Authors: Andrea Pinto, Tomer Galanti, Randall Balestriero
Abstract: Large Language Models (LLMs) are widely deployed in real-world applications, yet little is known about their training dynamics at the token level. Evaluation typically relies on aggregated training loss, measured at the batch level, which overlooks subtle per-token biases arising from (i) varying token-level dynamics and (ii) structural biases introduced by hyperparameters. While weight decay is commonly used to stabilize training, we reveal that it silently introduces performance biases detectable only at the token level. In fact, we empirically show across different dataset sizes, model architectures and sizes ranging from 270M to 3B parameters that as weight decay increases, low-frequency tokens are disproportionately depreciated. This is particularly concerning, as these neglected low-frequency tokens represent the vast majority of the token distribution in most languages, calling for novel regularization techniques that ensure fairness across all available tokens.
Authors: Shangqian Gao, Chi-Heng Lin, Ting Hua, Tang Zheng, Yilin Shen, Hongxia Jin, Yen-Chang Hsu
Abstract: Large Language Models (LLMs) have achieved remarkable success in various natural language processing tasks, including language modeling, understanding, and generation. However, the increased memory and computational costs associated with these models pose significant challenges for deployment on resource-limited devices. Structural pruning has emerged as a promising solution to reduce the costs of LLMs without requiring post-processing steps. Prior structural pruning methods either follow the dependence of structures at the cost of limiting flexibility, or introduce non-trivial additional parameters by incorporating different projection matrices. In this work, we propose a novel approach that relaxes the constraint imposed by regular structural pruning methods and eliminates the structural dependence along the embedding dimension. Our dimension-independent structural pruning method offers several benefits. Firstly, our method enables different blocks to utilize different subsets of the feature maps. Secondly, by removing structural dependence, we facilitate each block to possess varying widths along its input and output dimensions, thereby significantly enhancing the flexibility of structural pruning. We evaluate our method on various LLMs, including OPT, LLaMA, LLaMA-2, Phi-1.5, and Phi-2. Experimental results demonstrate that our approach outperforms other state-of-the-art methods, showing for the first time that structural pruning can achieve an accuracy similar to semi-structural pruning.
Authors: Seiji Maekawa, Hayate Iso, Nikita Bhutani
Abstract: The rapid increase in textual information means we need more efficient methods to sift through, organize, and understand it all. While retrieval-augmented generation (RAG) models excel in accessing information from large document collections, they struggle with complex tasks that require aggregation and reasoning over information spanning across multiple documents--what we call holistic reasoning. Long-context language models (LCLMs) have great potential for managing large-scale documents, but their holistic reasoning capabilities remain unclear. In this work, we introduce HoloBench, a novel framework that brings database reasoning operations into text-based contexts, making it easier to systematically evaluate how LCLMs handle holistic reasoning across large documents. Our approach adjusts key factors such as context length, information density, distribution of information, and query complexity to evaluate LCLMs comprehensively. Our experiments show that the amount of information in the context has a bigger influence on LCLM performance than the actual context length. Furthermore, the complexity of queries affects performance more than the amount of information, particularly for different types of queries. Interestingly, queries that involve finding maximum or minimum values are easier for LCLMs and are less affected by context length, even though they pose challenges for RAG systems. However, tasks requiring the aggregation of multiple pieces of information show a noticeable drop in accuracy as context length increases. Additionally, we find that while grouping relevant information generally improves performance, the optimal positioning varies across models. Our findings surface both the advancements and the ongoing challenges in achieving a holistic understanding of long contexts.
Authors: Shaun Ho
Abstract: This paper aims to offer AI & Law researchers and practitioners a more detailed understanding of whether and how continued pre-training and instruction fine-tuning (IFT) of large language models (LLMs) on legal corpora increases their utilization of human-defined legal concepts when developing global contextual representations of input sequences. We compared three models: Mistral 7B, SaulLM-7B-Base (Mistral 7B with continued pre-training on legal corpora), and SaulLM-7B-Instruct (with further IFT). This preliminary assessment examined 7 distinct text sequences from recent AI & Law literature, each containing a human-defined legal concept. We first compared the proportions of total attention the models allocated to subsets of tokens representing the legal concepts. We then visualized patterns of raw attention score alterations, evaluating whether legal training introduced novel attention patterns corresponding to structures of human legal knowledge. This inquiry revealed that (1) the impact of legal training was unevenly distributed across the various human-defined legal concepts, and (2) the contextual representations of legal knowledge learned during legal training did not coincide with structures of human-defined legal concepts. We conclude with suggestions for further investigation into the dynamics of legal LLM training.
Authors: Konstantin Yakovlev, Sergey Nikolenko, Andrey Bout
Abstract: The recently proposed ToolkenGPT tool learning paradigm demonstrates promising performance but suffers from two major issues: first, it cannot benefit from tool documentation, and second, it often makes mistakes in whether to use a tool at all. We introduce Toolken+ that mitigates the first problem by reranking top $k$ tools selected by ToolkenGPT and the second problem with a special "Reject" option such that the model will generate a vocabulary token if "Reject" is ranked first. We demonstrate the effectiveness of Toolken+ on multistep numerical reasoning and tool selection tasks.
Authors: Kushal Tatariya, Vladimir Araujo, Thomas Bauwens, Miryam de Lhoneux
Abstract: Pixel-based language models have emerged as a compelling alternative to subword-based language modelling, particularly because they can represent virtually any script. PIXEL, a canonical example of such a model, is a vision transformer that has been pre-trained on rendered text. While PIXEL has shown promising cross-script transfer abilities and robustness to orthographic perturbations, it falls short of outperforming monolingual subword counterparts like BERT in most other contexts. This discrepancy raises questions about the amount of linguistic knowledge learnt by these models and whether their performance in language tasks stems more from their visual capabilities than their linguistic ones. To explore this, we probe PIXEL using a variety of linguistic and visual tasks to assess its position on the vision-to-language spectrum. Our findings reveal a substantial gap between the model's visual and linguistic understanding. The lower layers of PIXEL predominantly capture superficial visual features, whereas the higher layers gradually learn more syntactic and semantic abstractions. Additionally, we examine variants of PIXEL trained with different text rendering strategies, discovering that introducing certain orthographic constraints at the input level can facilitate earlier learning of surface-level features. With this study, we hope to provide insights that aid the further development of pixel-based language models.
Authors: Yanyue Xie, Zhi Zhang, Ding Zhou, Cong Xie, Ziang Song, Xin Liu, Yanzhi Wang, Xue Lin, An Xu
Abstract: Mixture-of-Experts (MoE) architectures face challenges such as high memory consumption and redundancy in experts. Pruning MoE can reduce network weights while maintaining model performance. Motivated by the recent observation of emergent large magnitude features in Large Language Models (LLM) and MoE routing policy, we propose MoE-Pruner, a method that prunes weights with the smallest magnitudes multiplied by the corresponding input activations and router weights, on each output neuron. Our pruning method is one-shot, requiring no retraining or weight updates. We evaluate our method on Mixtral-8x7B and Mixtral-8x22B across multiple language benchmarks. Experimental results show that our pruning method significantly outperforms state-of-the-art LLM pruning methods. Furthermore, our pruned MoE models can benefit from a pretrained teacher model through expert-wise knowledge distillation, improving performance post-pruning. Experimental results demonstrate that the Mixtral-8x7B model with 50% sparsity maintains 99% of the performance of the original model after the expert-wise knowledge distillation.
Authors: David Bamman, Kent K. Chang, Li Lucy, Naitian Zhou
Abstract: In this work, we survey the way in which classification is used as a sensemaking practice in cultural analytics, and assess where large language models can fit into this landscape. We identify ten tasks supported by publicly available datasets on which we empirically assess the performance of LLMs compared to traditional supervised methods, and explore the ways in which LLMs can be employed for sensemaking goals beyond mere accuracy. We find that prompt-based LLMs are competitive with traditional supervised models for established tasks, but perform less well on de novo tasks. In addition, LLMs can assist sensemaking by acting as an intermediary input to formal theory testing.
Authors: Kaiqiao Han, Tianqing Fang, Zhaowei Wang, Yangqiu Song, Mark Steedman
Abstract: While Large Language Models (LLMs) have showcased remarkable proficiency in reasoning, there is still a concern about hallucinations and unreliable reasoning issues due to semantic associations and superficial logical chains. To evaluate the extent to which LLMs perform robust reasoning instead of relying on superficial logical chains, we propose a new evaluation dataset, the Concept-Reversed Winograd Schema Challenge (CR-WSC), based on the famous Winograd Schema Challenge (WSC) dataset. By simply reversing the concepts to those that are more associated with the wrong answer, we find that the performance of LLMs drops significantly despite the rationale of reasoning remaining the same. Furthermore, we propose Abstraction-of-Thought (AoT), a novel prompt method for recovering adversarial cases to normal cases using conceptual abstraction to improve LLMs' robustness and consistency in reasoning, as demonstrated by experiments on CR-WSC.
Authors: Yuanyuan Lei, Ruihong Huang
Abstract: Logical fallacy uses invalid or faulty reasoning in the construction of a statement. Despite the prevalence and harmfulness of logical fallacies, detecting and classifying logical fallacies still remains a challenging task. We observe that logical fallacies often use connective words to indicate an intended logical relation between two arguments, while the argument semantics does not actually support the logical relation. Inspired by this observation, we propose to build a logical structure tree to explicitly represent and track the hierarchical logic flow among relation connectives and their arguments in a statement. Specifically, this logical structure tree is constructed in an unsupervised manner guided by the constituency tree and a taxonomy of connectives for ten common logical relations, with relation connectives as non-terminal nodes and textual arguments as terminal nodes, and the latter are mostly elementary discourse units. We further develop two strategies to incorporate the logical structure tree into LLMs for fallacy reasoning. Firstly, we transform the tree into natural language descriptions and feed the textualized tree into LLMs as a part of the hard text prompt. Secondly, we derive a relation-aware tree embedding and insert the tree embedding into LLMs as a soft prompt. Experiments on benchmark datasets demonstrate that our approach based on logical structure tree significantly improves precision and recall for both fallacy detection and fallacy classification.
Authors: Hugo Abonizio, Thales Sales Almeida, Thiago Laitz, Roseval Malaquias Junior, Giovana Kerche Bon\'as, Rodrigo Nogueira, Ramon Pires
Abstract: This report presents Sabi\'a-3, our new flagship language model trained on a large brazilian-centric corpus. Evaluations across diverse professional and academic benchmarks show a strong performance on Portuguese and Brazil-related tasks. Sabi\'a-3 shows large improvements in comparison to our previous best of model, Sabi\'a-2 Medium, especially in reasoning-intensive tasks. Notably, Sabi\'a-3's average performance matches frontier LLMs, while it is offered at a three to four times lower cost per token, reinforcing the benefits of domain specialization.
Authors: Amirhossein Herandi, Yitao Li, Zhanlin Liu, Ximin Hu, Xiao Cai
Abstract: Accurate skill extraction from job descriptions is crucial in the hiring process but remains challenging. Named Entity Recognition (NER) is a common approach used to address this issue. With the demonstrated success of large language models (LLMs) in various NLP tasks, including NER, we propose fine-tuning a specialized Skill-LLM and a light weight model to improve the precision and quality of skill extraction. In our study, we evaluated the fine-tuned Skill-LLM and the light weight model using a benchmark dataset and compared its performance against state-of-the-art (SOTA) methods. Our results show that this approach outperforms existing SOTA techniques.
Authors: Giuseppe G. A. Celano
Abstract: This paper presents an experiment consisting in the comparison of six models to identify a state-of-the-art morphosyntactic parser and lemmatizer for Ancient Greek capable of annotating according to the Ancient Greek Dependency Treebank annotation scheme. A normalized version of the major collections of annotated texts was used to (i) train the baseline model Dithrax with randomly initialized character embeddings and (ii) fine-tune Trankit and four recent models pretrained on Ancient Greek texts, i.e., GreBERTa and PhilBERTa for morphosyntactic annotation and GreTA and PhilTa for lemmatization. A Bayesian analysis shows that Dithrax and Trankit annotate morphology practically equivalently, while syntax is best annotated by Trankit and lemmata by GreTa. The results of the experiment suggest that token embeddings are not sufficient to achieve high UAS and LAS scores unless they are coupled with a modeling strategy specifically designed to capture syntactic relationships. The dataset and best-performing models are made available online for reuse.
Authors: Cassandra L. Jacobs, Lo\"ic Grobol, Alvin Tsang
Abstract: In this work we compare the generative behavior at the next token prediction level in several language models by comparing them to human productions in the cloze task. We find that while large models trained for longer are typically better estimators of human productions, but they reliably under-estimate the probabilities of human responses, over-rank rare responses, under-rank top responses, and produce highly distinct semantic spaces. Altogether, this work demonstrates in a tractable, interpretable domain that LM generations can not be used as replacements of or models of the cloze task.
Authors: Ben Hagag, Liav Harpaz, Gil Semo, Dor Bernsohn, Rohit Saha, Pashootan Vaezipoor, Kyryl Truskovskyi, Gerasimos Spanakis
Abstract: This paper presents the results of the LegalLens Shared Task, focusing on detecting legal violations within text in the wild across two sub-tasks: LegalLens-NER for identifying legal violation entities and LegalLens-NLI for associating these violations with relevant legal contexts and affected individuals. Using an enhanced LegalLens dataset covering labor, privacy, and consumer protection domains, 38 teams participated in the task. Our analysis reveals that while a mix of approaches was used, the top-performing teams in both tasks consistently relied on fine-tuning pre-trained language models, outperforming legal-specific models and few-shot methods. The top-performing team achieved a 7.11% improvement in NER over the baseline, while NLI saw a more marginal improvement of 5.7%. Despite these gains, the complexity of legal texts leaves room for further advancements.
Authors: Sachita Nishal, Eric Lee, Nicholas Diakopoulos
Abstract: This study offers an initial evaluation of a human-in-the-loop system leveraging GPT-4 (a large language model or LLM), and Retrieval-Augmented Generation (RAG) to identify and define jargon terms in scientific abstracts, based on readers' self-reported knowledge. The system achieves fairly high recall in identifying jargon and preserves relative differences in readers' jargon identification, suggesting personalization as a feasible use-case for LLMs to support sense-making of complex information. Surprisingly, using only abstracts for context to generate definitions yields slightly more accurate and higher quality definitions than using RAG-based context from the fulltext of an article. The findings highlight the potential of generative AI for assisting science reporters, and can inform future work on developing tools to simplify dense documents.
Authors: Arushi Goel, Karan Sapra, Matthieu Le, Rafael Valle, Andrew Tao, Bryan Catanzaro
Abstract: Large Language Models (LLMs) have made significant strides in text generation and comprehension, with recent advancements extending into multimodal LLMs that integrate visual and audio inputs. However, these models continue to struggle with fine-grained, cross-modal temporal understanding, particularly when correlating events across audio and video streams. We address these challenges with two key contributions: a new dataset and model, called OCTAV and OMCAT respectively. OCTAV (Omni Context and Temporal Audio Video) is a novel dataset designed to capture event transitions across audio and video. Second, OMCAT (Omni Context Aware Transformer) is a powerful model that leverages RoTE (Rotary Time Embeddings), an innovative extension of RoPE, to enhance temporal grounding and computational efficiency in time-anchored tasks. Through a robust three-stage training pipeline-feature alignment, instruction tuning, and OCTAV-specific training-OMCAT excels in cross-modal temporal understanding. Our model demonstrates state-of-the-art performance on Audio-Visual Question Answering (AVQA) tasks and the OCTAV benchmark, showcasing significant gains in temporal reasoning and cross-modal alignment, as validated through comprehensive experiments and ablation studies. Our dataset and code will be made publicly available. The link to our demo page is https://om-cat.github.io.
Authors: Huiwen Wu, Xiaohan Li, Xiaogang Xu, Jiafei Wu, Deyi Zhang, Zhe Liu
Abstract: The development of Large Language Models (LLMs) has significantly advanced various AI applications in commercial and scientific research fields, such as scientific literature summarization, writing assistance, and knowledge graph construction. However, a significant challenge is the high risk of hallucination during LLM inference, which can lead to security concerns like factual inaccuracies, inconsistent information, and fabricated content. To tackle this issue, it is essential to develop effective methods for reducing hallucination while maintaining the original capabilities of the LLM. This paper introduces a novel approach called Iterative Model-level Contrastive Learning (Iter-AHMCL) to address hallucination. This method modifies the representation layers of pre-trained LLMs by using contrastive `positive' and `negative' models, trained on data with and without hallucinations. By leveraging the differences between these two models, we create a more straightforward pathway to eliminate hallucinations, and the iterative nature of contrastive learning further enhances performance. Experimental validation on four pre-trained foundation LLMs (LLaMA2, Alpaca, LLaMA3, and Qwen) finetuning with a specially designed dataset shows that our approach achieves an average improvement of 10.1 points on the TruthfulQA benchmark. Comprehensive experiments demonstrate the effectiveness of Iter-AHMCL in reducing hallucination while maintaining the general capabilities of LLMs.
Authors: Wachara Fungwacharakorn, Nguyen Ha Thanh, May Myo Zin, Ken Satoh
Abstract: This paper presents a novel approach termed Layer-of-Thoughts Prompting (LoT), which utilizes constraint hierarchies to filter and refine candidate responses to a given query. By integrating these constraints, our method enables a structured retrieval process that enhances explainability and automation. Existing methods have explored various prompting techniques but often present overly generalized frameworks without delving into the nuances of prompts in multi-turn interactions. Our work addresses this gap by focusing on the hierarchical relationships among prompts. We demonstrate that the efficacy of thought hierarchy plays a critical role in developing efficient and interpretable retrieval algorithms. Leveraging Large Language Models (LLMs), LoT significantly improves the accuracy and comprehensibility of information retrieval tasks.
Authors: Hai-Long Nguyen, Tan-Minh Nguyen, Duc-Minh Nguyen, Thi-Hai-Yen Vuong, Ha-Thanh Nguyen, Xuan-Hieu Phan
Abstract: Statutory law retrieval is a typical problem in legal language processing, that has various practical applications in law engineering. Modern deep learning-based retrieval methods have achieved significant results for this problem. However, retrieval systems relying on semantic and lexical correlations often exhibit limitations, particularly when handling queries that involve real-life scenarios, or use the vocabulary that is not specific to the legal domain. In this work, we focus on overcoming this weaknesses by utilizing the logical reasoning capabilities of large language models (LLMs) to identify relevant legal terms and facts related to the situation mentioned in the query. The proposed retrieval system integrates additional information from the term--based expansion and query reformulation to improve the retrieval accuracy. The experiments on COLIEE 2022 and COLIEE 2023 datasets show that extra knowledge from LLMs helps to improve the retrieval result of both lexical and semantic ranking models. The final ensemble retrieval system outperformed the highest results among all participating teams in the COLIEE 2022 and 2023 competitions.
Authors: Junjie Xing, Yeye He, Mengyu Zhou, Haoyu Dong, Shi Han, Dongmei Zhang, Surajit Chaudhuri
Abstract: In this work, we propose Table-LLM-Specialist, or Table-Specialist for short, as a new self-trained fine-tuning paradigm specifically designed for table tasks. Our insight is that for each table task, there often exist two dual versions of the same task, one generative and one classification in nature. Leveraging their duality, we propose a Generator-Validator paradigm, to iteratively generate-then-validate training data from language-models, to fine-tune stronger \sys models that can specialize in a given task, without requiring manually-labeled data. Our extensive evaluations suggest that our Table-Specialist has (1) \textit{strong performance} on diverse table tasks over vanilla language-models -- for example, Table-Specialist fine-tuned on GPT-3.5 not only outperforms vanilla GPT-3.5, but can often match or surpass GPT-4 level quality, (2) \textit{lower cost} to deploy, because when Table-Specialist fine-tuned on GPT-3.5 achieve GPT-4 level quality, it becomes possible to deploy smaller models with lower latency and inference cost, with comparable quality, and (3) \textit{better generalizability} when evaluated across multiple benchmarks, since \sys is fine-tuned on a broad range of training data systematically generated from diverse real tables. Our code and data will be available at https://github.com/microsoft/Table-LLM-Specialist.
Authors: Juan Manuel P\'erez, Paula Miguel, Viviana Cotik
Abstract: Hate speech detection deals with many language variants, slang, slurs, expression modalities, and cultural nuances. This outlines the importance of working with specific corpora, when addressing hate speech within the scope of Natural Language Processing, recently revolutionized by the irruption of Large Language Models. This work presents a brief analysis of the performance of large language models in the detection of Hate Speech for Rioplatense Spanish. We performed classification experiments leveraging chain-of-thought reasoning with ChatGPT 3.5, Mixtral, and Aya, comparing their results with those of a state-of-the-art BERT classifier. These experiments outline that, even if large language models show a lower precision compared to the fine-tuned BERT classifier and, in some cases, they find hard-to-get slurs or colloquialisms, they still are sensitive to highly nuanced cases (particularly, homophobic/transphobic hate speech). We make our code and models publicly available for future research.
Authors: Shiqi Qiao, Ning Xv, Biao Liu, Xin Geng
Abstract: Large language models have achieved remarkable capabilities, but aligning their outputs with human values and preferences remains a significant challenge. Existing alignment methods primarily focus on positive examples while overlooking the importance of negative responses in guiding models away from undesirable behaviors. For instance, the widely-used alignment datasets reveals a scarcity of explicit negative examples that contradict human values, hindering its ability to discourage harmful or biased outputs during training. To address this limitation, we propose NEAT, i.e., NEgative-prompt-driven AlignmenT, to introduce negative prompts to generate undesirable responses alongside positive examples during the optimization process. NEAT explicitly penalizes the model for producing harmful outputs, guiding it not only toward desirable behaviors but also steering it away from generating undesirable, biased responses. This dual feedback mechanism enables better alignment with human preferences, crucial in contexts where avoiding harm is paramount. Starting from a pre-trained language model, NEAT performs online alignment by incorporating a ranking loss derived from an expanded preference dataset containing both positive and negative examples. Extensive experiments validate NEAT's effectiveness in significantly enhancing language models' alignment with human values and preferences.
Authors: Harbani Jaggi, Kashyap Murali, Eve Fleisig, Erdem B{\i}y{\i}k
Abstract: When annotators disagree, predicting the labels given by individual annotators can capture nuances overlooked by traditional label aggregation. We introduce three approaches to predicting individual annotator ratings on the toxicity of text by incorporating individual annotator-specific information: a neural collaborative filtering (NCF) approach, an in-context learning (ICL) approach, and an intermediate embedding-based architecture. We also study the utility of demographic information for rating prediction. NCF showed limited utility; however, integrating annotator history, demographics, and survey information permits both the embedding-based architecture and ICL to substantially improve prediction accuracy, with the embedding-based architecture outperforming the other methods. We also find that, if demographics are predicted from survey information, using these imputed demographics as features performs comparably to using true demographic data. This suggests that demographics may not provide substantial information for modeling ratings beyond what is captured in survey responses. Our findings raise considerations about the relative utility of different types of annotator information and provide new approaches for modeling annotators in subjective NLP tasks.
Authors: Xiaonan Jing, Srinivas Billa, Danny Godbout
Abstract: Hallucination has been a popular topic in natural language generation (NLG). In real-world applications, unfaithful content can result in bad data quality or loss of trust from end users. Thus, it is crucial to fact-check before adopting NLG for production usage, which can be expensive if done manually. In this paper, we investigate automated faithfulness evaluation in guided NLG. We developed a rubrics template and use large language models (LLMs) to score the generation into quantifiable scales. We compared popular LLMs as well as the widely adopted natural language inference (NLI) models in scoring quality and sensitivity. In addition, we developed methods to generation synthetic unfaithful data, as well as a heuristics to quantify the percentage of hallucination. Our results on 4 travel-domain industry dataset show that GPT-4 can provide accurate judgement and explanation on whether a source and a generation are factually consistent. Furthermore, we found that tuning NLI models on synthetic data can improve performance. Lastly, we present insights on latency and cost for deploying such system.
Authors: Yulei Qian, Fengcun Li, Xiangyang Ji, Xiaoyu Zhao, Jianchao Tan, Kefeng Zhang, Xunliang Cai
Abstract: Large Language Model (LLM) has revolutionized the field of artificial intelligence, with their capabilities expanding rapidly due to advances in deep learning and increased computational resources. The mixture-of-experts (MoE) model has emerged as a prominent architecture in the field of LLM, better balancing the model performance and computational efficiency. MoE architecture allows for effective scaling and efficient parallel processing, but the GEMM (General Matrix Multiply) of MoE and the large parameters introduce challenges in terms of computation efficiency and communication overhead, which becomes the throughput bottleneck during inference. Applying a single parallelism strategy like EP, DP, PP, etc. to MoE architecture usually achieves sub-optimal inference throughput, the straightforward combinations of existing different parallelisms on MoE can not obtain optimal inference throughput yet. This paper introduces EPS-MoE, a novel expert pipeline scheduler for MoE that goes beyond the existing inference parallelism schemes. Our approach focuses on optimizing the computation of MoE FFN (FeedForward Network) modules by dynamically selecting the best kernel implementation of GroupGemm and DenseGemm for different loads and adaptively overlapping these computations with \textit{all2all} communication, leading to a substantial increase in throughput. Our experimental results demonstrate an average 21% improvement in prefill throughput over existing parallel inference methods. Specifically, we validated our method on DeepSeekV2, a highly optimized model claimed to achieve a prefill throughput of 100K tokens per second. By applying EPS-MoE, we further accelerated it to at least 120K tokens per second.
Authors: Jintao Liu, Ruixue Ding, Linhao Zhang, Pengjun Xie, Fie Huang
Abstract: Retrieval-Augmented Generation (RAG) aims to enhance large language models (LLMs) to generate more accurate and reliable answers with the help of the retrieved context from external knowledge sources, thereby reducing the incidence of hallucinations. Despite the advancements, evaluating these systems remains a crucial research area due to the following issues: (1) Limited data diversity: The insufficient diversity of knowledge sources and query types constrains the applicability of RAG systems; (2) Obscure problems location: Existing evaluation methods have difficulty in locating the stage of the RAG pipeline where problems occur; (3) Unstable retrieval evaluation: These methods often fail to effectively assess retrieval performance, particularly when the chunking strategy changes. To tackle these challenges, we propose a Comprehensive Full-chain Evaluation (CoFE-RAG) framework to facilitate thorough evaluation across the entire RAG pipeline, including chunking, retrieval, reranking, and generation. To effectively evaluate the first three phases, we introduce multi-granularity keywords, including coarse-grained and fine-grained keywords, to assess the retrieved context instead of relying on the annotation of golden chunks. Moreover, we release a holistic benchmark dataset tailored for diverse data scenarios covering a wide range of document formats and query types. We demonstrate the utility of the CoFE-RAG framework by conducting experiments to evaluate each stage of RAG systems. Our evaluation method provides unique insights into the effectiveness of RAG systems in handling diverse data scenarios, offering a more nuanced understanding of their capabilities and limitations.
Authors: Junjie Chen, Weihang Su, Zhumin Chu, Haitao Li, Qinyao Ai, Yiqun Liu, Min Zhang, Shaoping Ma
Abstract: With the rapid development of large language models (LLMs), how to efficiently evaluate them has become an important research question. Existing evaluation methods often suffer from high costs, limited test formats, the need of human references, and systematic evaluation biases. To address these limitations, our study introduces the Auto-PRE, an automatic LLM evaluation framework based on peer review. In contrast to previous studies that rely on human annotations, Auto-PRE selects evaluator LLMs automatically based on their inherent traits including consistency, self-confidence, and pertinence. We conduct extensive experiments on three tasks: summary generation, non-factoid question-answering, and dialogue generation. Experimental results indicate our Auto-PRE achieves state-of-the-art performance at a lower cost. Moreover, our study highlights the impact of prompt strategies and evaluation formats on evaluation performance, offering guidance for method optimization in the future.
Authors: Tim Hunter
Abstract: A central goal of linguistic theory is to find a precise characterization of the notion "possible human language", in the form of a computational device that is capable of describing all and only the languages that can be acquired by a typically developing human child. The success of recent large language models (LLMs) in NLP applications arguably raises the possibility that LLMs might be computational devices that meet this goal. This would only be the case if, in addition to succeeding in learning human languages, LLMs struggle to learn "impossible" human languages. Kallini et al. (2024; "Mission: Impossible Language Models", Proc. ACL) conducted experiments aiming to test this by training GPT-2 on a variety of synthetic languages, and found that it learns some more successfully than others. They present these asymmetries as support for the idea that LLMs' inductive biases align with what is regarded as "possible" for human languages, but the most significant comparison has a confound that makes this conclusion unwarranted. In this paper I explain the confound and suggest some ways forward towards constructing a comparison that appropriately tests the underlying issue.
Authors: Simeng Sun, Cheng-Ping Hsieh
Abstract: We analyze contextual representations in neural autoregressive language models, emphasizing long-range contexts that span several thousand tokens. Our methodology employs a perturbation setup and the metric \emph{Anisotropy-Calibrated Cosine Similarity}, to capture the degree of contextualization of long-range patterns from the perspective of representation geometry. We begin the analysis with a case study on standard decoder-only Transformers, demonstrating that similar perplexity can exhibit markedly different downstream task performance, which can be explained by the difference in contextualization of long-range content. Next, we extend the analysis to other models, covering recent novel architectural designs and various training configurations. The representation-level results illustrate a reduced capacity for high-complexity (i.e., less compressible) sequences across architectures, and that fully recurrent models rely heavily on local context, whereas hybrid models more effectively encode the entire sequence structure. Finally, preliminary analysis of model size and training configurations on the encoding of long-range context suggest potential directions for improving existing language models.
Authors: Lei Sun, Xinchen Wang, Youdi Li
Abstract: Large Language Models (LLMs) possess impressive reasoning abilities but are prone to generating incorrect information, often referred to as hallucinations. While incorporating external Knowledge Graphs (KGs) can partially mitigate this issue, existing methods primarily treat KGs as static knowledge repositories, overlooking the critical disparity between KG and LLM knowledge, and failing to fully exploit the reasoning capabilities inherent in KGs. To address these limitations, we propose Pyramid-Driven Alignment (PDA), a novel framework for seamlessly integrating LLMs with KGs. PDA utilizes Pyramid Principle analysis to construct a hierarchical pyramid structure. This structure is designed to reflect the input question and generate more validated deductive knowledge, thereby enhancing the alignment of LLMs and KGs and ensuring more cohesive integration. Furthermore, PDA employs a recursive mechanism to harness the underlying reasoning abilities of KGs, resulting in more accurate knowledge retrieval for question-answering tasks. Our experimental results reveal a substantial performance advantage of PDA over state-of-the-art baselines, with improvements reaching 26.70% and 26.78%.
Authors: Weixuan Wang, Jingyuan Yang, Wei Peng
Abstract: Large language models (LLMs) have achieved remarkable performance across many tasks, yet aligning them with desired behaviors remains challenging. Activation intervention has emerged as an effective and economical method to modify the behavior of LLMs. Despite considerable interest in this area, current intervention methods exclusively employ a fixed steering vector to modify model activations, lacking adaptability to diverse input semantics. To address this limitation, we propose Semantics-Adaptive Dynamic Intervention (SADI), a novel method that constructs a dynamic steering vector to intervene model activations at inference time. More specifically, SADI utilizes activation differences in contrastive pairs to precisely identify critical elements of an LLM (i.e., attention heads, hidden states, and neurons) for targeted intervention. During inference, SADI dynamically steers model behavior by scaling element-wise activations based on the directions of input semantics. Experimental results show that SADI outperforms established baselines by substantial margins, improving task performance without training. SADI's cost-effectiveness and generalizability across various LLM backbones and tasks highlight its potential as a versatile alignment technique. In addition, we release the code to foster research along this line:https://github.com/weixuan-wang123/SADI.
Authors: Siyi Liu, Qiang Ning, Kishaloy Halder, Wei Xiao, Zheng Qi, Phu Mon Htut, Yi Zhang, Neha Anna John, Bonan Min, Yassine Benajiba, Dan Roth
Abstract: Open domain question answering systems frequently rely on information retrieved from large collections of text (such as the Web) to answer questions. However, such collections of text often contain conflicting information, and indiscriminately depending on this information may result in untruthful and inaccurate answers. To understand the gravity of this problem, we collect a human-annotated dataset, Question Answering with Conflicting Contexts (QACC), and find that as much as 25% of unambiguous, open domain questions can lead to conflicting contexts when retrieved using Google Search. We evaluate and benchmark three powerful Large Language Models (LLMs) with our dataset QACC and demonstrate their limitations in effectively addressing questions with conflicting information. To explore how humans reason through conflicting contexts, we request our annotators to provide explanations for their selections of correct answers. We demonstrate that by finetuning LLMs to explain their answers, we can introduce richer information into their training that guide them through the process of reasoning with conflicting contexts.
Authors: Jiahao Yuan, Dehui Du, Hao Zhang, Zixiang Di, Usman Naseem
Abstract: Large language models (LLMs) have shown remarkable performance in reasoning tasks but face limitations in mathematical and complex logical reasoning. Existing methods to improve LLMs' logical capabilities either involve traceable or verifiable logical sequences that generate more reliable responses by constructing logical structures yet increase computational costs, or introduces rigid logic template rules, reducing flexibility. In this paper, we propose Reversal of Thought (RoT), a novel framework aimed at enhancing the logical reasoning abilities of LLMs. RoT utilizes a Preference-Guided Reverse Reasoning warm-up strategy, which integrates logical symbols for pseudocode planning through meta-cognitive mechanisms and pairwise preference self-evaluation to generate task-specific prompts solely through demonstrations, aligning with LLMs' cognitive preferences shaped by Reinforcement Learning with Human Feedback (RLHF). Through reverse reasoning, we ultilize a Cognitive Preference Manager to assess knowledge boundaries and further expand LLMs' reasoning capabilities by aggregating solution logic for known tasks and stylistic templates for unknown tasks. Experiments across various tasks demonstrate that RoT surpasses existing baselines in both reasoning accuracy and efficiency.
Authors: Kosuke Akimoto, Masafumi Oyamada
Abstract: In this paper, we address the challenge of optimizing training setups for Large Language Models (LLMs) of low-resource language with a limited amount of corpus. Existing works adopt multi-epoch, multi-lingual, and two-stage training to utilize the limited target language corpus efficiently. However, there is still a lack of understanding about the optimal hyperparameter setups for combining these three approaches to train LLMs. We exhaustively explore training setups for low-resource language LLM, combining these three approaches, and found the following insights for efficiently reducing the cost of hyperparameter search: (1) As the amount of target language corpus decreases, the optimal training approach shifts from monolingual single-stage training to multi-lingual two-stage training at a compute budget dependent threshold. (2) The optimal model scale remains stable regardless of the amount of target language corpus, allowing the use of the compute-optimal scale of monolingual training. (3) The optimal number of epochs can be extrapolated from smaller-scale experiments to larger scale using our proposed model. Also, we provide evidence that, in single-stage training, the target language validation loss follows a power law with respect to the target language ratio, with an exponent independent of the amount of data, model scale, and language pair.
Authors: Jia Deng, Tianyi Tang, Yanbin Yin, Wenhao Yang, Wayne Xin Zhao, Ji-Rong Wen
Abstract: Large language models (LLMs) have become increasingly proficient at simulating various personality traits, an important capability for supporting related applications (e.g., role-playing). To further improve this capacity, in this paper, we present a neuron-based approach for personality trait induction in LLMs, with three major technical contributions. First, we construct PersonalityBench, a large-scale dataset for identifying and evaluating personality traits in LLMs. This dataset is grounded in the Big Five personality traits from psychology and is designed to assess the generative capabilities of LLMs towards specific personality traits. Second, by leveraging PersonalityBench, we propose an efficient method for identifying personality-related neurons within LLMs by examining the opposite aspects of a given trait. Third, we develop a simple yet effective induction method that manipulates the values of these identified personality-related neurons. This method enables fine-grained control over the traits exhibited by LLMs without training and modifying model parameters. Extensive experiments validate the efficacy of our neuron identification and trait induction methods. Notably, our approach achieves comparable performance as fine-tuned models, offering a more efficient and flexible solution for personality trait induction in LLMs. We provide access to all the mentioned resources at https://github.com/RUCAIBox/NPTI.
Authors: Botian Jiang, Lei Li, Xiaonan Li, Zhaowei Li, Xiachong Feng, Lingpeng Kong, Qi Liu, Xipeng Qiu
Abstract: The rapid advancement of Multimodal Large Language Models (MLLMs) has been accompanied by the development of various benchmarks to evaluate their capabilities. However, the true nature of these evaluations and the extent to which they assess multimodal reasoning versus merely leveraging the underlying Large Language Model (LLM) backbone remain unclear. This paper presents a comprehensive investigation into the role of LLM backbones in MLLM evaluation, focusing on two critical aspects: the degree to which current benchmarks truly assess multimodal reasoning and the influence of LLM prior knowledge on performance. Specifically, we introduce a modified evaluation protocol to disentangle the contributions of the LLM backbone from multimodal integration, and an automatic knowledge identification technique for diagnosing whether LLMs equip the necessary knowledge for corresponding multimodal questions. Our study encompasses four diverse MLLM benchmarks and eight state-of-the-art MLLMs. Key findings reveal that some benchmarks allow high performance even without visual inputs and up to 50\% of error rates can be attributed to insufficient world knowledge in the LLM backbone, indicating a heavy reliance on language capabilities. To address knowledge deficiencies, we propose a knowledge augmentation pipeline that achieves significant performance gains, with improvements of up to 60\% on certain datasets, resulting in a approximately 4x increase in performance. Our work provides crucial insights into the role of the LLM backbone in MLLMs, and highlights the need for more nuanced benchmarking approaches.
Authors: Daniele Gambetta, Gizem Gezici, Fosca Giannotti, Dino Pedreschi, Alistair Knott, Luca Pappalardo
Abstract: Recent research has focused on the medium and long-term impacts of generative AI, posing scientific and societal challenges mainly due to the detection and reliability of machine-generated information, which is projected to form the major content on the Web soon. Prior studies show that LLMs exhibit a lower performance in generation tasks (model collapse) as they undergo a fine-tuning process across multiple generations on their own generated content (self-consuming loop). In this paper, we present a comprehensive simulation framework built upon the chat version of LLama2, focusing particularly on the linguistic aspects of the generated content, which has not been fully examined in existing studies. Our results show that the model produces less lexical rich content across generations, reducing diversity. The lexical richness has been measured using the linguistic measures of entropy and TTR as well as calculating the POSTags frequency. The generated content has also been examined with an $n$-gram analysis, which takes into account the word order, and semantic networks, which consider the relation between different words. These findings suggest that the model collapse occurs not only by decreasing the content diversity but also by distorting the underlying linguistic patterns of the generated text, which both highlight the critical importance of carefully choosing and curating the initial input text, which can alleviate the model collapse problem. Furthermore, we conduct a qualitative analysis of the fine-tuned models of the pipeline to compare their performances on generic NLP tasks to the original model. We find that autophagy transforms the initial model into a more creative, doubtful and confused one, which might provide inaccurate answers and include conspiracy theories in the model responses, spreading false and biased information on the Web.
Authors: Ali Gebe\c{s}\c{c}e, G\"ozde G\"ul \c{S}ahin
Abstract: Sophisticated grammatical error detection/correction tools are available for a small set of languages such as English and Chinese. However, it is not straightforward -- if not impossible -- to adapt them to morphologically rich languages with complex writing rules like Turkish which has more than 80 million speakers. Even though several tools exist for Turkish, they primarily focus on spelling errors rather than grammatical errors and lack features such as web interfaces, error explanations and feedback mechanisms. To fill this gap, we introduce GECTurk WEB, a light, open-source, and flexible web-based system that can detect and correct the most common forms of Turkish writing errors, such as the misuse of diacritics, compound and foreign words, pronouns, light verbs along with spelling mistakes. Our system provides native speakers and second language learners an easily accessible tool to detect/correct such mistakes and also to learn from their mistakes by showing the explanation for the violated rule(s). The proposed system achieves 88,3 system usability score, and is shown to help learn/remember a grammatical rule (confirmed by 80% of the participants). The GECTurk WEB is available both as an offline tool at https://github.com/GGLAB-KU/gecturkweb or online at www.gecturk.net.
Authors: Yejun Yoon, Jaeyoon Jung, Seunghyun Yoon, Kunwoo Park
Abstract: To tackle the AVeriTeC shared task hosted by the FEVER-24, we introduce a system that only employs publicly available large language models (LLMs) for each step of automated fact-checking, dubbed the Herd of Open LLMs for verifying real-world claims (HerO). HerO employs multiple LLMs for each step of automated fact-checking. For evidence retrieval, a language model is used to enhance a query by generating hypothetical fact-checking documents. We prompt pretrained and fine-tuned LLMs for question generation and veracity prediction by crafting prompts with retrieved in-context samples. HerO achieved 2nd place on the leaderboard with the AVeriTeC score of 0.57, suggesting the potential of open LLMs for verifying real-world claims. For future research, we make our code publicly available at https://github.com/ssu-humane/HerO.
Authors: Amin Abolghasemi, Leif Azzopardi, Seyyed Hadi Hashemi, Maarten de Rijke, Suzan Verberne
Abstract: Attributing answers to source documents is an approach used to enhance the verifiability of a model's output in retrieval augmented generation (RAG). Prior work has mainly focused on improving and evaluating the attribution quality of large language models (LLMs) in RAG, but this may come at the expense of inducing biases in the attribution of answers. We define and examine two aspects in the evaluation of LLMs in RAG pipelines, namely attribution sensitivity and bias with respect to authorship information. We explicitly inform an LLM about the authors of source documents, instruct it to attribute its answers, and analyze (i) how sensitive the LLM's output is to the author of source documents, and (ii) whether the LLM exhibits a bias towards human-written or AI-generated source documents. We design an experimental setup in which we use counterfactual evaluation to study three LLMs in terms of their attribution sensitivity and bias in RAG pipelines. Our results show that adding authorship information to source documents can significantly change the attribution quality of LLMs by 3% to 18%. Moreover, we show that LLMs can have an attribution bias towards explicit human authorship, which can serve as a competing hypothesis for findings of prior work that shows that LLM-generated content may be preferred over human-written contents. Our findings indicate that metadata of source documents can influence LLMs' trust, and how they attribute their answers. Furthermore, our research highlights attribution bias and sensitivity as a novel aspect of brittleness in LLMs.
Authors: Zongqian Li, Yinhong Liu, Yixuan Su, Nigel Collier
Abstract: Leveraging large language models (LLMs) for complex natural language tasks typically requires long-form prompts to convey detailed requirements and information, which results in increased memory usage and inference costs. To mitigate these challenges, multiple efficient methods have been proposed, with prompt compression gaining significant research interest. This survey provides an overview of prompt compression techniques, categorized into hard prompt methods and soft prompt methods. First, the technical approaches of these methods are compared, followed by an exploration of various ways to understand their mechanisms, including the perspectives of attention optimization, Parameter-Efficient Fine-Tuning (PEFT), modality fusion, and new synthetic language. We also examine the downstream adaptations of various prompt compression techniques. Finally, the limitations of current prompt compression methods are analyzed, and several future directions are outlined, such as optimizing the compression encoder, combining hard and soft prompts methods, and leveraging insights from multimodality.
Authors: Niels Horn, Desmond Elliott
Abstract: We study how features emerge, disappear, and persist across models fine-tuned on different domains of text. More specifically, we start from a base one-layer Transformer language model that is trained on a combination of the BabyLM corpus, and a collection of Python code from The Stack. This base model is adapted to two new domains of text: TinyStories, and the Lua programming language, respectively; and then these two models are merged using these two models using spherical linear interpolation. Our exploration aims to provide deeper insights into the stability and transformation of features across typical transfer-learning scenarios using small-scale models and sparse auto-encoders.
Authors: Jingming Zhuo, Songyang Zhang, Xinyu Fang, Haodong Duan, Dahua Lin, Kai Chen
Abstract: Large language models (LLMs) have demonstrated impressive capabilities across various tasks, but their performance is highly sensitive to the prompts utilized. This variability poses challenges for accurate assessment and user satisfaction. Current research frequently overlooks instance-level prompt variations and their implications on subjective evaluations. To address these shortcomings, we introduce ProSA, a framework designed to evaluate and comprehend prompt sensitivity in LLMs. ProSA incorporates a novel sensitivity metric, PromptSensiScore, and leverages decoding confidence to elucidate underlying mechanisms. Our extensive study, spanning multiple tasks, uncovers that prompt sensitivity fluctuates across datasets and models, with larger models exhibiting enhanced robustness. We observe that few-shot examples can alleviate this sensitivity issue, and subjective evaluations are also susceptible to prompt sensitivities, particularly in complex, reasoning-oriented tasks. Furthermore, our findings indicate that higher model confidence correlates with increased prompt robustness. We believe this work will serve as a helpful tool in studying prompt sensitivity of LLMs. The project is released at: https://github.com/open-compass/ProSA .
Authors: Giada Palmieri, Konstantinos Kogkalidis
Abstract: We discuss the open question of the relation between semantics and nominal class assignment in Swahili. We approach the problem from a computational perspective, aiming first to quantify the extent of this relation, and then to explicate its nature, taking extra care to suppress morphosyntactic confounds. Our results are the first of their kind, providing a quantitative evaluation of the semantic cohesion of each nominal class, as well as a nuanced taxonomic description of its semantic content.
Authors: Daichi Hayakawa, Issei Sato
Abstract: In this study, we provide constructive proof that Transformers can recognize and generate hierarchical language efficiently with respect to model size, even without the need for a specific positional encoding. Specifically, we show that causal masking and a starting token enable Transformers to compute positional information and depth within hierarchical structures. We demonstrate that Transformers without positional encoding can generate hierarchical languages. Furthermore, we suggest that explicit positional encoding might have a detrimental effect on generalization with respect to sequence length.
Authors: Xiaochen Zhu, Caiqi Zhang, Tom Stafford, Nigel Collier, Andreas Vlachos
Abstract: The conformity effect describes the tendency of individuals to align their responses with the majority. Studying this bias in large language models (LLMs) is crucial, as LLMs are increasingly used in various information-seeking and decision-making tasks as conversation partners to improve productivity. Thus, conformity to incorrect responses can compromise their effectiveness. In this paper, we adapt psychological experiments to examine the extent of conformity in state-of-the-art LLMs. Our findings reveal that all models tested exhibit varying levels of conformity toward the majority, regardless of their initial choice or correctness, across different knowledge domains. Notably, we are the first to show that LLMs are more likely to conform when they are more uncertain in their own prediction. We further explore factors that influence conformity, such as training paradigms and input characteristics, finding that instruction-tuned models are less susceptible to conformity, while increasing the naturalness of majority tones amplifies conformity. Finally, we propose two interventions--Devil's Advocate and Question Distillation--to mitigate conformity, providing insights into building more robust language models.
Authors: Mengze Hong, Yuanfeng Song, Di Jiang, Lu Wang, Zichang Guo, Chen Jason Zhang
Abstract: Reliable responses of service chatbots are often achieved by employing retrieval-based methods that restrict answers to a knowledge base comprising predefined question-answer pairs (QA pairs). To accommodate potential variations in how a customer's query may be expressed, it emerges as the favored solution to augment these QA pairs with similar questions that are possibly diverse while remaining semantic consistency. This augmentation task is known as Similar Question Generation (SQG). Traditional methods that heavily rely on human efforts or rule-based techniques suffer from limited diversity or significant semantic deviation from the source question, only capable of producing a finite number of useful questions. To address these limitations, we propose an SQG approach based on Large Language Models (LLMs), capable of producing a substantial number of diverse questions while maintaining semantic consistency to the source QA pair. This is achieved by leveraging LLMs' natural language understanding capability through fine-tuning with specially designed prompts. The experiments conducted on a real customer-service dataset demonstrate that our method surpasses baseline methods by a significant margin in terms of semantic diversity. Human evaluation further confirms that integrating the answer that reflects the customer's intention is crucial for increasing the number of generated questions that meet business requirements.
Authors: Hyeonwoo Kim, Dahyun Kim, Jihoo Kim, Sukyung Lee, Yungi Kim, Chanjun Park
Abstract: The Open Ko-LLM Leaderboard has been instrumental in benchmarking Korean Large Language Models (LLMs), yet it has certain limitations. Notably, the disconnect between quantitative improvements on the overly academic leaderboard benchmarks and the qualitative impact of the models should be addressed. Furthermore, the benchmark suite is largely composed of translated versions of their English counterparts, which may not fully capture the intricacies of the Korean language. To address these issues, we propose Open Ko-LLM Leaderboard2, an improved version of the earlier Open Ko-LLM Leaderboard. The original benchmarks are entirely replaced with new tasks that are more closely aligned with real-world capabilities. Additionally, four new native Korean benchmarks are introduced to better reflect the distinct characteristics of the Korean language. Through these refinements, Open Ko-LLM Leaderboard2 seeks to provide a more meaningful evaluation for advancing Korean LLMs.
Authors: Minghao Wu, Thuy-Trang Vu, Lizhen Qu, Gholamreza Haffari
Abstract: The performance of large language models (LLMs) in natural language processing (NLP) tasks is significantly influenced by the quality and diversity of data used for supervised fine-tuning (SFT). Current data selection methods often focus solely on quality or diversity, leading to underperforming models due to suboptimal training data. In this paper, we introduce GraphFilter, a novel method that represents the dataset as a bipartite graph, linking sentences to their constituent n-grams. This representation effectively captures the relationships between sentences and linguistic patterns, facilitating the selection of sentences that enhance n-gram diversity. To balance quality and diversity during selection, we propose a priority function that combines the quality metric with the diversity metric in a multiplicative manner. GraphFilter iteratively selects high-priority sentences, updates the bipartite graph by removing covered n-grams, and re-calculates priorities to reflect the evolving data landscape. We conduct extensive experiments using three model backbones across six widely used benchmarks. The results demonstrate that GraphFilter outperforms all nine baseline approaches, achieving superior model performance and computational efficiency. Our analyses validate the effectiveness of our design choices, examine the subsets selected by GraphFilter and other methods, highlight the importance of instruction diversity, and explore the role of quality and diversity in relation to subset sizes. GraphFilter establishes a new foundation for effective data selection strategies, encouraging further research in data selection for LLMs.
Authors: Weixuan Wang, Minghao Wu, Barry Haddow, Alexandra Birch
Abstract: Large Language Models (LLMs) have shown remarkable capabilities in natural language processing but exhibit significant performance gaps among different languages. Most existing approaches to address these disparities rely on pretraining or fine-tuning, which are resource-intensive. To overcome these limitations without incurring significant costs, we propose Inference-Time Cross-Lingual Intervention (INCLINE), a novel framework that enhances LLM performance on low-performing (source) languages by aligning their internal representations with those of high-performing (target) languages during inference. INCLINE initially learns alignment matrices using parallel sentences from source and target languages through a Least-Squares optimization, and then applies these matrices during inference to transform the low-performing language representations toward the high-performing language space. Extensive experiments on nine benchmarks with five LLMs demonstrate that INCLINE significantly improves performance across diverse tasks and languages, compared to recent strong baselines. Our analysis demonstrates that INCLINE is highly cost-effective and applicable to a wide range of applications. In addition, we release the code to foster research along this line: https://github.com/weixuan-wang123/INCLINE.
Authors: Leo Kohlenberg, Leonard Horns, Frederic Sadrieh, Nils Kiele, Matthis Clausen, Konstantin Ketterer, Avetis Navasardyan, Tamara Czinczoll, Gerard de Melo, Ralf Herbrich
Abstract: Annotating large datasets can be challenging. However, crowd-sourcing is often expensive and can lack quality, especially for non-trivial tasks. We propose a method of using LLMs as few-shot learners for annotating data in a complex natural language task where we learn a standalone model to predict usage options for products from customer reviews. We also propose a new evaluation metric for this scenario, HAMS4, that can be used to compare a set of strings with multiple reference sets. Learning a custom model offers individual control over energy efficiency and privacy measures compared to using the LLM directly for the sequence-to-sequence task. We compare this data annotation approach with other traditional methods and demonstrate how LLMs can enable considerable cost savings. We find that the quality of the resulting data exceeds the level attained by third-party vendor services and that GPT-4-generated labels even reach the level of domain experts. We make the code and generated labels publicly available.
Authors: Zerui Xu, Fang Wu, Tianfan Fu, Yue Zhao
Abstract: Machine learning (ML) exhibits promise in the clinical domain. However, it is constrained by data scarcity and ethical considerations, as the generation of clinical trials presents significant challenges due to stringent privacy regulations, high costs, and the extended duration required for conducting studies with human participants. Despite the advancements of large language models (LLMs) in general generation tasks, their potential in facilitating the generation of synthetic clinical trials is under-explored. To address this gap, we introduce a novel Retrieval-Reasoning few-shot framework that leverages LLMs to generate artificial yet realistic and diverse clinical trials with binary success/failure labels. Experiments conducted on real clinical trials from the \url{ClinicalTrials.gov} database demonstrate that our synthetic data can effectively augment real datasets. Furthermore, by fine-tuning a pre-trained model as a binary classifier on synthetic clinical trial datasets, we demonstrate that this augmentation enhances model training for downstream tasks such as trial outcome prediction. Our findings suggest that LLMs for synthetic clinical trial generation hold promise for accelerating clinical research and upholding ethical standards for patient privacy. The code is publicly available at https://anonymous.4open.science/r/Retrieval_Reasoning_Clinical_Trial_Generation-3EC4.
URLs: https://anonymous.4open.science/r/Retrieval_Reasoning_Clinical_Trial_Generation-3EC4.
Authors: Boyang Xue, Hongru Wang, Rui Wang, Sheng Wang, Zezhong Wang, Yiming Du, Bin Liang, Kam-Fai Wong
Abstract: The tendency of Large Language Models (LLMs) to generate hallucinations raises concerns regarding their reliability. Therefore, confidence estimations indicating the extent of trustworthiness of the generations become essential. However, current LLM confidence estimations in languages other than English remain underexplored. This paper addresses this gap by introducing a comprehensive investigation of Multilingual Confidence estimation (MlingConf) on LLMs, focusing on both language-agnostic (LA) and language-specific (LS) tasks to explore the performance and language dominance effects of multilingual confidence estimations on different tasks. The benchmark comprises four meticulously checked and human-evaluate high-quality multilingual datasets for LA tasks and one for the LS task tailored to specific social, cultural, and geographical contexts of a language. Our experiments reveal that on LA tasks English exhibits notable linguistic dominance in confidence estimations than other languages, while on LS tasks, using question-related language to prompt LLMs demonstrates better linguistic dominance in multilingual confidence estimations. The phenomena inspire a simple yet effective native-tone prompting strategy by employing language-specific prompts for LS tasks, effectively improving LLMs' reliability and accuracy on LS tasks.
Authors: Yongqin Xu, Huan Li, Ke Chen, Lidan Shou
Abstract: Schema and entity matching tasks are crucial for data integration and management. While large language models (LLMs) have shown promising results in these tasks, they suffer from hallucinations and confusion about task instructions. In this paper, we present the Knowledge-Compliant Matching Framework (KcMF), an LLM-based approach that addresses these issues without the need for domain-specific fine-tuning. KcMF employs a pseudo-code-based task decomposition strategy to adopt task-specific natural language statements that guide LLM reasoning and reduce confusion. We also propose two mechanisms, Dataset as Knowledge (DaK) and Example as Knowledge (EaK), to build domain knowledge sets when unstructured domain knowledge is lacking. Additionally, we introduce a result-ensembling strategy to leverage multiple knowledge sources and suppress poorly formatted outputs. Comprehensive evaluations on schema and entity matching tasks demonstrate that KcMF outperforms previous non-LLM state-of-the-art (SOTA) methods by an average F1 score of 22.9% and competes effectively with SOTA fine-tuned LLMs. Moreover, KcMF generalizes well across different LLMs.
Authors: Jared Joselowitz, Arjun Jagota, Satyapriya Krishna, Sonali Parbhoo
Abstract: Large language models (LLMs) trained with Reinforcement Learning from Human Feedback (RLHF) have demonstrated remarkable capabilities, but their underlying reward functions and decision-making processes remain opaque. This paper introduces a novel approach to interpreting LLMs by applying inverse reinforcement learning (IRL) to recover their implicit reward functions. We conduct experiments on toxicity-aligned LLMs of varying sizes, extracting reward models that achieve up to 80.40% accuracy in predicting human preferences. Our analysis reveals key insights into the non-identifiability of reward functions, the relationship between model size and interpretability, and potential pitfalls in the RLHF process. We demonstrate that IRL-derived reward models can be used to fine-tune new LLMs, resulting in comparable or improved performance on toxicity benchmarks. This work provides a new lens for understanding and improving LLM alignment, with implications for the responsible development and deployment of these powerful systems.
Authors: Nathan Cornille, Florian Mai, Jingyuan Sun, Marie-Francine Moens
Abstract: Through end-to-end training to predict the next token, LLMs have become valuable tools for various tasks. Enhancing their core training in language modeling can improve numerous downstream applications. A successful approach to enhance language modeling uses a separate planning module to predict abstract labels of future sentences and conditions the LM on these predictions. However, this method is non-differentiable, preventing joint end-to-end tuning of the planner with the LM. We propose an effective method to improve this approach by enabling joint fine-tuning of the planner and the LM. We show that a naive way of approximating the gradient of selecting a label via the straight-through estimator is not effective. Instead, we propose to use the predicted label probabilities as mixing weights to condition the LM on a weighted average of label embeddings in a differentiable manner. This not only enables joint fine-tuning of the planner and the LM, but also allows the LM to draw on the full label distribution predicted by the planner, retaining more information. Our experimental results show consistent improvements in perplexity.
Authors: Samee Arif, Zohaib Khan, Agha Ali Raza, Awais Athar
Abstract: This paper presents an analysis of biases in open-source Large Language Models (LLMs) across various genders, religions, and races. We introduce a methodology for generating a bias detection dataset using seven bias triggers: General Debate, Positioned Debate, Career Advice, Story Generation, Problem-Solving, Cover-Letter Writing, and CV Generation. We use GPT-4o to generate a diverse set of prompts for each trigger across various genders, religious and racial groups. We evaluate models from Llama and Gemma family on the generated dataset. We anonymise the LLM-generated text associated with each group using GPT-4o-mini and do a pairwise comparison using GPT-4o-as-a-Judge. To quantify bias in the LLM-generated text we use the number of wins and losses in the pairwise comparison. Our analysis spans three languages, English, German, and Arabic to explore how language influences bias manifestation. Our findings reveal that LLMs exhibit strong polarization toward certain groups across each category, with a notable consistency observed across models. However, when switching languages, variations and anomalies emerge, often attributable to cultural cues and contextual differences.
Authors: Fanny Jourdan
Abstract: The burgeoning field of Natural Language Processing (NLP) stands at a critical juncture where the integration of fairness within its frameworks has become an imperative. This PhD thesis addresses the need for equity and transparency in NLP systems, recognizing that fairness in NLP is not merely a technical challenge but a moral and ethical necessity, requiring a rigorous examination of how these technologies interact with and impact diverse human populations. Through this lens, this thesis undertakes a thorough investigation into the development of equitable NLP methodologies and the evaluation of biases that prevail in current systems. First, it introduces an innovative algorithm to mitigate biases in multi-class classifiers, tailored for high-risk NLP applications, surpassing traditional methods in both bias mitigation and prediction accuracy. Then, an analysis of the Bios dataset reveals the impact of dataset size on discriminatory biases and the limitations of standard fairness metrics. This awareness has led to explorations in the field of explainable AI, aiming for a more complete understanding of biases where traditional metrics are limited. Consequently, the thesis presents COCKATIEL, a model-agnostic explainability method that identifies and ranks concepts in Transformer models, outperforming previous approaches in sentiment analysis tasks. Finally, the thesis contributes to bridging the gap between fairness and explainability by introducing TaCo, a novel method to neutralize bias in Transformer model embeddings. In conclusion, this thesis constitutes a significant interdisciplinary endeavor that intertwines explicability and fairness to challenge and reshape current NLP paradigms. The methodologies and critiques presented contribute to the ongoing discourse on fairness in machine learning, offering actionable solutions for more equitable and responsible AI systems.
Authors: Akriti Jain, Saransh Sharma, Koyel Mukherjee, Soumyabrata Pal
Abstract: Auto-regressive Large Language Models (LLMs) demonstrate remarkable performance across domanins such as vision and language processing. However, due to sequential processing through a stack of transformer layers, autoregressive decoding faces significant computation/latency challenges, particularly in resource constrained environments like mobile and edge devices. Existing approaches in literature that aim to improve latency via skipping layers have two distinct flavors - 1) Early exit 2) Input-agnostic heuristics where tokens exit at pre-determined layers irrespective of input sequence. Both the above strategies have limitations - the former cannot be applied to handle KV Caching necessary for speed-ups in modern framework and the latter does not capture the variation in layer importance across tasks or more generally, across input sequences. To address both limitations, we propose FIRST, an algorithm that reduces inference latency by using layer-specific routers to select a subset of transformer layers adaptively for each input sequence - the prompt (during prefill stage) decides which layers will be skipped during decoding. FIRST preserves compatibility with KV caching enabling faster inference while being quality-aware. FIRST is model-agnostic and can be easily enabled on any pre-trained LLM. We further improve performance by incorporating LoRA adapters for fine-tuning on external datasets, enhancing task-specific accuracy while maintaining latency benefits. Our approach reveals that input adaptivity is critical - indeed, different task-specific middle layers play a crucial role in evolving hidden representations depending on task. Extensive experiments show that FIRST significantly reduces latency while retaining competitive performance (as compared to baselines), making our approach an efficient solution for LLM deployment in low-resource environments.
Authors: Jinjie Wei, Dingkang Yang, Yanshu Li, Qingyao Xu, Zhaoyu Chen, Mingcheng Li, Yue Jiang, Xiaolu Hou, Lihua Zhang
Abstract: Large Language Model (LLM)-driven interactive systems currently show potential promise in healthcare domains. Despite their remarkable capabilities, LLMs typically lack personalized recommendations and diagnosis analysis in sophisticated medical applications, causing hallucinations and performance bottlenecks. To address these challenges, this paper proposes MedAide, an LLM-based omni medical multi-agent collaboration framework for specialized healthcare services. Specifically, MedAide first performs query rewriting through retrieval-augmented generation to accomplish accurate medical intent understanding. Immediately, we devise a contextual encoder to obtain intent prototype embeddings, which are used to recognize fine-grained intents by similarity matching. According to the intent relevance, the activated agents collaborate effectively to provide integrated decision analysis. Extensive experiments are conducted on four medical benchmarks with composite intents. Experimental results from automated metrics and expert doctor evaluations show that MedAide outperforms current LLMs and improves their medical proficiency and strategic reasoning.
Authors: Andong Chen, Kehai Chen, Yang Xiang, Xuefeng Bai, Muyun Yang, Tiejun Zhao, Min zhang
Abstract: The remarkable understanding and generation capabilities of large language models (LLMs) have greatly improved translation performance. However, incorrect understanding of the sentence to be translated can degrade translation quality. To address this issue, we proposed a novel Iterative Bilingual Understanding Translation (IBUT) method based on the cross-lingual capabilities of LLMs and the dual characteristics of translation tasks. The cross-lingual capability of LLMs enables the generation of contextual understanding for both the source and target languages separately. Furthermore, the dual characteristics allow IBUT to generate effective cross-lingual feedback, iteratively refining contextual understanding, thereby reducing errors and improving translation performance. Experimental results showed that the proposed IBUT outperforms several strong comparison methods, especially being generalized to multiple domains (e.g., news, commonsense, and cultural translation benchmarks).
Authors: Zhihao Zhang, Yixing Fan, Ruqing Zhang, Jiafeng Guo
Abstract: The advancement of LLMs has significantly boosted the performance of complex long-form question answering tasks. However, one prominent issue of LLMs is the generated "hallucination" responses that are not factual. Consequently, attribution for each claim in responses becomes a common solution to improve the factuality and verifiability. Existing researches mainly focus on how to provide accurate citations for the response, which largely overlook the importance of identifying the claims or statements for each response. To bridge this gap, we introduce a new claim decomposition benchmark, which requires building system that can identify atomic and checkworthy claims for LLM responses. Specifically, we present the Chinese Atomic Claim Decomposition Dataset (CACDD), which builds on the WebCPM dataset with additional expert annotations to ensure high data quality. The CACDD encompasses a collection of 500 human-annotated question-answer pairs, including a total of 4956 atomic claims. We further propose a new pipeline for human annotation and describe the challenges of this task. In addition, we provide experiment results on zero-shot, few-shot and fine-tuned LLMs as baselines. The results show that the claim decomposition is highly challenging and requires further explorations. All code and data are publicly available at \url{https://github.com/FBzzh/CACDD}.
Authors: Yiming Lu, Yebowen Hu, Hassan Foroosh, Wei Jin, Fei Liu
Abstract: Countless decisions shape our daily lives, and it is paramount to understand the how and why behind these choices. In this paper, we introduce a new LLM decision-making framework called STRUX, which enhances LLM decision-making by providing structured explanations. These include favorable and adverse facts related to the decision, along with their respective strengths. STRUX begins by distilling lengthy information into a concise table of key facts. It then employs a series of self-reflection steps to determine which of these facts are pivotal, categorizing them as either favorable or adverse in relation to a specific decision. Lastly, we fine-tune an LLM to identify and prioritize these key facts to optimize decision-making. STRUX has been evaluated on the challenging task of forecasting stock investment decisions based on earnings call transcripts and demonstrated superior performance against strong baselines. It enhances decision transparency by allowing users to understand the impact of different factors, representing a meaningful step towards practical decision-making with LLMs.
Authors: Paul Youssef, Zhixue Zhao, J\"org Schl\"otterer, Christin Seifert
Abstract: In-context knowledge editing (IKE) enables efficient modification of large language model (LLM) outputs without parameter changes and at zero-cost. However, it can be misused to manipulate responses opaquely, e.g., insert misinformation or offensive content. Such malicious interventions could be incorporated into high-level wrapped APIs where the final input prompt is not shown to end-users. To address this issue, we investigate the detection and reversal of IKE-edits. First, we demonstrate that IKE-edits can be detected with high accuracy (F1 > 80\%) using only the top-10 output probabilities of the next token, even in a black-box setting, e.g. proprietary LLMs with limited output information. Further, we introduce the novel task of reversing IKE-edits using specially tuned reversal tokens. We explore using both continuous and discrete reversal tokens, achieving over 80\% accuracy in recovering original, unedited outputs across multiple LLMs. Our continuous reversal tokens prove particularly effective, with minimal impact on unedited prompts. Through analysis of output distributions, attention patterns, and token rankings, we provide insights into IKE's effects on LLMs and how reversal tokens mitigate them. This work represents a significant step towards enhancing LLM resilience against potential misuse of in-context editing, improving their transparency and trustworthiness.
Authors: Herun Wan, Minnan Luo, Zhixiong Su, Guang Dai, Xiang Zhao
Abstract: Evidence-enhanced detectors present remarkable abilities in identifying malicious social text with related evidence. However, the rise of large language models (LLMs) brings potential risks of evidence pollution to confuse detectors. This paper explores how to manipulate evidence, simulating potential misuse scenarios including basic pollution, and rephrasing or generating evidence by LLMs. To mitigate its negative impact, we propose three defense strategies from both the data and model sides, including machine-generated text detection, a mixture of experts, and parameter updating. Extensive experiments on four malicious social text detection tasks with ten datasets present that evidence pollution, especially the generate strategy, significantly compromises existing detectors. On the other hand, the defense strategies could mitigate evidence pollution, but they faced limitations for practical employment, such as the need for annotated data and huge inference costs. Further analysis illustrates that polluted evidence is of high quality, would compromise the model calibration, and could ensemble to amplify the negative impact.
Authors: Yixi Ding, Jiaying Wu, Tongyao Zhu, Yanxia Qin, Qian Liu, Min-Yen Kan
Abstract: To broaden the dissemination of scientific knowledge to diverse audiences, scientific document summarization must simultaneously control multiple attributes such as length and empirical focus. However, existing research typically focuses on controlling single attributes, leaving the compositional control of multiple attributes underexplored. To address this gap, we introduce CCSBench, a benchmark for compositional controllable summarization in the scientific domain. Our benchmark enables fine-grained control over both explicit attributes (e.g., length), which are objective and straightforward, and implicit attributes (e.g., empirical focus), which are more subjective and conceptual. We conduct extensive experiments on GPT-4, LLaMA2, and other popular LLMs under various settings. Our findings reveal significant limitations in large language models' ability to balance trade-offs between control attributes, especially implicit ones that require deeper understanding and abstract reasoning.
Authors: Vernon Y. H. Toh, Deepanway Ghosal, Soujanya Poria
Abstract: Large language models (LLMs) have shown increasing proficiency in solving mathematical reasoning problems. However, many current open-source LLMs often still make calculation and semantic understanding errors in their intermediate reasoning steps. In this work, we propose PROVE, a simple yet effective framework that uses program-based verification as a heuristic to filter out potentially incorrect reasoning paths before aggregating the final answers. Instead of relying on vanilla majority voting, our approach rejects solutions whose corresponding program outputs are inconsistent with the generated solution, aggregating only those validated by Python programs. We conducted extensive experiments on 13 open-source LLMs from various model families and sizes, ranging from 0.5B to 13B parameters, across seven math benchmarks. We demonstrate that PROVE consistently outperforms vanilla majority voting as a heuristic for solving mathematical reasoning tasks across all datasets and model sizes. Notably, PROVE increases accuracy on the GSM8K benchmark from 48.85% to 53.83% for Qwen2-0.5B-Instruct, from 65.66% to 73.01% for Llama-3.2-1B-Instruct, from 73.39% to 79.61% for Gemma-2-2b-it, and from 41.32% to 59.51% for Llama-2-7B-chat. Our codes are available at https://github.com/declare-lab/prove.
Authors: Yedi Hu, Yunzhi Yao, Ningyu Zhang, Shumin Deng, Huajun Chen
Abstract: Model merging has become one of the key technologies for enhancing the capabilities and efficiency of Large Language Models (LLMs). However, our understanding of the expected performance gains and principles when merging any two models remains limited. In this work, we introduce model kinship, the degree of similarity or relatedness between LLMs, analogous to biological evolution. With comprehensive empirical analysis, we find that there is a certain relationship between model kinship and the performance gains after model merging, which can help guide our selection of candidate models. Inspired by this, we propose a new model merging strategy: Top-k Greedy Merging with Model Kinship, which can yield better performance on benchmark datasets. Specifically, we discover that using model kinship as a criterion can assist us in continuously performing model merging, alleviating the degradation (local optima) in model evolution, whereas model kinship can serve as a guide to escape these traps. Code is available at https://github.com/zjunlp/ModelKinship.
Authors: Aaron Macks
Abstract: This paper describes a parsing/generation system for finite verbal forms in Akkadian, with the possible addition of suffixes, implemented in Prolog. The work described provides the framework and engine to interpret the D, N, and G stems along with accusative, dative and ventive endings.
Authors: Ruimeng Ye, Yang Xiao, Bo Hui
Abstract: As large language models (LLMs) continue to advance, ensuring their alignment with human values becomes increasingly critical. Traditional alignment methods heavily rely on human feedback to fine-tune models. With the emergence of superhuman models whose outputs may surpass human understanding, evaluating and aligning these models using human judgments poses significant challenges. To address the challenges, recent works use weak supervisors to elicit knowledge from much stronger models. However, there are important disanalogies between the empirical setup in the existing works and the genuine goal of alignment. We remark that existing works investigate the phenomenon of weak-to-strong generation in analogous setup (i.e., binary classification), rather than practical alignment-relevant tasks (e.g., safety). In this paper, we bridge this gap by extending weak-to-strong generation to the context of practical alignment. We empirically demonstrate the widespread phenomenon of weak-to-strong generation in three complicated alignment tasks: safety, toxicity, and legal reasoning}. Furthermore, we explore efficient strategies for improving alignment performance to enhance the quality of model outcomes. Lastly, we summarize and analyze the challenges and potential solutions in regard to specific alignment tasks, which we hope to catalyze the research progress on the topic of weak-to-strong generalization. Our code is released at https://github.com/yeruimeng/WTS.git.
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: Mete Ismayilzada, Defne Circi, Jonne S\"alev\"a, Hale Sirin, Abdullatif K\"oksal, Bhuwan Dhingra, Antoine Bosselut, Lonneke van der Plas, Duygu Ataman
Abstract: Large language models (LLMs) have demonstrated significant progress in various natural language generation and understanding tasks. However, their linguistic generalization capabilities remain questionable, raising doubts about whether these models learn language similarly to humans. While humans exhibit compositional generalization and linguistic creativity in language use, the extent to which LLMs replicate these abilities, particularly in morphology, is under-explored. In this work, we systematically investigate the morphological generalization abilities of LLMs through the lens of compositionality. We define morphemes as compositional primitives and design a novel suite of generative and discriminative tasks to assess morphological productivity and systematicity. Focusing on agglutinative languages such as Turkish and Finnish, we evaluate several state-of-the-art instruction-finetuned multilingual models, including GPT-4 and Gemini. Our analysis shows that LLMs struggle with morphological compositional generalization particularly when applied to novel word roots, with performance declining sharply as morphological complexity increases. While models can identify individual morphological combinations better than chance, their performance lacks systematicity, leading to significant accuracy gaps compared to humans.
Authors: Nedjma Ousidhoum, Meriem Beloucif, Saif M. Mohammad
Abstract: Language is a symbolic capital that affects people's lives in many ways (Bourdieu, 1977, 1991). It is a powerful tool that accounts for identities, cultures, traditions, and societies in general. Hence, data in a given language should be viewed as more than a collection of tokens. Good data collection and labeling practices are key to building more human-centered and socially aware technologies. While there has been a rising interest in mid- to low-resource languages within the NLP community, work in this space has to overcome unique challenges such as data scarcity and access to suitable annotators. In this paper, we collect feedback from those directly involved in and impacted by NLP artefacts for mid- to low-resource languages. We conduct a quantitative and qualitative analysis of the responses and highlight the main issues related to (1) data quality such as linguistic and cultural data suitability; and (2) the ethics of common annotation practices such as the misuse of online community services. Based on these findings, we make several recommendations for the creation of high-quality language artefacts that reflect the cultural milieu of its speakers, while simultaneously respecting the dignity and labor of data workers.
Authors: Genta Indra Winata, Frederikus Hudi, Patrick Amadeus Irawan, David Anugraha, Rifki Afina Putri, Yutong Wang, Adam Nohejl, Ubaidillah Ariq Prathama, Nedjma Ousidhoum, Afifa Amriani, Anar Rzayev, Anirban Das, Ashmari Pramodya, Aulia Adila, Bryan Wilie, Candy Olivia Mawalim, Ching Lam Cheng, Daud Abolade, Emmanuele Chersoni, Enrico Santus, Fariz Ikhwantri, Garry Kuwanto, Hanyang Zhao, Haryo Akbarianto Wibowo, Holy Lovenia, Jan Christian Blaise Cruz, Jan Wira Gotama Putra, Junho Myung, Lucky Susanto, Maria Angelica Riera Machin, Marina Zhukova, Michael Anugraha, Muhammad Farid Adilazuarda, Natasha Santosa, Peerat Limkonchotiwat, Raj Dabre, Rio Alexander Audino, Samuel Cahyawijaya, Shi-Xiong Zhang, Stephanie Yulia Salim, Yi Zhou, Yinxuan Gui, David Ifeoluwa Adelani, En-Shiun Annie Lee, Shogo Okada, Ayu Purwarianti, Alham Fikri Aji, Taro Watanabe, Derry Tanti Wijaya, Alice Oh, Chong-Wah Ngo
Abstract: Vision Language Models (VLMs) often struggle with culture-specific knowledge, particularly in languages other than English and in underrepresented cultural contexts. To evaluate their understanding of such knowledge, we introduce WorldCuisines, a massive-scale benchmark for multilingual and multicultural, visually grounded language understanding. This benchmark includes a visual question answering (VQA) dataset with text-image pairs across 30 languages and dialects, spanning 9 language families and featuring over 1 million data points, making it the largest multicultural VQA benchmark to date. It includes tasks for identifying dish names and their origins. We provide evaluation datasets in two sizes (12k and 60k instances) alongside a training dataset (1 million instances). Our findings show that while VLMs perform better with correct location context, they struggle with adversarial contexts and predicting specific regional cuisines and languages. To support future research, we release a knowledge base with annotated food entries and images along with the VQA data.
Authors: Jo\~ao Matos, Shan Chen, Siena Placino, Yingya Li, Juan Carlos Climent Pardo, Daphna Idan, Takeshi Tohyama, David Restrepo, Luis F. Nakayama, Jose M. M. Pascual-Leone, Guergana Savova, Hugo Aerts, Leo A. Celi, A. Ian Wong, Danielle S. Bitterman, Jack Gallifant
Abstract: Multimodal/vision language models (VLMs) are increasingly being deployed in healthcare settings worldwide, necessitating robust benchmarks to ensure their safety, efficacy, and fairness. Multiple-choice question and answer (QA) datasets derived from national medical examinations have long served as valuable evaluation tools, but existing datasets are largely text-only and available in a limited subset of languages and countries. To address these challenges, we present WorldMedQA-V, an updated multilingual, multimodal benchmarking dataset designed to evaluate VLMs in healthcare. WorldMedQA-V includes 568 labeled multiple-choice QAs paired with 568 medical images from four countries (Brazil, Israel, Japan, and Spain), covering original languages and validated English translations by native clinicians, respectively. Baseline performance for common open- and closed-source models are provided in the local language and English translations, and with and without images provided to the model. The WorldMedQA-V benchmark aims to better match AI systems to the diverse healthcare environments in which they are deployed, fostering more equitable, effective, and representative applications.
Authors: Grace Yang, Zhiyi Li, Yandong Liu, Jungyeul Park
Abstract: Named entity recognition (NER) is a crucial task that aims to identify structured information, which is often replete with complex, technical terms and a high degree of variability. Accurate and reliable NER can facilitate the extraction and analysis of important information. However, NER for other than English is challenging due to limited data availability, as the high expertise, time, and expenses are required to annotate its data. In this paper, by using the limited data, we explore various factors including model structure, corpus annotation scheme and data augmentation techniques to improve the performance of a NER model for French. Our experiments demonstrate that these approaches can significantly improve the model's F1 score from original CRF score of 62.41 to 79.39. Our findings suggest that considering different extrinsic factors and combining these techniques is a promising approach for improving NER performance where the size of data is limited.
Authors: Ajay Patel, Jiacheng Zhu, Justin Qiu, Zachary Horvitz, Marianna Apidianaki, Kathleen McKeown, Chris Callison-Burch
Abstract: Style representations aim to embed texts with similar writing styles closely and texts with different styles far apart, regardless of content. However, the contrastive triplets often used for training these representations may vary in both style and content, leading to potential content leakage in the representations. We introduce StyleDistance, a novel approach to training stronger content-independent style embeddings. We use a large language model to create a synthetic dataset of near-exact paraphrases with controlled style variations, and produce positive and negative examples across 40 distinct style features for precise contrastive learning. We assess the quality of our synthetic data and embeddings through human and automatic evaluations. StyleDistance enhances the content-independence of style embeddings, which generalize to real-world benchmarks and outperform leading style representations in downstream applications. Our model can be found at https://huggingface.co/StyleDistance/styledistance .
Authors: Hao-Yuan Chang, Kang L. Wang
Abstract: Recent developments in adversarial attacks on deep learning leave many mission-critical natural language processing (NLP) systems at risk of exploitation. To address the lack of computationally efficient adversarial defense methods, this paper reports a novel, universal technique that drastically improves the robustness of Bidirectional Encoder Representations from Transformers (BERT) by combining the unitary weights with the multi-margin loss. We discover that the marriage of these two simple ideas amplifies the protection against malicious interference. Our model, the unitary multi-margin BERT (UniBERT), boosts post-attack classification accuracies significantly by 5.3% to 73.8% while maintaining competitive pre-attack accuracies. Furthermore, the pre-attack and post-attack accuracy tradeoff can be adjusted via a single scalar parameter to best fit the design requirements for the target applications.
Authors: Yingya Li, Timothy Miller, Steven Bethard, Guergana Savova
Abstract: The success of multi-task learning can depend heavily on which tasks are grouped together. Naively grouping all tasks or a random set of tasks can result in negative transfer, with the multi-task models performing worse than single-task models. Though many efforts have been made to identify task groupings and to measure the relatedness among different tasks, it remains a challenging research topic to define a metric to identify the best task grouping out of a pool of many potential task combinations. We propose a metric of task relatedness based on task difficulty measured by pointwise V-usable information (PVI). PVI is a recently proposed metric to estimate how much usable information a dataset contains given a model. We hypothesize that tasks with not statistically different PVI estimates are similar enough to benefit from the joint learning process. We conduct comprehensive experiments to evaluate the feasibility of this metric for task grouping on 15 NLP datasets in the general, biomedical, and clinical domains. We compare the results of the joint learners against single learners, existing baseline methods, and recent large language models, including Llama 2 and GPT-4. The results show that by grouping tasks with similar PVI estimates, the joint learners yielded competitive results with fewer total parameters, with consistent performance across domains.
Authors: Jihao Zhao, Zhiyuan Ji, Pengnian Qi, Simin Niu, Bo Tang, Feiyu Xiong, Zhiyu Li
Abstract: Retrieval-Augmented Generation (RAG), while serving as a viable complement to large language models (LLMs), often overlooks the crucial aspect of text chunking within its pipeline, which impacts the quality of knowledge-intensive tasks. This paper introduces the concept of Meta-Chunking, which refers to a granularity between sentences and paragraphs, consisting of a collection of sentences within a paragraph that have deep linguistic logical connections. To implement Meta-Chunking, we designed two strategies based on LLMs: Margin Sampling Chunking and Perplexity Chunking. The former employs LLMs to perform binary classification on whether consecutive sentences need to be segmented, making decisions based on the probability difference obtained from margin sampling. The latter precisely identifies text chunk boundaries by analyzing the characteristics of perplexity distribution. Additionally, considering the inherent complexity of different texts, we propose a strategy that combines Meta-Chunking with dynamic merging to achieve a balance between fine-grained and coarse-grained text chunking. Experiments conducted on eleven datasets demonstrate that Meta-Chunking can more efficiently improve the performance of single-hop and multi-hop question answering based on RAG. For instance, on the 2WikiMultihopQA dataset, it outperforms similarity chunking by 1.32 while only consuming 45.8% of the time. Our code is available at https://github.com/IAAR-Shanghai/Meta-Chunking.
Authors: Ross Deans Kristensen-McLachlan, Rebecca M. M. Hicke, M\'arton Kardos, Mette Thun{\o}
Abstract: Does the People's Republic of China (PRC) interfere with European elections through ethnic Chinese diaspora media? This question forms the basis of an ongoing research project exploring how PRC narratives about European elections are represented in Chinese diaspora media, and thus the objectives of PRC news media manipulation. In order to study diaspora media efficiently and at scale, it is necessary to use techniques derived from quantitative text analysis, such as topic modelling. In this paper, we present a pipeline for studying information dynamics in Chinese media. Firstly, we present KeyNMF, a new approach to static and dynamic topic modelling using transformer-based contextual embedding models. We provide benchmark evaluations to demonstrate that our approach is competitive on a number of Chinese datasets and metrics. Secondly, we integrate KeyNMF with existing methods for describing information dynamics in complex systems. We apply this pipeline to data from five news sites, focusing on the period of time leading up to the 2024 European parliamentary elections. Our methods and results demonstrate the effectiveness of KeyNMF for studying information dynamics in Chinese media and lay groundwork for further work addressing the broader research questions.
Authors: Seongho Kim, Jihyun Moon, Juntaek Oh, Insu Choi, Joon-Sung Yang
Abstract: The advent of the Attention mechanism and Transformer architecture enables contextually natural text generation and compresses the burden of processing entire source information into singular vectors. Based on these two main ideas, model sizes gradually increases to accommodate more precise and comprehensive information, leading to the current state-of-the-art LLMs being very large, with parameters around 70 billion. As the model sizes are growing, the demand for substantial storage and computational capacity increases. This leads to the development of high-bandwidth memory and accelerators, as well as a variety of model architectures designed to meet these requirements. We note that LLM architectures have increasingly converged. This paper analyzes how these converged architectures perform in terms of layer configurations, operational mechanisms, and model sizes, considering various hyperparameter settings. In this paper, we conduct a concise survey of the history of LLMs by tracing the evolution of their operational improvements. Furthermore, we summarize the performance trends of LLMs under various hyperparameter settings using the RTX 6000, which features the state-of-the-art Ada Lovelace architecture. We conclude that even the same model can exhibit different behaviors depending on the hyperparameters or whether it is deployed in server or edge environments.
Authors: Tanwi Mallick, Orcun Yildiz, David Lenz, Tom Peterka
Abstract: We develop an iterative assistant we call ChatVis that can synthetically generate Python scripts for data analysis and visualization using a large language model (LLM). The assistant allows a user to specify the operations in natural language, attempting to generate a Python script for the desired operations, prompting the LLM to revise the script as needed until it executes correctly. The iterations include an error detection and correction mechanism that extracts error messages from the execution of the script and subsequently prompts LLM to correct the error. Our method demonstrates correct execution on five canonical visualization scenarios, comparing results with ground truth. We also compared our results with scripts generated by several other LLMs without any assistance. In every instance, ChatVis successfully generated the correct script, whereas the unassisted LLMs failed to do so. The code is available on GitHub: https://github.com/tanwimallick/ChatVis/.
Authors: Dancheng Liu, Jason Yang, Ishan Albrecht-Buehler, Helen Qin, Sophie Li, Yuting Hu, Amir Nassereldine, Jinjun Xiong
Abstract: Speech is a fundamental aspect of human life, crucial not only for communication but also for cognitive, social, and academic development. Children with speech disorders (SD) face significant challenges that, if unaddressed, can result in lasting negative impacts. Traditionally, speech and language assessments (SLA) have been conducted by skilled speech-language pathologists (SLPs), but there is a growing need for efficient and scalable SLA methods powered by artificial intelligence. This position paper presents a survey of existing techniques suitable for automating SLA pipelines, with an emphasis on adapting automatic speech recognition (ASR) models for children's speech, an overview of current SLAs and their automated counterparts to demonstrate the feasibility of AI-enhanced SLA pipelines, and a discussion of practical considerations, including accessibility and privacy concerns, associated with the deployment of AI-powered SLAs.
Authors: Erik Arakelyan, Pasquale Minervini, Pat Verga, Patrick Lewis, Isabelle Augenstein
Abstract: Modern Question Answering (QA) and Reasoning approaches based on Large Language Models (LLMs) commonly use prompting techniques, such as Chain-of-Thought (CoT), assuming the resulting generation will have a more granular exploration and reasoning over the question space and scope. However, such methods struggle with generating outputs that are faithful to the intermediate chain of reasoning produced by the model. On the other end of the spectrum, neuro-symbolic methods such as Faithful CoT (F-CoT) propose to combine LLMs with external symbolic solvers. While such approaches boast a high degree of faithfulness, they usually require a model trained for code generation and struggle with tasks that are ambiguous or hard to formalise strictly. We introduce \textbf{F}aithful \textbf{L}ogic-\textbf{A}ided \textbf{R}easoning and \textbf{E}xploration (\textbf{\ours}), a novel interpretable approach for traversing the problem space using task decompositions. We use the LLM to plan a solution, soft-formalise the query into facts and predicates using a logic programming code and simulate that code execution using an exhaustive multi-hop search over the defined space. Our method allows us to compute the faithfulness of the reasoning process w.r.t. the generated code and analyse the steps of the multi-hop search without relying on external solvers. Our methods achieve SOTA results on $\mathbf{7}$ out of $\mathbf{9}$ diverse reasoning benchmarks. We also show that model faithfulness positively correlates with overall performance and further demonstrate that {\textbf{\ours}} allows pinpointing the decisive factors sufficient for and leading to the correct answer with optimal reasoning during the multi-hop search.
Authors: Hyejun Jeong, Shiqing Ma, Amir Houmansadr
Abstract: Bias in machine learning models has been a chronic problem, especially as these models influence decision-making in human society. In generative AI, such as Large Language Models, the impact of bias is even more profound compared to the classification models. LLMs produce realistic and human-like content that users may unconsciously trust, which could perpetuate harmful stereotypes to the uncontrolled public. It becomes particularly concerning when utilized in journalism or education. While prior studies have explored and quantified bias in individual AI models, no work has yet compared bias similarity across different LLMs. To fill this gap, we take a comprehensive look at ten open- and closed-source LLMs from four model families, assessing the extent of biases through output distribution. Using two datasets-one containing 4k questions and another with one million questions for each of the four bias dimensions -- we measure functional similarity to understand how biases manifest across models. Our findings reveal that 1) fine-tuning does not significantly alter output distributions, which would limit its ability to mitigate bias, 2) LLMs within the same family tree do not produce similar output distributions, implying that addressing bias in one model could have limited implications for others in the same family, and 3) there is a possible risk of training data information leakage, raising concerns about privacy and data security. Our analysis provides insight into LLM behavior and highlights potential risks in real-world deployment.
Authors: Hazel Doughty, Fida Mohammad Thoker, Cees G. M. Snoek
Abstract: This paper strives for motion-focused video-language representations. Existing methods to learn video-language representations use spatial-focused data, where identifying the objects and scene is often enough to distinguish the relevant caption. We instead propose LocoMotion to learn from motion-focused captions that describe the movement and temporal progression of local object motions. We achieve this by adding synthetic motions to videos and using the parameters of these motions to generate corresponding captions. Furthermore, we propose verb-variation paraphrasing to increase the caption variety and learn the link between primitive motions and high-level verbs. With this, we are able to learn a motion-focused video-language representation. Experiments demonstrate our approach is effective for a variety of downstream tasks, particularly when limited data is available for fine-tuning. Code is available: https://hazeldoughty.github.io/Papers/LocoMotion/
Authors: Yilun Hao, Yang Zhang, Chuchu Fan
Abstract: While large language models (LLMs) have recently demonstrated strong potential in solving planning problems, there is a trade-off between flexibility and complexity. LLMs, as zero-shot planners themselves, are still not capable of directly generating valid plans for complex planning problems such as multi-constraint or long-horizon tasks. On the other hand, many frameworks aiming to solve complex planning problems often rely on task-specific preparatory efforts, such as task-specific in-context examples and pre-defined critics/verifiers, which limits their cross-task generalization capability. In this paper, we tackle these challenges by observing that the core of many planning problems lies in optimization problems: searching for the optimal solution (best plan) with goals subject to constraints (preconditions and effects of decisions). With LLMs' commonsense, reasoning, and programming capabilities, this opens up the possibilities of a universal LLM-based approach to planning problems. Inspired by this observation, we propose LLMFP, a general-purpose framework that leverages LLMs to capture key information from planning problems and formally formulate and solve them as optimization problems from scratch, with no task-specific examples needed. We apply LLMFP to 9 planning problems, ranging from multi-constraint decision making to multi-step planning problems, and demonstrate that LLMFP achieves on average 83.7% and 86.8% optimal rate across 9 tasks for GPT-4o and Claude 3.5 Sonnet, significantly outperforming the best baseline (direct planning with OpenAI o1-preview) with 37.6% and 40.7% improvements. We also validate components of LLMFP with ablation experiments and analyzed the underlying success and failure reasons.
Authors: Zifei Xu, Alexander Lan, Wanzin Yazar, Tristan Webb, Sayeh Sharify, Xin Wang
Abstract: Generalization abilities of well-trained large language models (LLMs) are known to scale predictably as a function of model size. In contrast to the existence of practical scaling laws governing pre-training, the quality of LLMs after post-training compression remains highly unpredictable, often requiring case-by-case validation in practice. In this work, we attempted to close this gap for post-training weight quantization of LLMs by conducting a systematic empirical study on multiple LLM families quantized to numerous low-precision tensor data types using popular weight quantization techniques. We identified key scaling factors pertaining to characteristics of the local loss landscape, based on which the performance of quantized LLMs can be reasonably well predicted by a statistical model.
Authors: Chaoqi Wang, Zhuokai Zhao, Chen Zhu, Karthik Abinav Sankararaman, Michal Valko, Xuefei Cao, Zhaorun Chen, Madian Khabsa, Yuxin Chen, Hao Ma, Sinong Wang
Abstract: Recent advancements in generative models, particularly large language models (LLMs) and diffusion models, have been driven by extensive pretraining on large datasets followed by post-training. However, current post-training methods such as reinforcement learning from human feedback (RLHF) and direct alignment from preference methods (DAP) primarily utilize single-sample comparisons. These approaches often fail to capture critical characteristics such as generative diversity and bias, which are more accurately assessed through multiple samples. To address these limitations, we introduce a novel approach that extends post-training to include multi-sample comparisons. To achieve this, we propose Multi-sample Direct Preference Optimization (mDPO) and Multi-sample Identity Preference Optimization (mIPO). These methods improve traditional DAP methods by focusing on group-wise characteristics. Empirically, we demonstrate that multi-sample comparison is more effective in optimizing collective characteristics~(e.g., diversity and bias) for generative models than single-sample comparison. Additionally, our findings suggest that multi-sample comparisons provide a more robust optimization framework, particularly for dataset with label noise.
Authors: Lichang Chen, Hexiang Hu, Mingda Zhang, Yiwen Chen, Zifeng Wang, Yandong Li, Pranav Shyam, Tianyi Zhou, Heng Huang, Ming-Hsuan Yang, Boqing Gong
Abstract: We introduce OmnixR, an evaluation suite designed to benchmark SoTA Omni-modality Language Models, such as GPT-4o and Gemini. Evaluating OLMs, which integrate multiple modalities such as text, vision, and audio, presents unique challenges. Particularly, the user message might often consist of multiple modalities, such that OLMs have to establish holistic understanding and reasoning across modalities to accomplish the task. Existing benchmarks are limited to single modality or dual-modality tasks, overlooking comprehensive multi-modal assessments of model reasoning. To address this, OmnixR offers two evaluation variants: (1)synthetic subset: a synthetic dataset generated automatically by translating text into multiple modalities--audio, images, video, and hybrids (Omnify). (2)realistic subset: a real-world dataset, manually curated and annotated by experts, for evaluating cross-modal reasoning in natural settings. OmnixR presents a unique evaluation towards assessing OLMs over a diverse mix of modalities, such as a question that involves video, audio, and text, providing a rigorous cross-modal reasoning testbed unlike any existing benchmarks. Our experiments find that all state-of-the-art OLMs struggle with OmnixR questions that require integrating information from multiple modalities to answer. Further analysis highlights differences in reasoning behavior, underscoring the challenges of omni-modal AI alignment.
Authors: Luyi Ma, Xiaohan Li, Zezhong Fan, Jianpeng Xu, Jason Cho, Praveen Kanumala, Kaushiki Nag, Sushant Kumar, Kannan Achan
Abstract: Integrating diverse data modalities is crucial for enhancing the performance of personalized recommendation systems. Traditional models, which often rely on singular data sources, lack the depth needed to accurately capture the multifaceted nature of item features and user behaviors. This paper introduces a novel framework for multi-behavior recommendations, leveraging the fusion of triple-modality, which is visual, textual, and graph data through alignment with large language models (LLMs). By incorporating visual information, we capture contextual and aesthetic item characteristics; textual data provides insights into user interests and item features in detail; and graph data elucidates relationships within the item-behavior heterogeneous graphs. Our proposed model called Triple Modality Fusion (TMF) utilizes the power of LLMs to align and integrate these three modalities, achieving a comprehensive representation of user behaviors. The LLM models the user's interactions including behaviors and item features in natural languages. Initially, the LLM is warmed up using only natural language-based prompts. We then devise the modality fusion module based on cross-attention and self-attention mechanisms to integrate different modalities from other models into the same embedding space and incorporate them into an LLM. Extensive experiments demonstrate the effectiveness of our approach in improving recommendation accuracy. Further ablation studies validate the effectiveness of our model design and benefits of the TMF.
Authors: Yong Xie, Karan Aggarwal, Aitzaz Ahmad, Stephen Lau
Abstract: We present a novel approach to automatically generate non-trivial task-specific synthetic datasets for hallucination detection. Our approach features a two-step generation-selection pipeline, using hallucination pattern guidance and a language style alignment during generation. Hallucination pattern guidance leverages the most important task-specific hallucination patterns while language style alignment aligns the style of the synthetic dataset with benchmark text. To obtain robust supervised detectors from synthetic datasets, we also adopt a data mixture strategy to improve performance robustness and generalization. Our results on three datasets show that our generated hallucination text is more closely aligned with non-hallucinated text versus baselines, to train hallucination detectors with better generalization. Our hallucination detectors trained on synthetic datasets outperform in-context-learning (ICL)-based detectors by a large margin of 32%. Our extensive experiments confirm the benefits of our approach with cross-task and cross-generator generalization. Our data-mixture-based training further improves the generalization and robustness of hallucination detection.
Authors: Christoph Minixhofer, Ondrej Klejch, Peter Bell
Abstract: Synthetically generated speech has rapidly approached human levels of naturalness. However, the paradox remains that ASR systems, when trained on TTS output that is judged as natural by humans, continue to perform badly on real speech. In this work, we explore whether this phenomenon is due to the oversmoothing behaviour of models commonly used in TTS, with a particular focus on the behaviour of TTS-for-ASR as the amount of TTS training data is scaled up. We systematically compare Denoising Diffusion Probabilistic Models (DDPM) to Mean Squared Error (MSE) based models for TTS, when used for ASR model training. We test the scalability of the two approaches, varying both the number hours, and the number of different speakers. We find that for a given model size, DDPM can make better use of more data, and a more diverse set of speakers, than MSE models. We achieve the best reported ratio between real and synthetic speech WER to date (1.46), but also find that a large gap remains.
Authors: Maxime Kayser, Bayar Menzat, Cornelius Emde, Bogdan Bercean, Alex Novak, Abdala Espinosa, Bartlomiej W. Papiez, Susanne Gaube, Thomas Lukasiewicz, Oana-Maria Camburu
Abstract: The growing capabilities of AI models are leading to their wider use, including in safety-critical domains. Explainable AI (XAI) aims to make these models safer to use by making their inference process more transparent. However, current explainability methods are seldom evaluated in the way they are intended to be used: by real-world end users. To address this, we conducted a large-scale user study with 85 healthcare practitioners in the context of human-AI collaborative chest X-ray analysis. We evaluated three types of explanations: visual explanations (saliency maps), natural language explanations, and a combination of both modalities. We specifically examined how different explanation types influence users depending on whether the AI advice and explanations are factually correct. We find that text-based explanations lead to significant over-reliance, which is alleviated by combining them with saliency maps. We also observe that the quality of explanations, that is, how much factually correct information they entail, and how much this aligns with AI correctness, significantly impacts the usefulness of the different explanation types.
Authors: Yuanning Cui, Zequn Sun, Wei Hu
Abstract: Extensive knowledge graphs (KGs) have been constructed to facilitate knowledge-driven tasks across various scenarios. However, existing work usually develops separate reasoning models for different KGs, lacking the ability to generalize and transfer knowledge across diverse KGs and reasoning settings. In this paper, we propose a prompt-based KG foundation model via in-context learning, namely KG-ICL, to achieve a universal reasoning ability. Specifically, we introduce a prompt graph centered with a query-related example fact as context to understand the query relation. To encode prompt graphs with the generalization ability to unseen entities and relations in queries, we first propose a unified tokenizer that maps entities and relations in prompt graphs to predefined tokens. Then, we propose two message passing neural networks to perform prompt encoding and KG reasoning, respectively. We conduct evaluation on 43 different KGs in both transductive and inductive settings. Results indicate that the proposed KG-ICL outperforms baselines on most datasets, showcasing its outstanding generalization and universal reasoning capabilities. The source code is accessible on GitHub: https://github.com/nju-websoft/KG-ICL.
Authors: Shashank Sonkar, Xinghe Chen, Naiming Liu, Richard G. Baraniuk, Mrinmaya Sachan
Abstract: Accurately modeling student cognition is crucial for developing effective AI-driven educational technologies. A key challenge is creating realistic student models that satisfy two essential properties: (1) accurately replicating specific misconceptions, and (2) correctly solving problems where these misconceptions are not applicable. This dual requirement reflects the complex nature of student understanding, where misconceptions coexist with correct knowledge. This paper investigates whether Large Language Models (LLMs) can be instruction-tuned to meet this dual requirement and effectively simulate student thinking in algebra. We introduce MalAlgoPy, a novel Python library that generates datasets reflecting authentic student solution patterns through a graph-based representation of algebraic problem-solving. Utilizing MalAlgoPy, we define and examine Cognitive Student Models (CSMs) - LLMs instruction tuned to faithfully emulate realistic student behavior. Our findings reveal that LLMs trained on misconception examples can efficiently learn to replicate errors. However, the training diminishes the model's ability to solve problems correctly, particularly for problem types where the misconceptions are not applicable, thus failing to satisfy second property of CSMs. We demonstrate that by carefully calibrating the ratio of correct to misconception examples in the training data - sometimes as low as 0.25 - it is possible to develop CSMs that satisfy both properties. Our insights enhance our understanding of AI-based student models and pave the way for effective adaptive learning systems.
Authors: Yaxi Lu, Shenzhi Yang, Cheng Qian, Guirong Chen, Qinyu Luo, Yesai Wu, Huadong Wang, Xin Cong, Zhong Zhang, Yankai Lin, Weiwen Liu, Yasheng Wang, Zhiyuan Liu, Fangming Liu, Maosong Sun
Abstract: Agents powered by large language models have shown remarkable abilities in solving complex tasks. However, most agent systems remain reactive, limiting their effectiveness in scenarios requiring foresight and autonomous decision-making. In this paper, we tackle the challenge of developing proactive agents capable of anticipating and initiating tasks without explicit human instructions. We propose a novel data-driven approach for this problem. Firstly, we collect real-world human activities to generate proactive task predictions. These predictions are then labeled by human annotators as either accepted or rejected. The labeled data is used to train a reward model that simulates human judgment and serves as an automatic evaluator of the proactiveness of LLM agents. Building on this, we develop a comprehensive data generation pipeline to create a diverse dataset, ProactiveBench, containing 6,790 events. Finally, we demonstrate that fine-tuning models with the proposed ProactiveBench can significantly elicit the proactiveness of LLM agents. Experimental results show that our fine-tuned model achieves an F1-Score of 66.47% in proactively offering assistance, outperforming all open-source and close-source models. These results highlight the potential of our method in creating more proactive and effective agent systems, paving the way for future advancements in human-agent collaboration.
Authors: Markus J. Buehler
Abstract: PRefLexOR (Preference-based Recursive Language Modeling for Exploratory Optimization of Reasoning) combines preference optimization with concepts from Reinforcement Learning to enable models to self-teach through iterative reasoning improvements. We propose a recursive learning approach that engages the model in multi-step reasoning, revisiting, and refining intermediate steps before producing a final output in training and inference phases. Through multiple training stages, the model first learns to align its reasoning with accurate decision paths by optimizing the log odds between preferred and non-preferred responses. During this process, PRefLexOR builds a dynamic knowledge graph by generating questions from random text chunks and retrieval-augmentation to contextualize relevant details from the entire training corpus. In the second stage, preference optimization enhances model performance by using rejection sampling to fine-tune reasoning quality by continually producing in-situ training data while masking the reasoning steps. Recursive optimization within a thinking token framework introduces iterative feedback loops, where the model refines reasoning, achieving deeper coherence, consistency, and adaptability. Implemented in small language models with only 3 billion parameters, we should that even tiny models can iteratively teach themselves to reason with greater depth and reflectivity. Our implementation is straightforward and can be incorporated into any existing pretrained LLM. We focus our examples on applications in biological materials science and demonstrate the method in a variety of case studies that range from in-domain to cross-domain applications. Using reasoning strategies that include thinking and reflection modalities we build a multi-agent recursive self-improving inference approach to successively improve responses via repeated sampling in inference time.
Authors: Aozhu Chen, Hazel Doughty, Xirong Li, Cees G. M. Snoek
Abstract: Video-text retrieval has seen significant advancements, yet the ability of models to discern subtle differences in captions still requires verification. In this paper, we introduce a new approach for fine-grained evaluation. Our approach can be applied to existing datasets by automatically generating hard negative test captions with subtle single-word variations across nouns, verbs, adjectives, adverbs, and prepositions. We perform comprehensive experiments using four state-of-the-art models across two standard benchmarks (MSR-VTT and VATEX) and two specially curated datasets enriched with detailed descriptions (VLN-UVO and VLN-OOPS), resulting in a number of novel insights: 1) our analyses show that the current evaluation benchmarks fall short in detecting a model's ability to perceive subtle single-word differences, 2) our fine-grained evaluation highlights the difficulty models face in distinguishing such subtle variations. To enhance fine-grained understanding, we propose a new baseline that can be easily combined with current methods. Experiments on our fine-grained evaluations demonstrate that this approach enhances a model's ability to understand fine-grained differences.
Authors: Jian Xie, Kexun Zhang, Jiangjie Chen, Siyu Yuan, Kai Zhang, Yikai Zhang, Lei Li, Yanghua Xiao
Abstract: Autonomous planning has been an ongoing pursuit since the inception of artificial intelligence. Based on curated problem solvers, early planning agents could deliver precise solutions for specific tasks but lacked generalization. The emergence of large language models (LLMs) and their powerful reasoning capabilities has reignited interest in autonomous planning by automatically generating reasonable solutions for given tasks. However, prior research and our experiments show that current language agents still lack human-level planning abilities. Even the state-of-the-art reasoning model, OpenAI o1, achieves only 15.6% on one of the complex real-world planning benchmarks. This highlights a critical question: What hinders language agents from achieving human-level planning? Although existing studies have highlighted weak performance in agent planning, the deeper underlying issues and the mechanisms and limitations of the strategies proposed to address them remain insufficiently understood. In this work, we apply the feature attribution study and identify two key factors that hinder agent planning: the limited role of constraints and the diminishing influence of questions. We also find that although current strategies help mitigate these challenges, they do not fully resolve them, indicating that agents still have a long way to go before reaching human-level intelligence.
Authors: Shicheng Xu, Liang Pang, Yunchang Zhu, Huawei Shen, Xueqi Cheng
Abstract: Vision-language alignment in Large Vision-Language Models (LVLMs) successfully enables LLMs to understand visual input. However, we find that existing vision-language alignment methods fail to transfer the existing safety mechanism for text in LLMs to vision, which leads to vulnerabilities in toxic image. To explore the cause of this problem, we give the insightful explanation of where and how the safety mechanism of LVLMs operates and conduct comparative analysis between text and vision. We find that the hidden states at the specific transformer layers play a crucial role in the successful activation of safety mechanism, while the vision-language alignment at hidden states level in current methods is insufficient. This results in a semantic shift for input images compared to text in hidden states, therefore misleads the safety mechanism. To address this, we propose a novel Text-Guided vision-language Alignment method (TGA) for LVLMs. TGA retrieves the texts related to input vision and uses them to guide the projection of vision into the hidden states space in LLMs. Experiments show that TGA not only successfully transfers the safety mechanism for text in basic LLMs to vision in vision-language alignment for LVLMs without any safety fine-tuning on the visual modality but also maintains the general performance on various vision tasks (Safe and Good).
Authors: Lingxiao Luo, Bingda Tang, Xuanzhong Chen, Rong Han, Ting Chen
Abstract: Recent advancements in Vision Language Models (VLMs) have demonstrated remarkable promise in generating visually grounded responses. However, their application in the medical domain is hindered by unique challenges. For instance, most VLMs rely on a single method of visual grounding, whereas complex medical tasks demand more versatile approaches. Additionally, while most VLMs process only 2D images, a large portion of medical images are 3D. The lack of medical data further compounds these obstacles. To address these challenges, we present VividMed, a vision language model with versatile visual grounding for medicine. Our model supports generating both semantic segmentation masks and instance-level bounding boxes, and accommodates various imaging modalities, including both 2D and 3D data. We design a three-stage training procedure and an automatic data synthesis pipeline based on open datasets and models. Besides visual grounding tasks, VividMed also excels in other common downstream tasks, including Visual Question Answering (VQA) and report generation. Ablation studies empirically show that the integration of visual grounding ability leads to improved performance on these tasks. Our code is publicly available at https://github.com/function2-llx/MMMM.
Authors: Lazar {\DJ}okovi\'c, Marko Robnik-\v{S}ikonja
Abstract: The sarcasm detection task in natural language processing tries to classify whether an utterance is sarcastic or not. It is related to sentiment analysis since it often inverts surface sentiment. Because sarcastic sentences are highly dependent on context, and they are often accompanied by various non-verbal cues, the task is challenging. Most of related work focuses on high-resourced languages like English. To build a sarcasm detection dataset for a less-resourced language, such as Slovenian, we leverage two modern techniques: a machine translation specific medium-size transformer model, and a very large generative language model. We explore the viability of translated datasets and how the size of a pretrained transformer affects its ability to detect sarcasm. We train ensembles of detection models and evaluate models' performance. The results show that larger models generally outperform smaller ones and that ensembling can slightly improve sarcasm detection performance. Our best ensemble approach achieves an $\text{F}_1$-score of 0.765 which is close to annotators' agreement in the source language.
Authors: Zhaoyang Wang, Weilei He, Zhiyuan Liang, Xuchao Zhang, Chetan Bansal, Ying Wei, Weitong Zhang, Huaxiu Yao
Abstract: Recent self-rewarding large language models (LLM) have successfully applied LLM-as-a-Judge to iteratively improve the alignment performance without the need of human annotations for preference data. These methods commonly utilize the same LLM to act as both the policy model (which generates responses) and the reward model (which scores and ranks those responses). The ranked responses are then used as preference pairs to train the LLM via direct alignment technologies (e.g. DPO). However, it is noteworthy that throughout this process, there is no guarantee of accuracy in the rewarding and ranking, which is critical for ensuring accurate rewards and high-quality preference data. Empirical results from relatively small LLMs (e.g., 7B parameters) also indicate that improvements from self-rewarding may diminish after several iterations in certain situations, which we hypothesize is due to accumulated bias in the reward system. This bias can lead to unreliable preference data for training the LLM. To address this issue, we first formulate and analyze the generalized iterative preference fine-tuning framework for self-rewarding language model. We then introduce the regularization to this generalized framework to mitigate the overconfident preference labeling in the self-rewarding process. Based on this theoretical insight, we propose a Consistency Regularized sElf-rewarding lAnguage Model (CREAM) that leverages the rewarding consistency across different iterations to regularize the self-rewarding training, helping the model to learn from more reliable preference data. With this explicit regularization, our empirical results demonstrate the superiority of CREAM in improving both reward consistency and alignment performance. The code is publicly available at https://github.com/Raibows/CREAM.
Authors: Hongcheng Gao, Tianyu Pang, Chao Du, Taihang Hu, Zhijie Deng, Min Lin
Abstract: With the rapid progress of diffusion-based content generation, significant efforts are being made to unlearn harmful or copyrighted concepts from pretrained diffusion models (DMs) to prevent potential model misuse. However, it is observed that even when DMs are properly unlearned before release, malicious finetuning can compromise this process, causing DMs to relearn the unlearned concepts. This occurs partly because certain benign concepts (e.g., "skin") retained in DMs are related to the unlearned ones (e.g., "nudity"), facilitating their relearning via finetuning. To address this, we propose meta-unlearning on DMs. Intuitively, a meta-unlearned DM should behave like an unlearned DM when used as is; moreover, if the meta-unlearned DM undergoes malicious finetuning on unlearned concepts, the related benign concepts retained within it will be triggered to self-destruct, hindering the relearning of unlearned concepts. Our meta-unlearning framework is compatible with most existing unlearning methods, requiring only the addition of an easy-to-implement meta objective. We validate our approach through empirical experiments on meta-unlearning concepts from Stable Diffusion models (SD-v1-4 and SDXL), supported by extensive ablation studies. Our code is available at https://github.com/sail-sg/Meta-Unlearning.
Authors: Yida Yin, Zekai Wang, Yuvan Sharma, Dantong Niu, Trevor Darrell, Roei Herzig
Abstract: Recently, Large Language Models (LLMs) have achieved remarkable success using in-context learning (ICL) in the language domain. However, leveraging the ICL capabilities within LLMs to directly predict robot actions remains largely unexplored. In this paper, we introduce RoboPrompt, a framework that enables off-the-shelf text-only LLMs to directly predict robot actions through ICL without training. Our approach first heuristically identifies keyframes that capture important moments from an episode. Next, we extract end-effector actions from these keyframes as well as the estimated initial object poses, and both are converted into textual descriptions. Finally, we construct a structured template to form ICL demonstrations from these textual descriptions and a task instruction. This enables an LLM to directly predict robot actions at test time. Through extensive experiments and analysis, RoboPrompt shows stronger performance over zero-shot and ICL baselines in simulated and real-world settings.
Authors: Sijun Tan, Siyuan Zhuang, Kyle Montgomery, William Y. Tang, Alejandro Cuadron, Chenguang Wang, Raluca Ada Popa, Ion Stoica
Abstract: LLM-based judges have emerged as a scalable alternative to human evaluation and are increasingly used to assess, compare, and improve models. However, the reliability of LLM-based judges themselves is rarely scrutinized. As LLMs become more advanced, their responses grow more sophisticated, requiring stronger judges to evaluate them. Existing benchmarks primarily focus on a judge's alignment with human preferences, but often fail to account for more challenging tasks where crowdsourced human preference is a poor indicator of factual and logical correctness. To address this, we propose a novel evaluation framework to objectively evaluate LLM-based judges. Based on this framework, we propose JudgeBench, a benchmark for evaluating LLM-based judges on challenging response pairs spanning knowledge, reasoning, math, and coding. JudgeBench leverages a novel pipeline for converting existing difficult datasets into challenging response pairs with preference labels reflecting objective correctness. Our comprehensive evaluation on a collection of prompted judges, fine-tuned judges, multi-agent judges, and reward models shows that JudgeBench poses a significantly greater challenge than previous benchmarks, with many strong models (e.g., GPT-4o) performing just slightly better than random guessing. Overall, JudgeBench offers a reliable platform for assessing increasingly advanced LLM-based judges. Data and code are available at https://github.com/ScalerLab/JudgeBench .
Authors: Milan Straka, Jana Strakov\'a
Abstract: We describe the winning submission to the CRAC 2022 Shared Task on Multilingual Coreference Resolution. Our system first solves mention detection and then coreference linking on the retrieved spans with an antecedent-maximization approach, and both tasks are fine-tuned jointly with shared Transformer weights. We report results of fine-tuning a wide range of pretrained models. The center of this contribution are fine-tuned multilingual models. We found one large multilingual model with sufficiently large encoder to increase performance on all datasets across the board, with the benefit not limited only to the underrepresented languages or groups of typologically relative languages. The source code is available at https://github.com/ufal/crac2022-corpipe.
Authors: Tarek Naous, Michael J. Ryan, Anton Lavrouk, Mohit Chandra, Wei Xu
Abstract: We present a comprehensive evaluation of large language models for multilingual readability assessment. Existing evaluation resources lack domain and language diversity, limiting the ability for cross-domain and cross-lingual analyses. This paper introduces ReadMe++, a multilingual multi-domain dataset with human annotations of 9757 sentences in Arabic, English, French, Hindi, and Russian, collected from 112 different data sources. This benchmark will encourage research on developing robust multilingual readability assessment methods. Using ReadMe++, we benchmark multilingual and monolingual language models in the supervised, unsupervised, and few-shot prompting settings. The domain and language diversity in ReadMe++ enable us to test more effective few-shot prompting, and identify shortcomings in state-of-the-art unsupervised methods. Our experiments also reveal exciting results of superior domain generalization and enhanced cross-lingual transfer capabilities by models trained on ReadMe++. We will make our data publicly available and release a python package tool for multilingual sentence readability prediction using our trained models at: https://github.com/tareknaous/readme
Authors: Nuo Chen, Zinan Zheng, Ning Wu, Ming Gong, Dongmei Zhang, Jia Li
Abstract: Existing research predominantly focuses on developing powerful language learning models (LLMs) for mathematical reasoning within monolingual languages, with few explorations in preserving efficacy in a multilingual context. To bridge this gap, this paper pioneers exploring and training powerful Multilingual Math Reasoning (xMR) LLMs. Firstly, by utilizing translation, we construct the first multilingual math reasoning instruction dataset, MGSM8KInstruct, encompassing ten distinct languages, thus addressing the issue of training data scarcity in xMR tasks. Based on the collected dataset, we propose different training strategies to build powerful xMR LLMs, named MathOctopus, notably outperform conventional open-source LLMs and exhibit superiority over ChatGPT in few-shot scenarios. Notably, MathOctopus-13B reaches 47.6% accuracy which exceeds ChatGPT 46.3% on MGSM testset. Beyond remarkable results, we unearth several pivotal observations and insights from extensive experiments: (1) When extending the rejection sampling strategy to the multilingual context, it proves effective for model performances, albeit limited. (2) Employing parallel corpora for math Supervised Fine-Tuning (SFT) across multiple languages not only significantly enhances model performance multilingually but also elevates their monolingual performance. This indicates that crafting multilingual corpora can be regarded as a vital strategy for enhancing model performance in a specific language, especially in mathematical reasoning tasks. For instance, MathOctopus-7B improves its counterparts that trained on English from 42.2% to 50.8% on GSM8K testset. Codes are available at https://github.com/microsoft/MathOctopus.
Authors: Xiaorong Wang, Clara Na, Emma Strubell, Sorelle Friedler, Sasha Luccioni
Abstract: Despite the popularity of the `pre-train then fine-tune' paradigm in the NLP community, existing work quantifying energy costs and associated carbon emissions has largely focused on language model pre-training. Although a single pre-training run draws substantially more energy than fine-tuning, fine-tuning is performed more frequently by many more individual actors, and thus must be accounted for when considering the energy and carbon footprint of NLP. In order to better characterize the role of fine-tuning in the landscape of energy and carbon emissions in NLP, we perform a careful empirical study of the computational costs of fine-tuning across tasks, datasets, hardware infrastructure and measurement modalities. Our experimental results allow us to place fine-tuning energy and carbon costs into perspective with respect to pre-training and inference, and outline recommendations to NLP researchers and practitioners who wish to improve their fine-tuning energy efficiency.
Authors: Milan Straka
Abstract: We present CorPipe, the winning entry to the CRAC 2023 Shared Task on Multilingual Coreference Resolution. Our system is an improved version of our earlier multilingual coreference pipeline, and it surpasses other participants by a large margin of 4.5 percent points. CorPipe first performs mention detection, followed by coreference linking via an antecedent-maximization approach on the retrieved spans. Both tasks are trained jointly on all available corpora using a shared pretrained language model. Our main improvements comprise inputs larger than 512 subwords and changing the mention decoding to support ensembling. The source code is available at https://github.com/ufal/crac2023-corpipe.
Authors: Fanny Jourdan, Louis B\'ethune, Agustin Picard, Laurent Risser, Nicholas Asher
Abstract: Ensuring fairness in NLP models is crucial, as they often encode sensitive attributes like gender and ethnicity, leading to biased outcomes. Current concept erasure methods attempt to mitigate this by modifying final latent representations to remove sensitive information without retraining the entire model. However, these methods typically rely on linear classifiers, which leave models vulnerable to non-linear adversaries capable of recovering sensitive information. We introduce Targeted Concept Erasure (TaCo), a novel approach that removes sensitive information from final latent representations, ensuring fairness even against non-linear classifiers. Our experiments show that TaCo outperforms state-of-the-art methods, achieving greater reductions in the prediction accuracy of sensitive attributes by non-linear classifier while preserving overall task performance. Code is available on https://github.com/fanny-jourdan/TaCo.
Authors: Somnath Banerjee, Amruit Sahoo, Sayan Layek, Avik Dutta, Rima Hazra, Animesh Mukherjee
Abstract: In the continuously advancing AI landscape, crafting context-rich and meaningful responses via Large Language Models (LLMs) is essential. Researchers are becoming more aware of the challenges that LLMs with fewer parameters encounter when trying to provide suitable answers to open-ended questions. To address these hurdles, the integration of cutting-edge strategies, augmentation of rich external domain knowledge to LLMs, offers significant improvements. This paper introduces a novel framework that combines graph-driven context retrieval in conjunction to knowledge graphs based enhancement, honing the proficiency of LLMs, especially in domain specific community question answering platforms like AskUbuntu, Unix, and ServerFault. We conduct experiments on various LLMs with different parameter sizes to evaluate their ability to ground knowledge and determine factual accuracy in answers to open-ended questions. Our methodology GraphContextGen consistently outperforms dominant text-based retrieval systems, demonstrating its robustness and adaptability to a larger number of use cases. This advancement highlights the importance of pairing context rich data retrieval with LLMs, offering a renewed approach to knowledge sourcing and generation in AI systems. We also show that, due to rich contextual data retrieval, the crucial entities, along with the generated answer, remain factually coherent with the gold answer.
Authors: Yijiang River Dong, Hongzhou Lin, Mikhail Belkin, Ramon Huerta, Ivan Vuli\'c
Abstract: Mitigating the retention of sensitive or private information in large language models is essential for enhancing privacy and safety. Existing unlearning methods, like Gradient Ascent and Negative Preference Optimization, directly tune models to remove unwanted information. However, these methods often become unstable because they fine-tune by maximizing cross-entropy loss, which is the opposite of traditional loss minimization in learning. This reversal creates instability, especially on larger datasets, as the model struggles to balance unlearning with maintaining language capacity, leading to over-unlearning. In this paper, we introduce UnDIAL (Unlearning via Self-Distillation on Adjusted Logits), a novel and robust unlearning method. Our approach leverages self-distillation to adjust logits and selectively reduce the influence of targeted tokens. This technique ensures smooth convergence and avoids catastrophic forgetting, even in challenging unlearning tasks with large datasets and sequential unlearning requests. Extensive experiments show that UnDIAL can achieve both robustness in unlearning and scalability while maintaining stable training dynamics and resilience to hyperparameter tuning.
Authors: Haoyu Wang, Ruirui Li, Haoming Jiang, Jinjin Tian, Zhengyang Wang, Chen Luo, Xianfeng Tang, Monica Cheng, Tuo Zhao, Jing Gao
Abstract: Retrieval-augmented Large Language Models (LLMs) offer substantial benefits in enhancing performance across knowledge-intensive scenarios. However, these methods often face challenges with complex inputs and encounter difficulties due to noisy knowledge retrieval, notably hindering model effectiveness. To address this issue, we introduce BlendFilter, a novel approach that elevates retrieval-augmented LLMs by integrating query generation blending with knowledge filtering. BlendFilter proposes the blending process through its query generation method, which integrates both external and internal knowledge augmentation with the original query, ensuring comprehensive information gathering. Additionally, our distinctive knowledge filtering module capitalizes on the intrinsic capabilities of the LLM, effectively eliminating extraneous data. We conduct extensive experiments on three open-domain question answering benchmarks, and the findings clearly indicate that our innovative BlendFilter surpasses state-of-the-art baselines significantly.
Authors: Fan Huang, Haewoon Kwak, Jisun An
Abstract: The robustness of AI-content detection models against sophisticated adversarial strategies, such as paraphrasing or word switching, is a rising concern in natural language generation (NLG) applications. This study proposes ToBlend, a novel token-level ensemble text generation method to challenge the robustness of current AI-content detection approaches by utilizing multiple sets of candidate generative large language models (LLMs). By randomly sampling token(s) from candidate LLMs sets, we find ToBlend significantly drops the performance of most mainstream AI-content detection methods. We evaluate the text quality produced under different ToBlend settings based on annotations from experienced human experts. We proposed a fine-tuned Llama3.1 model to distinguish the ToBlend generated text more accurately. Our findings underscore our proposed text generation approach's great potential in deceiving and improving detection models. Our datasets, codes, and annotations are open-sourced.
Authors: Baihan Lin, Djallel Bouneffouf, Yulia Landa, Rachel Jespersen, Cheryl Corcoran, Guillermo Cecchi
Abstract: The therapeutic working alliance is a critical factor in predicting the success of psychotherapy treatment. Traditionally, working alliance assessment relies on questionnaires completed by both therapists and patients. In this paper, we present COMPASS, a novel framework to directly infer the therapeutic working alliance from the natural language used in psychotherapy sessions. Our approach utilizes advanced large language models (LLMs) to analyze transcripts of psychotherapy sessions and compare them with distributed representations of statements in the working alliance inventory. Analyzing a dataset of over 950 sessions covering diverse psychiatric conditions including anxiety, depression, schizophrenia, and suicidal tendencies, we demonstrate the effectiveness of our method in providing fine-grained mapping of patient-therapist alignment trajectories and offering interpretability for clinical psychiatry and in identifying emerging patterns related to the condition being treated. By employing various deep learning-based topic modeling techniques in combination with prompting generative language models, we analyze the topical characteristics of different psychiatric conditions and their evolution at a turn-level resolution. This combined framework enhances the understanding of therapeutic interactions, enabling timely feedback for therapists regarding the quality of therapeutic relationships and providing interpretable insights to improve the effectiveness of psychotherapy.
Authors: Zexuan Qiu, Jingjing Li, Shijue Huang, Xiaoqi Jiao, Wanjun Zhong, Irwin King
Abstract: Developing Large Language Models (LLMs) with robust long-context capabilities has been the recent research focus, resulting in the emergence of long-context LLMs proficient in Chinese. However, the evaluation of these models remains underdeveloped due to a lack of benchmarks. To address this gap, we present CLongEval, a comprehensive Chinese benchmark for evaluating long-context LLMs. CLongEval is characterized by three key features: (1) Sufficient data volume, comprising 7 distinct tasks and 7,267 examples; (2) Broad applicability, accommodating to models with context windows size from 1K to 100K; (3) High quality, with over 2,000 manually annotated question-answer pairs in addition to the automatically constructed labels. With CLongEval, we undertake a comprehensive assessment of 6 open-source long-context LLMs and 2 leading commercial counterparts that feature both long-context abilities and proficiency in Chinese. We also provide in-depth analysis based on the empirical results, trying to shed light on the critical capabilities that present challenges in long-context settings. The dataset, evaluation scripts, and model outputs are released.
Authors: Saeel Sandeep Nachane, Ojas Gramopadhye, Prateek Chanda, Ganesh Ramakrishnan, Kshitij Sharad Jadhav, Yatin Nandwani, Dinesh Raghu, Sachindra Joshi
Abstract: In this paper, we propose a modified version of the MedQA-USMLE dataset, named MEDQA-OPEN, which contains open-ended medical questions without options to mimic clinical scenarios, along with clinician-approved reasoned answers. Additionally, we implement a prompt driven by Chain of Thought (CoT) reasoning, CLINICR, to mirror the prospective process of incremental reasoning, reaching a correct response to medical questions. We empirically demonstrate how CLINICR outperforms the state-of-the-art 5-shot CoT-based prompt (Li\'evin et al., 2022). We also present an approach that mirrors real-life clinical practice by first exploring multiple differential diagnoses through MCQ-CLINICR and subsequently narrowing down to a final diagnosis using MCQ-ELIMINATIVE. Finally, emphasizing the importance of response verification in medical settings, we utilize a reward model mechanism, replacing the elimination process performed by MCQ-ELIMINATIVE.
Authors: Guangming Huang, Yingya Li, Shoaib Jameel, Yunfei Long, Giorgos Papanastasiou
Abstract: Deep learning (DL) has substantially enhanced natural language processing (NLP) in healthcare research. However, the increasing complexity of DL-based NLP necessitates transparent model interpretability, or at least explainability, for reliable decision-making. This work presents a thorough scoping review of explainable and interpretable DL in healthcare NLP. The term "eXplainable and Interpretable Artificial Intelligence" (XIAI) is introduced to distinguish XAI from IAI. Different models are further categorized based on their functionality (model-, input-, output-based) and scope (local, global). Our analysis shows that attention mechanisms are the most prevalent emerging IAI technique. The use of IAI is growing, distinguishing it from XAI. The major challenges identified are that most XIAI does not explore "global" modelling processes, the lack of best practices, and the lack of systematic evaluation and benchmarks. One important opportunity is to use attention mechanisms to enhance multi-modal XIAI for personalized medicine. Additionally, combining DL with causal logic holds promise. Our discussion encourages the integration of XIAI in Large Language Models (LLMs) and domain-specific smaller models. In conclusion, XIAI adoption in healthcare requires dedicated in-house expertise. Collaboration with domain experts, end-users, and policymakers can lead to ready-to-use XIAI methods across NLP and medical tasks. While challenges exist, XIAI techniques offer a valuable foundation for interpretable NLP algorithms in healthcare.
Authors: Maksym Taranukhin, Sahithya Ravi, Gabor Lukacs, Evangelos Milios, Vered Shwartz
Abstract: The Canadian air travel sector has seen a significant increase in flight delays, cancellations, and other issues concerning passenger rights. Recognizing this demand, we present a chatbot to assist passengers and educate them about their rights. Our system breaks a complex user input into simple queries which are used to retrieve information from a collection of documents detailing air travel regulations. The most relevant passages from these documents are presented along with links to the original documents and the generated queries, enabling users to dissect and leverage the information for their unique circumstances. The system successfully overcomes two predominant challenges: understanding complex user inputs, and delivering accurate answers, free of hallucinations, that passengers can rely on for making informed decisions. A user study comparing the chatbot to a Google search demonstrated the chatbot's usefulness and ease of use. Beyond the primary goal of providing accurate and timely information to air passengers regarding their rights, we hope that this system will also enable further research exploring the tradeoff between the user-friendly conversational interface of chatbots and the accuracy of retrieval systems.
Authors: Yu Feng, Ben Zhou, Weidong Lin, Dan Roth
Abstract: Predictive models often need to work with incomplete information in real-world tasks. Consequently, they must provide reliable probability or confidence estimation, especially in large-scale decision making and planning tasks. Current large language models (LLM) are insufficient for such accurate estimations, but they can generate relevant factors that may affect the probabilities, produce coarse-grained probabilities when the information is more complete, and help determine which factors are relevant to specific downstream contexts. In this paper, we make use of these capabilities of LLMs to provide a significantly more accurate probabilistic estimation. We propose BIRD, a novel probabilistic inference framework that aligns a Bayesian network with LLM abductions and then estimates more accurate probabilities in a deduction step. We show BIRD provides reliable probability estimations that are 30\% better than those provided directly by LLM baselines. These estimates can further contribute to better and more trustworthy decision-making.
Authors: Brendan King, Jeffrey Flanigan
Abstract: Training task-oriented dialogue systems typically requires turn-level annotations for interacting with their APIs: e.g. a dialogue state and the system actions taken at each step. These annotations can be costly to produce, error-prone, and require both domain and annotation expertise. With advances in LLMs, we hypothesize that unlabeled data and a schema definition are sufficient for building a working task-oriented dialogue system, completely unsupervised. We consider a novel unsupervised setting of only (1) a well-defined API schema (2) a set of unlabeled dialogues between a user and agent. We propose an innovative approach using expectation-maximization (EM) that infers turn-level annotations as latent variables using a noisy channel model to build an end-to-end dialogue agent. Evaluating our approach on the MultiWOZ benchmark, our method more than doubles the dialogue success rate of a strong GPT-3.5 baseline.
Authors: Lin Ai, Zheng Hui, Zizhou Liu, Julia Hirschberg
Abstract: Machine Reading Comprehension (MRC) poses a significant challenge in the field of Natural Language Processing (NLP). While mainstream MRC methods predominantly leverage extractive strategies using encoder-only models such as BERT, generative approaches face the issue of out-of-control generation -- a critical problem where answers generated are often incorrect, irrelevant, or unfaithful to the source text. To address these limitations in generative models for MRC, we introduce the Question-Attended Span Extraction (QASE) module. Integrated during the fine-tuning phase of pre-trained generative language models (PLMs), QASE significantly enhances their performance, allowing them to surpass the extractive capabilities of advanced Large Language Models (LLMs) such as GPT-4 in few-shot settings. Notably, these gains in performance do not come with an increase in computational demands. The efficacy of the QASE module has been rigorously tested across various datasets, consistently achieving or even surpassing state-of-the-art (SOTA) results, thereby bridging the gap between generative and extractive models in extractive MRC tasks.
Authors: Fenglin Liu, Zheng Li, Hongjian Zhou, Qingyu Yin, Jingfeng Yang, Xianfeng Tang, Chen Luo, Ming Zeng, Haoming Jiang, Yifan Gao, Priyanka Nigam, Sreyashi Nag, Bing Yin, Yining Hua, Xuan Zhou, Omid Rohanian, Anshul Thakur, Lei Clifton, David A. Clifton
Abstract: The adoption of large language models (LLMs) to assist clinicians has attracted remarkable attention. Existing works mainly adopt the close-ended question-answering (QA) task with answer options for evaluation. However, many clinical decisions involve answering open-ended questions without pre-set options. To better understand LLMs in the clinic, we construct a benchmark ClinicBench. We first collect eleven existing datasets covering diverse clinical language generation, understanding, and reasoning tasks. Furthermore, we construct six novel datasets and clinical tasks that are complex but common in real-world practice, e.g., open-ended decision-making, long document processing, and emerging drug analysis. We conduct an extensive evaluation of twenty-two LLMs under both zero-shot and few-shot settings. Finally, we invite medical experts to evaluate the clinical usefulness of LLMs. The benchmark data is available at https://github.com/AI-in-Health/ClinicBench.
Authors: Arkadiy Saakyan, Shreyas Kulkarni, Tuhin Chakrabarty, Smaranda Muresan
Abstract: Large Vision-Language Models (VLMs) have demonstrated strong capabilities in tasks requiring a fine-grained understanding of literal meaning in images and text, such as visual question-answering or visual entailment. However, there has been little exploration of these models' capabilities when presented with images and captions containing figurative meaning, such as metaphors or humor. To close this gap, we propose a new task framing the figurative meaning understanding problem as an explainable visual entailment task, where the model has to predict whether the image (premise) entails a caption (hypothesis) and justify the predicted label with a textual explanation. The figurative phenomena can be present either in the image, the caption, or both. Utilizing a human-AI collaboration approach, we build the accompanying expert-verified dataset V-FLUTE, containing 6,027 {image, caption, label, explanation} instances spanning five diverse figurative phenomena: metaphors, similes, idioms, sarcasm, and humor. Through automatic evaluation, we find that VLMs struggle to generalize from literal to figurative meaning, particularly when it is present in images. Further, we identify common types of errors in VLM reasoning via human evaluation.
Authors: Letian Peng, Jingbo Shang
Abstract: Persona-driven role-playing (PRP) aims to build AI characters that can respond to user queries by faithfully sticking with all persona statements. Unfortunately, existing faithfulness criteria for PRP are limited to coarse-grained LLM-based scoring without a clear definition or formulation. This paper presents a pioneering exploration to quantify PRP faithfulness as a fine-grained and explainable criterion, which also serves as a reliable reference for optimization. Our criterion first discriminates persona statements into active and passive constraints by identifying the query-statement relevance. Then, we incorporate all constraints following the principle that the AI character's response should be (a) entailed by active (relevant) constraints and (b) not contradicted by passive (irrelevant) constraints. We translate this principle mathematically into a novel Active-Passive-Constraint (APC) score, a constraint-wise sum of natural language inference (NLI) scores weighted by relevance scores. In practice, we build the APC scoring system by symbolically distilling small discriminators from GPT-4 for efficiency. We validate the quality of the APC score against human evaluation based on example personas with tens of statements, and the results show a high correlation. We further leverage it as a reward system in direct preference optimization (DPO) for better AI characters. Our experiments offer a fine-grained and explainable comparison between existing PRP techniques, revealing their advantages and limitations. We further find APC-based DPO to be one of the most competitive techniques for sticking with all constraints and can be well incorporated with other techniques. We then extend the scale of the experiments to real persons with hundreds of statements and reach a consistent conclusion.
Authors: Demi Zhang, Bushi Xiao, Chao Gao, Sangpil Youm, Bonnie J Dorr
Abstract: This study evaluates the performance of Recurrent Neural Network (RNN) and Transformer models in replicating cross-language structural priming, a key indicator of abstract grammatical representations in human language processing. Focusing on Chinese-English priming, which involves two typologically distinct languages, we examine how these models handle the robust phenomenon of structural priming, where exposure to a particular sentence structure increases the likelihood of selecting a similar structure subsequently. Our findings indicate that transformers outperform RNNs in generating primed sentence structures, with accuracy rates that exceed 25.84\% to 33. 33\%. This challenges the conventional belief that human sentence processing primarily involves recurrent and immediate processing and suggests a role for cue-based retrieval mechanisms. This work contributes to our understanding of how computational models may reflect human cognitive processes across diverse language families.
Authors: Minjia Mao, Dongjun Wei, Zeyu Chen, Xiao Fang, Michael Chau
Abstract: Recent advancements in large language models (LLMs) have highlighted the risk of misusing them, raising the need for accurate detection of LLM-generated content. In response, a viable solution is to inject imperceptible identifiers into LLMs, known as watermarks. Previous work demonstrates that unbiased watermarks ensure unforgeability and preserve text quality by maintaining the expectation of the LLM output probability distribution. However, previous unbiased watermarking methods suffer from one or more of the following issues: (1) requiring access to white-box LLMs during detection, (2) incurring long detection time, (3) being not robust against simple watermarking attacks, (4) failing to provide statistical guarantees for the type II error of watermark detection, and (5) being not statistically unbiased for low-entropy scenarios, which hinder their deployment in practice. This study proposes the Sampling One Then Accepting (STA-1) method, a watermark that can address all of these issues. Moreover, we discuss the tradeoff between watermark strength and text quality for unbiased watermarks. We show that in low-entropy scenarios, unbiased watermarks face a tradeoff between watermark strength and the risk of unsatisfactory outputs. Experimental results on both low-entropy and high-entropy datasets demonstrate that STA-1 achieves text quality and watermark strength comparable to existing unbiased watermarks, with a low risk of unsatisfactory outputs. Implementation codes for this study are available online.
Authors: Yunzhi Yao, Ningyu Zhang, Zekun Xi, Mengru Wang, Ziwen Xu, Shumin Deng, Huajun Chen
Abstract: The remarkable capabilities of modern large language models are rooted in their vast repositories of knowledge encoded within their parameters, enabling them to perceive the world and engage in reasoning. The inner workings of how these models store knowledge have long been a subject of intense interest and investigation among researchers. To date, most studies have concentrated on isolated components within these models, such as the Multilayer Perceptrons and attention head. In this paper, we delve into the computation graph of the language model to uncover the knowledge circuits that are instrumental in articulating specific knowledge. The experiments, conducted with GPT2 and TinyLLAMA, have allowed us to observe how certain information heads, relation heads, and Multilayer Perceptrons collaboratively encode knowledge within the model. Moreover, we evaluate the impact of current knowledge editing techniques on these knowledge circuits, providing deeper insights into the functioning and constraints of these editing methodologies. Finally, we utilize knowledge circuits to analyze and interpret language model behaviors such as hallucinations and in-context learning. We believe the knowledge circuits hold potential for advancing our understanding of Transformers and guiding the improved design of knowledge editing. Code and data are available in https://github.com/zjunlp/KnowledgeCircuits.
Authors: Gon\c{c}alo R. A. Faria, Sweta Agrawal, Ant\'onio Farinhas, Ricardo Rei, Jos\'e G. C. de Souza, Andr\'e F. T. Martins
Abstract: An important challenge in machine translation (MT) is to generate high-quality and diverse translations. Prior work has shown that the estimated likelihood from the MT model correlates poorly with translation quality. In contrast, quality evaluation metrics (such as COMET or BLEURT) exhibit high correlations with human judgments, which has motivated their use as rerankers (such as quality-aware and minimum Bayes risk decoding). However, relying on a single translation with high estimated quality increases the chances of "gaming the metric''. In this paper, we address the problem of sampling a set of high-quality and diverse translations. We provide a simple and effective way to avoid over-reliance on noisy quality estimates by using them as the energy function of a Gibbs distribution. Instead of looking for a mode in the distribution, we generate multiple samples from high-density areas through the Metropolis-Hastings algorithm, a simple Markov chain Monte Carlo approach. The results show that our proposed method leads to high-quality and diverse outputs across multiple language pairs (English$\leftrightarrow${German, Russian}) with two strong decoder-only LLMs (Alma-7b, Tower-7b).
Authors: Haoyu Wang, Tianci Liu, Ruirui Li, Monica Cheng, Tuo Zhao, Jing Gao
Abstract: Pre-trained language models, trained on large-scale corpora, demonstrate strong generalizability across various NLP tasks. Fine-tuning these models for specific tasks typically involves updating all parameters, which is resource-intensive. Parameter-efficient fine-tuning (PEFT) methods, such as the popular LoRA family, introduce low-rank matrices to learn only a few parameters efficiently. However, during inference, the product of these matrices updates all pre-trained parameters, complicating tasks like knowledge editing that require selective updates. We propose a novel PEFT method, which conducts \textbf{r}ow and c\textbf{o}lumn-wise spar\textbf{se} \textbf{lo}w-\textbf{r}ank \textbf{a}daptation (RoseLoRA), to address this challenge. RoseLoRA identifies and updates only the most important parameters for a specific task, maintaining efficiency while preserving other model knowledge. By adding a sparsity constraint on the product of low-rank matrices and converting it to row and column-wise sparsity, we ensure efficient and precise model updates. Our theoretical analysis guarantees the lower bound of the sparsity with respective to the matrix product. Extensive experiments on five benchmarks across twenty datasets demonstrate that RoseLoRA outperforms baselines in both general fine-tuning and knowledge editing tasks.
Authors: Minda Hu, Licheng Zong, Hongru Wang, Jingyan Zhou, Jingjing Li, Yichen Gao, Kam-Fai Wong, Yu Li, Irwin King
Abstract: Large Language Models (LLMs) have shown great potential in the biomedical domain with the advancement of retrieval-augmented generation (RAG). However, existing retrieval-augmented approaches face challenges in addressing diverse queries and documents, particularly for medical knowledge queries, resulting in sub-optimal performance. To address these limitations, we propose a novel plug-and-play LLM-based retrieval method called Self-Rewarding Tree Search (SeRTS) based on Monte Carlo Tree Search (MCTS) and a self-rewarding paradigm. By combining the reasoning capabilities of LLMs with the effectiveness of tree search, SeRTS boosts the zero-shot performance of retrieving high-quality and informative results for RAG. We further enhance retrieval performance by fine-tuning LLMs with Proximal Policy Optimization (PPO) objectives using the trajectories collected by SeRTS as feedback. Controlled experiments using the BioASQ-QA dataset with GPT-3.5-Turbo and LLama2-7b demonstrate that our method significantly improves the performance of the BM25 retriever and surpasses the strong baseline of self-reflection in both efficiency and scalability. Moreover, SeRTS generates higher-quality feedback for PPO training than self-reflection. Our proposed method effectively adapts LLMs to document retrieval tasks, enhancing their ability to retrieve highly relevant documents for RAG in the context of medical knowledge queries. This work presents a significant step forward in leveraging LLMs for accurate and comprehensive biomedical question answering.
Authors: Shengkang Wang, Hongzhan Lin, Ziyang Luo, Zhen Ye, Guang Chen, Jing Ma
Abstract: Large vision-language models (LVLMs) have significantly improved multimodal reasoning tasks, such as visual question answering and image captioning. These models embed multimodal facts within their parameters, rather than relying on external knowledge bases to store factual information explicitly. However, the content discerned by LVLMs may deviate from actual facts due to inherent bias or incorrect inference. To address this issue, we introduce MFC-Bench, a rigorous and comprehensive benchmark designed to evaluate the factual accuracy of LVLMs across three stages of verdict prediction for MFC: Manipulation, Out-of-Context, and Veracity Classification. Through our evaluation on MFC-Bench, we benchmarked a dozen diverse and representative LVLMs, uncovering that current models still fall short in multimodal fact-checking and demonstrate insensitivity to various forms of manipulated content. We hope that MFC-Bench could raise attention to the trustworthy AI potentially assisted by LVLMs in the future. The MFC-Bench and accompanying resources are publicly accessible at https://github.com/wskbest/MFC-Bench, contributing to ongoing research in the multimodal fact-checking field.
Authors: Boxuan Lyu, Hidetaka Kamigaito, Kotaro Funakoshi, Manabu Okumura
Abstract: Maximum a posteriori decoding, a commonly used method for neural machine translation (NMT), aims to maximize the estimated posterior probability. However, high estimated probability does not always lead to high translation quality. Minimum Bayes Risk (MBR) decoding (\citealp{kumar2004minimum}) offers an alternative by seeking hypotheses with the highest expected utility. In this paper, we show that Quality Estimation (QE) reranking (\citealp{fernandes-etal-2022-quality}), which uses a QE model as a reranker, can be viewed as a variant of MBR. Inspired by this, we propose source-based MBR (sMBR) decoding, a novel approach that utilizes synthetic sources (generated via back-translation or paraphrasing) as ``support hypotheses'' and a reference-free quality estimation metric as the utility function, marking the first work to solely use sources in MBR decoding. Experiments show that sMBR outperforms QE reranking and the standard MBR decoding. Our findings suggest that sMBR is a promising approach for NMT decoding.
Authors: Ronny Luss, Erik Miehling, Amit Dhurandhar
Abstract: The advent of black-box deep neural network classification models has sparked the need to explain their decisions. However, in the case of generative AI, such as large language models (LLMs), there is no class prediction to explain. Rather, one can ask why an LLM output a particular response to a given prompt. In this paper, we answer this question by proposing, to the best of our knowledge, the first contrastive explanation methods requiring simply black-box/query access. Our explanations suggest that an LLM outputs a reply to a given prompt because if the prompt was slightly modified, the LLM would have given a different response that is either less preferable or contradicts the original response. The key insight is that contrastive explanations simply require a scoring function that has meaning to the user and not necessarily a specific real valued quantity (viz. class label). We offer two algorithms for finding contrastive explanations: i) A myopic algorithm, which although effective in creating contrasts, requires many model calls and ii) A budgeted algorithm, our main algorithmic contribution, which intelligently creates contrasts adhering to a query budget, necessary for longer contexts. We show the efficacy of these methods on diverse natural language tasks such as open-text generation, automated red teaming, and explaining conversational degradation.
Authors: Jie Wang
Abstract: We introduce the concept of AI-oracle machines for intelligent computing and outline several applications to demonstrate their potential. Following this, we advocate for the development of a comprehensive platform to streamline the implementation of AI-oracle machines.
Authors: Hawon Jeong, ChaeHun Park, Jimin Hong, Hojoon Lee, Jaegul Choo
Abstract: As large language models (LLMs) are increasingly used as evaluators for natural language generation tasks, ensuring unbiased assessments is essential. However, LLM evaluators often display biased preferences, such as favoring verbosity and authoritative tones. Our empirical analysis reveals that these biases are exacerbated in pairwise evaluation, where LLMs directly compare two outputs and easily prioritize superficial attributes. In contrast, pointwise evaluation, which assesses outputs independently, is less susceptible to such bias because each output is judged in isolation. To address the limitations of the pairwise evaluation, we introduce a novel evaluation method, PRePair, which integrates pointwise reasoning within a pairwise framework. PRePair effectively alleviates biased preference, improving performance on the adversarial benchmark (LLMBar) while outperforming pointwise evaluation on the standard benchmark (MT-Bench).
Authors: Akanksha Mehndiratta, Krishna Asawa
Abstract: Canonical Correlation Analysis (CCA) has been exploited immensely for learning latent representations in various fields. This study takes a step further by demonstrating the potential of CCA in identifying Elementary Discourse Units(EDUs) that captures the latent information within the textual data. The probabilistic interpretation of CCA discussed in this study utilizes the two-view nature of textual data, i.e. the consecutive sentences in a document or turns in a dyadic conversation, and has a strong theoretical foundation. Furthermore, this study proposes a model for Elementary Discourse Unit(EDU) segmentation that discovers EDUs in textual data without any supervision. To validate the model, the EDUs are utilized as textual unit for content selection in textual similarity task. Empirical results on Semantic Textual Similarity(STSB) and Mohler datasets confirm that, despite represented as a unigram, the EDUs deliver competitive results and can even beat various sophisticated supervised techniques. The model is simple, linear, adaptable and language independent making it an ideal baseline particularly when labeled training data is scarce or nonexistent.
Authors: Mahammed Kamruzzaman, Gene Louis Kim
Abstract: Persona assignment has become a common strategy for customizing LLM use to particular tasks and contexts. In this study, we explore how evaluation of different nations change when LLMs are assigned specific nationality personas. We assign 193 different nationality personas (e.g., an American person) to four LLMs and examine how the LLM evaluations (or ''perceptions'')of countries change. We find that all LLM-persona combinations tend to favor Western European nations, though nation-personas push LLM behaviors to focus more on and treat the nation-persona's own region more favorably. Eastern European, Latin American, and African nations are treated more negatively by different nationality personas. We additionally find that evaluations by nation-persona LLMs of other nations correlate with human survey responses but fail to match the values closely. Our study provides insight into how biases and stereotypes are realized within LLMs when adopting different national personas. In line with the ''Blueprint for an AI Bill of Rights'', our findings underscore the critical need for developing mechanisms to ensure that LLM outputs promote fairness and avoid over-generalization.
Authors: Hakaze Cho, Yoshihiro Sakai, Mariko Kato, Kenshiro Tanaka, Akira Ishii, Naoya Inoue
Abstract: In-Context Learning (ICL) typically utilizes classification criteria from output probabilities of manually selected label tokens. However, we argue that such token-based classification criteria lead to suboptimal decision boundaries, despite delicate calibrations through translation and constrained rotation applied. To address this problem, we propose Hidden Calibration, which renounces token probabilities and uses the nearest centroid classifier on the LM's last hidden states. In detail, we assign the label of the nearest centroid previously estimated from a calibration set to the test sample as the predicted label. Our experiments on 6 models and 10 classification datasets indicate that Hidden Calibration consistently outperforms current token-based baselines by about 20%~50%, achieving a strong state-of-the-art in ICL. Our further analysis demonstrates that Hidden Calibration finds better classification criteria with less inter-class overlap, and LMs provide linearly separable intra-class clusters with the help of demonstrations, which supports Hidden Calibration and gives new insights into the principle of ICL.
Authors: Yun-Shiuan Chuang, Krirk Nirunwiroj, Zach Studdiford, Agam Goyal, Vincent V. Frigo, Sijia Yang, Dhavan Shah, Junjie Hu, Timothy T. Rogers
Abstract: Creating human-like large language model (LLM) agents is crucial for faithful social simulation. Having LLMs role-play based on demographic information sometimes improves human likeness but often does not. This study assessed whether LLM alignment with human behavior can be improved by integrating information from empirically-derived human belief networks. Using data from a human survey, we estimated a belief network encompassing 64 topics loading on nine non-overlapping latent factors. We then seeded LLM-based agents with an opinion on one topic, and assessed the alignment of its expressed opinions on remaining test topics with corresponding human data. Role-playing based on demographic information alone did not align LLM and human opinions, but seeding the agent with a single belief greatly improved alignment for topics related in the belief network, and not for topics outside the network. These results suggest a novel path for human-LLM belief alignment in work seeking to simulate and understand patterns of belief distributions in society.
Authors: Amalie Brogaard Pauli, Isabelle Augenstein, Ira Assent
Abstract: We are exposed to much information trying to influence us, such as teaser messages, debates, politically framed news, and propaganda - all of which use persuasive language. With the recent interest in Large Language Models (LLMs), we study the ability of LLMs to produce persuasive text. As opposed to prior work which focuses on particular domains or types of persuasion, we conduct a general study across various domains to measure and benchmark to what degree LLMs produce persuasive language - both when explicitly instructed to rewrite text to be more or less persuasive and when only instructed to paraphrase. We construct the new dataset Persuasive-Pairs of pairs of a short text and its rewrite by an LLM to amplify or diminish persuasive language. We multi-annotate the pairs on a relative scale for persuasive language: a valuable resource in itself, and for training a regression model to score and benchmark persuasive language, including for new LLMs across domains. In our analysis, we find that different 'personas' in LLaMA3's system prompt change persuasive language substantially, even when only instructed to paraphrase.
Authors: Hanqi Yan, Yanzheng Xiang, Guangyi Chen, Yifei Wang, Lin Gui, Yulan He
Abstract: To better interpret the intrinsic mechanism of large language models (LLMs), recent studies focus on monosemanticity on its basic units. A monosemantic neuron is dedicated to a single and specific concept, which forms a one-to-one correlation between neurons and concepts. Despite extensive research in monosemanticity probing, it remains unclear whether monosemanticity is beneficial or harmful to model capacity. To explore this question, we revisit monosemanticity from the feature decorrelation perspective and advocate for its encouragement. We experimentally observe that the current conclusion by wang2024learning, which suggests that decreasing monosemanticity enhances model performance, does not hold when the model changes. Instead, we demonstrate that monosemanticity consistently exhibits a positive correlation with model capacity, in the preference alignment process. Consequently, we apply feature correlation as a proxy for monosemanticity and incorporate a feature decorrelation regularizer into the dynamic preference optimization process. The experiments show that our method not only enhances representation diversity and activation sparsity but also improves preference alignment performance.
Authors: Davide Venditti, Elena Sofia Ruzzetti, Giancarlo A. Xompero, Cristina Giannone, Andrea Favalli, Raniero Romagnoli, Fabio Massimo Zanzotto
Abstract: Large language models (LLMs) require a significant redesign in solutions to preserve privacy in data-intensive applications due to their text-generation capabilities. Indeed, LLMs tend to memorize and emit private information when maliciously prompted. In this paper, we introduce Private Association Editing (PAE) as a novel defense approach for private data leakage. PAE is designed to effectively remove Personally Identifiable Information (PII) without retraining the model. Experimental results demonstrate the effectiveness of PAE with respect to alternative baseline methods. We believe PAE will serve as a critical tool in the ongoing effort to protect data privacy in LLMs, encouraging the development of safer models for real-world applications.
Authors: Yi-Sheng Hsu, Nils Feldhus, Sherzod Hakimov
Abstract: Free-text rationales justify model decisions in natural language and thus become likable and accessible among approaches to explanation across many tasks. However, their effectiveness can be hindered by misinterpretation and hallucination. As a perturbation test, we investigate how large language models (LLMs) perform rationale generation under the effects of readability level control, i.e., being prompted for an explanation targeting a specific expertise level, such as sixth grade or college. We find that explanations are adaptable to such instruction, though the requested readability is often misaligned with the measured text complexity according to traditional readability metrics. Furthermore, the generated rationales tend to feature medium level complexity, which correlates with the measured quality using automatic metrics. Finally, our human annotators confirm a generally satisfactory impression on rationales at all readability levels, with high-school-level readability being most commonly perceived and favored.
Authors: Jiaru Zou, Mengyu Zhou, Tao Li, Shi Han, Dongmei Zhang
Abstract: Recent advances in fine-tuning large language models (LLMs) have greatly enhanced their usage in domain-specific tasks. Despite the success, fine-tuning continues to rely on repeated and lengthy prompts, which escalate computational expenses, require more resources, and lead to slower inference. In this paper, we present a novel approach, PromptIntern, which internalizes prompt knowledge during model fine-tuning to achieve efficient inference and save costs. Instead of compressing the prompts for a vanilla model, PromptIntern aims to embed the recurrent prompt directly into the model parameters. We design a fine-tuning pipeline that includes instruction template compression, few-shot example absorption, and a progressive internalization strategy, effectively diminishing the need for intricate prompts during inference. Comprehensive experiments on challenging NL2Code tasks demonstrate that our method reduces input tokens by more than 90%, accelerates inference by 4.2 times, and reduces monetary inference costs by 88.3%.
Authors: Zhengping Jiang, Jingyu Zhang, Nathaniel Weir, Seth Ebner, Miriam Wanner, Kate Sanders, Daniel Khashabi, Anqi Liu, Benjamin Van Durme
Abstract: Hallucinations pose a challenge to the application of large language models (LLMs) thereby motivating the development of metrics to evaluate factual precision. We observe that popular metrics using the Decompose-Then-Verify framework, such as \FActScore, can be manipulated by adding obvious or repetitive subclaims to artificially inflate scores. This observation motivates our new customizable plug-and-play subclaim selection component called Core, which filters down individual subclaims according to their uniqueness and informativeness. We show that many popular factual precision metrics augmented by Core are substantially more robust on a wide range of knowledge domains. We release an evaluation framework supporting easy and modular use of Core and various decomposition strategies, which we recommend adoption by the community. We also release an expansion of the FActScore biography dataset to facilitate further studies of decomposition-based factual precision evaluation.
Authors: Hung Phan, Anurag Acharya, Rounak Meyur, Sarthak Chaturvedi, Shivam Sharma, Mike Parker, Dan Nally, Ali Jannesari, Karl Pazdernik, Mahantesh Halappanavar, Sai Munikoti, Sameera Horawalavithana
Abstract: As LLMs become increasingly ubiquitous, researchers have tried various techniques to augment the knowledge provided to these models. Long context and retrieval-augmented generation (RAG) are two such methods that have recently gained popularity. In this work, we examine the benefits of both of these techniques by utilizing question answering (QA) task in a niche domain. While the effectiveness of LLM-based QA systems has already been established at an acceptable level in popular domains such as trivia and literature, it has not often been established in niche domains that traditionally require specialized expertise. We construct the NEPAQuAD1.0 benchmark to evaluate the performance of five long-context LLMs -- Claude Sonnet, Gemini, GPT-4, Llama 3.1, and Mistral -- when answering questions originating from Environmental Impact Statements prepared by U.S. federal government agencies in accordance with the National Environmental Environmental Act (NEPA). We specifically measure the ability of LLMs to understand the nuances of legal, technical, and compliance-related information present in NEPA documents in different contextual scenarios. We test the LLMs' internal prior NEPA knowledge by providing questions without any context, as well as assess how LLMs synthesize the contextual information present in long NEPA documents to facilitate the question/answering task. We compare the performance of the models in handling different types of questions (e.g., problem-solving, divergent, etc.). Our results suggest that RAG powered models significantly outperform those provided with only the PDF context in terms of answer accuracy, regardless of the choice of the LLM. Our further analysis reveals that many models perform better answering closed type questions (Yes/No) than divergent and problem-solving questions.
Authors: Wangtao Sun, Chenxiang Zhang, XueYou Zhang, Xuanqing Yu, Ziyang Huang, Pei Chen, Haotian Xu, Shizhu He, Jun Zhao, Kang Liu
Abstract: Although Large Language Models (LLMs) have demonstrated strong ability, they are further supposed to be controlled and guided by in real-world scenarios to be safe, accurate, and intelligent. This demands the possession of capability of LLMs. However, no prior work has made a clear evaluation of the inferential rule-following capability of LLMs. Previous studies that try to evaluate the inferential rule-following capability of LLMs fail to distinguish the inferential rule-following scenarios from the instruction-following scenarios. Therefore, this paper first clarifies the concept of inferential rule-following and proposes a comprehensive benchmark, RuleBench, to evaluate a diversified range of inferential rule-following abilities. Our experimental results on a variety of LLMs show that they are still limited in following rules. Our analysis based on the evaluation results provides insights into the improvements for LLMs toward a better inferential rule-following intelligent agent. We further propose Inferential Rule-Following Tuning (IRFT). The experimental results show that through IRFT, LLMs can learn abstract rule-following abilities from purely synthetic data and then generalize to RuleBench. The data and code can be found at: https://anonymous.4open.science/r/llm-rule-following-B3E3/
URLs: https://anonymous.4open.science/r/llm-rule-following-B3E3/
Authors: Sanchit Ahuja, Kumar Tanmay, Hardik Hansrajbhai Chauhan, Barun Patra, Kriti Aggarwal, Luciano Del Corro, Arindam Mitra, Tejas Indulal Dhamecha, Ahmed Awadallah, Monojit Choudhary, Vishrav Chaudhary, Sunayana Sitaram
Abstract: Despite the remarkable success of LLMs in English, there is a significant gap in performance in non-English languages. In order to address this, we introduce a novel recipe for creating a multilingual synthetic instruction tuning dataset, sPhinX, which is created by selectively translating instruction response pairs from English into 50 languages. We test the effectiveness of sPhinx by using it to fine-tune two state-of-the-art models, Mistral-7B and Phi-Small and then evaluating them across a comprehensive suite of multilingual benchmarks that test reasoning, question answering, reading comprehension and machine translation. Our results show that Mistral-7B and Phi-Small fine-tuned with sPhinX perform better on an average by 5%pt for both the models when compared to the base variants of these models. We also devise a strategy to incorporate N-shot examples in each fine-tuning sample which further boosts the performance of these models by 9%pt and 4%pt respectively respectively compared to vanilla fine-tuning. To show efficacy of our data curation approach, we also directly translate our original dataset to the target languages, and observe an increase of 7%pt and 4%pt on both the models respectively. sPhinX outperforms other multilingual instruction tuning datasets in both efficiency and diversity, reducing dataset creation costs. It also maintains strong performance on standard English LLM benchmarks, with minimal regression.
Authors: Donghoon Han, Eunhwan Park, Gisang Lee, Adam Lee, Nojun Kwak
Abstract: The rapid expansion of multimedia content has made accurately retrieving relevant videos from large collections increasingly challenging. Recent advancements in text-video retrieval have focused on cross-modal interactions, large-scale foundation model training, and probabilistic modeling, yet often neglect the crucial user perspective, leading to discrepancies between user queries and the content retrieved. To address this, we introduce MERLIN (Multimodal Embedding Refinement via LLM-based Iterative Navigation), a novel, training-free pipeline that leverages Large Language Models (LLMs) for iterative feedback learning. MERLIN refines query embeddings from a user perspective, enhancing alignment between queries and video content through a dynamic question answering process. Experimental results on datasets like MSR-VTT, MSVD, and ActivityNet demonstrate that MERLIN substantially improves Recall@1, outperforming existing systems and confirming the benefits of integrating LLMs into multimodal retrieval systems for more responsive and context-aware multimedia retrieval.
Authors: Takuma Udagawa, Masayuki Suzuki, Masayasu Muraoka, Gakuto Kurata
Abstract: Error correction (EC) based on large language models is an emerging technology to enhance the performance of automatic speech recognition (ASR) systems. Generally, training data for EC are collected by automatically pairing a large set of ASR hypotheses (as sources) and their gold references (as targets). However, the quality of such pairs is not guaranteed, and we observed various types of noise which can make the EC models brittle, e.g. inducing overcorrection in out-of-domain (OOD) settings. In this work, we propose two fundamental criteria that EC training data should satisfy: namely, EC targets should (1) improve linguistic acceptability over sources and (2) be inferable from the available context (e.g. source phonemes). Through these criteria, we identify low-quality EC pairs and train the models not to make any correction in such cases, the process we refer to as conservative data filtering. In our experiments, we focus on Japanese ASR using a strong Conformer-CTC as the baseline and finetune Japanese LLMs for EC. Through our evaluation on a suite of 21 internal benchmarks, we demonstrate that our approach can significantly reduce overcorrection and improve both the accuracy and quality of ASR results in the challenging OOD settings.
Authors: Zehui Wu, Ziwei Gong, Lin Ai, Pengyuan Shi, Kaan Donbekci, Julia Hirschberg
Abstract: Emotion recognition in speech is a challenging multimodal task that requires understanding both verbal content and vocal nuances. This paper introduces a novel approach to emotion detection using Large Language Models (LLMs), which have demonstrated exceptional capabilities in natural language understanding. To overcome the inherent limitation of LLMs in processing audio inputs, we propose SpeechCueLLM, a method that translates speech characteristics into natural language descriptions, allowing LLMs to perform multimodal emotion analysis via text prompts without any architectural changes. Our method is minimal yet impactful, outperforming baseline models that require structural modifications. We evaluate SpeechCueLLM on two datasets: IEMOCAP and MELD, showing significant improvements in emotion recognition accuracy, particularly for high-quality audio data. We also explore the effectiveness of various feature representations and fine-tuning strategies for different LLMs. Our experiments demonstrate that incorporating speech descriptions yields a more than 2% increase in the average weighted F1 score on IEMOCAP (from 70.111% to 72.596%).
Authors: Ying Zhang, Dongyuan Li, Manabu Okumura
Abstract: Learning token embeddings based on token co-occurrence statistics has proven effective for both pre-training and fine-tuning in natural language processing. However, recent studies have pointed out that the distribution of learned embeddings degenerates into anisotropy (i.e., non-uniform distribution), and even pre-trained language models (PLMs) suffer from a loss of semantics-related information in embeddings for low-frequency tokens. This study first analyzes the fine-tuning dynamics of encoder-based PLMs and demonstrates their robustness against degeneration. On the basis of this analysis, we propose DefinitionEMB, a method that utilizes definitions to re-construct isotropically distributed and semantics-related token embeddings for encoder-based PLMs while maintaining original robustness during fine-tuning. Our experiments demonstrate the effectiveness of leveraging definitions from Wiktionary to re-construct such embeddings for two encoder-based PLMs: RoBERTa-base and BART-large. Furthermore, the re-constructed embeddings for low-frequency tokens improve the performance of these models across various GLUE and four text summarization datasets.
Authors: Samee Arif, Sualeha Farid, Abdul Hameed Azeemi, Awais Athar, Agha Ali Raza
Abstract: This paper presents a novel methodology for generating synthetic Preference Optimization (PO) datasets using multi-agent workflows. We evaluate the effectiveness and potential of these workflows in automating and enhancing the dataset generation process. PO dataset generation requires two modules: (1) response evaluation, and (2) response generation. In the response evaluation module, the responses from Large Language Models (LLMs) are evaluated and ranked - a task typically carried out by human annotators that we automate using LLMs. We assess the response evaluation module in a 2 step process. In step 1, we assess LLMs as evaluators using three distinct prompting strategies. In step 2, we apply the winning prompting strategy to compare the performance of LLM-as-a-Judge, LLMs-as-a-Jury, and LLM Debate. Our evaluation shows that GPT-4o-as-a-Judge is more consistent across all datasets. For the response generation module, we use the identified LLM evaluator configuration and compare different configurations of the LLM Feedback Loop. We use the win rate to determine the best multi-agent configuration for generation. Experimenting with various configurations, we find that the LLM Feedback Loop, with Llama as the generator and Gemma as the reviewer, achieves a notable 71.8% and 73.8% win rate over single-agent Llama and Gemma, respectively. After identifying the best configurations for both modules, we generate our PO datasets using the above pipeline.
Authors: Yujie Feng, Bo Liu, Xiaoyu Dong, Zexin Lu, Li-Ming Zhan, Albert Y. S. Lam, Xiao-Ming Wu
Abstract: An ideal dialogue system requires continuous skill acquisition and adaptation to new tasks while retaining prior knowledge. Dialogue State Tracking (DST), vital in these systems, often involves learning new services and confronting catastrophic forgetting, along with a critical capability loss termed the "Value Selection Quandary." To address these challenges, we introduce the Reason-of-Select (RoS) distillation method by enhancing smaller models with a novel 'meta-reasoning' capability. Meta-reasoning employs an enhanced multi-domain perspective, combining fragments of meta-knowledge from domain-specific dialogues during continual learning. This transcends traditional single-perspective reasoning. The domain bootstrapping process enhances the model's ability to dissect intricate dialogues from multiple possible values. Its domain-agnostic property aligns data distribution across different domains, effectively mitigating forgetting. Additionally, two novel improvements, "multi-value resolution" strategy and Semantic Contrastive Reasoning Selection method, significantly enhance RoS by generating DST-specific selection chains and mitigating hallucinations in teachers' reasoning, ensuring effective and reliable knowledge transfer. Extensive experiments validate the exceptional performance and robust generalization capabilities of our method. The source code is provided for reproducibility.
Authors: Andong Chen, Lianzhang Lou, Kehai Chen, Xuefeng Bai, Yang Xiang, Muyun Yang, Tiejun Zhao, Min Zhang
Abstract: Large language models (LLMs) have shown remarkable performance in general translation tasks. However, the increasing demand for high-quality translations that are not only adequate but also fluent and elegant. To assess the extent to which current LLMs can meet these demands, we introduce a suitable benchmark for translating classical Chinese poetry into English. This task requires not only adequacy in translating culturally and historically significant content but also a strict adherence to linguistic fluency and poetic elegance. Our study reveals that existing LLMs fall short of this task. To address these issues, we propose RAT, a \textbf{R}etrieval-\textbf{A}ugmented machine \textbf{T}ranslation method that enhances the translation process by incorporating knowledge related to classical poetry. Additionally, we propose an automatic evaluation metric based on GPT-4, which better assesses translation quality in terms of adequacy, fluency, and elegance, overcoming the limitations of traditional metrics. Our dataset and code will be made available.
Authors: Haau-Sing Li, Patrick Fernandes, Iryna Gurevych, Andr\'e F. T. Martins
Abstract: Recently, a diverse set of decoding and reranking procedures have been shown effective for LLM-based code generation. However, a comprehensive framework that links and experimentally compares these methods is missing. We address this by proposing Decoding Objectives for Code Execution, a comprehensive framework that includes candidate generation, $n$-best reranking, minimum Bayes risk (MBR) decoding, and self-debugging as the core components. We then study the contributions of these components through execution-based evaluation metrics. Our findings highlight the importance of execution-based methods and the difference gap between execution-based and execution-free methods. Furthermore, we assess the impact of filtering based on trial unit tests, a simple and effective strategy that has been often overlooked in prior works. We also propose self-debugging on multiple candidates, obtaining state-of-the-art performance on reranking for code generation. We expect our framework to provide a solid guideline for future research on code generation.
Authors: Hyeryun Park, Jeongwon Kwak, Dongsuk Jang, Sumin Park, Jinwook Choi
Abstract: Fact-checking techniques can mitigate hallucinations in Large Language Models (LLMs), a prominent issue in specialized domains. As parameter-efficient techniques such as Low-Rank Adaptation (LoRA) can overcome substantial computational overhead, some studies have explored the integration of multiple LoRAs. While previous studies focus on parallel integration, this paper investigates methods to establish connections among multiple LoRAs. We create three reasoning datasets tailored to fact-checking and fine-tune individual LoRAs, allowing them to view and reason from diverse perspectives. Then, we explore strategies for allocating these reasoning LoRAs and introduce LoraMap, an approach to map connections between them. The results of the fact-checking task demonstrate that the performance of LoraMap is superior to LoraHub, an existing method for integrating LoRAs. LoraMap also outperforms with significantly fewer trainable parameters than LoraConcat, which concatenates LoRAs and further fine-tunes them.
Authors: Saeid Asgari Taghanaki, Aliasgahr Khani, Amir Khasahmadi
Abstract: Existing benchmarks for large language models (LLMs) increasingly struggle to differentiate between top-performing models, underscoring the need for more challenging evaluation frameworks. We introduce MMLU-Pro+, an enhanced benchmark building upon MMLU-Pro to assess shortcut learning and higher-order reasoning in LLMs. By incorporating questions with multiple correct answers across diverse domains, MMLU-Pro+ tests LLMs' ability to engage in complex reasoning and resist simplistic problem-solving strategies. Our results show that MMLU-Pro+ maintains MMLU-Pro's difficulty while providing a more rigorous test of model discrimination, particularly in multi-correct answer scenarios. We introduce novel metrics like shortcut selection ratio and correct pair identification ratio, offering deeper insights into model behavior and anchoring bias. Evaluations of six state-of-the-art LLMs reveal significant performance gaps, highlighting variations in reasoning abilities and bias susceptibility. We release the dataset and evaluation codes at \url{https://github.com/asgsaeid/mmlu-pro-plus}.
Authors: Xiaoyun Jin, Mirjam Ernestus, R. Harald Baayen
Abstract: In Mandarin, the tonal contours of monosyllabic words produced in isolation or in careful speech are characterized by four lexical tones: a high-level tone (T1), a rising tone (T2), a dipping tone (T3) and a falling tone (T4). However, in spontaneous speech, the actual tonal realization of monosyllabic words can deviate significantly from these canonical tones due to intra-syllabic co-articulation and inter-syllabic co-articulation with adjacent tones. In addition, Chuang et al. (2024) recently reported that the tonal contours of disyllabic Mandarin words with T2-T4 tone pattern are co-determined by their meanings. Following up on their research, we present a corpus-based investigation of how the pitch contours of monosyllabic words are realized in spontaneous conversational Mandarin, focusing on the effects of contextual predictors on the one hand, and the way in words' meanings co-determine pitch contours on the other hand. We analyze the F0 contours of 3824 tokens of 63 different word types in a spontaneous Taiwan Mandarin corpus, using the generalized additive (mixed) model to decompose a given observed pitch contour into a set of component pitch contours. We show that the tonal context substantially modify a word's canonical tone. Once the effect of tonal context is controlled for, T2 and T3 emerge as low flat tones, contrasting with T1 as a high tone, and with T4 as a high-to-mid falling tone. The neutral tone (T0), which in standard descriptions, is realized based on the preceding tone, emerges as a low tone in its own right, modified by the other predictors in the same way as the standard tones T1, T2, T3, and T4. We also show that word, and even more so, word sense, co-determine words' F0 contours. Analyses of variable importance using random forests further supported the substantial effect of tonal context and an effect of word sense.
Authors: Jing Ma
Abstract: Causal inference has been a pivotal challenge across diverse domains such as medicine and economics, demanding a complicated integration of human knowledge, mathematical reasoning, and data mining capabilities. Recent advancements in natural language processing (NLP), particularly with the advent of large language models (LLMs), have introduced promising opportunities for traditional causal inference tasks. This paper reviews recent progress in applying LLMs to causal inference, encompassing various tasks spanning different levels of causation. We summarize the main causal problems and approaches, and present a comparison of their evaluation results in different causal scenarios. Furthermore, we discuss key findings and outline directions for future research, underscoring the potential implications of integrating LLMs in advancing causal inference methodologies.
Authors: Heegyu Kim, Taeyang Jeon, Seunghwan Choi, Seungtaek Choi, Hyunsouk Cho
Abstract: Text-to-SQL systems have become crucial for translating natural language into SQL queries in various industries, enabling non-technical users to perform complex data operations. The need for accurate evaluation methods has increased as these systems have grown more sophisticated. However, the Execution Accuracy (EX), the most prevalent evaluation metric, still shows many false positives and negatives. Thus, this paper introduces FLEX (False-Less EXecution), a novel approach to evaluating text-to-SQL systems using large language models (LLMs) to emulate human expert-level evaluation of SQL queries. Our metric improves agreement with human experts (from 62 to 87.04 in Cohen's kappa) with comprehensive context and sophisticated criteria. Our extensive experiments yield several key insights: (1) Models' performance increases by over 2.6 points on average, substantially affecting rankings on Spider and BIRD benchmarks; (2) The underestimation of models in EX primarily stems from annotation quality issues; and (3) Model performance on particularly challenging questions tends to be overestimated. This work contributes to a more accurate and nuanced evaluation of text-to-SQL systems, potentially reshaping our understanding of state-of-the-art performance in this field.
Authors: Ishani Mondal, Zongxia Li, Yufang Hou, Anandhavelu Natarajan, Aparna Garimella, Jordan Boyd-Graber
Abstract: Automating the creation of scientific diagrams from academic papers can significantly streamline the development of tutorials, presentations, and posters, thereby saving time and accelerating the process. Current text-to-image models struggle with generating accurate and visually appealing diagrams from long-context inputs. We propose SciDoc2Diagram, a task that extracts relevant information from scientific papers and generates diagrams, along with a benchmarking dataset, SciDoc2DiagramBench. We develop a multi-step pipeline SciDoc2Diagrammer that generates diagrams based on user intentions using intermediate code generation. We observed that initial diagram drafts were often incomplete or unfaithful to the source, leading us to develop SciDoc2Diagrammer-Multi-Aspect-Feedback (MAF), a refinement strategy that significantly enhances factual correctness and visual appeal and outperforms existing models on both automatic and human judgement.
Authors: Linus Tze En Foo, Lynnette Hui Xian Ng
Abstract: Singlish, or formally Colloquial Singapore English, is an English-based creole language originating from the SouthEast Asian country Singapore. The language contains influences from Sinitic languages such as Chinese dialects, Malay, Tamil and so forth. A fundamental task to understanding Singlish is to first understand the pragmatic functions of its discourse particles, upon which Singlish relies heavily to convey meaning. This work offers a preliminary effort to disentangle the Singlish discourse particles (lah, meh and hor) with task-driven representation learning. After disentanglement, we cluster these discourse particles to differentiate their pragmatic functions, and perform Singlish-to-English machine translation. Our work provides a computational method to understanding Singlish discourse particles, and opens avenues towards a deeper comprehension of the language and its usage.
Authors: Nirmalya Thakur
Abstract: The work presented in this paper makes three scientific contributions with a specific focus on mining and analysis of COVID-19-related posts on Instagram. First, it presents a multilingual dataset of 500,153 Instagram posts about COVID-19 published between January 2020 and September 2024. This dataset, available at https://dx.doi.org/10.21227/d46p-v480, contains Instagram posts in 161 different languages as well as 535,021 distinct hashtags. After the development of this dataset, multilingual sentiment analysis was performed, which involved classifying each post as positive, negative, or neutral. The results of sentiment analysis are presented as a separate attribute in this dataset. Second, it presents the results of performing sentiment analysis per year from 2020 to 2024. The findings revealed the trends in sentiment related to COVID-19 on Instagram since the beginning of the pandemic. For instance, between 2020 and 2024, the sentiment trends show a notable shift, with positive sentiment decreasing from 38.35% to 28.69%, while neutral sentiment rising from 44.19% to 58.34%. Finally, the paper also presents findings of language-specific sentiment analysis. This analysis highlighted similar and contrasting trends of sentiment across posts published in different languages on Instagram. For instance, out of all English posts, 49.68% were positive, 14.84% were negative, and 35.48% were neutral. In contrast, among Hindi posts, 4.40% were positive, 57.04% were negative, and 38.56% were neutral, reflecting distinct differences in the sentiment distribution between these two languages.
Authors: Zhuochun Li, Yuelyu Ji, Rui Meng, Daqing He
Abstract: While reasoning capabilities typically emerge in large language models (LLMs) with tens of billions of parameters, recent research focuses on improving smaller open-source models through knowledge distillation (KD) from commercial LLMs. However, many of these studies rely solely on responses from a single LLM as the gold rationale, unlike the natural human learning process, which involves understanding both the correct answers and the reasons behind mistakes. In this paper, we introduce a novel Fault-Aware Distillation via Peer-Review (FAIR) approach: 1) Instead of merely obtaining gold rationales from teachers, our method asks teachers to identify and explain the student's mistakes, providing customized instruction learning data. 2) We design a simulated peer-review process between teacher LLMs, which selects only the generated rationales above the acceptance threshold. This reduces the chance of teachers guessing correctly with flawed rationale, improving instructional data quality. Comprehensive experiments and analysis on mathematical, commonsense, and logical reasoning tasks demonstrate the effectiveness of our method.
Authors: Ruizhe Chen, Xiaotian Zhang, Meng Luo, Wenhao Chai, Zuozhu Liu
Abstract: Aligning with personalized preferences, which vary significantly across cultural, educational, and political differences, poses a significant challenge due to the computational costs and data demands of traditional alignment methods. In response, this paper presents Personalized Alignment at Decoding-time (PAD), a novel framework designed to align LLM outputs with diverse personalized preferences during the inference phase, eliminating the need for additional training. By introducing a unique personalized reward modeling strategy, this framework decouples the text generation process from personalized preferences, facilitating the generation of generalizable token-level personalized rewards. The PAD algorithm leverages these rewards to guide the decoding process, dynamically tailoring the base model's predictions to personalized preferences. Extensive experimental results demonstrate that PAD not only outperforms existing training-based alignment methods in terms of aligning with diverse preferences but also shows significant generalizability to preferences unseen during training and scalability across different base models. This work advances the capability of LLMs to meet user needs in real-time applications, presenting a substantial step forward in personalized LLM alignment.
Authors: F{\i}rat \"Oncel, Matthias Bethge, Beyza Ermis, Mirco Ravanelli, Cem Subakan, \c{C}a\u{g}atay Y{\i}ld{\i}z
Abstract: In the last decade, the generalization and adaptation abilities of deep learning models were typically evaluated on fixed training and test distributions. Contrary to traditional deep learning, large language models (LLMs) are (i) even more overparameterized, (ii) trained on unlabeled text corpora curated from the Internet with minimal human intervention, and (iii) trained in an online fashion. These stark contrasts prevent researchers from transferring lessons learned on model generalization and adaptation in deep learning contexts to LLMs. To this end, our short paper introduces empirical observations that aim to shed light on further training of already pretrained language models. Specifically, we demonstrate that training a model on a text domain could degrade its perplexity on the test portion of the same domain. We observe with our subsequent analysis that the performance degradation is positively correlated with the similarity between the additional and the original pretraining dataset of the LLM. Our further token-level perplexity observations reveals that the perplexity degradation is due to a handful of tokens that are not informative about the domain. We hope these findings will guide us in determining when to adapt a model vs when to rely on its foundational capabilities.
Authors: Tim Schopf, Alexander Blatzheim, Nektarios Machner, Florian Matthes
Abstract: Scientific document classification is a critical task and often involves many classes. However, collecting human-labeled data for many classes is expensive and usually leads to label-scarce scenarios. Moreover, recent work has shown that sentence embedding model fine-tuning for few-shot classification is efficient, robust, and effective. In this work, we propose FusionSent (Fusion-based Sentence Embedding Fine-tuning), an efficient and prompt-free approach for few-shot classification of scientific documents with many classes. FusionSent uses available training examples and their respective label texts to contrastively fine-tune two different sentence embedding models. Afterward, the parameters of both fine-tuned models are fused to combine the complementary knowledge from the separate fine-tuning steps into a single model. Finally, the resulting sentence embedding model is frozen to embed the training instances, which are then used as input features to train a classification head. Our experiments show that FusionSent significantly outperforms strong baselines by an average of $6.0$ $F_{1}$ points across multiple scientific document classification datasets. In addition, we introduce a new dataset for multi-label classification of scientific documents, which contains 203,961 scientific articles and 130 classes from the arXiv category taxonomy. Code and data are available at https://github.com/sebischair/FusionSent.
Authors: Hemank Lamba, Anton Abilov, Ke Zhang, Elizabeth M. Olson, Henry k. Dambanemuya, Jo\~ao c. B\'arcia, David S. Batista, Christina Wille, Aoife Cahill, Joel Tetreault, Alex Jaimes
Abstract: Humanitarian organizations can enhance their effectiveness by analyzing data to discover trends, gather aggregated insights, manage their security risks, support decision-making, and inform advocacy and funding proposals. However, data about violent incidents with direct impact and relevance for humanitarian aid operations is not readily available. An automatic data collection and NLP-backed classification framework aligned with humanitarian perspectives can help bridge this gap. In this paper, we present HumVI - a dataset comprising news articles in three languages (English, French, Arabic) containing instances of different types of violent incidents categorized by the humanitarian sector they impact, e.g., aid security, education, food security, health, and protection. Reliable labels were obtained for the dataset by partnering with a data-backed humanitarian organization, Insecurity Insight. We provide multiple benchmarks for the dataset, employing various deep learning architectures and techniques, including data augmentation and mask loss, to address different task-related challenges, e.g., domain expansion. The dataset is publicly available at https://github.com/dataminr-ai/humvi-dataset.
Authors: Yanjun Chen, Dawei Zhu, Yirong Sun, Xinghao Chen, Wei Zhang, Xiaoyu Shen
Abstract: Reinforcement Learning from Human Feedback significantly enhances Natural Language Processing by aligning language models with human expectations. A critical factor in this alignment is the strength of reward models used during training. This study explores whether stronger reward models invariably lead to better language models. In this paper, through experiments on relevance, factuality, and completeness tasks using the QA-FEEDBACK dataset and reward models based on Longformer, we uncover a surprising paradox: language models trained with moderately accurate reward models outperform those guided by highly accurate ones. This challenges the widely held belief that stronger reward models always lead to better language models, and opens up new avenues for future research into the key factors driving model performance and how to choose the most suitable reward models. Code and additional details are available at https://github.com/EIT-NLP/AccuracyParadox-RLHF.
Authors: Gian Maria Campedelli, Nicol\`o Penzo, Massimo Stefan, Roberto Dess\`i, Marco Guerini, Bruno Lepri, Jacopo Staiano
Abstract: As Large Language Model (LLM)-based agents become increasingly autonomous and will more freely interact with each other, studying interactions between them becomes crucial to anticipate emergent phenomena and potential risks. Drawing inspiration from the widely popular Stanford Prison Experiment, we contribute to this line of research by studying interaction patterns of LLM agents in a context characterized by strict social hierarchy. We do so by specifically studying two types of phenomena: persuasion and anti-social behavior in simulated scenarios involving a guard and a prisoner agent who seeks to achieve a specific goal (i.e., obtaining additional yard time or escape from prison). Leveraging 200 experimental scenarios for a total of 2,000 machine-machine conversations across five different popular LLMs, we provide a set of noteworthy findings. We first document how some models consistently fail in carrying out a conversation in our multi-agent setup where power dynamics are at play. Then, for the models that were able to engage in successful interactions, we empirically show how the goal that an agent is set to achieve impacts primarily its persuasiveness, while having a negligible effect with respect to the agent's anti-social behavior. Third, we highlight how agents' personas, and particularly the guard's personality, drive both the likelihood of successful persuasion from the prisoner and the emergence of anti-social behaviors. Fourth, we show that even without explicitly prompting for specific personalities, anti-social behavior emerges by simply assigning agents' roles. These results bear implications for the development of interactive LLM agents as well as the debate on their societal impact.
Authors: Gleb Kuzmin, Petr Strepetov, Maksim Stankevich, Artem Shelmanov, Ivan Smirnov
Abstract: This paper compares the effectiveness of traditional machine learning methods, encoder-based models, and large language models (LLMs) on the task of detecting depression and anxiety. Five datasets were considered, each differing in format and the method used to define the target pathology class. We tested AutoML models based on linguistic features, several variations of encoder-based Transformers such as BERT, and state-of-the-art LLMs as pathology classification models. The results demonstrated that LLMs outperform traditional methods, particularly on noisy and small datasets where training examples vary significantly in text length and genre. However, psycholinguistic features and encoder-based models can achieve performance comparable to language models when trained on texts from individuals with clinically confirmed depression, highlighting their potential effectiveness in targeted clinical applications.
Authors: Ziyue Li, Tianyi Zhou
Abstract: While large language models (LLMs) excel on generation tasks, their decoder-only architecture often limits their potential as embedding models if no further representation finetuning is applied. Does this contradict their claim of generalists? To answer the question, we take a closer look at Mixture-of-Experts (MoE) LLMs. Our study shows that the expert routers in MoE LLMs can serve as an off-the-shelf embedding model with promising performance on a diverse class of embedding-focused tasks, without requiring any finetuning. Moreover, our extensive analysis shows that the MoE routing weights (RW) is complementary to the hidden state (HS) of LLMs, a widely-used embedding. Compared to HS, we find that RW is more robust to the choice of prompts and focuses on high-level semantics. Motivated by the analysis, we propose MoEE combining RW and HS, which achieves better performance than using either separately. Our exploration of their combination and prompting strategy shed several novel insights, e.g., a weighted sum of RW and HS similarities outperforms the similarity on their concatenation. Our experiments are conducted on 6 embedding tasks with 20 datasets from the Massive Text Embedding Benchmark (MTEB). The results demonstrate the significant improvement brought by MoEE to LLM-based embedding without further finetuning.
Authors: Chinmay Dandekar (University of California, Santa Barbara), Wenda Xu (University of California, Santa Barbara), Xi Xu (Carnegie Mellon University), Siqi Ouyang (Carnegie Mellon University), Lei Li (Carnegie Mellon University)
Abstract: With the rapid advancement of machine translation research, evaluation toolkits have become essential for benchmarking system progress. Tools like COMET and SacreBLEU offer single quality score assessments that are effective for pairwise system comparisons. However, these tools provide limited insights for fine-grained system-level comparisons and the analysis of instance-level defects. To address these limitations, we introduce Translation Canvas, an explainable interface designed to pinpoint and analyze translation systems' performance: 1) Translation Canvas assists machine translation researchers in comprehending system-level model performance by identifying common errors (their frequency and severity) and analyzing relationships between different systems based on various evaluation metrics. 2) It supports fine-grained analysis by highlighting error spans with explanations and selectively displaying systems' predictions. According to human evaluation, Translation Canvas demonstrates superior performance over COMET and SacreBLEU packages under enjoyability and understandability criteria.
Authors: Hemant Yadav, Rajiv Ratn Shah, Sunayana Sitaram
Abstract: Information in speech can be divided into two categories: what is being said (content) and how it is expressed (other). Current state-of-the-art (SOTA) techniques model speech at fixed segments, usually 10-25 ms, using a single embedding. Given the orthogonal nature of other and content information, attempting to optimize both within a single embedding results in suboptimal solutions. This approach divides the models capacity, limiting its ability to build complex hierarchical features effectively. In this work, we present an end-to-end speech representation learning framework designed to jointly optimize the other and content information (JOOCI) in speech. By using separate learnable parameters, JOOCI addresses this optimization challenge by modeling other and content information independently. Our results show that JOOCI consistently outperforms other SOTA models of similar size (100 million parameters) and pre-training data used (960 hours) by a significant margin when evaluated on a range of speech downstream tasks in the SUPERB benchmark, as shown in Table 1.
Authors: S. Tamang, D. J. Bora
Abstract: This paper introduces a centralized, open-source dataset repository designed to advance NLP and NMT for Assamese, a low-resource language. The repository, available at GitHub, supports various tasks like sentiment analysis, named entity recognition, and machine translation by providing both pre-training and fine-tuning corpora. We review existing datasets, highlighting the need for standardized resources in Assamese NLP, and discuss potential applications in AI-driven research, such as LLMs, OCR, and chatbots. While promising, challenges like data scarcity and linguistic diversity remain. The repository aims to foster collaboration and innovation, promoting Assamese language research in the digital age.
Authors: Seung-Woo Choi, Ga-Hyun Yoo, Jay-Yoon Lee
Abstract: Recent Large Language Models (LLMs) have demonstrated strong performance in translation without needing to be finetuned on additional parallel corpora. However, they still underperform for low-resource language pairs. Previous works have focused on mitigating this issue by leveraging relevant few-shot examples or external resources such as dictionaries or grammar books, making models heavily reliant on these nonparametric sources of information. In this paper, we propose a novel method named IntGrad MT that focuses on fully exploiting an LLM's inherent translation capability. IntGrad MT achieves this by constructing a chain of few-shot examples, each consisting of a source sentence and the model's own translation, that rise incrementally in difficulty. IntGrad MT employs two techniques: Sentence Interpolation, which generates a sequence of sentences that gradually change from an easy sentence to translate to a difficult one, and Gradual MT, which sequentially translates this chain using translations of earlier sentences as few-shot examples for the translation of subsequent ones. With this approach, we observe a substantial enhancement in the xCOMET scores of various LLMs for multiple languages, especially in low-resource languages such as Hindi(8.26), Swahili(7.10), Bengali(6.97) and Marathi(13.03). Our approach presents a practical way of enhancing LLMs' performance without extra training.
Authors: Jean Fromentin (LMPA)
Abstract: Defined on Birman-Ko-Lee monoids, the rotating normal form has strong connections with the Dehornoy's braid ordering. It can be seen as a process for selecting between all the representative words of a Birman-Ko-Lee braid a particular one, called rotating word. In this paper we construct, for all n 2, a finite-state automaton which recognizes rotating words on n strands, proving that the rotating normal form is regular. As a consequence we obtain the regularity of a $\sigma$-definite normal form defined on the whole braid group.
Authors: Florinel-Alin Croitoru, Vlad Hondru, Radu Tudor Ionescu, Mubarak Shah
Abstract: Text-to-image diffusion models have recently attracted the interest of many researchers, and inverting the diffusion process can play an important role in better understanding the generative process and how to engineer prompts in order to obtain the desired images. To this end, we study the task of predicting the prompt embedding given an image generated by a generative diffusion model. We consider a series of white-box and black-box models (with and without access to the weights of the diffusion network) to deal with the proposed task. We propose a novel learning framework comprising a joint prompt regression and multi-label vocabulary classification objective that generates improved prompts. To further improve our method, we employ a curriculum learning procedure that promotes the learning of image-prompt pairs with lower labeling noise (i.e. that are better aligned). We conduct experiments on the DiffusionDB data set, predicting text prompts from images generated by Stable Diffusion. In addition, we make an interesting discovery: training a diffusion model on the prompt generation task can make the model generate images that are much better aligned with the input prompts, when the model is directly reused for text-to-image generation. Our code is publicly available for download at https://github.com/CroitoruAlin/Reverse-Stable-Diffusion.
URLs: https://github.com/CroitoruAlin/Reverse-Stable-Diffusion.
Authors: Sunjae Lee, Junyoung Choi, Jungjae Lee, Munim Hasan Wasi, Hojun Choi, Steven Y. Ko, Sangeun Oh, Insik Shin
Abstract: The advent of large language models (LLMs) has opened up new opportunities in the field of mobile task automation. Their superior language understanding and reasoning capabilities allow users to automate complex and repetitive tasks. However, due to the inherent unreliability and high operational cost of LLMs, their practical applicability is quite limited. To address these issues, this paper introduces MobileGPT, an innovative LLM-based mobile task automator equipped with a human-like app memory. MobileGPT emulates the cognitive process of humans interacting with a mobile app -- explore, select, derive, and recall. This approach allows for a more precise and efficient learning of a task's procedure by breaking it down into smaller, modular sub-tasks that can be re-used, re-arranged, and adapted for various objectives. We implement MobileGPT using online LLMs services (GPT-3.5 and GPT-4) and evaluate its performance on a dataset of 185 tasks across 18 mobile apps. The results indicate that MobileGPT can automate and learn new tasks with 82.7% accuracy, and is able to adapt them to different contexts with near perfect (98.75%) accuracy while reducing both latency and cost by 62.5% and 68.8%, respectively, compared to the GPT-4 powered baseline.
Authors: Young-Min Cho, Dandan Pang, Stuti Thapa, Garrick Sherman, Lyle Ungar, Louis Tay, Sharath Chandra Guntuku
Abstract: While affective expressions on social media have been extensively studied, most research has focused on the Western context. This paper explores cultural differences in affective expressions by comparing valence and arousal on Twitter/X (geolocated to the US) and Sina Weibo (in Mainland China). Using the NRC-VAD lexicon to measure valence and arousal, we identify distinct patterns of emotional expression across both platforms. Our analysis reveals a functional representation between valence and arousal, showing a negative offset in contrast to traditional lab-based findings which suggest a positive offset. Furthermore, we uncover significant cross-cultural differences in arousal, with US users displaying higher emotional intensity than Chinese users, regardless of the valence of the content. Finally, we conduct a comprehensive language analysis correlating n-grams and LDA topics with affective dimensions to deepen our understanding of how language and culture shape emotional expression. These findings contribute to a more nuanced understanding of affective communication across cultural and linguistic contexts on social media.
Authors: Yihong Tang, Zhaokai Wang, Ao Qu, Yihao Yan, Zhaofeng Wu, Dingyi Zhuang, Jushi Kai, Kebing Hou, Xiaotong Guo, Jinhua Zhao, Zhan Zhao, Wei Ma
Abstract: Citywalk, a recently popular form of urban travel, requires genuine personalization and understanding of fine-grained requests compared to traditional itinerary planning. In this paper, we introduce the novel task of Open-domain Urban Itinerary Planning (OUIP), which generates personalized urban itineraries from user requests in natural language. We then present ITINERA, an OUIP system that integrates spatial optimization with large language models to provide customized urban itineraries based on user needs. This involves decomposing user requests, selecting candidate points of interest (POIs), ordering the POIs based on cluster-aware spatial optimization, and generating the itinerary. Experiments on real-world datasets and the performance of the deployed system demonstrate our system's capacity to deliver personalized and spatially coherent itineraries compared to current solutions. Source codes of ITINERA are available at https://github.com/YihongT/ITINERA.
Authors: Long Li, Xuzheng He, Haozhe Wang, Linlin Wang, Liang He
Abstract: Program-of-Thought (PoT) replaces natural language-based Chain-of-Thought (CoT) as the most popular method in Large Language Models (LLMs) mathematical reasoning tasks by utilizing external tool calls to circumvent computational errors. However, our evaluation of the GPT-4 and Llama series reveals that using PoT introduces more reasoning errors, such as incorrect formulas or flawed logic, compared to CoT. To address this issue, we propose Human-Think Language (HTL), which leverages a suite of strategies that help integrate PoT and CoT, encompassing: (1) a new generation paradigm that uses full CoT reasoning to control code generation. (2) Focus Attention, that directs model attention to the CoT reasoning during PoT to generate more logical code. (3) reinforcement learning that utilizes the accuracy of both CoT and PoT responses as rewards to prevent repetitive reasoning steps in LLMs when solving difficult math problems. Our method achieves an average improvement of 6.5% on the Llama-Base model and 4.3% on the Mistral-Base model across 8 mathematical calculation datasets. It also shows significant effectiveness on five out-of-domain datasets by controlling the model's information flow, exhibiting strong transferability. Additionally, HTL shows the most significant improvement in non-mathematical natural language inference task, contributing to a unified reasoning task framework
Authors: Yilun Hao, Yongchao Chen, Yang Zhang, Chuchu Fan
Abstract: Despite their recent advancements, Large Language Models (LLMs) still struggle to directly generate correct plans for complex multi-constraint planning problems, even with self-verification and self-critique. For example, a U.S. domestic travel planning benchmark TravelPlanner was proposed in Xie et al. (2024), where the best LLM OpenAI o1-preview can only find travel plans that satisfy user requirements with a 10% success rate given all needed information. In this work, we tackle this difficult problem by proposing an LLM-based planning framework that formalizes and solves complex multi-constraint planning problems as constrained satisfiability problems, which are further consumed by sound and complete satisfiability solvers. We start with TravelPlanner as the primary use case and achieve a success rate of 93.9%. We demonstrate our framework's robustness by showing its effectiveness in diverse paraphrased prompts. More importantly, our framework has strong zero-shot generalizability: It can successfully handle unseen constraints in a completely unseen international travel dataset we created, and it can even generalize well to new domains such as routing and task allocation problems in a zero-shot manner. Moreover, when user input queries are infeasible, our framework can identify the unsatisfiable core, provide failure reasons, and offers personalized modification suggestions to users according to diverse human preferences. We show that our framework can modify and solve for an average of 81.6% and 91.7% unsatisfiable queries from two datasets and prove with ablations that all key components of our framework are effective and necessary.
Authors: Leyi Pan, Aiwei Liu, Zhiwei He, Zitian Gao, Xuandong Zhao, Yijian Lu, Binglin Zhou, Shuliang Liu, Xuming Hu, Lijie Wen, Irwin King, Philip S. Yu
Abstract: LLM watermarking, which embeds imperceptible yet algorithmically detectable signals in model outputs to identify LLM-generated text, has become crucial in mitigating the potential misuse of large language models. However, the abundance of LLM watermarking algorithms, their intricate mechanisms, and the complex evaluation procedures and perspectives pose challenges for researchers and the community to easily experiment with, understand, and assess the latest advancements. To address these issues, we introduce MarkLLM, an open-source toolkit for LLM watermarking. MarkLLM offers a unified and extensible framework for implementing LLM watermarking algorithms, while providing user-friendly interfaces to ensure ease of access. Furthermore, it enhances understanding by supporting automatic visualization of the underlying mechanisms of these algorithms. For evaluation, MarkLLM offers a comprehensive suite of 12 tools spanning three perspectives, along with two types of automated evaluation pipelines. Through MarkLLM, we aim to support researchers while improving the comprehension and involvement of the general public in LLM watermarking technology, fostering consensus and driving further advancements in research and application. Our code is available at https://github.com/THU-BPM/MarkLLM.
Authors: Govind Ramesh, Yao Dou, Wei Xu
Abstract: Research on jailbreaking has been valuable for testing and understanding the safety and security issues of large language models (LLMs). In this paper, we introduce Iterative Refinement Induced Self-Jailbreak (IRIS), a novel approach that leverages the reflective capabilities of LLMs for jailbreaking with only black-box access. Unlike previous methods, IRIS simplifies the jailbreaking process by using a single model as both the attacker and target. This method first iteratively refines adversarial prompts through self-explanation, which is crucial for ensuring that even well-aligned LLMs obey adversarial instructions. IRIS then rates and enhances the output given the refined prompt to increase its harmfulness. We find that IRIS achieves jailbreak success rates of 98% on GPT-4, 92% on GPT-4 Turbo, and 94% on Llama-3.1-70B in under 7 queries. It significantly outperforms prior approaches in automatic, black-box, and interpretable jailbreaking, while requiring substantially fewer queries, thereby establishing a new standard for interpretable jailbreaking methods.
Authors: Bowen Jiang, Yangxinyu Xie, Xiaomeng Wang, Yuan Yuan, Zhuoqun Hao, Xinyi Bai, Weijie J. Su, Camillo J. Taylor, Tanwi Mallick
Abstract: Rationality is the quality of being guided by reason, characterized by decision-making that aligns with evidence and logical principles. It plays a crucial role in reliable problem-solving by ensuring well-grounded and consistent solutions. While large language models (LLMs) have made significant progress in generating human-like text, they still exhibit limitations such as bounded knowledge space and inconsistent outputs. In response, recent efforts have shifted toward developing multimodal and multi-agent systems, as well as integrating modules like external tools, programming codes, symbolic reasoners, utility function, and conformal risk controls rather than relying solely on a single LLM for decision-making. This paper surveys the state-of-the-art advancements in language and multimodal agents, evaluates how they contribute to make intelligent agents more rational, and identifies open challenges and future research directions. We maintain an open repository at https://github.com/bowen-upenn/Agent_Rationality.
Authors: Tongtong Wu, Weigang Wu, Xingyu Wang, Kang Xu, Suyu Ma, Bo Jiang, Ping Yang, Zhenchang Xing, Yuan-Fang Li, Gholamreza Haffari
Abstract: Large Language Models (LLMs) have made tremendous strides in code generation, but existing research fails to account for the dynamic nature of software development, marked by frequent library updates. This gap significantly limits LLMs' deployment in realistic settings. In this paper, we propose two novel tasks aimed at bridging this gap: version-specific code completion (VSCC) and version-aware code migration (VACM). In conjunction, we introduce VersiCode, a comprehensive Python dataset specifically designed to evaluate LLMs on these two tasks, together with a novel evaluation metric, Critical Diff Check (CDC@1), which assesses code generation against evolving API requirements. We conduct an extensive evaluation on VersiCode, which reveals that version-controllable code generation is indeed a significant challenge, even for GPT-4o and other strong frontier models. We believe the novel tasks, dataset, and metric open up a new, important research direction that will further enhance LLMs' real-world applicability. The code and resources can be found at https://github.com/wutong8023/VersiCode.
Authors: Xiaowen Sun, Xufeng Zhao, Jae Hee Lee, Wenhao Lu, Matthias Kerzel, Stefan Wermter
Abstract: The state of an object reflects its current status or condition and is important for a robot's task planning and manipulation. However, detecting an object's state and generating a state-sensitive plan for robots is challenging. Recently, pre-trained Large Language Models (LLMs) and Vision-Language Models (VLMs) have shown impressive capabilities in generating plans. However, to the best of our knowledge, there is hardly any investigation on whether LLMs or VLMs can also generate object state-sensitive plans. To study this, we introduce an Object State-Sensitive Agent (OSSA), a task-planning agent empowered by pre-trained neural networks. We propose two methods for OSSA: (i) a modular model consisting of a pre-trained vision processing module (dense captioning model, DCM) and a natural language processing model (LLM), and (ii) a monolithic model consisting only of a VLM. To quantitatively evaluate the performances of the two methods, we use tabletop scenarios where the task is to clear the table. We contribute a multimodal benchmark dataset that takes object states into consideration. Our results show that both methods can be used for object state-sensitive tasks, but the monolithic approach outperforms the modular approach. The code for OSSA is available at https://github.com/Xiao-wen-Sun/OSSA
Authors: Jiuding Yang, Weidong Guo, Kaitong Yang, Xiangyang Li, Yu Xu, Di Niu
Abstract: The effective alignment of Large Language Models (LLMs) with precise instructions is essential for their application in diverse real-world scenarios. Current methods focus on enhancing the diversity and complexity of training and evaluation samples, yet they fall short in accurately assessing LLMs' ability to follow similar instruction variants. We introduce an effective data augmentation technique DeMoRecon that decomposes complex instructions into simpler sub-components, modifies these, and reconstructs them into new variants, thereby preserves the original instruction's context and complexity while introducing variability, which is critical for training and evaluating LLMs' instruction-following precision. Based on DeMoRecon, we developed the FGIV dataset which contains fine-grained instruction variants of 1,773 seed instructions to both fine-tune and evaluate LLMs. Our findings show that LLMs fine-tuned with FGIV will gain significant performance boost on both ours and commonly used instructions-following benchmarks.
Authors: Tuan-Luc Huynh, Thuy-Trang Vu, Weiqing Wang, Yinwei Wei, Trung Le, Dragan Gasevic, Yuan-Fang Li, Thanh-Toan Do
Abstract: Differentiable Search Index (DSI) utilizes Pre-trained Language Models (PLMs) for efficient document retrieval without relying on external indexes. However, DSI needs full re-training to handle updates in dynamic corpora, causing significant computational inefficiencies. We introduce PromptDSI, a prompt-based rehearsal-free approach for instance-wise incremental learning document retrieval. PromptDSI attaches prompts to the frozen PLM's encoder of DSI, leveraging its powerful representation to efficiently index new corpora while maintaining a balance between stability and plasticity. We eliminate the initial forward pass of prompt-based continual learning methods that doubles training and inference time. Moreover, we propose a topic-aware prompt pool that employs neural topic embeddings as fixed keys. This strategy ensures diverse and effective prompt usage, addressing the challenge of parameter underutilization caused by the collapse of the query-key matching mechanism. Our empirical evaluations demonstrate that BERT-based PromptDSI matches IncDSI in managing forgetting while improving new corpora performance by more than 4% Hits@10 on NQ320k and upto 3% MRR@10 on MS MARCO 300k.
Authors: Yingyu Liang, Zhenmei Shi, Zhao Song, Chiwun Yang
Abstract: Prompting and context-based fine-tuning methods, which we call Prefix Learning, have been proposed to enhance the performance of language models on various downstream tasks. They are empirically efficient and effective, matching the performance of full parameter fine-tuning, but the theoretical understandings are limited. In this paper, we aim to address this limitation by studying their ability from the perspective of prefix length. In particular, we provide a convergence guarantee for training an ultra-long prefix in a stylized setting using the Neural Tangent Kernel (NTK) framework. Based on this strong theoretical guarantee, we design and implement an algorithm that only needs to introduce and fine-tune a few extra trainable parameters instead of an infinite-long prefix in each layer of a transformer, and can approximate the prefix attention to a guaranteed polynomial-small error. Preliminary experimental results on vision, natural language, and math data show that our method achieves superior or competitive performance compared to existing methods like full parameters fine-tuning, P-Tuning V2, and LoRA. This demonstrates our method is promising for parameter-efficient fine-tuning. Our code can be found at \url{https://github.com/ChristianYang37/chiwun/tree/main/src/NTK-Attention}.
URLs: https://github.com/ChristianYang37/chiwun/tree/main/src/NTK-Attention
Authors: Jingwei Ni, Tobias Schimanski, Meihong Lin, Mrinmaya Sachan, Elliott Ash, Markus Leippold
Abstract: Retrieval Augmented Generation (RAG) is widely employed to ground responses to queries on domain-specific documents. But do RAG implementations leave out important information when answering queries that need an integrated analysis of information (e.g., Tell me good news in the stock market today.)? To address these concerns, RAG developers need to annotate information retrieval (IR) data for their domain of interest, which is challenging because (1) domain-specific queries usually need nuanced definitions of relevance beyond shallow semantic relevance; and (2) human or GPT-4 annotation is costly and cannot cover all (query, document) pairs (i.e., annotation selection bias), thus harming the effectiveness in evaluating IR recall. To address these challenges, we propose DIRAS (Domain-specific Information Retrieval Annotation with Scalability), a manual-annotation-free schema that fine-tunes open-sourced LLMs to consider nuanced relevance definition and annotate (partial) relevance labels with calibrated relevance scores. Extensive evaluation shows that DIRAS enables smaller (8B) LLMs to achieve GPT-4-level performance on annotating and ranking unseen (query, document) pairs, and is helpful for real-world RAG development. All code, LLM generations, and human annotations can be found in \url{https://github.com/EdisonNi-hku/DIRAS}.
Authors: Kyle Wong, Alfonso Amayuelas, Liangming Pan, William Yang Wang
Abstract: Large language models (LLMs) have shown remarkable performance on code generation tasks. A recent use case is iterative code repair, where an LLM fixes an incorrect program by rationalizing about errors and generating new code. Recent works augment the code repair process by integrating modern techniques such as chain-of-thought reasoning or distillation, but only study their benefits on high-resource languages like Python, and ignore low-resource languages like Perl. To address this gap of knowledge, we investigate the benefits of distilling code repair for both high and low resource languages to determine if the techniques that are effective in a high resource setting are also applicable in a low resource setting. Our evaluation shows that distilling the ability to repair code has language dependent benefits. To explain this behavior, we perform a further analysis and find that contrary to preexisting beliefs, the correlation between reasoning ability and code correction ability is weak. We hypothesize this weak correlation is magnified in low-resource settings where base models lack deep knowledge of a programming language, leading to wavering benefits of code repair.
Authors: Jiaming Ji, Donghai Hong, Borong Zhang, Boyuan Chen, Josef Dai, Boren Zheng, Tianyi Qiu, Boxun Li, Yaodong Yang
Abstract: In this work, we introduce the PKU-SafeRLHF dataset, designed to promote research on safety alignment in large language models (LLMs). As a sibling project to SafeRLHF and BeaverTails, we separate annotations of helpfulness and harmlessness for question-answering pairs, providing distinct perspectives on these coupled attributes. Overall, we provide 44.6k refined prompts and 265k question-answer pairs with safety meta-labels for 19 harm categories and three severity levels ranging from minor to severe, with answers generated by Llama-family models. Based on this, we collected 166.8k preference data, including dual-preference (helpfulness and harmlessness decoupled) and single-preference data (trade-off the helpfulness and harmlessness from scratch), respectively. Using the large-scale annotation data, we further train severity-sensitive moderation for the risk control of LLMs and safety-centric RLHF algorithms for the safety alignment of LLMs. We believe this dataset will be a valuable resource for the community, aiding in the safe deployment of LLMs.
Authors: Mirco Ravanelli, Titouan Parcollet, Adel Moumen, Sylvain de Langen, Cem Subakan, Peter Plantinga, Yingzhi Wang, Pooneh Mousavi, Luca Della Libera, Artem Ploujnikov, Francesco Paissan, Davide Borra, Salah Zaiem, Zeyu Zhao, Shucong Zhang, Georgios Karakasidis, Sung-Lin Yeh, Pierre Champion, Aku Rouhe, Rudolf Braun, Florian Mai, Juan Zuluaga-Gomez, Seyed Mahed Mousavi, Andreas Nautsch, Xuechen Liu, Sangeet Sagar, Jarod Duret, Salima Mdhaffar, Gaelle Laperriere, Mickael Rouvier, Renato De Mori, Yannick Esteve
Abstract: SpeechBrain is an open-source Conversational AI toolkit based on PyTorch, focused particularly on speech processing tasks such as speech recognition, speech enhancement, speaker recognition, text-to-speech, and much more. It promotes transparency and replicability by releasing both the pre-trained models and the complete "recipes" of code and algorithms required for training them. This paper presents SpeechBrain 1.0, a significant milestone in the evolution of the toolkit, which now has over 200 recipes for speech, audio, and language processing tasks, and more than 100 models available on Hugging Face. SpeechBrain 1.0 introduces new technologies to support diverse learning modalities, Large Language Model (LLM) integration, and advanced decoding strategies, along with novel models, tasks, and modalities. It also includes a new benchmark repository, offering researchers a unified platform for evaluating models across diverse tasks.
Authors: Ashkan Golgoon, Khashayar Filom, Arjun Ravi Kannan
Abstract: Large Language Models such as GPTs (Generative Pre-trained Transformers) exhibit remarkable capabilities across a broad spectrum of applications. Nevertheless, due to their intrinsic complexity, these models present substantial challenges in interpreting their internal decision-making processes. This lack of transparency poses critical challenges when it comes to their adaptation by financial institutions, where concerns and accountability regarding bias, fairness, and reliability are of paramount importance. Mechanistic interpretability aims at reverse engineering complex AI models such as transformers. In this paper, we are pioneering the use of mechanistic interpretability to shed some light on the inner workings of large language models for use in financial services applications. We offer several examples of how algorithmic tasks can be designed for compliance monitoring purposes. In particular, we investigate GPT-2 Small's attention pattern when prompted to identify potential violation of Fair Lending laws. Using direct logit attribution, we study the contributions of each layer and its corresponding attention heads to the logit difference in the residual stream. Finally, we design clean and corrupted prompts and use activation patching as a causal intervention method to localize our task completion components further. We observe that the (positive) heads $10.2$ (head $2$, layer $10$), $10.7$, and $11.3$, as well as the (negative) heads $9.6$ and $10.6$ play a significant role in the task completion.
Authors: Keane Ong, Rui Mao, Ranjan Satapathy, Ricardo Shirota Filho, Erik Cambria, Johan Sulaeman, Gianmarco Mengaldo
Abstract: Sustainability commonly refers to entities, such as individuals, companies, and institutions, having a non-detrimental (or even positive) impact on the environment, society, and the economy. With sustainability becoming a synonym of acceptable and legitimate behaviour, it is being increasingly demanded and regulated. Several frameworks and standards have been proposed to measure the sustainability impact of corporations, including United Nations' sustainable development goals and the recently introduced global sustainability reporting framework, amongst others. However, the concept of corporate sustainability is complex due to the diverse and intricate nature of firm operations (i.e. geography, size, business activities, interlinks with other stakeholders). As a result, corporate sustainability assessments are plagued by subjectivity both within data that reflect corporate sustainability efforts (i.e. corporate sustainability disclosures) and the analysts evaluating them. This subjectivity can be distilled into distinct challenges, such as incompleteness, ambiguity, unreliability and sophistication on the data dimension, as well as limited resources and potential bias on the analyst dimension. Put together, subjectivity hinders effective cost attribution to entities non-compliant with prevailing sustainability expectations, potentially rendering sustainability efforts and its associated regulations futile. To this end, we argue that Explainable Natural Language Processing (XNLP) can significantly enhance corporate sustainability analysis. Specifically, linguistic understanding algorithms (lexical, semantic, syntactic), integrated with XAI capabilities (interpretability, explainability, faithfulness), can bridge gaps in analyst resources and mitigate subjectivity problems within data.
Authors: Mohamad Rida Rammal, Ruida Zhou, Suhas Diggavi
Abstract: Data valuation seeks to answer the important question, "How much is this data worth?" Existing data valuation methods have largely focused on discriminative models, primarily examining data value through the lens of its utility in training. However, with the push for ever-larger language models, relying on valuation methods that require training becomes increasingly expensive and dependent on specific techniques. We propose an alternative perspective on the data value problem for language models, centering around the plausibility of the data. We posit that data holds lesser value if it can be plausibly generated by the model itself. Starting from some intuitive criteria that align with our notions of valuable data, we develop a novel value function that is computationally tractable and derived from first principles with provable properties. We conduct a theoretical analysis of our value function and evaluate it across multiple scenarios and datasets.
Authors: Yu Zhang, Changhao Pan, Wenxiang Guo, Ruiqi Li, Zhiyuan Zhu, Jialei Wang, Wenhao Xu, Jingyu Lu, Zhiqing Hong, Chuxin Wang, LiChao Zhang, Jinzheng He, Ziyue Jiang, Yuxin Chen, Chen Yang, Jiecheng Zhou, Xinyu Cheng, Zhou Zhao
Abstract: The scarcity of high-quality and multi-task singing datasets significantly hinders the development of diverse controllable and personalized singing tasks, as existing singing datasets suffer from low quality, limited diversity of languages and singers, absence of multi-technique information and realistic music scores, and poor task suitability. To tackle these problems, we present GTSinger, a large global, multi-technique, free-to-use, high-quality singing corpus with realistic music scores, designed for all singing tasks, along with its benchmarks. Particularly, (1) we collect 80.59 hours of high-quality singing voices, forming the largest recorded singing dataset; (2) 20 professional singers across nine widely spoken languages offer diverse timbres and styles; (3) we provide controlled comparison and phoneme-level annotations of six commonly used singing techniques, helping technique modeling and control; (4) GTSinger offers realistic music scores, assisting real-world musical composition; (5) singing voices are accompanied by manual phoneme-to-audio alignments, global style labels, and 16.16 hours of paired speech for various singing tasks. Moreover, to facilitate the use of GTSinger, we conduct four benchmark experiments: technique-controllable singing voice synthesis, technique recognition, style transfer, and speech-to-singing conversion. The corpus and demos can be found at http://gtsinger.github.io. We provide the dataset and the code for processing data and conducting benchmarks at https://huggingface.co/datasets/GTSinger/GTSinger and https://github.com/GTSinger/GTSinger.
URLs: http://gtsinger.github.io., https://huggingface.co/datasets/GTSinger/GTSinger, https://github.com/GTSinger/GTSinger.
Authors: Yuzi Yan, Xingzhou Lou, Jialian Li, Yiping Zhang, Jian Xie, Chao Yu, Yu Wang, Dong Yan, Yuan Shen
Abstract: As Large Language Models (LLMs) continue to progress toward more advanced forms of intelligence, Reinforcement Learning from Human Feedback (RLHF) is increasingly seen as a key pathway toward achieving Artificial General Intelligence (AGI). However, the reliance on reward-model-based (RM-based) alignment methods introduces significant challenges due to the inherent instability and imperfections of Reward Models (RMs), which can lead to critical issues such as reward hacking and misalignment with human intentions. In this paper, we introduce a reward-robust RLHF framework aimed at addressing these fundamental challenges, paving the way for more reliable and resilient learning in LLMs. Our approach introduces a novel optimization objective that carefully balances performance and robustness by incorporating Bayesian Reward Model Ensembles (BRME) to model the uncertainty set of reward functions. This allows the framework to integrate both nominal performance and minimum reward signals, ensuring more stable learning even with imperfect RMs. Empirical results demonstrate that our framework consistently outperforms baselines across diverse benchmarks, showing improved accuracy and long-term stability. We also provide a theoretical analysis, demonstrating that reward-robust RLHF approaches the stability of constant reward settings, which proves to be acceptable even in a stochastic-case analysis. Together, these contributions highlight the framework potential to enhance both the performance and stability of LLM alignment.
Authors: Xuanming Zhang, Yuxuan Chen, Yuan Yuan, Minlie Huang
Abstract: In real world software development, improper or missing exception handling can severely impact the robustness and reliability of code. Exception handling mechanisms require developers to detect, capture, and manage exceptions according to high standards, but many developers struggle with these tasks, leading to fragile code. This problem is particularly evident in open source projects and impacts the overall quality of the software ecosystem. To address this challenge, we explore the use of large language models (LLMs) to improve exception handling in code. Through extensive analysis, we identify three key issues: Insensitive Detection of Fragile Code, Inaccurate Capture of Exception Types, and Distorted Handling Solutions. These problems are widespread across real world repositories, suggesting that robust exception handling practices are often overlooked or mishandled. In response, we propose Seeker, a multi agent framework inspired by expert developer strategies for exception handling. Seeker uses agents: Scanner, Detector, Predator, Ranker, and Handler to assist LLMs in detecting, capturing, and resolving exceptions more effectively. Our work is the first systematic study on leveraging LLMs to enhance exception handling practices, providing valuable insights for future improvements in code reliability.
Authors: Qidong Huang, Xiaoyi Dong, Pan Zhang, Yuhang Zang, Yuhang Cao, Jiaqi Wang, Dahua Lin, Weiming Zhang, Nenghai Yu
Abstract: We present the Modality Integration Rate (MIR), an effective, robust, and generalized metric to indicate the multi-modal pre-training quality of Large Vision Language Models (LVLMs). Large-scale pre-training plays a critical role in building capable LVLMs, while evaluating its training quality without the costly supervised fine-tuning stage is under-explored. Loss, perplexity, and in-context evaluation results are commonly used pre-training metrics for Large Language Models (LLMs), while we observed that these metrics are less indicative when aligning a well-trained LLM with a new modality. Due to the lack of proper metrics, the research of LVLMs in the critical pre-training stage is hindered greatly, including the training data choice, efficient module design, etc. In this paper, we propose evaluating the pre-training quality from the inter-modal distribution distance perspective and present MIR, the Modality Integration Rate, which is 1) \textbf{Effective} to represent the pre-training quality and show a positive relation with the benchmark performance after supervised fine-tuning. 2) \textbf{Robust} toward different training/evaluation data. 3) \textbf{Generalize} across training configurations and architecture choices. We conduct a series of pre-training experiments to explore the effectiveness of MIR and observe satisfactory results that MIR is indicative about training data selection, training strategy schedule, and model architecture design to get better pre-training results. We hope MIR could be a helpful metric for building capable LVLMs and inspire the following research about modality alignment in different areas. Our code is at: https://github.com/shikiw/Modality-Integration-Rate.
Authors: Eunji Kim, Kyuhong Shim, Simyung Chang, Sungroh Yoon
Abstract: A text encoder within Vision-Language Models (VLMs) like CLIP plays a crucial role in translating textual input into an embedding space shared with images, thereby facilitating the interpretative analysis of vision tasks through natural language. Despite the varying significance of different textual elements within a sentence depending on the context, efforts to account for variation of importance in constructing text embeddings have been lacking. We propose a framework of Semantic Token Reweighting to build Interpretable text embeddings (SToRI), which incorporates controllability as well. SToRI refines the text encoding process in CLIP by differentially weighting semantic elements based on contextual importance, enabling finer control over emphasis responsive to data-driven insights and user preferences. The efficacy of SToRI is demonstrated through comprehensive experiments on few-shot image classification and image retrieval tailored to user preferences.
Authors: Jun Luo, Chen Chen, Shandong Wu
Abstract: Prompt learning for pre-trained Vision-Language Models (VLMs) like CLIP has demonstrated potent applicability across diverse downstream tasks. This lightweight approach has quickly gained traction from federated learning (FL) researchers who seek to efficiently adapt VLMs to heterogeneous scenarios. However, current federated prompt learning methods are habitually restricted to the traditional FL paradigm, where the participating clients are generally only allowed to download a single globally aggregated model from the server. While justifiable for training full-sized models under federated settings, in this work, we argue that this paradigm is ill-suited for lightweight prompts. By facilitating the clients to download multiple pre-aggregated prompts as fixed non-local experts, we propose Personalized Federated Mixture of Adaptive Prompts (pFedMoAP), a novel FL framework that personalizes the prompt learning process through the lens of Mixture of Experts (MoE). pFedMoAP implements a local attention-based gating network that learns to generate enhanced text features for better alignment with local image data on the client, benefiting from both local and downloaded non-local adaptive prompt experts. The non-local experts are sparsely selected from a server-maintained pool, fostering collaborative learning across clients. To evaluate the proposed algorithm, we conduct extensive experiments across 9 datasets under various heterogeneous federated settings. The results show that pFedMoAP consistently outperforms the state-of-the-art alternatives, underscoring its efficacy in personalizing prompt learning for CLIP within the federated learning paradigm.
Authors: Junkang Wu, Xue Wang, Zhengyi Yang, Jiancan Wu, Jinyang Gao, Bolin Ding, Xiang Wang, Rong Jin, Xiangnan He
Abstract: Aligning large language models (LLMs) with human values and intentions is crucial for their utility, honesty, and safety. Reinforcement learning from human feedback (RLHF) is a popular approach to achieve this alignment, but it faces challenges in computational efficiency and training stability. Recent methods like Direct Preference Optimization (DPO) and Simple Preference Optimization (SimPO) have proposed offline alternatives to RLHF, simplifying the process by reparameterizing the reward function. However, DPO depends on a potentially suboptimal reference model, and SimPO's assumption of a fixed target reward margin may lead to suboptimal decisions in diverse data settings. In this work, we propose $\alpha$-DPO, an adaptive preference optimization algorithm designed to address these limitations by introducing a dynamic reward margin. Specifically, $\alpha$-DPO employs an adaptive preference distribution, balancing the policy model and the reference model to achieve personalized reward margins. We provide theoretical guarantees for $\alpha$-DPO, demonstrating its effectiveness as a surrogate optimization objective and its ability to balance alignment and diversity through KL divergence control. Empirical evaluations on AlpacaEval 2 and Arena-Hard show that $\alpha$-DPO consistently outperforms DPO and SimPO across various model settings, establishing it as a robust approach for fine-tuning LLMs. Our method achieves significant improvements in win rates, highlighting its potential as a powerful tool for LLM alignment. The code is available at https://github.com/junkangwu/alpha-DPO
Authors: Wanying Wang, Zeyu Ma, Pengfei Liu, Mingang Chen
Abstract: While various vertical domain large language models (LLMs) have been developed, the challenge of automatically evaluating their performance across different domains remains significant. Current benchmark-based evaluation methods exhibit rigid, aimless interactions and rely on pre-collected static datasets that are costly to build, inflexible across domains, and misaligned with practical user needs. To address this issue, we revisit the evaluation components and introduce two concepts: Benchmark+, which extends traditional question-answer benchmark into a more flexible "strategy-criterion" format; and Assessment+, which enhances the interaction process, enabling deeper exploration and supporting both quantitative metrics and qualitative insights. These concepts capture the nuanced behaviors of LLMs through richer, multi-turn interactions. We propose an agent-based evaluation framework called TestAgent, which implements these concepts through retrieval augmented generation and reinforcement learning. Experiments on tasks ranging from constructing vertical domain evaluation to activating existing benchmarks demonstrate the effectiveness of TestAgent across various scenarios. We believe this work offers an interesting perspective on automatic evaluation for LLMs.