Authors: Subhasis Dasgupta, Pratik Satpati, Ishika Choudhary, Jaydip Sen
Abstract: In the recent past, there were several works on the prediction of stock price using different methods. Sentiment analysis of news and tweets and relating them to the movement of stock prices have already been explored. But, when we talk about the news, there can be several topics such as politics, markets, sports etc. It was observed that most of the prior analyses dealt with news or comments associated with particular stock prices only or the researchers dealt with overall sentiment scores only. However, it is quite possible that different topics having different levels of impact on the movement of the stock price or an index. The current study focused on bridging this gap by analysing the movement of Nifty 50 index with respect to the sentiments associated with news items related to various different topic such as sports, politics, markets etc. The study established that sentiment scores of news items of different other topics also have a significant impact on the movement of the index.
Authors: Eyad Gomaa, Gomaa Salah
Abstract: We present Quasar-1, a novel architecture that introduces temperature-guided reasoning to large language models through the Token Temperature Mechanism (TTM) and Guided Sequence of Thought (GSoT). Our approach leverages the concept of hot and cold tokens, where hot tokens are prioritized for their contextual relevance, while cold tokens provide supplementary information. This dynamic modulation of token importance enables the model to achieve superior logical reasoning capabilities compared to traditional chain-of-thought approaches. Through rigorous mathematical analysis, we prove that our temperature-guided attention mechanism converges to optimal reasoning paths with exponential guarantees. Empirical results show significant improvements in reasoning accuracy and computational efficiency across a wide range of tasks, making advanced AI reasoning accessible to a broader range of applications.
Authors: Fang Zeng, Zhiliang Lyu, Quanzheng Li, Xiang Li
Abstract: This study introduces "RadCouncil," a multi-agent Large Language Model (LLM) framework designed to enhance the generation of impressions in radiology reports from the finding section. RadCouncil comprises three specialized agents: 1) a "Retrieval" Agent that identifies and retrieves similar reports from a vector database, 2) a "Radiologist" Agent that generates impressions based on the finding section of the given report plus the exemplar reports retrieved by the Retrieval Agent, and 3) a "Reviewer" Agent that evaluates the generated impressions and provides feedback. The performance of RadCouncil was evaluated using both quantitative metrics (BLEU, ROUGE, BERTScore) and qualitative criteria assessed by GPT-4, using chest X-ray as a case study. Experiment results show improvements in RadCouncil over the single-agent approach across multiple dimensions, including diagnostic accuracy, stylistic concordance, and clarity. This study highlights the potential of utilizing multiple interacting LLM agents, each with a dedicated task, to enhance performance in specialized medical tasks and the development of more robust and adaptable healthcare AI solutions.
Authors: Saipraneeth Devunuri, Lewis Lehe
Abstract: This paper introduces a framework that leverages Large Language Models (LLMs) to answer natural language queries about General Transit Feed Specification (GTFS) data. The framework is implemented in a chatbot called TransitGPT with open-source code. TransitGPT works by guiding LLMs to generate Python code that extracts and manipulates GTFS data relevant to a query, which is then executed on a server where the GTFS feed is stored. It can accomplish a wide range of tasks, including data retrieval, calculations, and interactive visualizations, without requiring users to have extensive knowledge of GTFS or programming. The LLMs that produce the code are guided entirely by prompts, without fine-tuning or access to the actual GTFS feeds. We evaluate TransitGPT using GPT-4o and Claude-3.5-Sonnet LLMs on a benchmark dataset of 100 tasks, to demonstrate its effectiveness and versatility. The results show that TransitGPT can significantly enhance the accessibility and usability of transit data.
Authors: Yuxiao Lu, Arunesh Sinha, Pradeep Varakantham
Abstract: Large Language Models (LLMs) generating unsafe responses to toxic prompts is a significant issue in their applications. While various efforts aim to address this safety concern, previous approaches often demand substantial human data collection or rely on the less dependable option of using another LLM to generate corrective data. In this paper, we aim to take this problem and overcome limitations of requiring significant high-quality human data. Our method requires only a small set of unsafe responses to toxic prompts, easily obtained from the unsafe LLM itself. By employing a semantic cost combined with a negative Earth Mover Distance (EMD) loss, we guide the LLM away from generating unsafe responses. Additionally, we propose a novel lower bound for EMD loss, enabling more efficient optimization. Our results demonstrate superior performance and data efficiency compared to baselines, and we further examine the nuanced effects of over-alignment and potential degradation of language capabilities when using contrastive data.
Authors: Pu Zhao, Xuan Shen, Zhenglun Kong, Yixin Shen, Sung-En Chang, Timothy Rupprecht, Lei Lu, Enfu Nan, Changdi Yang, Yumei He, Xingchen Xu, Yu Huang, Wei Wang, Yue Chen, Yong He, Yanzhi Wang
Abstract: Recently, Large Language Models (LLMs) have undergone a significant transformation, marked by a rapid rise in both their popularity and capabilities. Leading this evolution are proprietary LLMs like GPT-4 and GPT-o1, which have captured widespread attention in the AI community due to their remarkable performance and versatility. Simultaneously, open-source LLMs, such as LLaMA and Mistral, have made great contributions to the ever-increasing popularity of LLMs due to the ease to customize and deploy the models across diverse applications. Although open-source LLMs present unprecedented opportunities for innovation and research, the commercialization of LLMs has raised concerns about transparency, reproducibility, and safety. Many open-source LLMs fail to meet fundamental transparency requirements by withholding essential components like training code and data, and some use restrictive licenses whilst claiming to be "open-source," which may hinder further innovations on LLMs. To mitigate this issue, we introduce Moxin 7B, a fully open-source LLM developed in accordance with the Model Openness Framework (MOF), a ranked classification system that evaluates AI models based on model completeness and openness, adhering to principles of open science, open source, open data, and open access. Our model achieves the highest MOF classification level of "open science" through the comprehensive release of pre-training code and configurations, training and fine-tuning datasets, and intermediate and final checkpoints. Experiments show that our model achieves superior performance in zero-shot evaluation compared with popular 7B models and performs competitively in few-shot evaluation.
Authors: Lincan Li, Jiaqi Li, Catherine Chen, Fred Gui, Hongjia Yang, Chenxiao Yu, Zhengguang Wang, Jianing Cai, Junlong Aaron Zhou, Bolin Shen, Alex Qian, Weixin Chen, Zhongkai Xue, Lichao Sun, Lifang He, Hanjie Chen, Kaize Ding, Zijian Du, Fangzhou Mu, Jiaxin Pei, Jieyu Zhao, Swabha Swayamdipta, Willie Neiswanger, Hua Wei, Xiyang Hu, Shixiang Zhu, Tianlong Chen, Yingzhou Lu, Yang Shi, Lianhui Qin, Tianfan Fu, Zhengzhong Tu, Yuzhe Yang, Jaemin Yoo, Jiaheng Zhang, Ryan Rossi, Liang Zhan, Liang Zhao, Emilio Ferrara, Yan Liu, Furong Huang, Xiangliang Zhang, Lawrence Rothenberg, Shuiwang Ji, Philip S. Yu, Yue Zhao, Yushun Dong
Abstract: In recent years, large language models (LLMs) have been widely adopted in political science tasks such as election prediction, sentiment analysis, policy impact assessment, and misinformation detection. Meanwhile, the need to systematically understand how LLMs can further revolutionize the field also becomes urgent. In this work, we--a multidisciplinary team of researchers spanning computer science and political science--present the first principled framework termed Political-LLM to advance the comprehensive understanding of integrating LLMs into computational political science. Specifically, we first introduce a fundamental taxonomy classifying the existing explorations into two perspectives: political science and computational methodologies. In particular, from the political science perspective, we highlight the role of LLMs in automating predictive and generative tasks, simulating behavior dynamics, and improving causal inference through tools like counterfactual generation; from a computational perspective, we introduce advancements in data preparation, fine-tuning, and evaluation methods for LLMs that are tailored to political contexts. We identify key challenges and future directions, emphasizing the development of domain-specific datasets, addressing issues of bias and fairness, incorporating human expertise, and redefining evaluation criteria to align with the unique requirements of computational political science. Political-LLM seeks to serve as a guidebook for researchers to foster an informed, ethical, and impactful use of Artificial Intelligence in political science. Our online resource is available at: http://political-llm.org/.
Authors: Thomas Pouplin, Katarzyna Kobalczyk, Hao Sun, Mihaela van der Schaar
Abstract: To develop autonomous agents capable of executing complex, multi-step decision-making tasks as specified by humans in natural language, existing reinforcement learning approaches typically require expensive labeled datasets or access to real-time experimentation. Moreover, conventional methods often face difficulties in generalizing to unseen goals and states, thereby limiting their practical applicability. This paper presents TEDUO, a novel training pipeline for offline language-conditioned policy learning. TEDUO operates on easy-to-obtain, unlabeled datasets and is suited for the so-called in-the-wild evaluation, wherein the agent encounters previously unseen goals and states. To address the challenges posed by such data and evaluation settings, our method leverages the prior knowledge and instruction-following capabilities of large language models (LLMs) to enhance the fidelity of pre-collected offline data and enable flexible generalization to new goals and states. Empirical results demonstrate that the dual role of LLMs in our framework-as data enhancers and generalizers-facilitates both effective and data-efficient learning of generalizable language-conditioned policies.
Authors: Bharath Raj S, Garvit Suri, Vikrant Dewangan, Raghav Sonavane
Abstract: Traditional greedy tokenization methods have been a critical step in Natural Language Processing (NLP), influencing how text is converted into tokens and directly impacting model performance. While subword tokenizers like Byte-Pair Encoding (BPE) are widely used, questions remain about their optimality across model scales and languages. In this work, we demonstrate through extensive experiments that an optimal BPE configuration significantly reduces token count compared to greedy segmentation, yielding improvements in token-saving percentages and performance benefits, particularly for smaller models. We evaluate tokenization performance across various intrinsic and extrinsic tasks, including generation and classification. Our findings suggest that compression-optimized tokenization strategies could provide substantial advantages for multilingual and low-resource language applications, highlighting a promising direction for further research and inclusive NLP.
Authors: Rozina L. Myoya, Vukosi Marivate, Idris Abdulmumin
Abstract: Public transport systems in many Sub-Saharan countries often receive less attention compared to other sectors, underscoring the need for innovative solutions to improve the Quality of Service (QoS) and overall user experience. This study explored commuter opinion mining to understand sentiments toward existing public transport systems in Kenya, Tanzania, and South Africa. We used a qualitative research design, analysing data from X (formerly Twitter) to assess sentiments across rail, mini-bus taxis, and buses. By leveraging Multilingual Opinion Mining techniques, we addressed the linguistic diversity and code-switching present in our dataset, thus demonstrating the application of Natural Language Processing (NLP) in extracting insights from under-resourced languages. We employed PLMs such as AfriBERTa, AfroXLMR, AfroLM, and PuoBERTa to conduct the sentiment analysis. The results revealed predominantly negative sentiments in South Africa and Kenya, while the Tanzanian dataset showed mainly positive sentiments due to the advertising nature of the tweets. Furthermore, feature extraction using the Word2Vec model and K-Means clustering illuminated semantic relationships and primary themes found within the different datasets. By prioritising the analysis of user experiences and sentiments, this research paves the way for developing more responsive, user-centered public transport systems in Sub-Saharan countries, contributing to the broader goal of improving urban mobility and sustainability.
Authors: Yingyi Ma, Zhe Liu, Ozlem Kalinli
Abstract: The advent of Large Language Models (LLM) has reformed the Automatic Speech Recognition (ASR). Prompting LLM with audio embeddings to generate transcriptions becomes the new state-of-the-art ASR. Despite LLMs being trained with an extensive amount of text corpora, high-quality domain-specific text data can still significantly enhance ASR performance on domain adaptation tasks. Although LLM-based ASR can naturally incorporate more text corpora by fine-tuning the LLM decoder, fine-tuning such ASR on text-only data without paired prompts may diminish the effectiveness of domain-specific knowledge. To mitigate this issue, we propose a two-step soft prompt fine-tuning strategy that enhances domain-specific text adaptation. Experimental results show that text adaptation with our proposed method achieved a relative up to 9% Word Error Rate (WER) reduction and up to 18% Entity Error Rate (EER) reduction on the target domain compared to the baseline ASR. Combining this with domain-specific Language Model (LM) fusion can further improve the EER by a relative 2-5%
Authors: Arda Sevinc, Abdurrahman Gumus
Abstract: Chain of Thought (CoT) was introduced in recent research as a method for improving step-by-step reasoning in Large Language Models. However, CoT has limited applications such as its need for hand-crafted few-shot exemplar prompts and no capability to adjust itself to different queries. In this work, we propose a system to automatically generate rationales using CoT. Our method improves multi-step implicit reasoning capabilities by decomposing the implicit query into several explicit questions. This provides interpretability for the model, improving reasoning in weaker LLMs. We test our approach with two Q\&A datasets: StrategyQA and HotpotQA. We show an increase in accuracy with both, especially on StrategyQA. To facilitate further research in this field, the complete source code for this study has been made publicly available on GitHub: https://github.com/miralab-ai/autoreason.
Authors: In Gim, Seung-seob Lee, Lin Zhong
Abstract: Large language models (LLMs) use function calls to interface with external tools and data source. However, the current approach to LLM function calling is inherently synchronous, where each call blocks LLM inference, limiting LLM operation and concurrent function execution. In this work, we propose AsyncLM, a system for asynchronous LLM function calling. AsyncLM improves LLM's operational efficiency by enabling LLMs to generate and execute function calls concurrently. Instead of waiting for each call's completion, AsyncLM introduces an interrupt mechanism to asynchronously notify the LLM in-flight when function calls return. We design an in-context protocol for function calls and interrupts, provide fine-tuning strategy to adapt LLMs to the interrupt semantics, and implement these mechanisms efficiently on LLM inference process. We demonstrate that AsyncLM can reduce end-to-end task completion latency from 1.6x-5.4x compared to synchronous function calling on a set of benchmark tasks in the Berkeley function calling leaderboard (BFCL). Furthermore, we discuss how interrupt mechanisms can be extended to enable novel human-LLM or LLM-LLM interactions.
Authors: Bohan Jiang, Dawei Li, Zhen Tan, Xinyi Zhou, Ashwin Rao, Kristina Lerman, H. Russell Bernard, Huan Liu
Abstract: Measuring the relative impact of CTs is important for prioritizing responses and allocating resources effectively, especially during crises. However, assessing the actual impact of CTs on the public poses unique challenges. It requires not only the collection of CT-specific knowledge but also diverse information from social, psychological, and cultural dimensions. Recent advancements in large language models (LLMs) suggest their potential utility in this context, not only due to their extensive knowledge from large training corpora but also because they can be harnessed for complex reasoning. In this work, we develop datasets of popular CTs with human-annotated impacts. Borrowing insights from human impact assessment processes, we then design tailored strategies to leverage LLMs for performing human-like CT impact assessments. Through rigorous experiments, we textit{discover that an impact assessment mode using multi-step reasoning to analyze more CT-related evidence critically produces accurate results; and most LLMs demonstrate strong bias, such as assigning higher impacts to CTs presented earlier in the prompt, while generating less accurate impact assessments for emotionally charged and verbose CTs.
Authors: Amirhossein Abaskohi, Spandana Gella, Giuseppe Carenini, Issam H. Laradji
Abstract: Multimodal multihop question answering is a complex task that requires reasoning over multiple sources of information, such as images and text, to answer questions. While there has been significant progress in visual question answering, the multihop setting remains unexplored due to the lack of high-quality datasets. Current methods focus on single-hop question answering or a single modality, which makes them unsuitable for real-world scenarios such as analyzing multimodal educational materials, summarizing lengthy academic articles, or interpreting scientific studies that combine charts, images, and text. To address this gap, we propose a novel methodology, introducing the first framework for creating a high-quality dataset that enables training models for multimodal multihop question answering. Our approach consists of a 5-stage pipeline that involves acquiring relevant multimodal documents from Wikipedia, synthetically generating high-level questions and answers, and validating them through rigorous criteria to ensure quality data. We evaluate our methodology by training models on our synthesized dataset and testing on two benchmarks, our results demonstrate that, with an equal sample size, models trained on our synthesized data outperform those trained on human-collected data by 1.9 in exact match (EM) on average. We believe our data synthesis method will serve as a strong foundation for training and evaluating multimodal multihop question answering models.
Authors: Wangli Yang, Jie Yang, Yi Guo, Johan Barthelemy
Abstract: The field of textual adversarial defenses has gained considerable attention in recent years due to the increasing vulnerability of natural language processing (NLP) models to adversarial attacks, which exploit subtle perturbations in input text to deceive models. This paper introduces the Defensive Dual Masking (DDM) algorithm, a novel approach designed to enhance model robustness against such attacks. DDM utilizes a unique adversarial training strategy where [MASK] tokens are strategically inserted into training samples to prepare the model to handle adversarial perturbations more effectively. During inference, potentially adversarial tokens are dynamically replaced with [MASK] tokens to neutralize potential threats while preserving the core semantics of the input. The theoretical foundation of our approach is explored, demonstrating how the selective masking mechanism strengthens the model's ability to identify and mitigate adversarial manipulations. Our empirical evaluation across a diverse set of benchmark datasets and attack mechanisms consistently shows that DDM outperforms state-of-the-art defense techniques, improving model accuracy and robustness. Moreover, when applied to Large Language Models (LLMs), DDM also enhances their resilience to adversarial attacks, providing a scalable defense mechanism for large-scale NLP applications.
Authors: Shiyue Zhang, David Wan, Arie Cattan, Ayal Klein, Ido Dagan, Mohit Bansal
Abstract: How to properly conduct human evaluations for text summarization is a longstanding challenge. The Pyramid human evaluation protocol, which assesses content selection by breaking the reference summary into sub-units and verifying their presence in the system summary, has been widely adopted. However, it suffers from a lack of systematicity in the definition and granularity of the sub-units. We address these problems by proposing QAPyramid, which decomposes each reference summary into finer-grained question-answer (QA) pairs according to the QA-SRL framework. We collect QA-SRL annotations for reference summaries from CNN/DM and evaluate 10 summarization systems, resulting in 8.9K QA-level annotations. We show that, compared to Pyramid, QAPyramid provides more systematic and fine-grained content selection evaluation while maintaining high inter-annotator agreement without needing expert annotations. Furthermore, we propose metrics that automate the evaluation pipeline and achieve higher correlations with QAPyramid than other widely adopted metrics, allowing future work to accurately and efficiently benchmark summarization systems.
Authors: Zijiang Yang
Abstract: Natural Language Inference (NLI) is the task of inferring whether the hypothesis can be justified by the given premise. Basically, we classify the hypothesis into three labels(entailment, neutrality and contradiction) given the premise. NLI was well studied by the previous researchers. A number of models, especially the transformer based ones, have achieved significant improvement on these tasks. However, it is reported that these models are suffering when they are dealing with hard datasets. Particularly, they perform much worse when dealing with unseen out-of-distribution premise and hypothesis. They may not understand the semantic content but learn the spurious correlations. In this work, we propose the data augmentation and preprocessing methods to solve the word overlap, numerical reasoning and length mismatch problems. These methods are general methods that do not rely on the distribution of the testing data and they help improve the robustness of the models.
Authors: Bo-Wen Zhang, Yan Yan, Boxiang Yang, Yifei Xue, Guang Liu
Abstract: While scaling laws optimize training configurations for large language models (LLMs) through experiments on smaller or early-stage models, they fail to predict emergent abilities due to the absence of such capabilities in these models. To address this, we propose a method that predicts emergent abilities by leveraging proxy tasks. We begin by establishing relevance metrics between the target task and candidate tasks based on performance differences across multiple models. These candidate tasks are then validated for robustness with small model ensembles, leading to the selection of the most appropriate proxy tasks. The predicted performance on the target task is then derived by integrating the evaluation results of these proxies. In a case study on tool utilization capabilities, our method demonstrated a strong correlation between predicted and actual performance, confirming its effectiveness.
Authors: Dongjun Kim, Minhyuk Kim, YongChan Chun, Chanjun Park, Heuiseok Lim
Abstract: Large Language Models (LLMs) have demonstrated notable proficiency in both code generation and comprehension across multiple programming languages. However, the mechanisms underlying this proficiency remain underexplored, particularly with respect to whether distinct programming languages are processed independently or within a shared parametric region. Drawing an analogy to the specialized regions of the brain responsible for distinct cognitive functions, we introduce the concept of Coding Spot, a specialized parametric region within LLMs that facilitates coding capabilities. Our findings identify this Coding Spot and show that targeted modifications to this subset significantly affect performance on coding tasks, while largely preserving non-coding functionalities. This compartmentalization mirrors the functional specialization observed in cognitive neuroscience, where specific brain regions are dedicated to distinct tasks, suggesting that LLMs may similarly employ specialized parameter regions for different knowledge domains.
Authors: Haoran Lian, Junmin Chen, Wei Huang, Yizhe Xiong, Wenping Hu, Guiguang Ding, Hui Chen, Jianwei Niu, Zijia Lin, Fuzheng Zhang, Di Zhang
Abstract: Recently, Large language models (LLMs) have revolutionized Natural Language Processing (NLP). Pretrained LLMs, due to limited training context size, struggle with handling long token sequences, limiting their performance on various downstream tasks. Current solutions toward long context modeling often employ multi-stage continual pertaining, which progressively increases the effective context length through several continual pretraining stages. However, those approaches require extensive manual tuning and human expertise. In this paper, we introduce a novel single-stage continual pretraining method, Head-Adaptive Rotary Position Encoding (HARPE), to equip LLMs with long context modeling capabilities while simplifying the training process. Our HARPE leverages different Rotary Position Encoding (RoPE) base frequency values across different attention heads and directly trains LLMs on the target context length. Extensive experiments on 4 language modeling benchmarks, including the latest RULER benchmark, demonstrate that HARPE excels in understanding and integrating long-context tasks with single-stage training, matching and even outperforming existing multi-stage methods. Our results highlight that HARPE successfully breaks the stage barrier for training LLMs with long context modeling capabilities.
Authors: Nimisha Ghosh, Daniele Santoni, Indrajit Saha, Giovanni Felici
Abstract: In recent times, Transformer-based language models are making quite an impact in the field of natural language processing. As relevant parallels can be drawn between biological sequences and natural languages, the models used in NLP can be easily extended and adapted for various applications in bioinformatics. In this regard, this paper introduces the major developments of Transformer-based models in the recent past in the context of nucleotide sequences. We have reviewed and analysed a large number of application-based papers on this subject, giving evidence of the main characterizing features and to different approaches that may be adopted to customize such powerful computational machines. We have also provided a structured description of the functioning of Transformers, that may enable even first time users to grab the essence of such complex architectures. We believe this review will help the scientific community in understanding the various applications of Transformer-based language models to nucleotide sequences. This work will motivate the readers to build on these methodologies to tackle also various other problems in the field of bioinformatics.
Authors: Bo Li, Di Liang, Zixin Zhang
Abstract: The Transformer-based model have made significant strides in semantic matching tasks by capturing connections between phrase pairs. However, to assess the relevance of sentence pairs, it is insufficient to just examine the general similarity between the sentences. It is crucial to also consider the tiny subtleties that differentiate them from each other. Regrettably, attention softmax operations in transformers tend to miss these subtle differences. To this end, in this work, we propose a novel semantic sentence matching model named Combined Attention Network based on Transformer model (Comateformer). In Comateformer model, we design a novel transformer-based quasi-attention mechanism with compositional properties. Unlike traditional attention mechanisms that merely adjust the weights of input tokens, our proposed method learns how to combine, subtract, or resize specific vectors when building a representation. Moreover, our proposed approach builds on the intuition of similarity and dissimilarity (negative affinity) when calculating dual affinity scores. This allows for a more meaningful representation of relationships between sentences. To evaluate the performance of our proposed model, we conducted extensive experiments on ten public real-world datasets and robustness testing. Experimental results show that our method achieves consistent improvements.
Authors: Hanlin Wu, Zhenguang G. Cai
Abstract: The identity of a speaker significantly influences spoken language comprehension by affecting both perception and expectation. This review explores speaker effects, focusing on how speaker information impacts language processing. We propose an integrative model featuring the interplay between bottom-up perception-based processes driven by acoustic details and top-down expectation-based processes driven by a speaker model. The acoustic details influence lower-level perception, while the speaker model modulates both lower-level and higher-level processes such as meaning interpretation and pragmatic inferences. We define speaker-idiosyncrasy and speaker-demographics effects and demonstrate how bottom-up and top-down processes interact at various levels in different scenarios. This framework contributes to psycholinguistic theory by offering a comprehensive account of how speaker information interacts with linguistic content to shape message construction. We suggest that speaker effects can serve as indices of a language learner's proficiency and an individual's characteristics of social cognition. We encourage future research to extend these findings to AI speakers, probing the universality of speaker effects across humans and artificial agents.
Authors: Ruiheng Liu, Jinyu Zhang, Yanqi Song, Yu Zhang, Bailong Yang
Abstract: Continual Semantic Parsing (CSP) aims to train parsers to convert natural language questions into SQL across tasks with limited annotated examples, adapting to the real-world scenario of dynamically updated databases. Previous studies mitigate this challenge by replaying historical data or employing parameter-efficient tuning (PET), but they often violate data privacy or rely on ideal continual learning settings. To address these problems, we propose a new Large Language Model (LLM)-Enhanced Continuous Semantic Parsing method, named LECSP, which alleviates forgetting while encouraging generalization, without requiring real data replay or ideal settings. Specifically, it first analyzes the commonalities and differences between tasks from the SQL syntax perspective to guide LLMs in reconstructing key memories and improving memory accuracy through a calibration strategy. Then, it uses a task-aware dual-teacher distillation framework to promote the accumulation and transfer of knowledge during sequential training. Experimental results on two CSP benchmarks show that our method significantly outperforms existing methods, even those utilizing data replay or ideal settings. Additionally, we achieve generalization performance beyond the upper limits, better adapting to unseen tasks.
Authors: Xiaonan Wang, Jinyoung Yeo, Joon-Ho Lim, Hansaem Kim
Abstract: Large language models have exhibited significant enhancements in performance across various tasks. However, the complexity of their evaluation increases as these models generate more fluent and coherent content. Current multilingual benchmarks often use translated English versions, which may incorporate Western cultural biases that do not accurately assess other languages and cultures. To address this research gap, we introduce KULTURE Bench, an evaluation framework specifically designed for Korean culture that features datasets of cultural news, idioms, and poetry. It is designed to assess language models' cultural comprehension and reasoning capabilities at the word, sentence, and paragraph levels. Using the KULTURE Bench, we assessed the capabilities of models trained with different language corpora and analyzed the results comprehensively. The results show that there is still significant room for improvement in the models' understanding of texts related to the deeper aspects of Korean culture.
Authors: Qinhong Lin, Linna Zhou, Zhongliang Yang, Yuang Cai
Abstract: Large Language Models (LLMs) display formidable capabilities in generative tasks but also pose potential risks due to their tendency to generate hallucinatory responses. Uncertainty Quantification (UQ), the evaluation of model output reliability, is crucial for ensuring the safety and robustness of AI systems. Recent studies have concentrated on model uncertainty by analyzing the relationship between output entropy under various sampling conditions and the corresponding labels. However, these methods primarily focus on measuring model entropy with precision to capture response characteristics, often neglecting the uncertainties associated with greedy decoding results-the sources of model labels, which can lead to biased classification outcomes. In this paper, we explore the biases introduced by greedy decoding and propose a label-confidence-aware (LCA) uncertainty estimation based on Kullback-Leibler (KL) divergence bridging between samples and label source, thus enhancing the reliability and stability of uncertainty assessments. Our empirical evaluations across a range of popular LLMs and NLP datasets reveal that different label sources can indeed affect classification, and that our approach can effectively capture differences in sampling results and label sources, demonstrating more effective uncertainty estimation.
Authors: Romain Stora\"i, Seung-won Hwang
Abstract: This paper aims to improve the performance of large language models by addressing the variable computational demands in inference steps, where some tokens require more computational resources than others. We present HARP, a simple modification to "off-the-shelf" Transformer forward pass. Drawing from hesitation and the framing effect in decision-making, HARP selectively applies additional computation when the model encounters uncertainty during token generation. Our method mimics human cognitive processes by pausing at difficult decision points and reframing inputs for a different perspective. Unlike other approaches, HARP is model-agnostic, training-free, and easy to implement. We thoroughly evaluate our method across various downstream tasks and model sizes, demonstrating performance improvements up to +5.16%. Notably, HARP achieves these gains while maintaining inference times twice faster than beam search. Simple and yet with significant gains, HARP offers a practical solution for enhancing the performance of Transformer-based language models with minimal computational impact.
Authors: Yongqi Li, Xin Miao, Shen Zhou, Mayi Xu, Yuyang Ren, Tieyun Qian
Abstract: Despite the rapid progress that existing automated feedback methods have made in correcting the output of large language models (LLMs), these methods cannot be well applied to the relation extraction (RE) task due to their designated feedback objectives and correction manner. To address this problem, we propose a novel automated feedback framework for RE, which presents a rationale supervisor to verify the rationale and provide re-selected demonstrations as feedback to correct the initial prediction. Specifically, we first design a causal intervention and observation method for to collect biased/unbiased rationales for contrastive training the rationale supervisor. Then, we present a verification-feedback-correction procedure to iteratively enhance LLMs' capability of handling the RE task. Extensive experiments prove that our proposed framework significantly outperforms existing methods.
Authors: Jiawei Chen, Wentao Chen, Jing Su, Jingjing Xu, Hongyu Lin, Mengjie Ren, Yaojie Lu, Xianpei Han, Le Sun
Abstract: Large language models (LLMs) have shown significant multilingual capabilities. However, the mechanisms underlying the development of these capabilities during pre-training are not well understood. In this paper, we use code LLMs as an experimental platform to explore the evolution of multilingual capabilities in LLMs during the pre-training process. Based on our observations, we propose the Babel Tower Hypothesis, which describes the entire process of LLMs acquiring new language capabilities. During the learning process, multiple languages initially share a single knowledge system dominated by the primary language and gradually develop language-specific knowledge systems. We then validate the above hypothesis by tracking the internal states of the LLMs through identifying working languages and language transferring neurons. Experimental results show that the internal state changes of the LLM are consistent with our Babel Tower Hypothesis. Building on these insights, we propose a novel method to construct an optimized pre-training corpus for multilingual code LLMs, which significantly outperforms LLMs trained on the original corpus. The proposed Babel Tower Hypothesis provides new insights into designing pre-training data distributions to achieve optimal multilingual capabilities in LLMs.
Authors: Lance Calvin Lim Gamboa, Mark Lee
Abstract: Bias studies on multilingual models confirm the presence of gender-related stereotypes in masked models processing languages with high NLP resources. We expand on this line of research by introducing Filipino CrowS-Pairs and Filipino WinoQueer: benchmarks that assess both sexist and anti-queer biases in pretrained language models (PLMs) handling texts in Filipino, a low-resource language from the Philippines. The benchmarks consist of 7,074 new challenge pairs resulting from our cultural adaptation of English bias evaluation datasets, a process that we document in detail to guide similar forthcoming efforts. We apply the Filipino benchmarks on masked and causal multilingual models, including those pretrained on Southeast Asian data, and find that they contain considerable amounts of bias. We also find that for multilingual models, the extent of bias learned for a particular language is influenced by how much pretraining data in that language a model was exposed to. Our benchmarks and insights can serve as a foundation for future work analyzing and mitigating bias in multilingual models.
Authors: Pedro H. V. Valois, Lincon S. Souza, Erica K. Shimomoto, Kazuhiro Fukui
Abstract: Interpretability is a key challenge in fostering trust for Large Language Models (LLMs), which stems from the complexity of extracting reasoning from model's parameters. We present the Frame Representation Hypothesis, a theoretically robust framework grounded in the Linear Representation Hypothesis (LRH) to interpret and control LLMs by modeling multi-token words. Prior research explored LRH to connect LLM representations with linguistic concepts, but was limited to single token analysis. As most words are composed of several tokens, we extend LRH to multi-token words, thereby enabling usage on any textual data with thousands of concepts. To this end, we propose words can be interpreted as frames, ordered sequences of vectors that better capture token-word relationships. Then, concepts can be represented as the average of word frames sharing a common concept. We showcase these tools through Top-k Concept-Guided Decoding, which can intuitively steer text generation using concepts of choice. We verify said ideas on Llama 3.1, Gemma 2, and Phi 3 families, demonstrating gender and language biases, exposing harmful content, but also potential to remediate them, leading to safer and more transparent LLMs. Code is available at https://github.com/phvv-me/frame-representation-hypothesis.git
URLs: https://github.com/phvv-me/frame-representation-hypothesis.git
Authors: Jian Liao, Yu Feng, Xiaoyu Wang, Suge Wang, Jianxing Zheng, Deyu Li
Abstract: In implicit emotion analysis (IEA), the subtlety of emotional expressions makes it particularly sensitive to user-specific characteristics. Existing studies often inject personalization into the analysis by focusing on the authorial dimension of the emotional text. However, these methods overlook the potential influence of the intended reader on the reaction of implicit emotions. In this paper, we refine the IEA task to Personalized Implicit Emotion Analysis (PIEA) and introduce the RAPPIE model, a novel framework designed to address the issue of missing user information within this task. In particular, 1) we create reader agents based on the Large Language Model to simulate reader reactions, to address challenges of the spiral of silence and data incompleteness encountered when acquiring reader feedback information. 2) We establish a reader propagation role system and develop a role-aware emotion propagation multi-view graph learning model, which effectively deals with the sparsity of reader information by utilizing the distribution of propagation roles. 3) We annotate two Chinese PIEA datasets with detailed user metadata, thereby addressing the limitation of prior datasets that primarily focus on textual content annotation. Extensive experiments on these datasets indicate that the RAPPIE model outperforms current state-of-the-art baselines, highlighting the significance and efficacy of incorporating reader feedback into the PIEA process.
Authors: Bo Lv, Chen Tang, Yanan Zhang, Xin Liu, Yue Yu, Ping Luo
Abstract: Ensembles of generative large language models (LLMs) can integrate the strengths of different LLMs to compensate for the limitations of individual models. However, recent work has focused on training an additional fusion model to combine complete responses from multiple LLMs, failing to tap into their collaborative potential to generate higher-quality responses. Moreover, as the additional fusion model is trained on a specialized dataset, these methods struggle with generalizing to open-domain queries from online users. In this paper, we propose SpecFuse, a novel ensemble framework that outputs the fused result by iteratively producing the next segment through collaboration among LLMs. This is achieved through cyclic execution of its inference and verification components. In each round, the inference component invokes each base LLM to generate candidate segments in parallel, and the verify component calls these LLMs again to predict the ranking of the segments. The top-ranked segment is then broadcast to all LLMs, encouraging them to generate higher-quality segments in the next round. This approach also allows the base LLMs to be plug-and-play, without any training or adaptation, avoiding generalization limitations. Furthermore, to conserve computational resources, we propose a model exit mechanism that dynamically excludes models exhibiting poor performance in previous rounds during each query response. In this way, it effectively reduces the number of model calls while maintaining overall performance.
Authors: Alan Sun, Ethan Sun, Warren Shepard
Abstract: Zero-shot capabilities of large language models make them powerful tools for solving a range of tasks without explicit training. It remains unclear, however, how these models achieve such performance, or why they can zero-shot some tasks but not others. In this paper, we shed some light on this phenomenon by defining and investigating algorithmic stability in language models -- changes in problem-solving strategy employed by the model as a result of changes in task specification. We focus on a task where algorithmic stability is needed for generalization: two-operand arithmetic. Surprisingly, we find that Gemma-2-2b employs substantially different computational models on closely related subtasks, i.e. four-digit versus eight-digit addition. Our findings suggest that algorithmic instability may be a contributing factor to language models' poor zero-shot performance across certain logical reasoning tasks, as they struggle to abstract different problem-solving strategies and smoothly transition between them.
Authors: Willemijn Klaassen, Bram van Dijk, Marco Spruit
Abstract: Artificially intelligent systems optimized for speech conversation are appearing at a fast pace. Such models are interesting from a healthcare perspective, as these voice-controlled assistants may support the elderly and enable remote health monitoring. The bottleneck for efficacy, however, is how well these devices work in practice and how the elderly experience them, but research on this topic is scant. We review elderly use of voice-controlled AI and highlight various user- and technology-centered issues, that need to be considered before effective speech-controlled AI for elderly care can be realized.
Authors: Dongfang Li, Zetian Sun, Xinshuo Hu, Baotian Hu, Min Zhang
Abstract: Large Language Models (LLMs) need to adapt to the continuous changes in data, tasks, and user preferences. Due to their massive size and the high costs associated with training, LLMs are not suitable for frequent retraining. However, updates are necessary to keep them in sync with rapidly evolving human knowledge. To address these challenges, this paper proposes the Compression Memory Training (CMT) method, an efficient and effective online adaptation framework for LLMs that features robust knowledge retention capabilities. Inspired by human memory mechanisms, CMT compresses and extracts information from new documents to be stored in a memory bank. When answering to queries related to these new documents, the model aggregates these document memories from the memory bank to better answer user questions. The parameters of the LLM itself do not change during training and inference, reducing the risk of catastrophic forgetting. To enhance the encoding, retrieval, and aggregation of memory, we further propose three new general and flexible techniques, including memory-aware objective, self-matching and top-aggregation. Extensive experiments conducted on three continual learning datasets (i.e., StreamingQA, SQuAD and ArchivalQA) demonstrate that the proposed method improves model adaptability and robustness across multiple base LLMs (e.g., +4.07 EM & +4.19 F1 in StreamingQA with Llama-2-7b).
Authors: Ahan Bhatt, Nandan Vaghela, Kush Dudhia
Abstract: Knowledge Graphs (KGs) are essential for the functionality of GraphRAGs, a form of Retrieval-Augmented Generative Systems (RAGs) that excel in tasks requiring structured reasoning and semantic understanding. However, creating KGs for GraphRAGs remains a significant challenge due to accuracy and scalability limitations of traditional methods. This paper introduces a novel approach leveraging large language models (LLMs) like GPT-4, LLaMA 2 (13B), and BERT to generate KGs directly from unstructured data, bypassing traditional pipelines. Using metrics such as Precision, Recall, F1-Score, Graph Edit Distance, and Semantic Similarity, we evaluate the models' ability to generate high-quality KGs. Results demonstrate that GPT-4 achieves superior semantic fidelity and structural accuracy, LLaMA 2 excels in lightweight, domain-specific graphs, and BERT provides insights into challenges in entity-relationship modeling. This study underscores the potential of LLMs to streamline KG creation and enhance GraphRAG accessibility for real-world applications, while setting a foundation for future advancements.
Authors: Philippe Blache, Emmanuele Chersoni, Giulia Rambelli, Alessandro Lenci
Abstract: The mechanisms of comprehension during language processing remains an open question. Classically, building the meaning of a linguistic utterance is said to be incremental, step-by-step, based on a compositional process. However, many different works have shown for a long time that non-compositional phenomena are also at work. It is therefore necessary to propose a framework bringing together both approaches. We present in this paper an approach based on Construction Grammars and completing this framework in order to account for these different mechanisms. We propose first a formal definition of this framework by completing the feature structure representation proposed in Sign-Based Construction Grammars. In a second step, we present a general representation of the meaning based on the interaction of constructions, frames and events. This framework opens the door to a processing mechanism for building the meaning based on the notion of activation evaluated in terms of similarity and unification. This new approach integrates features from distributional semantics into the constructionist framework, leading to what we call Distributional Construction Grammars.
Authors: Philipp Christmann, Gerhard Weikum
Abstract: This article presents the QUASAR system for question answering over unstructured text, structured tables, and knowledge graphs, with unified treatment of all sources. The system adopts a RAG-based architecture, with a pipeline of evidence retrieval followed by answer generation, with the latter powered by a moderate-sized language model. Additionally and uniquely, QUASAR has components for question understanding, to derive crisper input for evidence retrieval, and for re-ranking and filtering the retrieved evidence before feeding the most informative pieces into the answer generation. Experiments with three different benchmarks demonstrate the high answering quality of our approach, being on par with or better than large GPT models, while keeping the computational cost and energy consumption orders of magnitude lower.
Authors: Javad Seraj, Mohammad Mahdi Mohajeri, Mohammad Javad Dousti, Majid Nili Ahmadabadi
Abstract: Automatic evaluation by large language models (LLMs) is a prominent topic today; however, judgment and evaluation tasks are often subjective and influenced by various factors, making adaptation challenging. While many studies demonstrate the capabilities of state-of-the-art proprietary LLMs in comparison to human evaluators, they often struggle to adapt to reference evaluators over time, a requirement for achieving personalized judgment. Additionally, numerous works have attempted to apply open LLMs as judges or evaluators, but these efforts frequently overlook the limitations of working with scarce data. Personalized judgment is inherently associated with limited data scenarios, which are common in many real-world problems. Our work aims to present a data augmentation technique to select a more effective sample from limited data in order to align an open LLM with human preference. Our work achieves approximately 7% improvements in Pearson correlation with a reference judge over the baseline,and 30% improvement over the base model (Llama3.1-8B-Instruct) in the mathematical reasoning evaluation task. demonstrating that augmenting selecting more effective preference data enables our approach to surpass baseline methods.
Authors: Kinshuk Vasisht, Navreet Kaur, Danish Pruthi
Abstract: To deploy language models safely, it is crucial that they abstain from responding to inappropriate requests. Several prior studies test the safety promises of models based on their effectiveness in blocking malicious requests. In this work, we focus on evaluating the underlying techniques that cause models to abstain. We create SELECT, a benchmark derived from a set of benign concepts (e.g., "rivers") from a knowledge graph. The nature of SELECT enables us to isolate the effects of abstention techniques from other safety training procedures, as well as evaluate their generalization and specificity. Using SELECT, we benchmark different abstention techniques over six open-weight and closed-source models. We find that the examined techniques indeed cause models to abstain with over $80\%$ abstention rates. However, these techniques are not as effective for descendants of the target concepts, with refusal rates declining by $19\%$. We also characterize the generalization-vs-specificity trade-offs for different techniques. Overall, no single technique is invariably better than the others. Our findings call for a careful evaluation of different aspects of abstention, and hopefully inform practitioners of various trade-offs involved.
Authors: Ehsan Lotfi, Nikolay Banar, Nerses Yuzbashyan, Walter Daelemans
Abstract: Statutory article retrieval plays a crucial role in making legal information more accessible to both laypeople and legal professionals. Multilingual countries like Belgium present unique challenges for retrieval models due to the need for handling legal issues in multiple languages. Building on the Belgian Statutory Article Retrieval Dataset (BSARD) in French, we introduce the bilingual version of this dataset, bBSARD. The dataset contains parallel Belgian statutory articles in both French and Dutch, along with legal questions from BSARD and their Dutch translation. Using bBSARD, we conduct extensive benchmarking of retrieval models available for Dutch and French. Our benchmarking setup includes lexical models, zero-shot dense models, and fine-tuned small foundation models. Our experiments show that BM25 remains a competitive baseline compared to many zero-shot dense models in both languages. We also observe that while proprietary models outperform open alternatives in the zero-shot setting, they can be matched or surpassed by fine-tuning small language-specific models. Our dataset and evaluation code are publicly available.
Authors: Sebastian Steindl, Ulrich Sch\"afer, Bernd Ludwig
Abstract: Large-scale Wizard-Of-Oz dialogue datasets have enabled the training of deep learning-based dialogue systems. While they are successful as benchmark datasets, they lack certain types of utterances, which would make them more realistic. In this work, we investigate the creation of synthetic communication errors in an automatic pipeline. Based on linguistic theory, we propose and follow a simple error taxonomy. We focus on three types of miscommunications that could happen in real-world dialogues but are underrepresented in the benchmark dataset: misunderstandings, non-understandings and vaguely related questions. Our two-step approach uses a state-of-the-art Large Language Model (LLM) to first create the error and secondly the repairing utterance. We perform Language Model-based evaluation to ensure the quality of the generated utterances. We apply the method to the MultiWOZ dataset and evaluate it both qualitatively and empirically as well as with human judges. Our results indicate that current LLMs can aid in adding post-hoc miscommunications to benchmark datasets as a form of data augmentation. We publish the resulting dataset, in which nearly 1900 dialogues have been modified, as CoPrUS-MultiWOZ to facilitate future work on dialogue systems.
Authors: Ellen Yi-Ge, Jiechao Gao, Wei Han, Wei Zhu
Abstract: Recently, large language models (LLMs) have demonstrated impressive capabilities in dealing with new tasks with the help of in-context learning (ICL). In the study of Large Vision-Language Models (LVLMs), when implementing ICL, researchers usually adopts the naive strategies like fixed demonstrations across different samples, or selecting demonstrations directly via a visual-language embedding model. These methods does not guarantee the configured demonstrations fit the need of the LVLMs. To address this issue, we now propose a novel framework, \underline{d}emonstration \underline{r}etriever for large m\underline{u}lti-modal \underline{m}odel (DRUM), which fine-tunes the visual-language embedding model to better meet the LVLM's needs. First, we discuss the retrieval strategies for a visual-language task, assuming an embedding model is given. And we propose to concate the image and text embeddings to enhance the retrieval performance. Second, we propose to re-rank the demonstrations retrieved by the embedding model via the LVLM's feedbacks, and calculate a list-wise ranking loss for training the embedding model. Third, we propose an iterative demonstration mining strategy to improve the training of the embedding model. Through extensive experiments on 3 types of visual-language tasks, 7 benchmark datasets, our DRUM framework is proven to be effective in boosting the LVLM's in-context learning performance via retrieving more proper demonstrations.
Authors: Wonjin Lee, Kyumin Kim, Sungjae Lee, Jihun Lee, Kwang In KIm
Abstract: Applying language models (LMs) to tables is challenging due to the inherent structural differences between two-dimensional tables and one-dimensional text for which the LMs were originally designed. Furthermore, when applying linearized tables to LMs, the maximum token lengths often imposed in self-attention calculations make it difficult to comprehensively understand the context spread across large tables. To address these challenges, we present PieTa (Piece of Table), a new framework for sub-table-based question answering (QA). PieTa operates through an iterative process of dividing tables into smaller windows, using LMs to select relevant cells within each window, and merging these cells into a sub-table. This multi-resolution approach captures dependencies across multiple rows and columns while avoiding the limitations caused by long context inputs. Instantiated as a simple iterative sub-table union algorithm, PieTa demonstrates improved performance over previous sub-table-based QA approaches.
Authors: Matthieu Meeus, Anthony Rath\'e, Fran\c{c}ois Remy, Pieter Delobelle, Jens-Joris Decorte, Thomas Demeester
Abstract: While Large Language Models (LLMs) have shown remarkable capabilities in natural language understanding and generation, their performance often lags in lower-resource, non-English languages due to biases in the training data. In this work, we explore strategies for adapting the primarily English LLMs (Llama-2 and Llama-3) to Dutch, a language spoken by 30 million people worldwide yet often underrepresented in LLM development. We collect 104GB of Dutch text ($32$B tokens) from various sources to first apply continued pretraining using low-rank adaptation (LoRA), complemented with Dutch posttraining strategies provided by prior work. For Llama-2, we consider using (i) the tokenizer of the original model, and (ii) training a new, Dutch-specific tokenizer combined with embedding reinitialization. We evaluate our adapted models, ChocoLlama-2, both on standard benchmarks and a novel Dutch benchmark, ChocoLlama-Bench. Our results demonstrate that LoRA can effectively scale for language adaptation, and that tokenizer modification with careful weight reinitialization can improve performance. Notably, Llama-3 was released during the course of this project and, upon evaluation, demonstrated superior Dutch capabilities compared to our Dutch-adapted versions of Llama-2. We hence apply the same adaptation technique to Llama-3, using its original tokenizer. While our adaptation methods enhanced Llama-2's Dutch capabilities, we found limited gains when applying the same techniques to Llama-3. This suggests that for ever improving, multilingual foundation models, language adaptation techniques may benefit more from focusing on language-specific posttraining rather than on continued pretraining. We hope this work contributes to the broader understanding of adapting LLMs to lower-resource languages, and to the development of Dutch LLMs in particular.
Authors: Tom Kouwenhoven, Max Peeperkorn, Tessa Verhoef
Abstract: Human languages have evolved to be structured through repeated language learning and use. These processes introduce biases that operate during language acquisition and shape linguistic systems toward communicative efficiency. In this paper, we investigate whether the same happens if artificial languages are optimised for implicit biases of Large Language Models (LLMs). To this end, we simulate a classical referential game in which LLMs learn and use artificial languages. Our results show that initially unstructured holistic languages are indeed shaped to have some structural properties that allow two LLM agents to communicate successfully. Similar to observations in human experiments, generational transmission increases the learnability of languages, but can at the same time result in non-humanlike degenerate vocabularies. Taken together, this work extends experimental findings, shows that LLMs can be used as tools in simulations of language evolution, and opens possibilities for future human-machine experiments in this field.
Authors: Shuo Yang, Bardh Prenkaj, Gjergji Kasneci
Abstract: Despite the widespread use of LLMs due to their superior performance in various tasks, their high computational costs often lead potential users to opt for the pretraining-finetuning pipeline. However, biases prevalent in manually constructed datasets can introduce spurious correlations between tokens and labels, creating so-called shortcuts and hindering the generalizability of fine-tuned models. Existing debiasing methods often rely on prior knowledge of specific dataset biases, which is challenging to acquire a priori. We propose RAZOR (Rewriting And Zero-bias Optimization Refinement), a novel, unsupervised, and data-focused debiasing approach based on text rewriting for shortcut mitigation. RAZOR leverages LLMs to iteratively rewrite potentially biased text segments by replacing them with heuristically selected alternatives in a shortcut space defined by token statistics and positional information. This process aims to align surface-level text features more closely with diverse label distributions, thereby promoting the learning of genuine linguistic patterns. Compared with unsupervised SoTA models, RAZOR improves by 3.5% on the FEVER and 6.5% on MNLI and SNLI datasets according to the F1 score. Additionally, RAZOR effectively mitigates specific known biases, reducing bias-related terms by x2 without requiring prior bias information, a result that is on par with SoTA models that leverage prior information. Our work prioritizes data manipulation over architectural modifications, emphasizing the pivotal role of data quality in enhancing model performance and fairness. This research contributes to developing more robust evaluation benchmarks for debiasing methods by incorporating metrics for bias reduction and overall model efficacy.
Authors: Wang Liang
Abstract: Multilingual transfer ability, which reflects how well models fine-tuned on one source language can be applied to other languages, has been well studied in multilingual pre-trained models. However, the existence of such capability transfer between natural language and gene sequences/languages remains underexplored.This study addresses this gap by drawing inspiration from the sentence-pair classification task used for evaluating sentence similarity in natural language. We constructed two analogous tasks: DNA-pair classification(DNA sequence similarity) and DNA-protein-pair classification(gene coding determination). These tasks were designed to validate the transferability of capabilities from natural language to gene sequences. Even a small-scale pre-trained model like GPT-2-small, which was pre-trained on English, achieved an accuracy of 78% on the DNA-pair classification task after being fine-tuned on English sentence-pair classification data(XTREME PAWS-X). While training a BERT model on multilingual text, the precision reached 82%.On the more complex DNA-protein-pair classification task, however, the model's output was barely distinguishable from random output.Experiments suggest that there may be a capability transfer from natural language to genetic language, but further task testing is needed to confirm this.
Authors: Alfredo Garrach\'on Ruiz, Tom\'as de la Rosa, Daniel Borrajo
Abstract: The inference cost of Large Language Models (LLMs) is a significant challenge due to their computational demands, specially on tasks requiring long outputs. However, natural language often contains redundancy, which presents an opportunity for optimization. We have observed that LLMs can generate distilled language-concise outputs that retain essential meaning, when prompted appropriately. We propose a framework for saving computational cost, in which a shorter distilled output from the LLM is reconstructed into a full narrative by a smaller model with lower inference costs. Our experiments show promising results, particularly in general knowledge domains with 20.58% saved tokens on average with tiny decrease in evaluation metrics, hinting that this approach can effectively balance efficiency and accuracy in language processing tasks.
Authors: Inkit Padhi, Manish Nagireddy, Giandomenico Cornacchia, Subhajit Chaudhury, Tejaswini Pedapati, Pierre Dognin, Keerthiram Murugesan, Erik Miehling, Mart\'in Santill\'an Cooper, Kieran Fraser, Giulio Zizzo, Muhammad Zaid Hameed, Mark Purcell, Michael Desmond, Qian Pan, Inge Vejsbjerg, Elizabeth M. Daly, Michael Hind, Werner Geyer, Ambrish Rawat, Kush R. Varshney, Prasanna Sattigeri
Abstract: We introduce the Granite Guardian models, a suite of safeguards designed to provide risk detection for prompts and responses, enabling safe and responsible use in combination with any large language model (LLM). These models offer comprehensive coverage across multiple risk dimensions, including social bias, profanity, violence, sexual content, unethical behavior, jailbreaking, and hallucination-related risks such as context relevance, groundedness, and answer relevance for retrieval-augmented generation (RAG). Trained on a unique dataset combining human annotations from diverse sources and synthetic data, Granite Guardian models address risks typically overlooked by traditional risk detection models, such as jailbreaks and RAG-specific issues. With AUC scores of 0.871 and 0.854 on harmful content and RAG-hallucination-related benchmarks respectively, Granite Guardian is the most generalizable and competitive model available in the space. Released as open-source, Granite Guardian aims to promote responsible AI development across the community. https://github.com/ibm-granite/granite-guardian
Authors: Zijian Chen, John-Michael Gamble, Micaela Jantzi, John P. Hirdes, Jimmy Lin
Abstract: Manual assignment of Anatomical Therapeutic Chemical (ATC) codes to prescription records is a significant bottleneck in healthcare research and operations at Ontario Health and InterRAI Canada, requiring extensive expert time and effort. To automate this process while maintaining data privacy, we develop a practical approach using locally deployable large language models (LLMs). Inspired by recent advances in automatic International Classification of Diseases (ICD) coding, our method frames ATC coding as a hierarchical information extraction task, guiding LLMs through the ATC ontology level by level. We evaluate our approach using GPT-4o as an accuracy ceiling and focus development on open-source Llama models suitable for privacy-sensitive deployment. Testing across Health Canada drug product data, the RABBITS benchmark, and real clinical notes from Ontario Health, our method achieves 78% exact match accuracy with GPT-4o and 60% with Llama 3.1 70B. We investigate knowledge grounding through drug definitions, finding modest improvements in accuracy. Further, we show that fine-tuned Llama 3.1 8B matches zero-shot Llama 3.1 70B accuracy, suggesting that effective ATC coding is feasible with smaller models. Our results demonstrate the feasibility of automatic ATC coding in privacy-sensitive healthcare environments, providing a foundation for future deployments.
Authors: Michael Iannelli, Sneha Kuchipudi, Vera Dvorak
Abstract: Retrieval Augmented Generation (RAG) enables Large Language Models (LLMs) to generalize to new information by decoupling reasoning capabilities from static knowledge bases. Traditional RAG enhancements have explored vertical scaling -- assigning subtasks to specialized modules -- and horizontal scaling -- replicating tasks across multiple agents -- to improve performance. However, real-world applications impose diverse Service Level Agreements (SLAs) and Quality of Service (QoS) requirements, involving trade-offs among objectives such as reducing cost, ensuring answer quality, and adhering to specific operational constraints. In this work, we present a systems-oriented approach to multi-agent RAG tailored for real-world Question Answering (QA) applications. By integrating task-specific non-functional requirements -- such as answer quality, cost, and latency -- into the system, we enable dynamic reconfiguration to meet diverse SLAs. Our method maps these Service Level Objectives (SLOs) to system-level parameters, allowing the generation of optimal results within specified resource constraints. We conduct a case study in the QA domain, demonstrating how dynamic re-orchestration of a multi-agent RAG system can effectively manage the trade-off between answer quality and cost. By adjusting the system based on query intent and operational conditions, we systematically balance performance and resource utilization. This approach allows the system to meet SLOs for various query types, showcasing its practicality for real-world applications.
Authors: Haotong Yang, Xiyuan Wang, Qian Tao, Shuxian Hu, Zhouchen Lin, Muhan Zhang
Abstract: Recent research on integrating Large Language Models (LLMs) with Graph Neural Networks (GNNs) typically follows two approaches: LLM-centered models, which convert graph data into tokens for LLM processing, and GNN-centered models, which use LLMs to encode text features into node and edge representations for GNN input. LLM-centered models often struggle to capture graph structures effectively, while GNN-centered models compress variable-length textual data into fixed-size vectors, limiting their ability to understand complex semantics. Additionally, GNN-centered approaches require converting tasks into a uniform, manually-designed format, restricting them to classification tasks and preventing language output. To address these limitations, we introduce a new architecture that deeply integrates GNN with LLM, featuring three key innovations: (1) Structure-Aware Transformers, which incorporate GNN's message-passing capabilities directly into LLM's transformer layers, allowing simultaneous processing of textual and structural information and generating outputs from both GNN and LLM; (2) Graph-Text Cross-Attention, which processes full, uncompressed text from graph nodes and edges, ensuring complete semantic integration; and (3) GNN-LLM Twin Predictor, enabling LLM's flexible autoregressive generation alongside GNN's scalable one-pass prediction. GL-Fusion achieves outstand performance on various tasks. Notably, it achieves state-of-the-art performance on OGBN-Arxiv and OGBG-Code2.
Authors: Dongwei Wang, Huanrui Yang
Abstract: Quantization is a critical step to enable efficient LLM serving under limited resource. However, previous research observes that certain weights in the LLM, known as outliers, are significantly sensitive to quantization noises. Existing quantization methods leave these outliers as floating points or higher precisions to retain performance, posting challenges on the efficient hardware deployment of the mixed-precision model. This work investigates an alternative way to tame the sensitive weights' impact on the quantization error, by reducing the loss Hessian trace with respect to outliers through an efficient fine-tuning process. We propose Noise Perturbation Fine-tuning (NPFT), which identifies outlier weights and add random weight perturbations on the outliers as the model going through a PEFT optimization. NPFT tames the sensitivity of outlier weights so that the quantized model performance can be improved without special treatment to the outliers. When applied to OPT and LLaMA models, our NPFT method achieves stable performance improvements for both uniform and non-uniform quantizers, while also offering better inference efficiency. Notably, the simplest RTN can achieve performance on par with GPTQ using our NPFT on LLaMA2-7B-4bits benchmark.
Authors: Biman Barua, M. Shamim Kaiser
Abstract: The rapid growth of the travel industry has increased the need for real-time optimization in reservation systems that could take care of huge data and transaction volumes. This study proposes a hybrid framework that ut folds an Artificial Intelligence and a Microservices approach for the performance optimization of the system. The AI algorithms forecast demand patterns, optimize the allocation of resources, and enhance decision-making driven by Microservices architecture, hence decentralizing system components for scalability, fault tolerance, and reduced downtime. The model provided focuses on major problems associated with the travel reservation systems such as latency of systems, load balancing and data consistency. It endows the systems with predictive models based on AI improved ability to forecast user demands. Microservices would also take care of different scales during uneven traffic patterns. Hence, both aspects ensure better handling of peak loads and spikes while minimizing delays and ensuring high service quality. A comparison was made between traditional reservation models, which are monolithic and the new model of AI-Microservices. Comparatively, the analysis results state that there is a drastic improvement in processing times where the system uptime and resource utilization proved the capability of AI and the microservices in transforming the travel industry in terms of reservation. This research work focused on AI and Microservices towards real-time optimization, providing critical insight into how to move forward with practical recommendations for upgrading travel reservation systems with this technology.
Authors: Nahid Alam, Karthik Reddy Kanjula, Surya Guthikonda, Timothy Chung, Bala Krishna S Vegesna, Abhipsha Das, Anthony Susevski, Ryan Sze-Yin Chan, S M Iftekhar Uddin, Shayekh Bin Islam, Roshan Santhosh, Snegha A, Drishti Sharma, Chen Liu, Isha Chaturvedi, Genta Indra Winata, Ashvanth. S, Snehanshu Mukherjee, Alham Fikri Aji
Abstract: The rapid development of large Vision-Language Models (VLMs) has led to impressive results on academic benchmarks, primarily in widely spoken languages. However, significant gaps remain in the ability of current VLMs to handle low-resource languages and varied cultural contexts, largely due to a lack of high-quality, diverse, and safety-vetted data. Consequently, these models often struggle to understand low-resource languages and cultural nuances in a manner free from toxicity. To address these limitations, we introduce Maya, an open-source Multimodal Multilingual model. Our contributions are threefold: 1) a multilingual image-text pretraining dataset in eight languages, based on the LLaVA pretraining dataset; 2) a thorough analysis of toxicity within the LLaVA dataset, followed by the creation of a novel toxicity-free version across eight languages; and 3) a multilingual image-text model supporting these languages, enhancing cultural and linguistic comprehension in vision-language tasks. Code available at https://github.com/nahidalam/maya.
Authors: Fei Ma, Yukan Li, Yifan Xie, Ying He, Yi Zhang, Hongwei Ren, Zhou Liu, Wei Yao, Fuji Ren, Fei Richard Yu, Shiguang Ni
Abstract: Human emotion synthesis is a crucial aspect of affective computing. It involves using computational methods to mimic and convey human emotions through various modalities, with the goal of enabling more natural and effective human-computer interactions. Recent advancements in generative models, such as Autoencoders, Generative Adversarial Networks, Diffusion Models, Large Language Models, and Sequence-to-Sequence Models, have significantly contributed to the development of this field. However, there is a notable lack of comprehensive reviews in this field. To address this problem, this paper aims to address this gap by providing a thorough and systematic overview of recent advancements in human emotion synthesis based on generative models. Specifically, this review will first present the review methodology, the emotion models involved, the mathematical principles of generative models, and the datasets used. Then, the review covers the application of different generative models to emotion synthesis based on a variety of modalities, including facial images, speech, and text. It also examines mainstream evaluation metrics. Additionally, the review presents some major findings and suggests future research directions, providing a comprehensive understanding of the role of generative technology in the nuanced domain of emotion synthesis.
Authors: Zirun Guo, Tao Jin, Wenlong Xu, Wang Lin, Yangyang Wu
Abstract: Multimodal sentiment analysis (MSA) is an emerging research topic that aims to understand and recognize human sentiment or emotions through multiple modalities. However, in real-world dynamic scenarios, the distribution of target data is always changing and different from the source data used to train the model, which leads to performance degradation. Common adaptation methods usually need source data, which could pose privacy issues or storage overheads. Therefore, test-time adaptation (TTA) methods are introduced to improve the performance of the model at inference time. Existing TTA methods are always based on probabilistic models and unimodal learning, and thus can not be applied to MSA which is often considered as a multimodal regression task. In this paper, we propose two strategies: Contrastive Adaptation and Stable Pseudo-label generation (CASP) for test-time adaptation for multimodal sentiment analysis. The two strategies deal with the distribution shifts for MSA by enforcing consistency and minimizing empirical risk, respectively. Extensive experiments show that CASP brings significant and consistent improvements to the performance of the model across various distribution shift settings and with different backbones, demonstrating its effectiveness and versatility. Our codes are available at https://github.com/zrguo/CASP.
Authors: Hao Li, Ruoyuan Gong, Hao Jiang
Abstract: Predicting roll call votes through modeling political actors has emerged as a focus in quantitative political science and computer science. Widely used embedding-based methods generate vectors for legislators from diverse data sets to predict legislative behaviors. However, these methods often contend with challenges such as the need for manually predefined features, reliance on extensive training data, and a lack of interpretability. Achieving more interpretable predictions under flexible conditions remains an unresolved issue. This paper introduces the Political Actor Agent (PAA), a novel agent-based framework that utilizes Large Language Models to overcome these limitations. By employing role-playing architectures and simulating legislative system, PAA provides a scalable and interpretable paradigm for predicting roll-call votes. Our approach not only enhances the accuracy of predictions but also offers multi-view, human-understandable decision reasoning, providing new insights into political actor behaviors. We conducted comprehensive experiments using voting records from the 117-118th U.S. House of Representatives, validating the superior performance and interpretability of PAA. This study not only demonstrates PAA's effectiveness but also its potential in political science research.
Authors: Sayak Chakrabarty, Souradip Pal
Abstract: This paper introduces Multiple Choice Reasoning via. Process of Elimination using Multi-Modal models, herein referred to as Multi-Modal Process of Elimination (MM-PoE). This novel methodology is engineered to augment the efficacy of Vision-Language Models (VLMs) in multiple-choice visual reasoning tasks. Diverging from conventional approaches that evaluate each option independently, MM-PoE employs a dual-step scoring paradigm that initially identifies and excludes implausible choices, subsequently concentrating on the most probable remaining options. This method emulates human test-taking strategies, where individuals typically eliminate clearly incorrect answers prior to selecting the optimal response. Our empirical evaluations, conducted across three benchmark datasets, reveal that MM-PoE significantly improves both zero-shot and few-shot performance of contemporary state-of-the-art VLMs. Critically, this approach not only broadens the application of the elimination process to multi-modal contexts but also allows few-shot experiments, thereby addressing two principal limitations concerning usage of PoE only in zero-shot settings and only with a language-only framework. As a result, MM-PoE not only refines the reasoning capabilities of VLMs but also broadens their applicability to complex visual question-answering scenarios. All code and documentation supporting our work are available at https://pypi.org/project/mm-poe/, enabling researchers and practitioners to easily integrate and further develop these techniques.
Authors: Kaixun Yang, Mladen Rakovi\'c, Zhiping Liang, Lixiang Yan, Zijie Zeng, Yizhou Fan, Dragan Ga\v{s}evi\'c, Guanliang Chen
Abstract: Students are increasingly relying on Generative AI (GAI) to support their writing-a key pedagogical practice in education. In GAI-assisted writing, students can delegate core cognitive tasks (e.g., generating ideas and turning them into sentences) to GAI while still producing high-quality essays. This creates new challenges for teachers in assessing and supporting student learning, as they often lack insight into whether students are engaging in meaningful cognitive processes during writing or how much of the essay's quality can be attributed to those processes. This study aimed to help teachers better assess and support student learning in GAI-assisted writing by examining how different writing behaviors, especially those indicative of meaningful learning versus those that are not, impact essay quality. Using a dataset of 1,445 GAI-assisted writing sessions, we applied the cutting-edge method, X-Learner, to quantify the causal impact of three GAI-assisted writing behavioral patterns (i.e., seeking suggestions but not accepting them, seeking suggestions and accepting them as they are, and seeking suggestions and accepting them with modification) on four measures of essay quality (i.e., lexical sophistication, syntactic complexity, text cohesion, and linguistic bias). Our analysis showed that writers who frequently modified GAI-generated text-suggesting active engagement in higher-order cognitive processes-consistently improved the quality of their essays in terms of lexical sophistication, syntactic complexity, and text cohesion. In contrast, those who often accepted GAI-generated text without changes, primarily engaging in lower-order processes, saw a decrease in essay quality. Additionally, while human writers tend to introduce linguistic bias when writing independently, incorporating GAI-generated text-even without modification-can help mitigate this bias.
Authors: Fuhai Chen, Pengpeng Huang, Xuri Ge, Jie Huang, Zishuo Bao
Abstract: With the rapid development of multimedia, the shift from unimodal textual sentiment analysis to multimodal image-text sentiment analysis has obtained academic and industrial attention in recent years. However, multimodal sentiment analysis is affected by unimodal data bias, e.g., text sentiment is misleading due to explicit sentiment semantic, leading to low accuracy in the final sentiment classification. In this paper, we propose a novel CounterFactual Multimodal Sentiment Analysis framework (CF-MSA) using causal counterfactual inference to construct multimodal sentiment causal inference. CF-MSA mitigates the direct effect from unimodal bias and ensures heterogeneity across modalities by differentiating the treatment variables between modalities. In addition, considering the information complementarity and bias differences between modalities, we propose a new optimisation objective to effectively integrate different modalities and reduce the inherent bias from each modality. Experimental results on two public datasets, MVSA-Single and MVSA-Multiple, demonstrate that the proposed CF-MSA has superior debiasing capability and achieves new state-of-the-art performances. We will release the code and datasets to facilitate future research.
Authors: Andrea Caria
Abstract: This perspective article aims at providing an outline of the state of the art and future developments towards the integration of cutting-edge predictive language models with BCI. A synthetic overview of early and more recent linguistic models, from natural language processing (NLP) models to recent LLM, that to a varying extent improved predictive writing systems, is first provided. Second, a summary of previous BCI implementations integrating language models is presented. The few preliminary studies investigating the possible combination of LLM with BCI spellers to efficiently support fast communication and control are then described. Finally, current challenges and limitations towards the full integration of LLM with BCI systems are discussed. Recent investigations suggest that the combination of LLM with BCI might drastically improve human-computer interaction in patients with motor or language disorders as well as in healthy individuals. In particular, the pretrained autoregressive transformer models, such as GPT, that capitalize from parallelization, learning through pre-training and fine-tuning, promise a substantial improvement of BCI for communication with respect to previous systems incorporating simpler language models. Indeed, among various models, the GPT-2 was shown to represent an excellent candidate for its integration into BCI although testing was only perfomed on simulated conversations and not on real BCI scenarios. Prospectively, the full integration of LLM with advanced BCI systems might lead to a big leap forward towards fast, efficient and user-adaptive neurotechnology.
Authors: Raanan Y. Rohekar, Yaniv Gurwicz, Sungduk Yu, Vasudev Lal
Abstract: Are generative pre-trained transformer (GPT) models only trained to predict the next token, or do they implicitly learn a world model from which a sequence is generated one token at a time? We examine this question by deriving a causal interpretation of the attention mechanism in GPT, and suggesting a causal world model that arises from this interpretation. Furthermore, we propose that GPT-models, at inference time, can be utilized for zero-shot causal structure learning for in-distribution sequences. Empirical evaluation is conducted in a controlled synthetic environment using the setup and rules of the Othello board game. A GPT, pre-trained on real-world games played with the intention of winning, is tested on synthetic data that only adheres to the game rules. We find that the GPT model tends to generate next moves that adhere to the game rules for sequences for which the attention mechanism encodes a causal structure with high confidence. In general, in cases for which the GPT model generates moves that do not adhere to the game rules, it also fails to capture any causal structure.
Authors: Youchao Zhou, Heyan Huang, Zhijing Wu, Yuhang Liu, Xinglin Wang
Abstract: Long-form document matching aims to judge the relevance between two documents and has been applied to various scenarios. Most existing works utilize hierarchical or long context models to process documents, which achieve coarse understanding but may ignore details. Some researchers construct a document view with similar sentences about aligned document subtopics to focus on detailed matching signals. However, a long document generally contains multiple subtopics. The matching signals are heterogeneous from multiple topics. Considering only the homologous aligned subtopics may not be representative enough and may cause biased modeling. In this paper, we introduce a new framework to model representative matching signals. First, we propose to capture various matching signals through subtopics of document pairs. Next, We construct multiple document views based on subtopics to cover heterogeneous and valuable details. However, existing spatial aggregation methods like attention, which integrate all these views simultaneously, are hard to integrate heterogeneous information. Instead, we propose temporal aggregation, which effectively integrates different views gradually as the training progresses. Experimental results show that our learning framework is effective on several document-matching tasks, including news duplication and legal case retrieval.
Authors: Xiaqiang Tang, Jian Li, Nan Du, Sihong Xie
Abstract: Despite the superior performance of Large language models on many NLP tasks, they still face significant limitations in memorizing extensive world knowledge. Recent studies have demonstrated that leveraging the Retrieval-Augmented Generation (RAG) framework, combined with Knowledge Graphs that encapsulate extensive factual data in a structured format, robustly enhances the reasoning capabilities of LLMs. However, deploying such systems in real-world scenarios presents challenges: the continuous evolution of non-stationary environments may lead to performance degradation and user satisfaction requires a careful balance of performance and responsiveness. To address these challenges, we introduce a Multi-objective Multi-Armed Bandit enhanced RAG framework, supported by multiple retrieval methods with diverse capabilities under rich and evolving retrieval contexts in practice. Within this framework, each retrieval method is treated as a distinct ``arm''. The system utilizes real-time user feedback to adapt to dynamic environments, by selecting the appropriate retrieval method based on input queries and the historical multi-objective performance of each arm. Extensive experiments conducted on two benchmark KGQA datasets demonstrate that our method significantly outperforms baseline methods in non-stationary settings while achieving state-of-the-art performance in stationary environments. Code and data are available at https://github.com/FUTUREEEEEE/Dynamic-RAG.git
Authors: Bocheng Chen, Hanqing Guo, Qiben Yan
Abstract: Defense in large language models (LLMs) is crucial to counter the numerous attackers exploiting these systems to generate harmful content through manipulated prompts, known as jailbreak attacks. Although many defense strategies have been proposed, they often require access to the model's internal structure or need additional training, which is impractical for service providers using LLM APIs, such as OpenAI APIs or Claude APIs. In this paper, we propose a moving target defense approach that alters decoding hyperparameters to enhance model robustness against various jailbreak attacks. Our approach does not require access to the model's internal structure and incurs no additional training costs. The proposed defense includes two key components: (1) optimizing the decoding strategy by identifying and adjusting decoding hyperparameters that influence token generation probabilities, and (2) transforming the decoding hyperparameters and model system prompts into dynamic targets, which are continuously altered during each runtime. By continuously modifying decoding strategies and prompts, the defense effectively mitigates the existing attacks. Our results demonstrate that our defense is the most effective against jailbreak attacks in three of the models tested when using LLMs as black-box APIs. Moreover, our defense offers lower inference costs and maintains comparable response quality, making it a potential layer of protection when used alongside other defense methods.
Authors: Le Li, Jiale Wei, Pai Peng, Qiyuan Chen, Benjamin Guedj, Bo Cai
Abstract: Data scarcity and data imbalance have attracted a lot of attention in many fields. Data augmentation, explored as an effective approach to tackle them, can improve the robustness and efficiency of classification models by generating new samples. This paper presents REPRINT, a simple and effective hidden-space data augmentation method for imbalanced data classification. Given hidden-space representations of samples in each class, REPRINT extrapolates, in a randomized fashion, augmented examples for target class by using subspaces spanned by principal components to summarize distribution structure of both source and target class. Consequently, the examples generated would diversify the target while maintaining the original geometry of target distribution. Besides, this method involves a label refinement component which allows to synthesize new soft labels for augmented examples. Compared with different NLP data augmentation approaches under a range of data imbalanced scenarios on four text classification benchmark, REPRINT shows prominent improvements. Moreover, through comprehensive ablation studies, we show that label refinement is better than label-preserving for augmented examples, and that our method suggests stable and consistent improvements in terms of suitable choices of principal components. Moreover, REPRINT is appealing for its easy-to-use since it contains only one hyperparameter determining the dimension of subspace and requires low computational resource.
Authors: Dongqi Liu, Yifan Wang, Vera Demberg
Abstract: For text summarization, the role of discourse structure is pivotal in discerning the core content of a text. Regrettably, prior studies on incorporating Rhetorical Structure Theory (RST) into transformer-based summarization models only consider the nuclearity annotation, thereby overlooking the variety of discourse relation types. This paper introduces the 'RSTformer', a novel summarization model that comprehensively incorporates both the types and uncertainty of rhetorical relations. Our RST-attention mechanism, rooted in document-level rhetorical structure, is an extension of the recently devised Longformer framework. Through rigorous evaluation, the model proposed herein exhibits significant superiority over state-of-the-art models, as evidenced by its notable performance on several automatic metrics and human evaluation.
Authors: Dongqi Liu, Vera Demberg
Abstract: Large-scale language models, like ChatGPT, have garnered significant media attention and stunned the public with their remarkable capacity for generating coherent text from short natural language prompts. In this paper, we aim to conduct a systematic inspection of ChatGPT's performance in two controllable generation tasks, with respect to ChatGPT's ability to adapt its output to different target audiences (expert vs. layman) and writing styles (formal vs. informal). Additionally, we evaluate the faithfulness of the generated text, and compare the model's performance with human-authored texts. Our findings indicate that the stylistic variations produced by humans are considerably larger than those demonstrated by ChatGPT, and the generated texts diverge from human samples in several characteristics, such as the distribution of word types. Moreover, we observe that ChatGPT sometimes incorporates factual errors or hallucinations when adapting the text to suit a specific style.
Authors: Zhenhua Wang, Simin He, Guang Xu, Ming Ren
Abstract: Nowadays, the omnipresence of the Internet has fostered a subculture that congregates around the contemporary milieu. The subculture artfully articulates the intricacies of human feelings by ardently pursuing the allure of novelty, a fact that cannot be disregarded in the sentiment analysis. This paper aims to enrich data through the lens of subculture, to address the insufficient training data faced by sentiment analysis. To this end, a new approach of subculture-based data augmentation (SCDA) is proposed, which engenders enhanced texts for each training text by leveraging the creation of specific subcultural expression generators. The extensive experiments attest to the effectiveness and potential of SCDA. The results also shed light on the phenomenon that disparate subcultural expressions elicit varying degrees of sentiment stimulation. Moreover, an intriguing conjecture arises, suggesting the linear reversibility of certain subcultural expressions.
Authors: Wen Liang, Youzhi Liang
Abstract: BERT (Bidirectional Encoder Representations from Transformers) has revolutionized the field of natural language processing through its exceptional performance on numerous tasks. Yet, the majority of researchers have mainly concentrated on enhancements related to the model structure, such as relative position embedding and more efficient attention mechanisms. Others have delved into pretraining tricks associated with Masked Language Modeling, including whole word masking. DeBERTa introduced an enhanced decoder adapted for BERT's encoder model for pretraining, proving to be highly effective. We argue that the design and research around enhanced masked language modeling decoders have been underappreciated. In this paper, we propose several designs of enhanced decoders and introduce BPDec (BERT Pretraining Decoder), a novel method for modeling training. Typically, a pretrained BERT model is fine-tuned for specific Natural Language Understanding (NLU) tasks. In our approach, we utilize the original BERT model as the encoder, making only changes to the decoder without altering the encoder. This approach does not necessitate extensive modifications to the encoder architecture and can be seamlessly integrated into existing fine-tuning pipelines and services, offering an efficient and effective enhancement strategy. Compared to other methods, while we also incur a moderate training cost for the decoder during the pretraining process, our approach does not introduce additional training costs during the fine-tuning phase. We test multiple enhanced decoder structures after pretraining and evaluate their performance on the GLUE tasks and SQuAD tasks. Our results demonstrate that BPDec, having only undergone subtle refinements to the model structure during pretraining, significantly enhances model performance without escalating the finetuning cost, inference time and serving budget.
Authors: Yiming Ai, Zhiwei He, Ziyin Zhang, Wenhong Zhu, Hongkun Hao, Kai Yu, Lingjun Chen, Rui Wang
Abstract: In this study, we delve into the validity of conventional personality questionnaires in capturing the human-like personality traits of Large Language Models (LLMs). Our objective is to assess the congruence between the personality traits LLMs claim to possess and their demonstrated tendencies in real-world scenarios. By conducting an extensive examination of LLM outputs against observed human response patterns, we aim to understand the disjunction between self-knowledge and action in LLMs.
Authors: Asad Aali, Dave Van Veen, Yamin Ishraq Arefeen, Jason Hom, Christian Bluethgen, Eduardo Pontes Reis, Sergios Gatidis, Namuun Clifford, Joseph Daws, Arash S. Tehrani, Jangwon Kim, Akshay S. Chaudhari
Abstract: Brief hospital course (BHC) summaries are clinical documents that summarize a patient's hospital stay. While large language models (LLMs) depict remarkable capabilities in automating real-world tasks, their capabilities for healthcare applications such as synthesizing BHCs from clinical notes have not been shown. We introduce a novel pre-processed dataset, the MIMIC-IV-BHC, encapsulating clinical note and brief hospital course (BHC) pairs to adapt LLMs for BHC synthesis. Furthermore, we introduce a benchmark of the summarization performance of two general-purpose LLMs and three healthcare-adapted LLMs. Using clinical notes as input, we apply prompting-based (using in-context learning) and fine-tuning-based adaptation strategies to three open-source LLMs (Clinical-T5-Large, Llama2-13B, FLAN-UL2) and two proprietary LLMs (GPT-3.5, GPT-4). We evaluate these LLMs across multiple context-length inputs using natural language similarity metrics. We further conduct a clinical study with five clinicians, comparing clinician-written and LLM-generated BHCs across 30 samples, focusing on their potential to enhance clinical decision-making through improved summary quality. We observe that the Llama2-13B fine-tuned LLM outperforms other domain-adapted models given quantitative evaluation metrics of BLEU and BERT-Score. GPT-4 with in-context learning shows more robustness to increasing context lengths of clinical note inputs than fine-tuned Llama2-13B. Despite comparable quantitative metrics, the reader study depicts a significant preference for summaries generated by GPT-4 with in-context learning compared to both Llama2-13B fine-tuned summaries and the original summaries, highlighting the need for qualitative clinical evaluation.
Authors: Jiaxing Sun, Weiquan Huang, Jiang Wu, Chenya Gu, Wei Li, Songyang Zhang, Hang Yan, Conghui He
Abstract: We introduce CHARM, the first benchmark for comprehensively and in-depth evaluating the commonsense reasoning ability of large language models (LLMs) in Chinese, which covers both globally known and Chinese-specific commonsense. We evaluated 7 English and 12 Chinese-oriented LLMs on CHARM, employing 5 representative prompt strategies for improving LLMs' reasoning ability, such as Chain-of-Thought. Our findings indicate that the LLM's language orientation and the task's domain influence the effectiveness of the prompt strategy, which enriches previous research findings. We built closely-interconnected reasoning and memorization tasks, and found that some LLMs struggle with memorizing Chinese commonsense, affecting their reasoning ability, while others show differences in reasoning despite similar memorization performance. We also evaluated the LLMs' memorization-independent reasoning abilities and analyzed the typical errors. Our study precisely identified the LLMs' strengths and weaknesses, providing the clear direction for optimization. It can also serve as a reference for studies in other fields. We will release CHARM at https://github.com/opendatalab/CHARM .
Authors: Dongqi Liu, Yifan Wang, Jia Loy, Vera Demberg
Abstract: Scientific news reports serve as a bridge, adeptly translating complex research articles into reports that resonate with the broader public. The automated generation of such narratives enhances the accessibility of scholarly insights. In this paper, we present a new corpus to facilitate this paradigm development. Our corpus comprises a parallel compilation of academic publications and their corresponding scientific news reports across nine disciplines. To demonstrate the utility and reliability of our dataset, we conduct an extensive analysis, highlighting the divergences in readability and brevity between scientific news narratives and academic manuscripts. We benchmark our dataset employing state-of-the-art text generation models. The evaluation process involves both automatic and human evaluation, which lays the groundwork for future explorations into the automated generation of scientific news reports. The dataset and code related to this work are available at https://dongqi.me/projects/SciNews.
Authors: Ye Tian, Baolin Peng, Linfeng Song, Lifeng Jin, Dian Yu, Haitao Mi, Dong Yu
Abstract: Despite the impressive capabilities of Large Language Models (LLMs) on various tasks, they still struggle with scenarios that involves complex reasoning and planning. Recent work proposed advanced prompting techniques and the necessity of fine-tuning with high-quality data to augment LLMs' reasoning abilities. However, these approaches are inherently constrained by data availability and quality. In light of this, self-correction and self-learning emerge as viable solutions, employing strategies that allow LLMs to refine their outputs and learn from self-assessed rewards. Yet, the efficacy of LLMs in self-refining its response, particularly in complex reasoning and planning task, remains dubious. In this paper, we introduce AlphaLLM for the self-improvements of LLMs, which integrates Monte Carlo Tree Search (MCTS) with LLMs to establish a self-improving loop, thereby enhancing the capabilities of LLMs without additional annotations. Drawing inspiration from the success of AlphaGo, AlphaLLM addresses the unique challenges of combining MCTS with LLM for self-improvement, including data scarcity, the vastness search spaces of language tasks, and the subjective nature of feedback in language tasks. AlphaLLM is comprised of prompt synthesis component, an efficient MCTS approach tailored for language tasks, and a trio of critic models for precise feedback. Our experimental results in mathematical reasoning tasks demonstrate that AlphaLLM significantly enhances the performance of LLMs without additional annotations, showing the potential for self-improvement in LLMs.
Authors: Dongqi Liu, Vera Demberg
Abstract: For long document summarization, discourse structure is important to discern the key content of the text and the differences in importance level between sentences. Unfortunately, the integration of rhetorical structure theory (RST) into parameter-efficient fine-tuning strategies for long document summarization remains unexplored. Therefore, this paper introduces RST-LoRA and proposes four RST-aware variants to explicitly incorporate RST into the LoRA model. Our empirical evaluation demonstrates that incorporating the type and uncertainty of rhetorical relations can complementarily enhance the performance of LoRA in summarization tasks. Furthermore, the best-performing variant we introduced outperforms the vanilla LoRA and full-parameter fine-tuning models, as confirmed by multiple automatic and human evaluations, and even surpasses previous state-of-the-art methods.
Authors: Jiahao Zhao, Jingwei Zhu, Minghuan Tan, Min Yang, Renhao Li, Di Yang, Chenhao Zhang, Guancheng Ye, Chengming Li, Xiping Hu, Derek F. Wong
Abstract: In this paper, we introduce a novel psychological benchmark, CPsyExam, constructed from questions sourced from Chinese language examinations. CPsyExam is designed to prioritize psychological knowledge and case analysis separately, recognizing the significance of applying psychological knowledge to real-world scenarios. From the pool of 22k questions, we utilize 4k to create the benchmark that offers balanced coverage of subjects and incorporates a diverse range of case analysis techniques.Furthermore, we evaluate a range of existing large language models~(LLMs), spanning from open-sourced to API-based models. Our experiments and analysis demonstrate that CPsyExam serves as an effective benchmark for enhancing the understanding of psychology within LLMs and enables the comparison of LLMs across various granularities.
Authors: Reid McIlroy-Young, Katrina Brown, Conlan Olson, Linjun Zhang, Cynthia Dwork
Abstract: The development of generative language models that can create long and coherent textual outputs via autoregression has lead to a proliferation of uses and a corresponding sweep of analyses as researches work to determine the limitations of this new paradigm. Unlike humans, these 'Large Language Models' (LLMs) are highly sensitive to small changes in their inputs, leading to unwanted inconsistency in their behavior. One problematic inconsistency when LLMs are used to answer multiple-choice questions or analyze multiple inputs is order dependency: the output of an LLM can (and often does) change significantly when sub-sequences are swapped, despite both orderings being semantically identical. In this paper we present Set-Based Prompting, a technique that guarantees the output of an LLM will not have order dependence on a specified set of sub-sequences. We show that this method provably eliminates order dependency, and that it can be applied to any transformer-based LLM to enable text generation that is unaffected by re-orderings. Delving into the implications of our method, we show that, despite our inputs being out of distribution, the impact on expected accuracy is small, where the expectation is over the order of uniformly chosen shuffling of the candidate responses, and usually significantly less in practice. Thus, Set-Based Prompting can be used as a 'dropped-in' method on fully trained models. Finally, we discuss how our method's success suggests that other strong guarantees can be obtained on LLM performance via modifying the input representations.
Authors: Tao Zhang, Ziqian Zeng, Yuxiang Xiao, Huiping Zhuang, Cen Chen, James Foulds, Shimei Pan
Abstract: Large Language Models (LLMs) are prone to generating content that exhibits gender biases, raising significant ethical concerns. Alignment, the process of fine-tuning LLMs to better align with desired behaviors, is recognized as an effective approach to mitigate gender biases. Although proprietary LLMs have made significant strides in mitigating gender bias, their alignment datasets are not publicly available. The commonly used and publicly available alignment dataset, HH-RLHF, still exhibits gender bias to some extent. There is a lack of publicly available alignment datasets specifically designed to address gender bias. Hence, we developed a new dataset named GenderAlign, aiming at mitigating a comprehensive set of gender biases in LLMs. This dataset comprises 8k single-turn dialogues, each paired with a "chosen" and a "rejected" response. Compared to the "rejected" responses, the "chosen" responses demonstrate lower levels of gender bias and higher quality. Furthermore, we categorized the gender biases in the "rejected" responses of GenderAlign into 4 principal categories. The experimental results show the effectiveness of GenderAlign in reducing gender bias in LLMs.
Authors: Federico Ruggeri, Eleonora Misino, Arianna Muti, Katerina Korre, Paolo Torroni, Alberto Barr\'on-Cede\~no
Abstract: We introduce the Guideline-Centered Annotation Methodology (GCAM), a novel data annotation methodology designed to report the annotation guidelines associated with each data sample. Our approach addresses three key limitations of the standard prescriptive annotation methodology by reducing the information loss during annotation and ensuring adherence to guidelines. Furthermore, GCAM enables the efficient reuse of annotated data across multiple tasks. We evaluate GCAM in two ways: (i) through a human annotation study and (ii) an experimental evaluation with several machine learning models. Our results highlight the advantages of GCAM from multiple perspectives, demonstrating its potential to improve annotation quality and error analysis.
Authors: Jiale Cheng, Yida Lu, Xiaotao Gu, Pei Ke, Xiao Liu, Yuxiao Dong, Hongning Wang, Jie Tang, Minlie Huang
Abstract: Although Large Language Models (LLMs) are becoming increasingly powerful, they still exhibit significant but subtle weaknesses, such as mistakes in instruction-following or coding tasks. As these unexpected errors could lead to severe consequences in practical deployments, it is crucial to investigate the limitations within LLMs systematically. Traditional benchmarking approaches cannot thoroughly pinpoint specific model deficiencies, while manual inspections are costly and not scalable. In this paper, we introduce a unified framework, AutoDetect, to automatically expose weaknesses in LLMs across various tasks. Inspired by the educational assessment process that measures students' learning outcomes, AutoDetect consists of three LLM-powered agents: Examiner, Questioner, and Assessor. The collaboration among these three agents is designed to realize comprehensive and in-depth weakness identification. Our framework demonstrates significant success in uncovering flaws, with an identification success rate exceeding 30% in prominent models such as ChatGPT and Claude. More importantly, these identified weaknesses can guide specific model improvements, proving more effective than untargeted data augmentation methods like Self-Instruct. Our approach has led to substantial enhancements in popular LLMs, including the Llama series and Mistral-7b, boosting their performance by over 10% across several benchmarks. Code and data are publicly available at https://github.com/thu-coai/AutoDetect.
Authors: Seungju Han, Kavel Rao, Allyson Ettinger, Liwei Jiang, Bill Yuchen Lin, Nathan Lambert, Yejin Choi, Nouha Dziri
Abstract: We introduce WildGuard -- an open, light-weight moderation tool for LLM safety that achieves three goals: (1) identifying malicious intent in user prompts, (2) detecting safety risks of model responses, and (3) determining model refusal rate. Together, WildGuard serves the increasing needs for automatic safety moderation and evaluation of LLM interactions, providing a one-stop tool with enhanced accuracy and broad coverage across 13 risk categories. While existing open moderation tools such as Llama-Guard2 score reasonably well in classifying straightforward model interactions, they lag far behind a prompted GPT-4, especially in identifying adversarial jailbreaks and in evaluating models' refusals, a key measure for evaluating safety behaviors in model responses. To address these challenges, we construct WildGuardMix, a large-scale and carefully balanced multi-task safety moderation dataset with 92K labeled examples that cover vanilla (direct) prompts and adversarial jailbreaks, paired with various refusal and compliance responses. WildGuardMix is a combination of WildGuardTrain, the training data of WildGuard, and WildGuardTest, a high-quality human-annotated moderation test set with 5K labeled items covering broad risk scenarios. Through extensive evaluations on WildGuardTest and ten existing public benchmarks, we show that WildGuard establishes state-of-the-art performance in open-source safety moderation across all the three tasks compared to ten strong existing open-source moderation models (e.g., up to 26.4% improvement on refusal detection). Importantly, WildGuard matches and sometimes exceeds GPT-4 performance (e.g., up to 3.9% improvement on prompt harmfulness identification). WildGuard serves as a highly effective safety moderator in an LLM interface, reducing the success rate of jailbreak attacks from 79.8% to 2.4%.
Authors: Erik Henriksson, Amanda Myntti, Saara Hellstr\"om, Anni Eskelinen, Selcen Erten-Johansson, Veronika Laippala
Abstract: This article investigates how well deep learning models can identify web registers -- text varieties such as news reports and discussion forums -- across 16 languages. We introduce the Multilingual CORE corpora, which contain 72,504 documents annotated with a hierarchical taxonomy of 25 registers designed to cover the entire open web. Our multilingual models achieve state-of-the-art results (79% F1 score) using multi-label classification. This performance matches or exceeds previous studies that used simpler classification schemes, showing that models can perform well even with a complex register scheme at a massively multilingual scale. However, we observe a consistent performance ceiling around 77-80% F1 score across all models and configurations. When we remove documents with uncertain labels through data pruning, performance increases to over 90% F1, suggesting that this ceiling stems from inherent ambiguity in web registers rather than model limitations. Analysis of hybrid documents -- texts combining multiple registers -- reveals that the main challenge is not in classifying hybrids themselves, but in distinguishing between hybrid and non-hybrid documents. Multilingual models consistently outperform monolingual ones, particularly helping languages with limited training data. While zero-shot performance drops by an average of 7% on unseen languages, this decrease varies substantially between languages (from 3% to 20%), indicating that while registers share many features across languages, they also maintain language-specific characteristics.
Authors: Sadia Alam, Md Farhan Ishmam, Navid Hasin Alvee, Md Shahnewaz Siddique, Md Azam Hossain, Abu Raihan Mostofa Kamal
Abstract: The widespread availability of code-mixed data can provide valuable insights into low-resource languages like Bengali, which have limited datasets. Sentiment analysis has been a fundamental text classification task across several languages for code-mixed data. However, there has yet to be a large-scale and diverse sentiment analysis dataset on code-mixed Bengali. We address this limitation by introducing BnSentMix, a sentiment analysis dataset on code-mixed Bengali consisting of 20,000 samples with 4 sentiment labels from Facebook, YouTube, and e-commerce sites. We ensure diversity in data sources to replicate realistic code-mixed scenarios. Additionally, we propose 14 baseline methods including novel transformer encoders further pre-trained on code-mixed Bengali-English, achieving an overall accuracy of 69.8% and an F1 score of 69.1% on sentiment classification tasks. Detailed analyses reveal variations in performance across different sentiment labels and text types, highlighting areas for future improvement.
Authors: Takumi Goto, Hiroyoshi Nagao, Yuta Koreeda
Abstract: Using token representation from bidirectional language models (LMs) such as BERT is still a widely used approach for token-classification tasks. Even though there exist much larger unidirectional LMs such as Llama-2, they are rarely used to replace the token representation of bidirectional LMs. In this work, we hypothesize that their lack of bidirectionality is keeping them behind. To that end, we propose to newly train a small backward LM and concatenate its representations to those of existing LM for downstream tasks. Through experiments in named entity recognition, we demonstrate that introducing backward model improves the benchmark performance more than 10 points. Furthermore, we show that the proposed method is especially effective for rare domains and in few-shot learning settings.
Authors: Leyi Pan, Aiwei Liu, Yijian Lu, Zitian Gao, Yichen Di, Shiyu Huang, Lijie Wen, Irwin King, Philip S. Yu
Abstract: Watermarking algorithms for large language models (LLMs) have attained high accuracy in detecting LLM-generated text. However, existing methods primarily focus on distinguishing fully watermarked text from non-watermarked text, overlooking real-world scenarios where LLMs generate only small sections within large documents. In this scenario, balancing time complexity and detection performance poses significant challenges. This paper presents WaterSeeker, a novel approach to efficiently detect and locate watermarked segments amid extensive natural text. It first applies an efficient anomaly extraction method to preliminarily locate suspicious watermarked regions. Following this, it conducts a local traversal and performs full-text detection for more precise verification. Theoretical analysis and experimental results demonstrate that WaterSeeker achieves a superior balance between detection accuracy and computational efficiency. Moreover, its localization capability lays the foundation for building interpretable AI detection systems. Our code is available at https://github.com/THU-BPM/WaterSeeker.
Authors: Yuxin Wang, Minghua Ma, Zekun Wang, Jingchang Chen, Huiming Fan, Liping Shan, Qing Yang, Dongliang Xu, Ming Liu, Bing Qin
Abstract: The colossal parameters and computational overhead of Large Language Models (LLMs) challenge their real-world applications. Network pruning, which targets unstructured or structured sparsity by removing redundant parameters, has recently been explored for LLM acceleration. Existing LLM pruning works focus on unstructured pruning, which typically requires special hardware support for a practical speed-up. In contrast, structured pruning can reduce latency on general devices. However, it remains a challenge to perform structured pruning efficiently and maintain performance, especially at high sparsity ratios. To this end, we introduce an efficient structured pruning framework named CFSP, which leverages both Coarse (interblock) and Fine-grained (intrablock) activation information as an importance criterion to guide pruning. The pruning is highly efficient, as it only requires one forward pass to compute feature activations. Specifically, we first allocate the sparsity budget across blocks based on their importance and then retain important weights within each block. In addition, we introduce a recovery fine-tuning strategy that adaptively allocates training overhead based on coarse-grained importance to further improve performance. Experimental results demonstrate that CFSP outperforms existing methods on diverse models across various sparsity budgets. Our code will be available at https://github.com/wyxscir/CFSP.
Authors: Lukas Birkenmaier, Matthias Roth, Indira Sen
Abstract: Computational text classification is a challenging task, especially for multi-dimensional social constructs. Recently, there has been increasing discussion that synthetic training data could enhance classification by offering examples of how these constructs are represented in texts. In this paper, we systematically examine the potential of theory-driven synthetic training data for improving the measurement of social constructs. In particular, we explore how researchers can transfer established knowledge from measurement instruments in the social sciences, such as survey scales or annotation codebooks, into theory-driven generation of synthetic data. Using two studies on measuring sexism and political topics, we assess the added value of synthetic training data for fine-tuning text classification models. Although the results of the sexism study were less promising, our findings demonstrate that synthetic data can be highly effective in reducing the need for labeled data in political topic classification. With only a minimal drop in performance, synthetic data allows for substituting large amounts of labeled data. Furthermore, theory-driven synthetic data performed markedly better than data generated without conceptual information in mind.
Authors: Shantanu Thorat, Tianbao Yang
Abstract: As LLMs increase in accessibility, LLM-generated texts have proliferated across several fields, such as scientific, academic, and creative writing. However, LLMs are not created equally; they may have different architectures and training datasets. Thus, some LLMs may be more challenging to detect than others. Using two datasets spanning four total writing domains, we train AI-generated (AIG) text classifiers using the LibAUC library - a deep learning library for training classifiers with imbalanced datasets. Our results in the Deepfake Text dataset show that AIG-text detection varies across domains, with scientific writing being relatively challenging. In the Rewritten Ivy Panda (RIP) dataset focusing on student essays, we find that the OpenAI family of LLMs was substantially difficult for our classifiers to distinguish from human texts. Additionally, we explore possible factors that could explain the difficulties in detecting OpenAI-generated texts.
Authors: Krishna Sayana, Raghavendra Vasudeva, Yuri Vasilevski, Kun Su, Liam Hebert, James Pine, Hubert Pham, Ambarish Jash, Sukhdeep Sodhi
Abstract: The recent advances in Large Language Model's generation and reasoning capabilities present an opportunity to develop truly conversational recommendation systems. However, effectively integrating recommender system knowledge into LLMs for natural language generation which is tailored towards recommendation tasks remains a challenge. This paper addresses this challenge by making two key contributions. First, we introduce a new dataset (REGEN) for natural language generation tasks in conversational recommendations. REGEN (Reviews Enhanced with GEnerative Narratives) extends the Amazon Product Reviews dataset with rich user narratives, including personalized explanations of product preferences, product endorsements for recommended items, and summaries of user purchase history. REGEN is made publicly available to facilitate further research. Furthermore, we establish benchmarks using well-known generative metrics, and perform an automated evaluation of the new dataset using a rater LLM. Second, the paper introduces a fusion architecture (CF model with an LLM) which serves as a baseline for REGEN. And to the best of our knowledge, represents the first attempt to analyze the capabilities of LLMs in understanding recommender signals and generating rich narratives. We demonstrate that LLMs can effectively learn from simple fusion architectures utilizing interaction-based CF embeddings, and this can be further enhanced using the metadata and personalization data associated with items. Our experiments show that combining CF and content embeddings leads to improvements of 4-12% in key language metrics compared to using either type of embedding individually. We also provide an analysis to interpret how CF and content embeddings contribute to this new generative task.
Authors: Tyler A. Chang, Dheeraj Rajagopal, Tolga Bolukbasi, Lucas Dixon, Ian Tenney
Abstract: Training data attribution (TDA) methods aim to attribute model outputs back to specific training examples, and the application of these methods to large language model (LLM) outputs could significantly advance model transparency and data curation. However, it has been challenging to date to apply these methods to the full scale of LLM pretraining. In this paper, we refine existing gradient-based methods to work effectively at scale, allowing us to retrieve influential examples for an 8B-parameter language model from a pretraining corpus of over 160B tokens with no need for subsampling or pre-filtering. Our method combines several techniques, including optimizer state correction, a task-specific Hessian approximation, and normalized encodings, which we find to be critical for performance at scale. In quantitative evaluations on a fact tracing task, our method performs best at identifying examples that influence model predictions, but classical, model-agnostic retrieval methods such as BM25 still perform better at finding passages which explicitly contain relevant facts. These results demonstrate a misalignment between factual *attribution* and causal *influence*. With increasing model size and training tokens, we find that influence more closely aligns with factual attribution. Finally, we examine different types of examples identified as influential by our method, finding that while many directly entail a particular fact, others support the same output by reinforcing priors on relation types, common entities, and names. We release our prompt set and model outputs, along with a web-based visualization tool to explore influential examples for factual predictions, commonsense reasoning, arithmetic, and open-ended generation for an 8B-parameter LLM.
Authors: Haitao Li, Qian Dong, Junjie Chen, Huixue Su, Yujia Zhou, Qingyao Ai, Ziyi Ye, Yiqun Liu
Abstract: The rapid advancement of Large Language Models (LLMs) has driven their expanding application across various fields. One of the most promising applications is their role as evaluators based on natural language responses, referred to as ''LLMs-as-judges''. This framework has attracted growing attention from both academia and industry due to their excellent effectiveness, ability to generalize across tasks, and interpretability in the form of natural language. This paper presents a comprehensive survey of the LLMs-as-judges paradigm from five key perspectives: Functionality, Methodology, Applications, Meta-evaluation, and Limitations. We begin by providing a systematic definition of LLMs-as-Judges and introduce their functionality (Why use LLM judges?). Then we address methodology to construct an evaluation system with LLMs (How to use LLM judges?). Additionally, we investigate the potential domains for their application (Where to use LLM judges?) and discuss methods for evaluating them in various contexts (How to evaluate LLM judges?). Finally, we provide a detailed analysis of the limitations of LLM judges and discuss potential future directions. Through a structured and comprehensive analysis, we aim aims to provide insights on the development and application of LLMs-as-judges in both research and practice. We will continue to maintain the relevant resource list at https://github.com/CSHaitao/Awesome-LLMs-as-Judges.
Authors: Yilong Chen, Junyuan Shang, Zhengyu Zhang, Jiawei Sheng, Tingwen Liu, Shuohuan Wang, Yu Sun, Hua Wu, Haifeng Wang
Abstract: Transformer models encounter challenges in scaling hidden dimensions efficiently, as uniformly increasing them inflates computational and memory costs while failing to emphasize the most relevant features for each token. For further understanding, we study hidden dimension sparsity and observe that trained Transformers utilize only a small fraction of token dimensions, revealing an "activation flow" pattern. Notably, there are shared sub-dimensions with sustained activation across multiple consecutive tokens and specialized sub-dimensions uniquely activated for each token. To better model token-relevant sub-dimensions, we propose MoHD (Mixture of Hidden Dimensions), a sparse conditional activation architecture. Particularly, MoHD employs shared sub-dimensions for common token features and a routing mechanism to dynamically activate specialized sub-dimensions. To mitigate potential information loss from sparsity, we design activation scaling and group fusion mechanisms to preserve activation flow. In this way, MoHD expands hidden dimensions with negligible increases in computation or parameters, efficient training and inference while maintaining performance. Evaluations across 10 NLP tasks show that MoHD surpasses Vanilla Transformers in parameter efficiency and task performance. It achieves 1.7% higher performance with 50% fewer activation parameters and 3.7% higher performance with a 3x parameter expansion at constant activation cost. MOHD offers a new perspective for scaling the model, showcasing the potential of hidden dimension sparsity to boost efficiency
Authors: Qing Zhang, Haocheng Lv, Jie Liu, Zhiyun Chen, Jianyong Duan, Hao Wang, Li He, Mingying Xv
Abstract: With the rise of large-scale language models (LLMs), it is currently popular and effective to convert multimodal information into text descriptions for multimodal multi-hop question answering. However, we argue that the current methods of multi-modal multi-hop question answering still mainly face two challenges: 1) The retrieved evidence containing a large amount of redundant information, inevitably leads to a significant drop in performance due to irrelevant information misleading the prediction. 2) The reasoning process without interpretable reasoning steps makes the model difficult to discover the logical errors for handling complex questions. To solve these problems, we propose a unified LLMs-based approach but without heavily relying on them due to the LLM's potential errors, and innovatively treat multimodal multi-hop question answering as a joint entailment tree generation and question answering problem. Specifically, we design a multi-task learning framework with a focus on facilitating common knowledge sharing across interpretability and prediction tasks while preventing task-specific errors from interfering with each other via mixture of experts. Afterward, we design an iterative feedback mechanism to further enhance both tasks by feeding back the results of the joint training to the LLM for regenerating entailment trees, aiming to iteratively refine the potential answer. Notably, our method has won the first place in the official leaderboard of WebQA (since April 10, 2024), and achieves competitive results on MultimodalQA.
Authors: Min Zeng, Caiquan Liu, Shiqi Zhang, Li Xie, Chen Sang, Xiaoxin Chen
Abstract: In recent years, the use of large language models (LLMs) for text classification has attracted widespread attention. Despite this, the classification accuracy of LLMs has not yet universally surpassed that of smaller models. LLMs can enhance their performance in text classification through fine-tuning. However, existing data quality research based on LLMs is challenging to apply directly to solve text classification problems. To further improve the performance of LLMs in classification tasks, this paper proposes a data quality enhancement (DQE) method for text classification based on LLMs. This method starts by using a greedy algorithm to select data, dividing the dataset into sampled and unsampled subsets, and then performing fine-tuning of the LLMs using the sampled data. Subsequently, this model is used to predict the outcomes for the unsampled data, categorizing incorrectly predicted data into uncovered, difficult, and noisy data. Experimental results demonstrate that our method effectively enhances the performance of LLMs in text classification tasks and significantly improves training efficiency, saving nearly half of the training time. Our method has achieved state-of-the-art performance in several open-source classification tasks.
Authors: Jake Chanenson, Madison Pickering, Noah Apthorpe
Abstract: Identifying contextual integrity (CI) and governing knowledge commons (GKC) parameters in privacy policy texts can facilitate normative privacy analysis. However, GKC-CI annotation has heretofore required manual or crowdsourced effort. This paper demonstrates that high-accuracy GKC-CI parameter annotation of privacy policies can be performed automatically using large language models. We fine-tune 50 open-source and proprietary models on 21,588 ground truth GKC-CI annotations from 16 privacy policies. Our best performing model has an accuracy of 90.65%, which is comparable to the accuracy of experts on the same task. We apply our best performing model to 456 privacy policies from a variety of online services, demonstrating the effectiveness of scaling GKC-CI annotation for privacy policy exploration and analysis. We publicly release our model training code, training and testing data, an annotation visualizer, and all annotated policies for future GKC-CI research.
Authors: Matthew Gwilliam, Michael Cogswell, Meng Ye, Karan Sikka, Abhinav Shrivastava, Ajay Divakaran
Abstract: Existing long video retrieval systems are trained and tested in the paragraph-to-video retrieval regime, where every long video is described by a single long paragraph. This neglects the richness and variety of possible valid descriptions of a video, which could range anywhere from moment-by-moment detail to a single phrase summary. To provide a more thorough evaluation of the capabilities of long video retrieval systems, we propose a pipeline that leverages state-of-the-art large language models to carefully generate a diverse set of synthetic captions for long videos. We validate this pipeline's fidelity via rigorous human inspection. We use synthetic captions from this pipeline to perform a benchmark of a representative set of video language models using long video datasets, and show that the models struggle on shorter captions. We show that finetuning on this data can both mitigate these issues (+2.8% R@1 over SOTA on ActivityNet with diverse captions), and even improve performance on standard paragraph-to-video retrieval (+1.0% R@1 on ActivityNet). We also use synthetic data from our pipeline as query expansion in the zero-shot setting (+3.4% R@1 on ActivityNet). We derive insights by analyzing failure cases for retrieval with short captions. For data access and other details, please refer to our project website at https://mgwillia.github.io/10k-words.
Authors: Xisen Jin, Xiang Ren
Abstract: Language models deployed in the wild make errors. However, simply updating the model with the corrected error instances causes catastrophic forgetting -- the updated model makes errors on instances learned during the instruction tuning or upstream training phase. Randomly replaying upstream data yields unsatisfactory performance and often comes with high variance and poor controllability. To this end, we try to forecast upstream examples that will be forgotten due to a model update for improved controllability of the replay process and interpretability. We train forecasting models given a collection of online learned examples and corresponding forgotten upstream pre-training examples. We propose a partially interpretable forecasting model based on the observation that changes in pre-softmax logit scores of pretraining examples resemble that of online learned examples, which performs decently on BART but fails on T5 models. We further show a black-box classifier based on inner products of example representations achieves better forecasting performance over a series of setups. Finally, we show that we reduce forgetting of upstream pretraining examples by replaying examples that are forecasted to be forgotten, demonstrating the practical utility of forecasting example forgetting.
Authors: Zhiqing Sun, Longhui Yu, Yikang Shen, Weiyang Liu, Yiming Yang, Sean Welleck, Chuang Gan
Abstract: Current AI alignment methodologies rely on human-provided demonstrations or judgments, and the learned capabilities of AI systems would be upper-bounded by human capabilities as a result. This raises a challenging research question: How can we keep improving the systems when their capabilities have surpassed the levels of humans? This paper answers this question in the context of tackling hard reasoning tasks (e.g., level 4-5 MATH problems) via learning from human annotations on easier tasks (e.g., level 1-3 MATH problems), which we term as easy-to-hard generalization. Our key insight is that an evaluator (reward model) trained on supervisions for easier tasks can be effectively used for scoring candidate solutions of harder tasks and hence facilitating easy-to-hard generalization over different levels of tasks. Based on this insight, we propose a novel approach to scalable alignment, which firstly trains the (process-supervised) reward models on easy problems (e.g., level 1-3), and then uses them to evaluate the performance of policy models on hard problems. We show that such easy-to-hard generalization from evaluators can enable easy-to-hard generalizations in generators either through re-ranking or reinforcement learning (RL). Notably, our process-supervised 7b RL model and 34b model (reranking@1024) achieves an accuracy of 34.0% and 52.5% on MATH500, respectively, despite only using human supervision on easy problems. Our approach suggests a promising path toward AI systems that advance beyond the frontier of human supervision.
Authors: Zhaolin Gao, Jonathan D. Chang, Wenhao Zhan, Owen Oertell, Gokul Swamy, Kiant\'e Brantley, Thorsten Joachims, J. Andrew Bagnell, Jason D. Lee, Wen Sun
Abstract: While originally developed for continuous control problems, Proximal Policy Optimization (PPO) has emerged as the work-horse of a variety of reinforcement learning (RL) applications, including the fine-tuning of generative models. Unfortunately, PPO requires multiple heuristics to enable stable convergence (e.g. value networks, clipping), and is notorious for its sensitivity to the precise implementation of these components. In response, we take a step back and ask what a minimalist RL algorithm for the era of generative models would look like. We propose REBEL, an algorithm that cleanly reduces the problem of policy optimization to regressing the relative reward between two completions to a prompt in terms of the policy, enabling strikingly lightweight implementation. In theory, we prove that fundamental RL algorithms like Natural Policy Gradient can be seen as variants of REBEL, which allows us to match the strongest known theoretical guarantees in terms of convergence and sample complexity in the RL literature. REBEL can also cleanly incorporate offline data and be extended to handle the intransitive preferences we frequently see in practice. Empirically, we find that REBEL provides a unified approach to language modeling and image generation with stronger or similar performance as PPO and DPO, all while being simpler to implement and more computationally efficient than PPO. When fine-tuning Llama-3-8B-Instruct, REBEL achieves strong performance in AlpacaEval 2.0, MT-Bench, and Open LLM Leaderboard.
Authors: Siyan Zhao, Tung Nguyen, Aditya Grover
Abstract: In-context learning is a key paradigm in large language models (LLMs) that enables them to generalize to new tasks and domains by simply prompting these models with a few exemplars without explicit parameter updates. Many attempts have been made to understand in-context learning in LLMs as a function of model scale, pretraining data, and other factors. In this work, we propose a new mechanism to probe and understand in-context learning from the lens of decision boundaries for in-context binary classification. Decision boundaries are straightforward to visualize and provide important information about the qualitative behavior of the inductive biases of standard classifiers. To our surprise, we find that the decision boundaries learned by current LLMs in simple binary classification tasks are often irregular and non-smooth, regardless of linear separability in the underlying task. This paper investigates the factors influencing these decision boundaries and explores methods to enhance their generalizability. We assess various approaches, including training-free and fine-tuning methods for LLMs, the impact of model architecture, and the effectiveness of active prompting techniques for smoothing decision boundaries in a data-efficient manner. Our findings provide a deeper understanding of in-context learning dynamics and offer practical improvements for enhancing robustness and generalizability of in-context learning.
Authors: Taejin Park, Ivan Medennikov, Kunal Dhawan, Weiqing Wang, He Huang, Nithin Rao Koluguri, Krishna C. Puvvada, Jagadeesh Balam, Boris Ginsburg
Abstract: We propose Sortformer, a novel neural model for speaker diarization, trained with unconventional objectives compared to existing end-to-end diarization models. The permutation problem in speaker diarization has long been regarded as a critical challenge. Most prior end-to-end diarization systems employ permutation invariant loss (PIL), which optimizes for the permutation that yields the lowest error. In contrast, we introduce Sort Loss, which enables a diarization model to autonomously resolve permutation, with or without PIL. We demonstrate that combining Sort Loss and PIL achieves performance competitive with state-of-the-art end-to-end diarization models trained exclusively with PIL. Crucially, we present a streamlined multispeaker ASR architecture that leverages Sortformer as a speaker supervision model, embedding speaker label estimation within the ASR encoder state using a sinusoidal kernel function. This approach resolves the speaker permutation problem through sorted objectives, effectively bridging speaker-label timestamps and speaker tokens. In our experiments, we show that the proposed multispeaker ASR architecture, enhanced with speaker supervision, improves performance via adapter techniques. Code and trained models will be made publicly available via the NVIDIA NeMo framework.
Authors: Neelabh Sinha, Vinija Jain, Aman Chadha
Abstract: Visual Question-Answering (VQA) has become key to user experience, particularly after improved generalization capabilities of Vision-Language Models (VLMs). But evaluating VLMs for an application requirement using a standardized framework in practical settings is still challenging. This paper aims to solve that using an end-to-end framework. We present VQA360 - a novel dataset derived from established VQA benchmarks, annotated with task types, application domains, and knowledge types, for a comprehensive evaluation. We also introduce GoEval, a multimodal evaluation metric developed using GPT-4o, achieving a correlation factor of 56.71% with human judgments. Our experiments with state-of-the-art VLMs reveal that no single model excels universally, thus, making a right choice a key design decision. Proprietary models such as Gemini-1.5-Pro and GPT-4o-mini generally outperform others, but open-source models like InternVL-2-8B and CogVLM-2-Llama-3-19B also demonstrate competitive strengths, while providing additional advantages. Our framework can also be extended to other tasks.
Authors: Qian Niu, Keyu Chen, Ming Li, Pohsun Feng, Ziqian Bi, Lawrence KQ Yan, Yichao Zhang, Caitlyn Heqi Yin, Cheng Fei, Junyu Liu, Benji Peng, Tianyang Wang, Yunze Wang, Silin Chen, Ming Liu
Abstract: Large Language Models (LLMs) have rapidly evolved from text-based systems to multimodal platforms, significantly impacting various sectors including healthcare. This comprehensive review explores the progression of LLMs to Multimodal Large Language Models (MLLMs) and their growing influence in medical practice. We examine the current landscape of MLLMs in healthcare, analyzing their applications across clinical decision support, medical imaging, patient engagement, and research. The review highlights the unique capabilities of MLLMs in integrating diverse data types, such as text, images, and audio, to provide more comprehensive insights into patient health. We also address the challenges facing MLLM implementation, including data limitations, technical hurdles, and ethical considerations. By identifying key research gaps, this paper aims to guide future investigations in areas such as dataset development, modality alignment methods, and the establishment of ethical guidelines. As MLLMs continue to shape the future of healthcare, understanding their potential and limitations is crucial for their responsible and effective integration into medical practice.
Authors: Mingliang Liang, Martha Larson
Abstract: We propose Word-Frequency-based Image-Text Pair Pruning (WFPP), a novel data pruning method that improves the efficiency of VLMs. Unlike MetaCLIP, our method does not need metadata for pruning, but selects text-image pairs to prune based on the content of the text. Specifically, WFPP prunes text-image pairs containing high-frequency words across the entire training dataset. The effect of WFPP is to reduce the dominance of frequent words. The result a better balanced word-frequency distribution in the dataset, which is known to improve the training of word embedding models. After pre-training on the pruned subset, we fine-tuned the model on the entire dataset for one additional epoch to achieve better performance. Our experiments demonstrate that applying WFPP when training a CLIP model improves performance on a wide range of downstream tasks. WFPP also provides the advantage of speeding up pre-training by using fewer samples. Additionally, we analyze the training data before and after pruning to visualize how WFPP changes the balance of word frequencies. We hope our work encourages researchers to consider the distribution of words in the training data when pre-training VLMs, not limited to CLIP.
Authors: Yuchen Cao, Jianglai Dai, Zhongyan Wang, Yeyubei Zhang, Xiaorui Shen, Yunchong Liu, Yexin Tian
Abstract: The global rise in depression necessitates innovative detection methods for early intervention. Social media provides a unique opportunity to identify depression through user-generated posts. This systematic review evaluates machine learning (ML) models for depression detection on social media, focusing on biases and methodological challenges throughout the ML lifecycle. A search of PubMed, IEEE Xplore, and Google Scholar identified 47 relevant studies published after 2010. The Prediction model Risk Of Bias ASsessment Tool (PROBAST) was utilized to assess methodological quality and risk of bias. Significant biases impacting model reliability and generalizability were found. There is a predominant reliance on Twitter (63.8%) and English-language content (over 90%), with most studies focusing on users from the United States and Europe. Non-probability sampling methods (approximately 80%) limit representativeness. Only 23% of studies explicitly addressed linguistic nuances like negations, crucial for accurate sentiment analysis. Inconsistent hyperparameter tuning was observed, with only 27.7% properly tuning models. About 17% did not adequately partition data into training, validation, and test sets, risking overfitting. While 74.5% used appropriate evaluation metrics for imbalanced data, others relied on accuracy without addressing class imbalance, potentially skewing results. Reporting transparency varied, often lacking critical methodological details. These findings highlight the need to diversify data sources, standardize preprocessing protocols, ensure consistent model development practices, address class imbalance, and enhance reporting transparency. By overcoming these challenges, future research can develop more robust and generalizable ML models for depression detection on social media, contributing to improved mental health outcomes globally.
Authors: Juyong Lee, Dongyoon Hahm, June Suk Choi, W. Bradley Knox, Kimin Lee
Abstract: Autonomous agents powered by large language models (LLMs) show promising potential in assistive tasks across various domains, including mobile device control. As these agents interact directly with personal information and device settings, ensuring their safe and reliable behavior is crucial to prevent undesirable outcomes. However, no benchmark exists for standardized evaluation of the safety of mobile device-control agents. In this work, we introduce MobileSafetyBench, a benchmark designed to evaluate the safety of device-control agents within a realistic mobile environment based on Android emulators. We develop a diverse set of tasks involving interactions with various mobile applications, including messaging and banking applications, challenging agents with managing risks encompassing misuse and negative side effects. These tasks include tests to evaluate the safety of agents in daily scenarios as well as their robustness against indirect prompt injection attacks. Our experiments demonstrate that baseline agents, based on state-of-the-art LLMs, often fail to effectively prevent harm while performing the tasks. To mitigate these safety concerns, we propose a prompting method that encourages agents to prioritize safety considerations. While this method shows promise in promoting safer behaviors, there is still considerable room for improvement to fully earn user trust. This highlights the urgent need for continued research to develop more robust safety mechanisms in mobile environments. We open-source our benchmark at: https://mobilesafetybench.github.io/.
Authors: Dawei Li, Bohan Jiang, Liangjie Huang, Alimohammad Beigi, Chengshuai Zhao, Zhen Tan, Amrita Bhattacharjee, Yuxuan Jiang, Canyu Chen, Tianhao Wu, Kai Shu, Lu Cheng, Huan Liu
Abstract: Assessment and evaluation have long been critical challenges in artificial intelligence (AI) and natural language processing (NLP). However, traditional methods, whether matching-based or embedding-based, often fall short of judging subtle attributes and delivering satisfactory results. Recent advancements in Large Language Models (LLMs) inspire the "LLM-as-a-judge" paradigm, where LLMs are leveraged to perform scoring, ranking, or selection across various tasks and applications. This paper provides a comprehensive survey of LLM-based judgment and assessment, offering an in-depth overview to advance this emerging field. We begin by giving detailed definitions from both input and output perspectives. Then we introduce a comprehensive taxonomy to explore LLM-as-a-judge from three dimensions: what to judge, how to judge and where to judge. Finally, we compile benchmarks for evaluating LLM-as-a-judge and highlight key challenges and promising directions, aiming to provide valuable insights and inspire future research in this promising research area. Paper list and more resources about LLM-as-a-judge can be found at \url{https://github.com/llm-as-a-judge/Awesome-LLM-as-a-judge} and \url{https://llm-as-a-judge.github.io}.
URLs: https://github.com/llm-as-a-judge/Awesome-LLM-as-a-judge, https://llm-as-a-judge.github.io
Authors: Alexander Capstick, Rahul G. Krishnan, Payam Barnaghi
Abstract: Large language models (LLMs), trained on diverse data effectively acquire a breadth of information across various domains. However, their computational complexity, cost, and lack of transparency hinder their direct application for specialised tasks. In fields such as clinical research, acquiring expert annotations or prior knowledge about predictive models is often costly and time-consuming. This study proposes the use of LLMs to elicit expert prior distributions for predictive models. This approach also provides an alternative to in-context learning, where language models are tasked with making predictions directly. In this work, we compare LLM-elicited and uninformative priors, evaluate whether LLMs truthfully generate parameter distributions, and propose a model selection strategy for in-context learning and prior elicitation. Our findings show that LLM-elicited prior parameter distributions significantly reduce predictive error compared to uninformative priors in low-data settings. Applied to clinical problems, this translates to fewer required biological samples, lowering cost and resources. Prior elicitation also consistently outperforms and proves more reliable than in-context learning at a lower cost, making it a preferred alternative in our setting. We demonstrate the utility of this method across various use cases, including clinical applications. For infection prediction, using LLM-elicited priors reduced the number of required labels to achieve the same accuracy as an uninformative prior by 55%, 200 days earlier in the study.
Authors: Alex Havrilla, Andrew Dai, Laura O'Mahony, Koen Oostermeijer, Vera Zisler, Alon Albalak, Fabrizio Milo, Sharath Chandra Raparthy, Kanishk Gandhi, Baber Abbasi, Duy Phung, Maia Iyer, Dakota Mahan, Chase Blagden, Srishti Gureja, Mohammed Hamdy, Wen-Ding Li, Giovanni Paolini, Pawan Sasanka Ammanamanchi, Elliot Meyerson
Abstract: Synthetic data generation with Large Language Models is a promising paradigm for augmenting natural data over a nearly infinite range of tasks. Given this variety, direct comparisons among synthetic data generation algorithms are scarce, making it difficult to understand where improvement comes from and what bottlenecks exist. We propose to evaluate algorithms via the makeup of synthetic data generated by each algorithm in terms of data quality, diversity, and complexity. We choose these three characteristics for their significance in open-ended processes and the impact each has on the capabilities of downstream models. We find quality to be essential for in-distribution model generalization, diversity to be essential for out-of-distribution generalization, and complexity to be beneficial for both. Further, we emphasize the existence of Quality-Diversity trade-offs in training data and the downstream effects on model performance. We then examine the effect of various components in the synthetic data pipeline on each data characteristic. This examination allows us to taxonomize and compare synthetic data generation algorithms through the components they utilize and the resulting effects on data QDC composition. This analysis extends into a discussion on the importance of balancing QDC in synthetic data for efficient reinforcement learning and self-improvement algorithms. Analogous to the QD trade-offs in training data, often there exist trade-offs between model output quality and output diversity which impact the composition of synthetic data. We observe that many models are currently evaluated and optimized only for output quality, thereby limiting output diversity and the potential for self-improvement. We argue that balancing these trade-offs is essential to the development of future self-improvement algorithms and highlight a number of works making progress in this direction.
Authors: Chujie Zheng, Zhenru Zhang, Beichen Zhang, Runji Lin, Keming Lu, Bowen Yu, Dayiheng Liu, Jingren Zhou, Junyang Lin
Abstract: As language models regularly make mistakes when solving math problems, automated identification of errors in the reasoning process becomes increasingly significant for their scalable oversight. In this paper, we introduce ProcessBench for measuring the ability to identify erroneous steps in mathematical reasoning. It consists of 3,400 test cases, primarily focused on competition- and Olympiad-level math problems. Each test case contains a step-by-step solution with error location annotated by human experts. Models are required to identify the earliest step that contains an error, or conclude that all steps are correct. We conduct extensive evaluation on ProcessBench, involving two types of models: process reward models (PRMs) and critic models, where for the latter we prompt general language models to critique each solution step by step. We draw two main observations: (1) Existing PRMs typically fail to generalize to more challenging math problems beyond GSM8K and MATH. They underperform both critic models (i.e., prompted general language models) and our own trained PRM that is straightforwardly fine-tuned on the PRM800K dataset. (2) The best open-source model, QwQ-32B-Preview, has demonstrated the critique capability competitive with the proprietary model GPT-4o, despite that it still lags behind the reasoning-specialized o1-mini. We hope ProcessBench can foster future research in reasoning process assessment, paving the way toward scalable oversight of language models.
Authors: Hongze Mi, Jinyuan Li, Xuying Zhang, Haoran Cheng, Jiahao Wang, Di Sun, Gang Pan
Abstract: Multimodal Entity Linking (MEL) is extensively utilized in the domains of information retrieval. However, existing MEL methods typically utilize mention words as mentions for retrieval. This results in a significant dependence of MEL on mention words, thereby constraining its capacity to effectively leverage information from both images and text. In situations where mention words are absent, MEL methods struggle to leverage image-text pairs for entity linking. To solve these issues, we introduce a Visual Prompts guided Multimodal Entity Linking (VP-MEL) task. VP-MEL directly marks specific regions within the image. These markers are referred to as visual prompts in VP-MEL. Without mention words, VP-MEL aims to utilize marked image-text pairs to align visual prompts with specific entities in the knowledge bases. A new dataset for the VP-MEL task, VPWiki, is proposed in this paper. Moreover, we propose a framework named FBMEL, which enhances the significance of visual prompts and fully leverages the information in image-text pairs. Experimental results on the VPWiki dataset demonstrate that FBMEL outperforms baseline methods across multiple benchmarks for the VP-MEL task.