balance both modalities and ensure stable training for billion-scale models. In this paper, we introduce the first open-source training recipe for developing SOLO, an open-source 7B LVLM using moderate academic resources. The training recipe involves initializing from LLMs, sequential pre-training on ImageNet and web-scale data, and instruction fine-tuning on our curated high-quality datasets. On extensive evaluation, SOLO demonstrates performance comparable to LLaVA-v1.5-7B, particularly excelling in visual mathematical reasoning.
Authors: Melissa Dell
Abstract: Deep learning provides powerful methods to impute structured information from large-scale, unstructured text and image datasets. For example, economists might wish to detect the presence of economic activity in satellite images, or to measure the topics or entities mentioned in social media, the congressional record, or firm filings. This review introduces deep neural networks, covering methods such as classifiers, regression models, generative AI, and embedding models. Applications include classification, document digitization, record linkage, and methods for data exploration in massive scale text and image corpora. When suitable methods are used, deep learning models can be cheap to tune and can scale affordably to problems involving millions or billions of data points.. The review is accompanied by a companion website, EconDL, with user-friendly demo notebooks, software resources, and a knowledge base that provides technical details and additional applications.
Authors: Yuhao Li, Ruoran Bai, Haiping Huang
Abstract: Large language models show a surprising in-context learning ability -- being able to use a prompt to form a prediction for a query, yet without additional training, in stark contrast to old-fashioned supervised learning. Providing a mechanistic interpretation and linking the empirical phenomenon to physics are thus challenging and remain unsolved. We study a simple yet expressive transformer with linear attention and map this structure to a spin glass model with real-valued spins, where the couplings and fields explain the intrinsic disorder in data. The spin glass model explains how the weight parameters interact with each other during pre-training, and further clarifies why an unseen function can be predicted by providing only a prompt yet without further training. Our theory reveals that for single-instance learning, increasing the task diversity leads to the emergence of in-context learning, by allowing the Boltzmann distribution to converge to a unique correct solution of weight parameters. Therefore the pre-trained transformer displays a prediction power in a novel prompt setting. The proposed analytically tractable model thus offers a promising avenue for thinking about how to interpret many intriguing but puzzling properties of large language models.
Authors: David Wood, Boris Lublinsky, Alexy Roytman, Shivdeep Singh, Constantin Adam, Abdulhamid Adebayo, Sungeun An, Yuan Chi Chang, Xuan-Hong Dang, Nirmit Desai, Michele Dolfi, Hajar Emami-Gohari, Revital Eres, Takuya Goto, Dhiraj Joshi, Yan Koyfman, Mohammad Nassar, Hima Patel, Paramesvaran Selvam, Yousaf Shah, Saptha Surendran, Daiki Tsuzuku, Petros Zerfos, Shahrokh Daijavad
Abstract: Data preparation is the first and a very important step towards any Large Language Model (LLM) development. This paper introduces an easy-to-use, extensible, and scale-flexible open-source data preparation toolkit called Data Prep Kit (DPK). DPK is architected and designed to enable users to scale their data preparation to their needs. With DPK they can prepare data on a local machine or effortlessly scale to run on a cluster with thousands of CPU Cores. DPK comes with a highly scalable, yet extensible set of modules that transform natural language and code data. If the user needs additional transforms, they can be easily developed using extensive DPK support for transform creation. These modules can be used independently or pipelined to perform a series of operations. In this paper, we describe DPK architecture and show its performance from a small scale to a very large number of CPUs. The modules from DPK have been used for the preparation of Granite Models [1] [2]. We believe DPK is a valuable contribution to the AI community to easily prepare data to enhance the performance of their LLM models or to fine-tune models with Retrieval-Augmented Generation (RAG).
Authors: Muzhi Li, Cehao Yang, Chengjin Xu, Zixing Song, Xuhui Jiang, Jian Guo, Ho-fung Leung, Irwin King
Abstract: Inductive knowledge graph completion (KGC) aims to predict missing triples with unseen entities. Recent works focus on modeling reasoning paths between the head and tail entity as direct supporting evidence. However, these methods depend heavily on the existence and quality of reasoning paths, which limits their general applicability in different scenarios. In addition, we observe that latent type constraints and neighboring facts inherent in KGs are also vital in inferring missing triples. To effectively utilize all useful information in KGs, we introduce CATS, a novel context-aware inductive KGC solution. With sufficient guidance from proper prompts and supervised fine-tuning, CATS activates the strong semantic understanding and reasoning capabilities of large language models to assess the existence of query triples, which consist of two modules. First, the type-aware reasoning module evaluates whether the candidate entity matches the latent entity type as required by the query relation. Then, the subgraph reasoning module selects relevant reasoning paths and neighboring facts, and evaluates their correlation to the query triple. Experiment results on three widely used datasets demonstrate that CATS significantly outperforms state-of-the-art methods in 16 out of 18 transductive, inductive, and few-shot settings with an average absolute MRR improvement of 7.2%.