Authors: Jillian Ross, Yoon Kim, Andrew W. Lo
Abstract: Humans are not homo economicus (i.e., rational economic beings). As humans, we exhibit systematic behavioral biases such as loss aversion, anchoring, framing, etc., which lead us to make suboptimal economic decisions. Insofar as such biases may be embedded in text data on which large language models (LLMs) are trained, to what extent are LLMs prone to the same behavioral biases? Understanding these biases in LLMs is crucial for deploying LLMs to support human decision-making. We propose utility theory-a paradigm at the core of modern economic theory-as an approach to evaluate the economic biases of LLMs. Utility theory enables the quantification and comparison of economic behavior against benchmarks such as perfect rationality or human behavior. To demonstrate our approach, we quantify and compare the economic behavior of a variety of open- and closed-source LLMs. We find that the economic behavior of current LLMs is neither entirely human-like nor entirely economicus-like. We also find that most current LLMs struggle to maintain consistent economic behavior across settings. Finally, we illustrate how our approach can measure the effect of interventions such as prompting on economic biases.
Authors: Adil Soubki, Shyne Choi, Owen Rambow
Abstract: We propose a framework for analyzing discourse by combining two interdependent concepts from sociolinguistic theory: face acts and politeness. While politeness has robust existing tools and data, face acts are less resourced. We introduce a new corpus created by annotating Wikipedia talk pages with face acts and we use this to train a face act tagger. We then employ our framework to study how face and politeness interact with gender and power in discussions between Wikipedia editors. Among other findings, we observe that female Wikipedians are not only more polite, which is consistent with prior studies, but that this difference corresponds with significantly more language directed at humbling aspects of their own face. Interestingly, the distinction nearly vanishes once limiting to editors with administrative power.
Authors: Ryan Aponte (Carnegie Mellon University), Ryan A. Rossi (Adobe Research), Shunan Guo (Adobe Research), Franck Dernoncourt (Adobe Research), Tong Yu (Adobe Research), Xiang Chen (Adobe Research), Subrata Mitra (Adobe Research), Nedim Lipka (Adobe Research)
Abstract: Large language models (LLMs) have been applied to a wide range of tasks, including text summarization, web navigation, and chatbots. They have benefitted from supervised fine-tuning (SFT) and reinforcement learning from human feedback (RLHF) following an unsupervised pretraining. These datasets can be difficult to collect, limited in scope, and vary in sample quality. Additionally, datasets can vary extensively in supervision format, from numerical to binary as well as multi-dimensional with many different values. We present a framework for fine-tuning LLMs using heterogeneous feedback, which has two main components. First, we combine the heterogeneous feedback data into a single supervision format, compatible with methods like SFT and RLHF. Next, given this unified feedback dataset, we extract a high-quality and diverse subset to obtain performance increases potentially exceeding the full dataset. We conduct extensive experiments to understand the effectiveness of these techniques for incorporating heterogeneous feedback, and demonstrate improvements from using a high-quality and diverse subset of the data. We find that our framework is able to improve models in multiple areas simultaneously, such as in instruction following and bias reduction.
Authors: Takehiro Takayanagi, Hiroki Sakaji, Kiyoshi Izumi
Abstract: Stock embedding is a method for vector representation of stocks. There is a growing demand for vector representations of stock, i.e., stock embedding, in wealth management sectors, and the method has been applied to various tasks such as stock price prediction, portfolio optimization, and similar fund identifications. Stock embeddings have the advantage of enabling the quantification of relative relationships between stocks, and they can extract useful information from unstructured data such as text and network data. In this study, we propose stock embedding enhanced with textual and network information (SETN) using a domain-adaptive pre-trained transformer-based model to embed textual information and a graph neural network model to grasp network information. We evaluate the performance of our proposed model on related company information extraction tasks. We also demonstrate that stock embeddings obtained from the proposed model perform better in creating thematic funds than those obtained from baseline methods, providing a promising pathway for various applications in the wealth management industry.
Authors: Tiezheng Guo, Chen Wang, Yanyi Liu, Jiawei Tang, Pan Li, Sai Xu, Qingwen Yang, Xianlin Gao, Zhi Li, Yingyou Wen
Abstract: Retrieving external knowledge and prompting large language models with relevant information is an effective paradigm to enhance the performance of question-answering tasks. Previous research typically handles paragraphs from external documents in isolation, resulting in a lack of context and ambiguous references, particularly in multi-document and complex tasks. To overcome these challenges, we propose a new retrieval framework IIER, that leverages Inter-chunk Interactions to Enhance Retrieval. This framework captures the internal connections between document chunks by considering three types of interactions: structural, keyword, and semantic. We then construct a unified Chunk-Interaction Graph to represent all external documents comprehensively. Additionally, we design a graph-based evidence chain retriever that utilizes previous paths and chunk interactions to guide the retrieval process. It identifies multiple seed nodes based on the target question and iteratively searches for relevant chunks to gather supporting evidence. This retrieval process refines the context and reasoning chain, aiding the large language model in reasoning and answer generation. Extensive experiments demonstrate that IIER outperforms strong baselines across four datasets, highlighting its effectiveness in improving retrieval and reasoning capabilities.
Authors: Heidi C. Zhang, Shabnam Behzad, Kawin Ethayarajh, Dan Jurafsky
Abstract: Model checklists (Ribeiro et al., 2020) have emerged as a useful tool for understanding the behavior of LLMs, analogous to unit-testing in software engineering. However, despite datasets being a key determinant of model behavior, evaluating datasets, e.g., for the existence of annotation artifacts, is largely done ad hoc, once a problem in model behavior has already been found downstream. In this work, we take a more principled approach to unit-testing datasets by proposing a taxonomy based on the V-information literature. We call a collection of such unit tests a data checklist. Using a checklist, not only are we able to recover known artifacts in well-known datasets such as SNLI, but we also discover previously unknown artifacts in preference datasets for LLM alignment. Data checklists further enable a new kind of data filtering, which we use to improve the efficacy and data efficiency of preference alignment.
Authors: Atsushi Kojima
Abstract: We propose the intermediate direct preference optimization (DPO) method to calculate the DPO loss at selected intermediate layers as an auxiliary loss for finetuning large language models (LLMs). The conventional DPO method fine-tunes a supervised fine-tuning (SFT) model by calculating the DPO loss using logits from the final layer. In our intermediate DPO approach, DPO losses are calculated using the logits from K-selected intermediate layers and averaged to obtain the intermediate DPO loss. For training the intermediate DPO model, the final loss is obtained by calculating the weighted sum of the DPO and intermediate DPO losses. During inference, the intermediate DPO model decodes using the final layer logits similarly to the conventional DPO model. In experiments using the ultrafeedback dataset, the performance of the intermediate DPO model was evaluated using GPT-4. As a result, the intermediate DPO model trained using the intermediate DPO loss calculated at the 22nd layer of a 32-layer SFT model achieved win rates of 52.5% and 67.5% against the conventional DPO and SFT models, respectively, demonstrating the effectiveness of the proposed method. Furthermore, we report the relationships among the position of the selected intermediate layers, the number of layers, and performance.
Authors: Atsushi Kojima
Abstract: Self-supervised learning, such as with the wav2vec 2.0 framework significantly improves the accuracy of end-to-end automatic speech recognition (ASR). Wav2vec 2.0 has been applied to single-channel end-to-end ASR models. In this work, we explored a self-supervised learning method for a multi-channel end-to-end ASR model based on the wav2vec 2.0 framework. As the multi-channel end-to-end ASR model, we focused on a multi-channel neural transducer. In pre-training, we compared three different methods for feature quantization to train a multi-channel conformer audio encoder: joint quantization, feature-wise quantization and channel-wise quantization. In fine-tuning, we trained the multi-channel conformer-transducer. All experiments were conducted using the far-field in-house and CHiME-4 datasets. The results of the experiments showed that feature-wise quantization was the most effective among the methods. We observed a 66% relative reduction in character error rate compared with the model without any pre-training for the far-field in-house dataset.
Authors: Da Ju, Karen Ulrich, Adina Williams
Abstract: People tend to use language to mention surprising properties of events: for example, when a banana is blue, we are more likely to mention color than when it is yellow. This fact is taken to suggest that yellowness is somehow a typical feature of bananas, and blueness is exceptional. Similar to how a yellow color is typical of bananas, there may also be genders that are typical of occupations. In this work, we explore this question using information theoretic techniques coupled with corpus statistic analysis. In two distinct large corpora, we do not find strong evidence that occupations and gender display the same patterns of mentioning as do bananas and color. Instead, we find that gender mentioning is correlated with femaleness of occupation in particular, suggesting perhaps that woman-dominated occupations are seen as somehow ``more gendered'' than male-dominated ones, and thereby they encourage more gender mentioning overall.
Authors: Iman Azimi, Mohan Qi, Li Wang, Amir M. Rahmani, Youlin Li
Abstract: Large language models (LLMs) are fundamentally transforming human-facing applications in the health and well-being domains: boosting patient engagement, accelerating clinical decision-making, and facilitating medical education. Although state-of-the-art LLMs have shown superior performance in several conversational applications, evaluations within nutrition and diet applications are still insufficient. In this paper, we propose to employ the Registered Dietitian (RD) exam to conduct a standard and comprehensive evaluation of state-of-the-art LLMs, GPT-4o, Claude 3.5 Sonnet, and Gemini 1.5 Pro, assessing both accuracy and consistency in nutrition queries. Our evaluation includes 1050 RD exam questions encompassing several nutrition topics and proficiency levels. In addition, for the first time, we examine the impact of Zero-Shot (ZS), Chain of Thought (CoT), Chain of Thought with Self Consistency (CoT-SC), and Retrieval Augmented Prompting (RAP) on both accuracy and consistency of the responses. Our findings revealed that while these LLMs obtained acceptable overall performance, their results varied considerably with different prompts and question domains. GPT-4o with CoT-SC prompting outperformed the other approaches, whereas Gemini 1.5 Pro with ZS recorded the highest consistency. For GPT-4o and Claude 3.5, CoT improved the accuracy, and CoT-SC improved both accuracy and consistency. RAP was particularly effective for GPT-4o to answer Expert level questions. Consequently, choosing the appropriate LLM and prompting technique, tailored to the proficiency level and specific domain, can mitigate errors and potential risks in diet and nutrition chatbots.
Authors: Zhaopeng Feng, Zijie Meng, Zuozhu Liu
Abstract: Large language models (LLMs) have attracted considerable attention in various fields for their cost-effective solutions to diverse challenges, especially with advancements in instruction tuning and quantization. E-commerce, with its complex tasks and extensive product-user interactions, presents a promising application area for LLMs. However, the domain-specific concepts and knowledge inherent in e-commerce pose significant challenges for adapting general LLMs. To address this issue, we developed EC-Guide \href{https://github.com/fzp0424/EC-Guide-KDDUP-2024}, a comprehensive e-commerce guide for instruction tuning and quantization of LLMs. We also heuristically integrated Chain-of-Thought (CoT) during inference to enhance arithmetic performance. Our approach achieved the 2nd place in Track 2 and 5th place in Track 5 at the Amazon KDD Cup'24 \href{https://www.aicrowd.com/challenges/amazon-kdd-cup-2024-multi-task-online-shopping-challenge-for-llms}. Additionally, our solution is model-agnostic, enabling effective scalability across larger systems.
URLs: https://github.com/fzp0424/EC-Guide-KDDUP-2024, https://www.aicrowd.com/challenges/amazon-kdd-cup-2024-multi-task-online-shopping-challenge-for-llms
Authors: Hui Ma, Bo Zhang, Bo Xu, Jian Wang, Hongfei Lin, Xiao Sun
Abstract: Empathetic response generation, aiming at understanding the user's situation and feelings and respond empathically, is crucial in building human-like dialogue systems. Previous methods mainly focus on using maximum likelihood estimation as the optimization objective for training response generation models, without taking into account the empathy level alignment between generated responses and target responses. To this end, we propose an empathetic response generation using reinforcement learning (EmpRL) framework. The framework designs an effective empathy reward function and generates empathetic responses by maximizing the expected reward through reinforcement learning. Given the powerful text generation capability of pre-trained language models, EmpRL utilizes the pre-trained T5 model as the generator and conducts further training to initialize the policy. To align the empathy level between generated responses and target responses in the context, an empathy reward function containing three empathy communication mechanisms, i.e., emotional reaction, interpretation, and exploration, is constructed using pre-designed and pre-trained empathy identifiers. Finally, the proximal policy optimization algorithm is used to further train the policy to produce empathetic responses. Both automatic and manual evaluations demonstrate that the proposed EmpRL framework can improve the quality of generated responses, enhance the empathy level similarity between generated and target responses, and produce empathetic responses covering both affective and cognitive aspects.
Authors: Daniel Steinigen, Roman Teucher, Timm Heine Ruland, Max Rudat, Nicolas Flores-Herr, Peter Fischer, Nikola Milosevic, Christopher Schymura, Angelo Ziletti
Abstract: Recent advancements in Large Language Models (LLMs) have showcased their proficiency in answering natural language queries. However, their effectiveness is hindered by limited domain-specific knowledge, raising concerns about the reliability of their responses. We introduce a hybrid system that augments LLMs with domain-specific knowledge graphs (KGs), thereby aiming to enhance factual correctness using a KG-based retrieval approach. We focus on a medical KG to demonstrate our methodology, which includes (1) pre-processing, (2) Cypher query generation, (3) Cypher query processing, (4) KG retrieval, and (5) LLM-enhanced response generation. We evaluate our system on a curated dataset of 69 samples, achieving a precision of 78\% in retrieving correct KG nodes. Our findings indicate that the hybrid system surpasses a standalone LLM in accuracy and completeness, as verified by an LLM-as-a-Judge evaluation method. This positions the system as a promising tool for applications that demand factual correctness and completeness, such as target identification -- a critical process in pinpointing biological entities for disease treatment or crop enhancement. Moreover, its intuitive search interface and ability to provide accurate responses within seconds make it well-suited for time-sensitive, precision-focused research contexts. We publish the source code together with the dataset and the prompt templates used.
Authors: Elvys Linhares Pontes, Carlos-Emiliano Gonz\'alez-Gallardo, Mohamed Benjannet, Caryn Qu, Antoine Doucet
Abstract: This article details our participation (L3iTC) in the FinLLM Challenge Task 2024, focusing on two key areas: Task 1, financial text classification, and Task 2, financial text summarization. To address these challenges, we fine-tuned several large language models (LLMs) to optimize performance for each task. Specifically, we used 4-bit quantization and LoRA to determine which layers of the LLMs should be trained at a lower precision. This approach not only accelerated the fine-tuning process on the training data provided by the organizers but also enabled us to run the models on low GPU memory. Our fine-tuned models achieved third place for the financial classification task with an F1-score of 0.7543 and secured sixth place in the financial summarization task on the official test datasets.
Authors: Pegah Ramezani, Achim Schilling, Patrick Krauss
Abstract: Understanding how language and linguistic constructions are processed in the brain is a fundamental question in cognitive computational neuroscience. In this study, we explore the representation and processing of Argument Structure Constructions (ASCs) in a recurrent neural language model. We trained a Long Short-Term Memory (LSTM) network on a custom-made dataset consisting of 2000 sentences, generated using GPT-4, representing four distinct ASCs: transitive, ditransitive, caused-motion, and resultative constructions. We analyzed the internal activations of the LSTM model's hidden layers using Multidimensional Scaling (MDS) and t-Distributed Stochastic Neighbor Embedding (t-SNE) to visualize the sentence representations. The Generalized Discrimination Value (GDV) was calculated to quantify the degree of clustering within these representations. Our results show that sentence representations form distinct clusters corresponding to the four ASCs across all hidden layers, with the most pronounced clustering observed in the last hidden layer before the output layer. This indicates that even a relatively simple, brain-constrained recurrent neural network can effectively differentiate between various construction types. These findings are consistent with previous studies demonstrating the emergence of word class and syntax rule representations in recurrent language models trained on next word prediction tasks. In future work, we aim to validate these results using larger language models and compare them with neuroimaging data obtained during continuous speech perception. This study highlights the potential of recurrent neural language models to mirror linguistic processing in the human brain, providing valuable insights into the computational and neural mechanisms underlying language understanding.
Authors: David Kletz (Lattice, LLF - UMR7110, UPCit\'e), Marie Candito (UPCit\'e, LLF - UMR7110), Pascal Amsili (Lattice)
Abstract: Contradictory results about the encoding of the semantic impact of negation in pretrained language models (PLMs). have been drawn recently (e.g. Kassner and Sch{\"u}tze (2020); Gubelmann and Handschuh (2022)). In this paper we focus rather on the way PLMs encode negation and its formal impact, through the phenomenon of the Negative Polarity Item (NPI) licensing in English. More precisely, we use probes to identify which contextual representations best encode 1) the presence of negation in a sentence, and 2) the polarity of a neighboring masked polarity item. We find that contextual representations of tokens inside the negation scope do allow for (i) a better prediction of the presence of not compared to those outside the scope and (ii) a better prediction of the right polarity of a masked polarity item licensed by not, although the magnitude of the difference varies from PLM to PLM. Importantly, in both cases the trend holds even when controlling for distance to not. This tends to indicate that the embeddings of these models do reflect the notion of negation scope, and do encode the impact of negation on NPI licensing. Yet, further control experiments reveal that the presence of other lexical items is also better captured when using the contextual representation of a token within the same syntactic clause than outside from it, suggesting that PLMs simply capture the more general notion of syntactic clause.
Authors: Amelie Robrecht, Judith Sieker, Clara Lachenmaier, Sina Zarie{\ss}, Stefan Kopp
Abstract: In this work, we want to give an overview on which pragmatic abilities have been tested in LLMs so far and how these tests have been carried out. To do this, we first discuss the scope of the field of pragmatics and suggest a subdivision into discourse pragmatics and interactional pragmatics. We give a non-exhaustive overview of the phenomena of those two subdomains and the methods traditionally used to analyze them. We subsequently consider the resulting heterogeneous set of phenomena and methods as a starting point for our survey of work on discourse pragmatics and interactional pragmatics in the context of LLMs.
Authors: Jinglong Gao, Chen Lu, Xiao Ding, Zhongyang Li, Ting Liu, Bing Qin
Abstract: Event Causality Extraction (ECE) aims at extracting causal event pairs from texts. Despite ChatGPT's recent success, fine-tuning small models remains the best approach for the ECE task. However, existing fine-tuning based ECE methods cannot address all three key challenges in ECE simultaneously: 1) Complex Causality Extraction, where multiple causal-effect pairs occur within a single sentence; 2) Subtask~ Interaction, which involves modeling the mutual dependence between the two subtasks of ECE, i.e., extracting events and identifying the causal relationship between extracted events; and 3) Knowledge Fusion, which requires effectively fusing the knowledge in two modalities, i.e., the expressive pretrained language models and the structured knowledge graphs. In this paper, we propose a unified ECE framework (UniCE to address all three issues in ECE simultaneously. Specifically, we design a subtask interaction mechanism to enable mutual interaction between the two ECE subtasks. Besides, we design a knowledge fusion mechanism to fuse knowledge in the two modalities. Furthermore, we employ separate decoders for each subtask to facilitate complex causality extraction. Experiments on three benchmark datasets demonstrate that our method achieves state-of-the-art performance and outperforms ChatGPT with a margin of at least 30% F1-score. More importantly, our model can also be used to effectively improve the ECE performance of ChatGPT via in-context learning.
Authors: Le Yu, Bowen Yu, Haiyang Yu, Fei Huang, Yongbin Li
Abstract: Merging Large Language Models (LLMs) aims to amalgamate multiple homologous LLMs into one with all the capabilities. Ideally, any LLMs sharing the same backbone should be mergeable, irrespective of whether they are Fine-Tuned (FT) with minor parameter changes or Pre-Trained (PT) with substantial parameter shifts. However, existing methods often manually assign the model importance, rendering them feasible only for LLMs with similar parameter alterations, such as multiple FT LLMs. The diverse parameter changed ranges between FT and PT LLMs pose challenges for current solutions in empirically determining the optimal combination. In this paper, we make a pioneering effort to broaden the applicability of merging techniques from FT to PT LLMs. We initially examine the efficacy of current methods in merging FT and PT LLMs, discovering that they struggle to deal with PT LLMs. Subsequently, we introduce an approach based on WeIght DisENtanglement (WIDEN) to effectively extend the merging scope, which first disentangles model weights into magnitude and direction components, and then performs adaptive fusion by considering their respective contributions. In the experiments, we merge Qwen1.5-Chat (an FT LLM with instruction-following skills) with Sailor (a PT LLM with multilingual abilities) across 7B and 14B model scales. Results reveal that: (1) existing solutions usually fail when merging Sailor, either losing both abilities or only retaining instruction-following skills; (2) WIDEN successfully injects the multilingual abilities of Sailor into Qwen1.5-Chat and make it proficient in Southeast Asian languages, achieving enhancements in the fundamental capabilities. In light of previous research, we also merge multiple 13B FT LLMs and observe that WIDEN achieves a balanced amalgamation of instruction following, mathematical reasoning, and code generation skills.
Authors: Zongqian Li, Yixuan Su, Nigel Collier
Abstract: Prompt compression is crucial for enhancing inference speed, reducing costs, and improving user experience. However, current methods face challenges such as low compression ratios and potential data leakage during evaluation. To address these issues, we propose 500xCompressor, a method that compresses extensive natural language contexts into a minimum of one single special token. The 500xCompressor introduces approximately 0.3% additional parameters and achieves compression ratios ranging from 6x to 480x. It is designed to compress any text, answer various types of questions, and could be utilized by the original large language model (LLM) without requiring fine-tuning. Initially, 500xCompressor was pretrained on the Arxiv Corpus, followed by fine-tuning on the ArxivQA dataset, and subsequently evaluated on strictly unseen and classical question answering (QA) datasets. The results demonstrate that the LLM retained 62.26-72.89% of its capabilities compared to using non-compressed prompts. This study also shows that not all the compressed tokens are equally utilized and that K V values have significant advantages over embeddings in preserving information at high compression ratios. The highly compressive nature of natural language prompts, even for fine-grained complex information, suggests promising potential for future applications and further research into developing a new LLM language.
Authors: Johannes Schneider
Abstract: Large language models (LLM)'s are increasingly used for topic modeling outperforming classical topic models such as LDA. Commonly, pre-trained LLM encoders such as BERT are used out-of-the-box despite the fact that fine-tuning is known to improve LLMs considerably. The challenge lies in obtaining a suitable (labeled) dataset for fine-tuning. In this paper, we use the recent idea to use bag of sentences as the elementary unit in computing topics. In turn, we derive an approach FT-Topic to perform unsupervised fine-tuning relying primarily on two steps for constructing a training dataset in an automatic fashion. First, a heuristic method to identifies pairs of sentence groups that are either assumed to be of the same or different topics. Second, we remove sentence pairs that are likely labeled incorrectly. The dataset is then used to fine-tune an encoder LLM, which can be leveraged by any topic modeling approach using embeddings. However, in this work, we demonstrate its effectiveness by deriving a novel state-of-the-art topic modeling method called SenClu, which achieves fast inference through an expectation-maximization algorithm and hard assignments of sentence groups to a single topic, while giving users the possibility to encode prior knowledge on the topic-document distribution. Code is at \url{https://github.com/JohnTailor/FT-Topic}
Authors: Yan Huang, Wei Liu
Abstract: In recent years, with the rapid development of deep learning technology, large language models (LLMs) such as BERT and GPT have achieved breakthrough results in natural language processing tasks. Machine translation (MT), as one of the core tasks of natural language processing, has also benefited from the development of large language models and achieved a qualitative leap. Despite the significant progress in translation performance achieved by large language models, machine translation still faces many challenges. Therefore, in this paper, we construct the dataset Euas-20 to evaluate the performance of large language models on translation tasks, the translation ability on different languages, and the effect of pre-training data on the translation ability of LLMs for researchers and developers.
Authors: Rajvee Sheth, Shubh Nisar, Heenaben Prajapati, Himanshu Beniwal, Mayank Singh
Abstract: As the NLP community increasingly addresses challenges associated with multilingualism, robust annotation tools are essential to handle multilingual datasets efficiently. In this paper, we introduce a code-mixed multilingual text annotation framework, COMMENTATOR, specifically designed for annotating code-mixed text. The tool demonstrates its effectiveness in token-level and sentence-level language annotation tasks for Hinglish text. We perform robust qualitative human-based evaluations to showcase COMMENTATOR led to 5x faster annotations than the best baseline. Our code is publicly available at \url{https://github.com/lingo-iitgn/commentator}. The demonstration video is available at \url{https://bit.ly/commentator_video}.
URLs: https://github.com/lingo-iitgn/commentator, https://bit.ly/commentator_video
Authors: Artur Guimar\~aes, Bruno Martins, Jo\~ao Magalh\~aes
Abstract: This paper describes our approach to the SemEval-2024 safe biomedical Natural Language Inference for Clinical Trials (NLI4CT) task, which concerns classifying statements about Clinical Trial Reports (CTRs). We explored the capabilities of Mistral-7B, a generalist open-source Large Language Model (LLM). We developed a prompt for the NLI4CT task, and fine-tuned a quantized version of the model using an augmented version of the training dataset. The experimental results show that this approach can produce notable results in terms of the macro F1-score, while having limitations in terms of faithfulness and consistency. All the developed code is publicly available on a GitHub repository
Authors: Leo Donisch, Sigurd Schacht, Carsten Lanquillon
Abstract: Large language models are ubiquitous in natural language processing because they can adapt to new tasks without retraining. However, their sheer scale and complexity present unique challenges and opportunities, prompting researchers and practitioners to explore novel model training, optimization, and deployment methods. This literature review focuses on various techniques for reducing resource requirements and compressing large language models, including quantization, pruning, knowledge distillation, and architectural optimizations. The primary objective is to explore each method in-depth and highlight its unique challenges and practical applications. The discussed methods are categorized into a taxonomy that presents an overview of the optimization landscape and helps navigate it to understand the research trajectory better.
Authors: Charles Brazier, Jean-Luc Rouas
Abstract: Large Language Models (LLMs) have shown remarkable performance in Natural Language Processing tasks, including Machine Translation (MT). In this work, we propose a novel MT pipeline that integrates emotion information extracted from a Speech Emotion Recognition (SER) model into LLMs to enhance translation quality. We first fine-tune five existing LLMs on the Libri-trans dataset and select the most performant model. Subsequently, we augment LLM prompts with different dimensional emotions and train the selected LLM under these different configurations. Our experiments reveal that integrating emotion information, especially arousal, into LLM prompts leads to notable improvements in translation quality.
Authors: Pranita Deshmukh, Nikita Kulkarni, Sanhita Kulkarni, Kareena Manghani, Raviraj Joshi
Abstract: With the surge in digital content in low-resource languages, there is an escalating demand for advanced Natural Language Processing (NLP) techniques tailored to these languages. BERT (Bidirectional Encoder Representations from Transformers), serving as the foundational framework for numerous NLP architectures and language models, is increasingly employed for the development of low-resource NLP models. Parameter Efficient Fine-Tuning (PEFT) is a method for fine-tuning Large Language Models (LLMs) and reducing the training parameters to some extent to decrease the computational costs needed for training the model and achieve results comparable to a fully fine-tuned model. In this work, we present a study of PEFT methods for the Indic low-resource language Marathi. We conduct a comprehensive analysis of PEFT methods applied to various monolingual and multilingual Marathi BERT models. These approaches are evaluated on prominent text classification datasets like MahaSent, MahaHate, and MahaNews. The incorporation of PEFT techniques is demonstrated to significantly expedite the training speed of the models, addressing a critical aspect of model development and deployment. In this study, we explore Low-Rank Adaptation of Large Language Models (LoRA) and adapter methods for low-resource text classification. We show that these methods are competitive with full fine-tuning and can be used without loss in accuracy. This study contributes valuable insights into the effectiveness of Marathi BERT models, offering a foundation for the continued advancement of NLP capabilities in Marathi and similar Indic languages.
Authors: Zifeng Cheng, Zhiwei Jiang, Yafeng Yin, Zhaoling Chen, Cong Wang, Shiping Ge, Qiguo Huang, Qing Gu
Abstract: Multi-Label Text Classification (MLTC) is a practical yet challenging task that involves assigning multiple non-exclusive labels to each document. Previous studies primarily focus on capturing label correlations to assist label prediction by introducing special labeling schemes, designing specific model structures, or adding auxiliary tasks. Recently, the $k$ Nearest Neighbor ($k$NN) framework has shown promise by retrieving labeled samples as references to mine label co-occurrence information in the embedding space. However, two critical biases, namely embedding alignment bias and confidence estimation bias, are often overlooked, adversely affecting prediction performance. In this paper, we introduce a DEbiased Nearest Neighbors (DENN) framework for MLTC, specifically designed to mitigate these biases. To address embedding alignment bias, we propose a debiased contrastive learning strategy, enhancing neighbor consistency on label co-occurrence. For confidence estimation bias, we present a debiased confidence estimation strategy, improving the adaptive combination of predictions from $k$NN and inductive binary classifications. Extensive experiments conducted on four public benchmark datasets (i.e., AAPD, RCV1-V2, Amazon-531, and EUR-LEX57K) showcase the effectiveness of our proposed method. Besides, our method does not introduce any extra parameters.
Authors: Yanyang Li, Shuo Liang, Michael R. Lyu, Liwei Wang
Abstract: Recent advancements in long-context modeling have enhanced language models (LMs) for complex tasks across multiple NLP applications. Despite this progress, we find that these models struggle with multi-hop reasoning and exhibit decreased performance in the presence of noisy contexts. In this paper, we introduce Reasoning with Attributions, a novel approach that prompts LMs to supply attributions for each assertion during their reasoning. We validate our approach through experiments on three multi-hop datasets, employing both proprietary and open-source models, and demonstrate its efficacy and resilience. Furthermore, we explore methods to augment reasoning capabilities via fine-tuning and offer an attribution-annotated dataset and a specialized training strategy. Our fine-tuned model achieves competitive performance on multi-hop reasoning benchmarks, closely paralleling proprietary LMs such as ChatGPT and Claude-instant.
Authors: Yifei Wang, Yuheng Chen, Wanting Wen, Yu Sheng, Linjing Li, Daniel Dajun Zeng
Abstract: In this paper, we investigate whether Large Language Models (LLMs) actively recall or retrieve their internal repositories of factual knowledge when faced with reasoning tasks. Through an analysis of LLMs' internal factual recall at each reasoning step via Knowledge Neurons, we reveal that LLMs fail to harness the critical factual associations under certain circumstances. Instead, they tend to opt for alternative, shortcut-like pathways to answer reasoning questions. By manually manipulating the recall process of parametric knowledge in LLMs, we demonstrate that enhancing this recall process directly improves reasoning performance whereas suppressing it leads to notable degradation. Furthermore, we assess the effect of Chain-of-Thought (CoT) prompting, a powerful technique for addressing complex reasoning tasks. Our findings indicate that CoT can intensify the recall of factual knowledge by encouraging LLMs to engage in orderly and reliable reasoning. Furthermore, we explored how contextual conflicts affect the retrieval of facts during the reasoning process to gain a comprehensive understanding of the factual recall behaviors of LLMs. Code and data will be available soon.
Authors: Jiaxi Yang, Binyuan Hui, Min Yang, Jian Yang, Junyang Lin, Chang Zhou
Abstract: The capability gap between open-source and closed-source large language models (LLMs) remains a challenge in text-to-SQL tasks. In this paper, we introduce a synthetic data approach that combines data produced by larger, more powerful models (strong models) with error information data generated by smaller, not well-aligned models (weak models). The method not only enhances the domain generalization of text-to-SQL models but also explores the potential of error data supervision through preference learning. Furthermore, we employ the synthetic data approach for instruction tuning on open-source LLMs, resulting SENSE, a specialized text-to-SQL model. The effectiveness of SENSE is demonstrated through state-of-the-art results on the SPIDER and BIRD benchmarks, bridging the performance gap between open-source models and methods prompted by closed-source models.
Authors: Boxi Cao, Mengjie Ren, Hongyu Lin, Xianpei Han, Feng Zhang, Junfeng Zhan, Le Sun
Abstract: Evaluation is the baton for the development of large language models. Current evaluations typically employ a single-item assessment paradigm for each atomic test objective, which struggles to discern whether a model genuinely possesses the required capabilities or merely memorizes/guesses the answers to specific questions. To this end, we propose a novel evaluation framework referred to as StructEval. Starting from an atomic test objective, StructEval deepens and broadens the evaluation by conducting a structured assessment across multiple cognitive levels and critical concepts, and therefore offers a comprehensive, robust and consistent evaluation for LLMs. Experiments on three widely-used benchmarks demonstrate that StructEval serves as a reliable tool for resisting the risk of data contamination and reducing the interference of potential biases, thereby providing more reliable and consistent conclusions regarding model capabilities. Our framework also sheds light on the design of future principled and trustworthy LLM evaluation protocols.
Authors: Ruizhe Zhang, Yongxin Xu, Yuzhen Xiao, Runchuan Zhu, Xinke Jiang, Xu Chu, Junfeng Zhao, Yasha Wang
Abstract: By integrating external knowledge, Retrieval-Augmented Generation (RAG) has become an effective strategy for mitigating the hallucination problems that large language models (LLMs) encounter when dealing with knowledge-intensive tasks. However, in the process of integrating external non-parametric supporting evidence with internal parametric knowledge, inevitable knowledge conflicts may arise, leading to confusion in the model's responses. To enhance the knowledge selection of LLMs in various contexts, some research has focused on refining their behavior patterns through instruction-tuning. Nonetheless, due to the absence of explicit negative signals and comparative objectives, models fine-tuned in this manner may still exhibit undesirable behaviors in the intricate and realistic retrieval scenarios. To this end, we propose a Knowledge-aware Preference Optimization, dubbed KaPO, aimed at achieving controllable knowledge selection in real retrieval scenarios. Concretely, we explore and simulate error types across diverse context combinations and learn how to avoid these negative signals through preference optimization methods. Simultaneously, by adjusting the balance between response length and the proportion of preference data representing different behavior patterns, we enhance the adherence capabilities and noise robustness of LLMs in a balanced manner. Experimental results show that KaPO outperforms previous methods for handling knowledge conflicts by over 37%, while also exhibiting robust generalization across various out-of-distribution datasets.
Authors: Amie J. Paige, Adil Soubki, John Murzaku, Owen Rambow, Susan E. Brennan
Abstract: Hedges allow speakers to mark utterances as provisional, whether to signal non-prototypicality or "fuzziness", to indicate a lack of commitment to an utterance, to attribute responsibility for a statement to someone else, to invite input from a partner, or to soften critical feedback in the service of face-management needs. Here we focus on hedges in an experimentally parameterized corpus of 63 Roadrunner cartoon narratives spontaneously produced from memory by 21 speakers for co-present addressees, transcribed to text (Galati and Brennan, 2010). We created a gold standard of hedges annotated by human coders (the Roadrunner-Hedge corpus) and compared three LLM-based approaches for hedge detection: fine-tuning BERT, and zero and few-shot prompting with GPT-4o and LLaMA-3. The best-performing approach was a fine-tuned BERT model, followed by few-shot GPT-4o. After an error analysis on the top performing approaches, we used an LLM-in-the-Loop approach to improve the gold standard coding, as well as to highlight cases in which hedges are ambiguous in linguistically interesting ways that will guide future research. This is the first step in our research program to train LLMs to interpret and generate collateral signals appropriately and meaningfully in conversation.
Authors: Alon Jacovi, Moran Ambar, Eyal Ben-David, Uri Shaham, Amir Feder, Mor Geva, Dror Marcus, Avi Caciularu
Abstract: There is a growing line of research on verifying the correctness of language models' outputs. At the same time, LMs are being used to tackle complex queries that require reasoning. We introduce CoverBench, a challenging benchmark focused on verifying LM outputs in complex reasoning settings. Datasets that can be used for this purpose are often designed for other complex reasoning tasks (e.g., QA) targeting specific use-cases (e.g., financial tables), requiring transformations, negative sampling and selection of hard examples to collect such a benchmark. CoverBench provides a diversified evaluation for complex claim verification in a variety of domains, types of reasoning, relatively long inputs, and a variety of standardizations, such as multiple representations for tables where available, and a consistent schema. We manually vet the data for quality to ensure low levels of label noise. Finally, we report a variety of competitive baseline results to show CoverBench is challenging and has very significant headroom. The data is available at https://huggingface.co/datasets/google/coverbench .
Authors: Hassan S. Shavarani, Anoop Sarkar
Abstract: The similarity between the question and indexed documents is a crucial factor in document retrieval for retrieval-augmented question answering. Although this is typically the only method for obtaining the relevant documents, it is not the sole approach when dealing with entity-centric questions. In this study, we propose Entity Retrieval, a novel retrieval method which rather than relying on question-document similarity, depends on the salient entities within the question to identify the retrieval documents. We conduct an in-depth analysis of the performance of both dense and sparse retrieval methods in comparison to Entity Retrieval. Our findings reveal that our method not only leads to more accurate answers to entity-centric questions but also operates more efficiently.
Authors: Eduardo Sanchez-Karhunen, Jose F. Quesada-Moreno, Miguel A. Guti\'errez-Naranjo
Abstract: Intent detection is a text classification task whose aim is to recognize and label the semantics behind a users query. It plays a critical role in various business applications. The output of the intent detection module strongly conditions the behavior of the whole system. This sequence analysis task is mainly tackled using deep learning techniques. Despite the widespread use of these techniques, the internal mechanisms used by networks to solve the problem are poorly understood. Recent lines of work have analyzed the computational mechanisms learned by RNNs from a dynamical systems perspective. In this work, we investigate how different RNN architectures solve the SNIPS intent detection problem. Sentences injected into trained networks can be interpreted as trajectories traversing a hidden state space. This space is constrained to a low-dimensional manifold whose dimensionality is related to the embedding and hidden layer sizes. To generate predictions, RNN steers the trajectories towards concrete regions, spatially aligned with the output layer matrix rows directions. Underlying the system dynamics, an unexpected fixed point topology has been identified with a limited number of attractors. Our results provide new insights into the inner workings of networks that solve the intent detection task.
Authors: Zihan Li, Diping Song, Zefeng Yang, Deming Wang, Fei Li, Xiulan Zhang, Paul E. Kinahan, Yu Qiao
Abstract: The need for improved diagnostic methods in ophthalmology is acute, especially in the less developed regions with limited access to specialists and advanced equipment. Therefore, we introduce VisionUnite, a novel vision-language foundation model for ophthalmology enhanced with clinical knowledge. VisionUnite has been pretrained on an extensive dataset comprising 1.24 million image-text pairs, and further refined using our proposed MMFundus dataset, which includes 296,379 high-quality fundus image-text pairs and 889,137 simulated doctor-patient dialogue instances. Our experiments indicate that VisionUnite outperforms existing generative foundation models such as GPT-4V and Gemini Pro. It also demonstrates diagnostic capabilities comparable to junior ophthalmologists. VisionUnite performs well in various clinical scenarios including open-ended multi-disease diagnosis, clinical explanation, and patient interaction, making it a highly versatile tool for initial ophthalmic disease screening. VisionUnite can also serve as an educational aid for junior ophthalmologists, accelerating their acquisition of knowledge regarding both common and rare ophthalmic conditions. VisionUnite represents a significant advancement in ophthalmology, with broad implications for diagnostics, medical education, and understanding of disease mechanisms.
Authors: Taichi Nishimura, Shota Nakada, Hokuto Munakata, Tatsuya Komatsu
Abstract: We propose Lighthouse, a user-friendly library for reproducible video moment retrieval and highlight detection (MR-HD). Although researchers proposed various MR-HD approaches, the research community holds two main issues. The first is a lack of comprehensive and reproducible experiments across various methods, datasets, and video-text features. This is because no unified training and evaluation codebase covers multiple settings. The second is user-unfriendly design. Because previous works use different libraries, researchers set up individual environments. In addition, most works release only the training codes, requiring users to implement the whole inference process of MR-HD. Lighthouse addresses these issues by implementing a unified reproducible codebase that includes six models, three features, and five datasets. In addition, it provides an inference API and web demo to make these methods easily accessible for researchers and developers. Our experiments demonstrate that Lighthouse generally reproduces the reported scores in the reference papers. The code is available at https://github.com/line/lighthouse.
Authors: Yuxin Wang, Duanyu Feng, Yongfu Dai, Zhengyu Chen, Jimin Huang, Sophia Ananiadou, Qianqian Xie, Hao Wang
Abstract: Data serves as the fundamental foundation for advancing deep learning, particularly tabular data presented in a structured format, which is highly conducive to modeling. However, even in the era of LLM, obtaining tabular data from sensitive domains remains a challenge due to privacy or copyright concerns. Hence, exploring how to effectively use models like LLMs to generate realistic and privacy-preserving synthetic tabular data is urgent. In this paper, we take a step forward to explore LLMs for tabular data synthesis and privacy protection, by introducing a new framework HARMONIC for tabular data generation and evaluation. In the tabular data generation of our framework, unlike previous small-scale LLM-based methods that rely on continued pre-training, we explore the larger-scale LLMs with fine-tuning to generate tabular data and enhance privacy. Based on idea of the k-nearest neighbors algorithm, an instruction fine-tuning dataset is constructed to inspire LLMs to discover inter-row relationships. Then, with fine-tuning, LLMs are trained to remember the format and connections of the data rather than the data itself, which reduces the risk of privacy leakage. In the evaluation part of our framework, we develop specific privacy risk metrics DLT for LLM synthetic data generation, as well as performance evaluation metrics LLE for downstream LLM tasks. Our experiments find that this tabular data generation framework achieves equivalent performance to existing methods with better privacy, which also demonstrates our evaluation framework for the effectiveness of synthetic data and privacy risks in LLM scenarios.
Authors: Qiang Sun, Yuanyi Luo, Sirui Li, Wenxiao Zhang, Wei Liu
Abstract: Multimodal conversational agents are highly desirable because they offer natural and human-like interaction. However, there is a lack of comprehensive end-to-end solutions to support collaborative development and benchmarking. While proprietary systems like GPT-4o and Gemini demonstrating impressive integration of audio, video, and text with response times of 200-250ms, challenges remain in balancing latency, accuracy, cost, and data privacy. To better understand and quantify these issues, we developed OpenOmni, an open-source, end-to-end pipeline benchmarking tool that integrates advanced technologies such as Speech-to-Text, Emotion Detection, Retrieval Augmented Generation, Large Language Models, along with the ability to integrate customized models. OpenOmni supports local and cloud deployment, ensuring data privacy and supporting latency and accuracy benchmarking. This flexible framework allows researchers to customize the pipeline, focusing on real bottlenecks and facilitating rapid proof-of-concept development. OpenOmni can significantly enhance applications like indoor assistance for visually impaired individuals, advancing human-computer interaction. Our demonstration video is available https://www.youtube.com/watch?v=zaSiT3clWqY, demo is available via https://openomni.ai4wa.com, code is available via https://github.com/AI4WA/OpenOmniFramework.
URLs: https://www.youtube.com/watch?v=zaSiT3clWqY,, https://openomni.ai4wa.com,, https://github.com/AI4WA/OpenOmniFramework.
Authors: Yanghai Zhang, Ye Liu, Shiwei Wu, Kai Zhang, Xukai Liu, Qi Liu, Enhong Chen
Abstract: The rapid increase in multimedia data has spurred advancements in Multimodal Summarization with Multimodal Output (MSMO), which aims to produce a multimodal summary that integrates both text and relevant images. The inherent heterogeneity of content within multimodal inputs and outputs presents a significant challenge to the execution of MSMO. Traditional approaches typically adopt a holistic perspective on coarse image-text data or individual visual objects, overlooking the essential connections between objects and the entities they represent. To integrate the fine-grained entity knowledge, we propose an Entity-Guided Multimodal Summarization model (EGMS). Our model, building on BART, utilizes dual multimodal encoders with shared weights to process text-image and entity-image information concurrently. A gating mechanism then combines visual data for enhanced textual summary generation, while image selection is refined through knowledge distillation from a pre-trained vision-language model. Extensive experiments on public MSMO dataset validate the superiority of the EGMS method, which also prove the necessity to incorporate entity information into MSMO problem.
Authors: Jihao Gu, Shuai Chen, Zelin Wang, Yibo Zhang, Ping Gong
Abstract: With the increasing number of parameters in large pre-trained models, LoRA as a parameter-efficient fine-tuning(PEFT) method is widely used for not adding inference overhead. The LoRA method assumes that weight changes during fine-tuning can be approximated by low-rank matrices. However, the rank values need to be manually verified to match different downstream tasks, and they cannot accommodate the varying importance of different layers in the model. In this work, we first analyze the relationship between the performance of different layers and their ranks using SVD. Based on this, we design the Singular-Value Based Adaptive Low-Rank Adaption(SARA), which adaptively finds the rank during initialization by performing SVD on the pre-trained weights. Additionally, we explore the Mixture-of-SARA(Mo-SARA), which significantly reduces the number of parameters by fine-tuning only multiple parallel sets of singular values controlled by a router. Extensive experiments on various complex tasks demonstrate the simplicity and parameter efficiency of our methods. They can effectively and adaptively find the most suitable rank for each layer of each model.
Authors: Charlie Snell, Jaehoon Lee, Kelvin Xu, Aviral Kumar
Abstract: Enabling LLMs to improve their outputs by using more test-time computation is a critical step towards building generally self-improving agents that can operate on open-ended natural language. In this paper, we study the scaling of inference-time computation in LLMs, with a focus on answering the question: if an LLM is allowed to use a fixed but non-trivial amount of inference-time compute, how much can it improve its performance on a challenging prompt? Answering this question has implications not only on the achievable performance of LLMs, but also on the future of LLM pretraining and how one should tradeoff inference-time and pre-training compute. Despite its importance, little research attempted to understand the scaling behaviors of various test-time inference methods. Moreover, current work largely provides negative results for a number of these strategies. In this work, we analyze two primary mechanisms to scale test-time computation: (1) searching against dense, process-based verifier reward models; and (2) updating the model's distribution over a response adaptively, given the prompt at test time. We find that in both cases, the effectiveness of different approaches to scaling test-time compute critically varies depending on the difficulty of the prompt. This observation motivates applying a "compute-optimal" scaling strategy, which acts to most effectively allocate test-time compute adaptively per prompt. Using this compute-optimal strategy, we can improve the efficiency of test-time compute scaling by more than 4x compared to a best-of-N baseline. Additionally, in a FLOPs-matched evaluation, we find that on problems where a smaller base model attains somewhat non-trivial success rates, test-time compute can be used to outperform a 14x larger model.
Authors: Bo Li, Yuanhan Zhang, Dong Guo, Renrui Zhang, Feng Li, Hao Zhang, Kaichen Zhang, Yanwei Li, Ziwei Liu, Chunyuan Li
Abstract: We present LLaVA-OneVision, a family of open large multimodal models (LMMs) developed by consolidating our insights into data, models, and visual representations in the LLaVA-NeXT blog series. Our experimental results demonstrate that LLaVA-OneVision is the first single model that can simultaneously push the performance boundaries of open LMMs in three important computer vision scenarios: single-image, multi-image, and video scenarios. Importantly, the design of LLaVA-OneVision allows strong transfer learning across different modalities/scenarios, yielding new emerging capabilities. In particular, strong video understanding and cross-scenario capabilities are demonstrated through task transfer from images to videos.
Authors: Hassan Haji Mohammadi, Alireza Talebpour, Ahmad Mahmoudi Aznaveh, Samaneh Yazdani
Abstract: Coreference resolution (CR), identifying expressions referring to the same real-world entity, is a fundamental challenge in natural language processing (NLP). This paper explores the latest advancements in CR, spanning coreference and anaphora resolution. We critically analyze the diverse corpora that have fueled CR research, highlighting their strengths, limitations, and suitability for various tasks. We examine the spectrum of evaluation metrics used to assess CR systems, emphasizing their advantages, disadvantages, and the need for more nuanced, task-specific metrics. Tracing the evolution of CR algorithms, we provide a detailed overview of methodologies, from rule-based approaches to cutting-edge deep learning architectures. We delve into mention-pair, entity-based, cluster-ranking, sequence-to-sequence, and graph neural network models, elucidating their theoretical foundations and performance on benchmark datasets. Recognizing the unique challenges of Persian CR, we dedicate a focused analysis to this under-resourced language. We examine existing Persian CR systems and highlight the emergence of end-to-end neural models leveraging pre-trained language models like ParsBERT. This review is an essential resource for researchers and practitioners, offering a comprehensive overview of the current state-of-the-art in CR, identifying key challenges, and charting a course for future research in this rapidly evolving field.
Authors: Yujia Qin, Shengding Hu, Yankai Lin, Weize Chen, Ning Ding, Ganqu Cui, Zheni Zeng, Yufei Huang, Chaojun Xiao, Chi Han, Yi Ren Fung, Yusheng Su, Huadong Wang, Cheng Qian, Runchu Tian, Kunlun Zhu, Shihao Liang, Xingyu Shen, Bokai Xu, Zhen Zhang, Yining Ye, Bowen Li, Ziwei Tang, Jing Yi, Yuzhang Zhu, Zhenning Dai, Lan Yan, Xin Cong, Yaxi Lu, Weilin Zhao, Yuxiang Huang, Junxi Yan, Xu Han, Xian Sun, Dahai Li, Jason Phang, Cheng Yang, Tongshuang Wu, Heng Ji, Zhiyuan Liu, Maosong Sun
Abstract: Humans possess an extraordinary ability to create and utilize tools, allowing them to overcome physical limitations and explore new frontiers. With the advent of foundation models, AI systems have the potential to be equally adept in tool use as humans. This paradigm, i.e., tool learning with foundation models, combines the strengths of specialized tools and foundation models to achieve enhanced accuracy, efficiency, and automation in problem-solving. Despite its immense potential, there is still a lack of a comprehensive understanding of key challenges, opportunities, and future endeavors in this field. To this end, we present a systematic investigation of tool learning in this paper. We first introduce the background of tool learning, including its cognitive origins, the paradigm shift of foundation models, and the complementary roles of tools and models. Then we recapitulate existing tool learning research into tool-augmented and tool-oriented learning. We formulate a general tool learning framework: starting from understanding the user instruction, models should learn to decompose a complex task into several subtasks, dynamically adjust their plan through reasoning, and effectively conquer each sub-task by selecting appropriate tools. We also discuss how to train models for improved tool-use capabilities and facilitate the generalization in tool learning. Considering the lack of a systematic tool learning evaluation in prior works, we experiment with 18 representative tools and show the potential of current foundation models in skillfully utilizing tools. Finally, we discuss several open problems that require further investigation for tool learning. In general, we hope this paper could inspire future research in integrating tools with foundation models.
Authors: Marcos Gomez-Vazquez, Jordi Cabot, Robert Claris\'o
Abstract: Tabular data is the most common format to publish and exchange structured data online. A clear example is the growing number of open data portals published by public administrations. However, exploitation of these data sources is currently limited to technical people able to programmatically manipulate and digest such data. As an alternative, we propose the use of chatbots to offer a conversational interface to facilitate the exploration of tabular data sources, including support for data analytics questions that are responded via charts rendered by the chatbot. Moreover, our chatbots are automatically generated from the data source itself thanks to the instantiation of a configurable collection of conversation patterns matched to the chatbot intents and entities.
Authors: Yan Zhuang, Qi Liu, Yuting Ning, Weizhe Huang, Zachary A. Pardos, Patrick C. Kyllonen, Jiyun Zu, Qingyang Mao, Rui Lv, Zhenya Huang, Guanhao Zhao, Zheng Zhang, Shijin Wang, Enhong Chen
Abstract: As AI systems continue to grow, particularly generative models like Large Language Models (LLMs), their rigorous evaluation is crucial for development and deployment. To determine their adequacy, researchers have developed various large-scale benchmarks against a so-called gold-standard test set and report metrics averaged across all items. However, this static evaluation paradigm increasingly shows its limitations, including high computational costs, data contamination, and the impact of low-quality or erroneous items on evaluation reliability and efficiency. In this Perspective, drawing from human psychometrics, we discuss a paradigm shift from static evaluation methods to adaptive testing. This involves estimating the characteristics and value of each test item in the benchmark and dynamically adjusting items in real-time, tailoring the evaluation based on the model's ongoing performance instead of relying on a fixed test set. This paradigm not only provides a more robust ability estimation but also significantly reduces the number of test items required. We analyze the current approaches, advantages, and underlying reasons for adopting psychometrics in AI evaluation. We propose that adaptive testing will become the new norm in AI model evaluation, enhancing both the efficiency and effectiveness of assessing advanced intelligence systems.
Authors: Mike D'Arcy, Alexis Ross, Erin Bransom, Bailey Kuehl, Jonathan Bragg, Tom Hope, Doug Downey
Abstract: We introduce the task of automatically revising scientific papers based on peer feedback and release ARIES, a dataset of review comments and their corresponding paper edits. The data is drawn from real reviewer-author interactions from computer science, and we provide labels linking each reviewer comment to the specific paper edits made by the author in response. We automatically create a high-precision silver training set, as well as an expert-labeled test set that shows high inter-annotator agreement. In experiments with 10 models covering the state of the art, we find that they struggle even to identify which edits correspond to a comment -- especially when the relationship between the edit and the comment is indirect and requires reasoning to uncover. We also extensively analyze GPT-4's ability to generate edits given a comment and the original paper. We find that it often succeeds on a superficial level, but tends to rigidly follow the wording of the feedback rather than the underlying intent, and lacks technical details compared to human-written edits.
Authors: Andrei C. Coman, Christos Theodoropoulos, Marie-Francine Moens, James Henderson
Abstract: Document-level relation extraction typically relies on text-based encoders and hand-coded pooling heuristics to aggregate information learned by the encoder. In this paper, we leverage the intrinsic graph processing capabilities of the Transformer model and propose replacing hand-coded pooling methods with new tokens in the input, which are designed to aggregate information via explicit graph relations in the computation of attention weights. We introduce a joint text-graph Transformer model and a graph-assisted declarative pooling (GADePo) specification of the input, which provides explicit and high-level instructions for information aggregation. GADePo allows the pooling process to be guided by domain-specific knowledge or desired outcomes but still learned by the Transformer, leading to more flexible and customisable pooling strategies. We evaluate our method across diverse datasets and models and show that our approach yields promising results that are consistently better than those achieved by the hand-coded pooling functions.
Authors: Tianlong Li, Shihan Dou, Wenhao Liu, Muling Wu, Changze Lv, Rui Zheng, Xiaoqing Zheng, Xuanjing Huang
Abstract: The recent surge in jailbreaking methods has revealed the vulnerability of Large Language Models (LLMs) to malicious inputs. While earlier research has primarily concentrated on increasing the success rates of jailbreaking attacks, the underlying mechanism for safeguarding LLMs remains underexplored. This study investigates the vulnerability of safety-aligned LLMs by uncovering specific activity patterns within the representation space generated by LLMs. Such ``safety patterns'' can be identified with only a few pairs of contrastive queries in a simple method and function as ``keys'' (used as a metaphor for security defense capability) that can be used to open or lock Pandora's Box of LLMs. Extensive experiments demonstrate that the robustness of LLMs against jailbreaking can be lessened or augmented by attenuating or strengthening the identified safety patterns. These findings deepen our understanding of jailbreaking phenomena and call for the LLM community to address the potential misuse of open-source LLMs.
Authors: Jimin Hong, Gibbeum Lee, Jaewoong Cho
Abstract: Recent advancements in large language models (LLMs) have remarkably enhanced performances on a variety of tasks in multiple languages. However, tokenizers in LLMs trained primarily on English-centric corpora often overly fragment a text into character or Unicode-level tokens in non-Roman alphabetic languages, leading to inefficient text generation. We introduce a simple yet effective framework to accelerate text generation in such languages. Our approach involves employing a new language model head with a vocabulary set tailored to a specific target language for a pre-trained LLM. This is followed by fine-tuning the new head while incorporating a verification step to ensure the model's performance is preserved. We show that this targeted fine-tuning, while freezing other model parameters, effectively reduces token fragmentation for the target language. Our extensive experiments demonstrate that the proposed framework increases the generation speed by a factor of 1.7 while maintaining the performance of pre-trained multilingual models on target monolingual tasks.
Authors: Bo Pan, Zheng Zhang, Yifei Zhang, Yuntong Hu, Liang Zhao
Abstract: Text-Attributed Graphs (TAGs) are graphs of connected textual documents. Graph models can efficiently learn TAGs, but their training heavily relies on human-annotated labels, which are scarce or even unavailable in many applications. Large language models (LLMs) have recently demonstrated remarkable capabilities in few-shot and zero-shot TAG learning, but they suffer from scalability, cost, and privacy issues. Therefore, in this work, we focus on synergizing LLMs and graph models with their complementary strengths by distilling the power of LLMs to a local graph model on TAG learning. To address the inherent gaps between LLMs (generative models for texts) and graph models (discriminative models for graphs), we propose first to let LLMs teach an interpreter with rich textual rationale and then let a student model mimic the interpreter's reasoning without LLMs' textual rationale. Extensive experiments validate the efficacy of our proposed framework.
Authors: Yi Zong, Xipeng Qiu
Abstract: The Large Vision-Language Models (LVLMs) have demonstrated great abilities in image perception and language understanding. However, existing multimodal benchmarks focus on primary perception abilities and commonsense knowledge which are insufficient to reflect the comprehensive capabilities of LVLMs. We propose GAOKAO-MM, a multimodal benchmark based on the Chinese College Entrance Examination (GAOKAO), comprising of 8 subjects and 12 types of images, such as diagrams, function graphs, maps and photos. GAOKAO-MM derives from native Chinese context and sets human-level requirements for the model's abilities, including perception, understanding, knowledge and reasoning. We evaluate 10 LVLMs and find that the accuracies of all of them are lower than 50%, with GPT-4-Vison (48.1%), Qwen-VL-Plus (41.2%) and Gemini-Pro-Vision (35.1%) ranking in the top three positions. The results of our multi-dimension analysis indicate that LVLMs have moderate distance towards Artificial General Intelligence (AGI) and provide insights facilitating the development of multilingual LVLMs.
Authors: Xuannan Liu, Peipei Li, Huaibo Huang, Zekun Li, Xing Cui, Jiahao Liang, Lixiong Qin, Weihong Deng, Zhaofeng He
Abstract: The massive generation of multimodal fake news involving both text and images exhibits substantial distribution discrepancies, prompting the need for generalized detectors. However, the insulated nature of training restricts the capability of classical detectors to obtain open-world facts. While Large Vision-Language Models (LVLMs) have encoded rich world knowledge, they are not inherently tailored for combating fake news and struggle to comprehend local forgery details. In this paper, we propose FKA-Owl, a novel framework that leverages forgery-specific knowledge to augment LVLMs, enabling them to reason about manipulations effectively. The augmented forgery-specific knowledge includes semantic correlation between text and images, and artifact trace in image manipulation. To inject these two kinds of knowledge into the LVLM, we design two specialized modules to establish their representations, respectively. The encoded knowledge embeddings are then incorporated into LVLMs. Extensive experiments on the public benchmark demonstrate that FKA-Owl achieves superior cross-domain performance compared to previous methods. Code is publicly available at https://liuxuannan.github.io/FKA_Owl.github.io/.
Authors: Zhenrong Cheng, Jiayan Guo, Hao Sun, Yan Zhang
Abstract: Current disfluency detection methods heavily rely on costly and scarce human-annotated data. To tackle this issue, some approaches employ heuristic or statistical features to generate disfluent sentences, partially improving detection performance. However, these sentences often deviate from real-life scenarios, constraining overall model enhancement. In this study, we propose a lightweight data augmentation approach for disfluency detection, utilizing the superior generative and semantic understanding capabilities of large language model (LLM) to generate disfluent sentences as augmentation data. We leverage LLM to generate diverse and more realistic sentences guided by specific prompts, without the need for fine-tuning the LLM. Subsequently, we apply an uncertainty-aware data filtering approach to improve the quality of the generated sentences, utilized in training a small detection model for improved performance. Experiments using enhanced data yielded state-of-the-art results. The results showed that using a small amount of LLM-generated enhanced data can significantly improve performance, thereby further enhancing cost-effectiveness. Our code is available here.
Authors: Philipp Mondorf, Barbara Plank
Abstract: Large language models (LLMs) have recently shown impressive performance on tasks involving reasoning, leading to a lively debate on whether these models possess reasoning capabilities similar to humans. However, despite these successes, the depth of LLMs' reasoning abilities remains uncertain. This uncertainty partly stems from the predominant focus on task performance, measured through shallow accuracy metrics, rather than a thorough investigation of the models' reasoning behavior. This paper seeks to address this gap by providing a comprehensive review of studies that go beyond task accuracy, offering deeper insights into the models' reasoning processes. Furthermore, we survey prevalent methodologies to evaluate the reasoning behavior of LLMs, emphasizing current trends and efforts towards more nuanced reasoning analyses. Our review suggests that LLMs tend to rely on surface-level patterns and correlations in their training data, rather than on sophisticated reasoning abilities. Additionally, we identify the need for further research that delineates the key differences between human and LLM-based reasoning. Through this survey, we aim to shed light on the complex reasoning processes within LLMs.
Authors: Marco Bronzini, Carlo Nicolini, Bruno Lepri, Jacopo Staiano, Andrea Passerini
Abstract: Large Language Models (LLMs) demonstrate an impressive capacity to recall a vast range of factual knowledge. However, understanding their underlying reasoning and internal mechanisms in exploiting this knowledge remains a key research area. This work unveils the factual information an LLM represents internally for sentence-level claim verification. We propose an end-to-end framework to decode factual knowledge embedded in token representations from a vector space to a set of ground predicates, showing its layer-wise evolution using a dynamic knowledge graph. Our framework employs activation patching, a vector-level technique that alters a token representation during inference, to extract encoded knowledge. Accordingly, we neither rely on training nor external models. Using factual and common-sense claims from two claim verification datasets, we showcase interpretability analyses at local and global levels. The local analysis highlights entity centrality in LLM reasoning, from claim-related information and multi-hop reasoning to representation errors causing erroneous evaluation. On the other hand, the global reveals trends in the underlying evolution, such as word-based knowledge evolving into claim-related facts. By interpreting semantics from LLM latent representations and enabling graph-related analyses, this work enhances the understanding of the factual knowledge resolution process.
Authors: Tim Baumg\"artner, Yang Gao, Dana Alon, Donald Metzler
Abstract: Reinforcement Learning from Human Feedback (RLHF) is a popular method for aligning Language Models (LM) with human values and preferences. RLHF requires a large number of preference pairs as training data, which are often used in both the Supervised Fine-Tuning and Reward Model training and therefore publicly available datasets are commonly used. In this work, we study to what extent a malicious actor can manipulate the LMs generations by poisoning the preferences, i.e., injecting poisonous preference pairs into these datasets and the RLHF training process. We propose strategies to build poisonous preference pairs and test their performance by poisoning two widely used preference datasets. Our results show that preference poisoning is highly effective: injecting a small amount of poisonous data (1-5\% of the original dataset), we can effectively manipulate the LM to generate a target entity in a target sentiment (positive or negative). The findings from our experiments also shed light on strategies to defend against the preference poisoning attack.
Authors: Milan Bhan, Jean-Noel Vittaut, Nina Achache, Victor Legrand, Nicolas Chesneau, Annabelle Blangero, Juliette Murris, Marie-Jeanne Lesot
Abstract: Toxicity mitigation consists in rephrasing text in order to remove offensive or harmful meaning. Neural natural language processing (NLP) models have been widely used to target and mitigate textual toxicity. However, existing methods fail to detoxify text while preserving the initial non-toxic meaning at the same time. In this work, we propose to apply counterfactual generation methods from the eXplainable AI (XAI) field to target and mitigate textual toxicity. In particular, we perform text detoxification by applying local feature importance and counterfactual generation methods to a toxicity classifier distinguishing between toxic and non-toxic texts. We carry out text detoxification through counterfactual generation on three datasets and compare our approach to three competitors. Automatic and human evaluations show that recently developed NLP counterfactual generators can mitigate toxicity accurately while better preserving the meaning of the initial text as compared to classical detoxification methods. Finally, we take a step back from using automated detoxification tools, and discuss how to manage the polysemous nature of toxicity and the risk of malicious use of detoxification tools. This work is the first to bridge the gap between counterfactual generation and text detoxification and paves the way towards more practical application of XAI methods.
Authors: David Stap, Eva Hasler, Bill Byrne, Christof Monz, Ke Tran
Abstract: Fine-tuning large language models (LLMs) for machine translation has shown improvements in overall translation quality. However, it is unclear what is the impact of fine-tuning on desirable LLM behaviors that are not present in neural machine translation models, such as steerability, inherent document-level translation abilities, and the ability to produce less literal translations. We perform an extensive translation evaluation on the LLaMA and Falcon family of models with model size ranging from 7 billion up to 65 billion parameters. Our results show that while fine-tuning improves the general translation quality of LLMs, several abilities degrade. In particular, we observe a decline in the ability to perform formality steering, to produce technical translations through few-shot examples, and to perform document-level translation. On the other hand, we observe that the model produces less literal translations after fine-tuning on parallel data. We show that by including monolingual data as part of the fine-tuning data we can maintain the abilities while simultaneously enhancing overall translation quality. Our findings emphasize the need for fine-tuning strategies that preserve the benefits of LLMs for machine translation.
Authors: Karuna Bhaila, Minh-Hao Van, Xintao Wu
Abstract: The widespread popularity of Large Language Models (LLMs), partly due to their unique ability to perform in-context learning, has also brought to light the importance of ethical and safety considerations when deploying these pre-trained models. In this work, we focus on investigating machine unlearning for LLMs motivated by data protection regulations. In contrast to the growing literature on fine-tuning methods to achieve unlearning, we focus on a comparatively lightweight alternative called soft prompting to realize the unlearning of a subset of training data. With losses designed to enforce forgetting as well as utility preservation, our framework \textbf{S}oft \textbf{P}rompting for \textbf{U}n\textbf{l}earning (SPUL) learns prompt tokens that can be appended to an arbitrary query to induce unlearning of specific examples at inference time without updating LLM parameters. We conduct a rigorous evaluation of the proposed method and our results indicate that SPUL can significantly improve the trade-off between utility and forgetting in the context of text classification and question answering with LLMs. We further validate our method using multiple LLMs to highlight the scalability of our framework and provide detailed insights into the choice of hyperparameters and the influence of the size of unlearning data. Our implementation is available at \url{https://github.com/karuna-bhaila/llm_unlearning}.
Authors: Chenlong Deng, Kelong Mao, Yuyao Zhang, Zhicheng Dou
Abstract: Legal judgment prediction is essential for enhancing judicial efficiency. In this work, we identify that existing large language models (LLMs) underperform in this domain due to challenges in understanding case complexities and distinguishing between similar charges. To adapt LLMs for effective legal judgment prediction, we introduce the Ask-Discriminate-Predict (ADAPT) reasoning framework inspired by human judicial reasoning. ADAPT involves decomposing case facts, discriminating among potential charges, and predicting the final judgment. We further enhance LLMs through fine-tuning with multi-task synthetic trajectories to improve legal judgment prediction accuracy and efficiency under our ADAPT framework. Extensive experiments conducted on two widely-used datasets demonstrate the superior performance of our framework in legal judgment prediction, particularly when dealing with complex and confusing charges.
Authors: Shengjie Ma, Chengjin Xu, Xuhui Jiang, Muzhi Li, Huaren Qu, Jian Guo
Abstract: Retrieval-augmented generation (RAG) has significantly advanced large language models (LLMs) by enabling dynamic information retrieval to mitigate knowledge gaps and hallucinations in generated content. However, these systems often falter with complex reasoning and consistency across diverse queries. In this work, we present Think-on-Graph 2.0, an enhanced RAG framework that aligns questions with the knowledge graph and uses it as a navigational tool, which deepens and refines the RAG paradigm for information collection and integration. The KG-guided navigation fosters deep and long-range associations to uphold logical consistency and optimize the scope of retrieval for precision and interoperability. In conjunction, factual consistency can be better ensured through semantic similarity guided by precise directives. ToG${2.0}$ not only improves the accuracy and reliability of LLMs' responses but also demonstrates the potential of hybrid structured knowledge systems to significantly advance LLM reasoning, aligning it closer to human-like performance. We conducted extensive experiments on four public datasets to demonstrate the advantages of our method compared to the baseline.
Authors: Yukio Ohsawa, Dingming Xue, Kaira Sekiguchi
Abstract: The senses of words evolve. The sense of the same word may change from today to tomorrow, and multiple senses of the same word may be the result of the evolution of each other, that is, they may be parents and children. If we view Juba as an evolving ecosystem, the paradigm of learning the correct answer, which does not move with the sense of a word, is no longer valid. This paper is a case study that shows that word polysemy is an evolutionary consequence of the modification of Semantic Cells, which has al-ready been presented by the author, by introducing a small amount of diversity in its initial state as an example of analyzing the current set of short sentences. In particular, the analysis of a sentence sequence of 1000 sentences in some order for each of the four senses of the word Spring, collected using Chat GPT, shows that the word acquires the most polysemy monotonically in the analysis when the senses are arranged in the order in which they have evolved. In other words, we present a method for analyzing the dynamism of a word's acquiring polysemy with evolution and, at the same time, a methodology for viewing polysemy from an evolutionary framework rather than a learning-based one.
Authors: Bowen Wang, Jiuyang Chang, Yiming Qian, Guoxin Chen, Junhao Chen, Zhouqiang Jiang, Jiahao Zhang, Yuta Nakashima, Hajime Nagahara
Abstract: Large language models (LLMs) have recently showcased remarkable capabilities, spanning a wide range of tasks and applications, including those in the medical domain. Models like GPT-4 excel in medical question answering but may face challenges in the lack of interpretability when handling complex tasks in real clinical settings. We thus introduce the diagnostic reasoning dataset for clinical notes (DiReCT), aiming at evaluating the reasoning ability and interpretability of LLMs compared to human doctors. It contains 511 clinical notes, each meticulously annotated by physicians, detailing the diagnostic reasoning process from observations in a clinical note to the final diagnosis. Additionally, a diagnostic knowledge graph is provided to offer essential knowledge for reasoning, which may not be covered in the training data of existing LLMs. Evaluations of leading LLMs on DiReCT bring out a significant gap between their reasoning ability and that of human doctors, highlighting the critical need for models that can reason effectively in real-world clinical scenarios.
Authors: Jian Gu, Aldeida Aleti, Chunyang Chen, Hongyu Zhang
Abstract: Large Language Models (LLMs) have already gained widespread adoption in software engineering, particularly in code generation tasks. However, updating these models with new knowledge can be prohibitively expensive, yet it is essential to maximize their utility, such as implementing a hotfix technique to address urgent or critical LLM errors. In this paper, we propose \textsc{MENT}, a novel and effective model editing approach to repair LLMs in coding tasks. \textsc{MENT} is effective, efficient, and reliable, capable of correcting a neural model by patching just one or two neurons. As pioneering work on neuron-level model editing of generative models, we formalize the editing process and introduce the involved concepts. We also introduce new measures to evaluate its generalization ability and establish a benchmark for further study. Our approach is evaluated on three coding tasks: line-level code generation, shellcode generation, and intent-to-bash translation. The experimental results demonstrate that the proposed approach significantly outperforms the state-of-the-art in both effectiveness and efficiency measures. Furthermore, we showcase the applications of \textsc{MENT} for LLM reasoning in software engineering. By editing LLM knowledge, the directly or indirectly dependent behaviors of API invocation in the chain-of-thought change accordingly. This illustrates the significance of repairing LLMs in the context of software engineering.
Authors: Jirayu Burapacheep, Ishan Gaur, Agam Bhatia, Tristan Thrush
Abstract: This paper introduces the ColorSwap dataset, designed to assess and improve the proficiency of multimodal models in matching objects with their colors. The dataset is comprised of 2,000 unique image-caption pairs, grouped into 1,000 examples. Each example includes a caption-image pair, along with a ``color-swapped'' pair. We follow the Winoground schema: the two captions in an example have the same words, but the color words have been rearranged to modify different objects. The dataset was created through a novel blend of automated caption and image generation with humans in the loop. We evaluate image-text matching (ITM) and visual language models (VLMs) and find that even the latest ones are still not robust at this task. GPT-4V and LLaVA score 72% and 42% on our main VLM metric, although they may improve with more advanced prompting techniques. On the main ITM metric, contrastive models such as CLIP and SigLIP perform close to chance (at 12% and 30%, respectively), although the non-contrastive BLIP ITM model is stronger (87%). We also find that finetuning on fewer than 2,000 examples yields significant performance gains on this out-of-distribution word-order understanding task. The dataset is here: https://github.com/Top34051/colorswap and here: https://huggingface.co/datasets/stanfordnlp/colorswap.
URLs: https://github.com/Top34051/colorswap, https://huggingface.co/datasets/stanfordnlp/colorswap.
Authors: Changmeng Zheng, Dayong Liang, Wengyu Zhang, Xiao-Yong Wei, Tat-Seng Chua, Qing Li
Abstract: This paper presents a pilot study aimed at introducing multi-agent debate into multimodal reasoning. The study addresses two key challenges: the trivialization of opinions resulting from excessive summarization and the diversion of focus caused by distractor concepts introduced from images. These challenges stem from the inductive (bottom-up) nature of existing debating schemes. To address the issue, we propose a deductive (top-down) debating approach called Blueprint Debate on Graphs (BDoG). In BDoG, debates are confined to a blueprint graph to prevent opinion trivialization through world-level summarization. Moreover, by storing evidence in branches within the graph, BDoG mitigates distractions caused by frequent but irrelevant concepts. Extensive experiments validate that BDoG is able to achieve state-of-the-art results in ScienceQA and MMBench with significant improvements over previous methods. The source code can be accessed at https://github.com/thecharm/BDoG.
Authors: Mahjabin Nahar, Haeseung Seo, Eun-Ju Lee, Aiping Xiong, Dongwon Lee
Abstract: The widespread adoption and transformative effects of large language models (LLMs) have sparked concerns regarding their capacity to produce inaccurate and fictitious content, referred to as `hallucinations'. Given the potential risks associated with hallucinations, humans should be able to identify them. This research aims to understand the human perception of LLM hallucinations by systematically varying the degree of hallucination (genuine, minor hallucination, major hallucination) and examining its interaction with warning (i.e., a warning of potential inaccuracies: absent vs. present). Participants (N=419) from Prolific rated the perceived accuracy and engaged with content (e.g., like, dislike, share) in a Q/A format. Participants ranked content as truthful in the order of genuine, minor hallucination, and major hallucination, and user engagement behaviors mirrored this pattern. More importantly, we observed that warning improved the detection of hallucination without significantly affecting the perceived truthfulness of genuine content. We conclude by offering insights for future tools to aid human detection of hallucinations. All survey materials, demographic questions, and post-session questions are available at: https://github.com/MahjabinNahar/fakes-of-varying-shades-survey-materials
URLs: https://github.com/MahjabinNahar/fakes-of-varying-shades-survey-materials
Authors: Wenyan Li, Jiaang Li, Rita Ramos, Raphael Tang, Desmond Elliott
Abstract: Recent advances in retrieval-augmented models for image captioning highlight the benefit of retrieving related captions for efficient, lightweight models with strong domain-transfer capabilities. While these models demonstrate the success of retrieval augmentation, retrieval models are still far from perfect in practice: the retrieved information can sometimes mislead the model, resulting in incorrect generation and worse performance. In this paper, we analyze the robustness of a retrieval-augmented captioning model SmallCap. Our analysis shows that the model is sensitive to tokens that appear in the majority of the retrieved captions, and the input attribution shows that those tokens are likely copied into the generated output. Given these findings, we propose to train the model by sampling retrieved captions from more diverse sets. This decreases the chance that the model learns to copy majority tokens, and improves both in-domain and cross-domain performance.
Authors: Laura Weidinger, John Mellor, Bernat Guillen Pegueroles, Nahema Marchal, Ravin Kumar, Kristian Lum, Canfer Akbulut, Mark Diaz, Stevie Bergman, Mikel Rodriguez, Verena Rieser, William Isaac
Abstract: This research introduces STAR, a sociotechnical framework that improves on current best practices for red teaming safety of large language models. STAR makes two key contributions: it enhances steerability by generating parameterised instructions for human red teamers, leading to improved coverage of the risk surface. Parameterised instructions also provide more detailed insights into model failures at no increased cost. Second, STAR improves signal quality by matching demographics to assess harms for specific groups, resulting in more sensitive annotations. STAR further employs a novel step of arbitration to leverage diverse viewpoints and improve label reliability, treating disagreement not as noise but as a valuable contribution to signal quality.
Authors: Minqian Liu, Zhiyang Xu, Zihao Lin, Trevor Ashby, Joy Rimchala, Jiaxin Zhang, Lifu Huang
Abstract: Interleaved text-and-image generation has been an intriguing research direction, where the models are required to generate both images and text pieces in an arbitrary order. Despite the emerging advancements in interleaved generation, the progress in its evaluation still significantly lags behind. Existing evaluation benchmarks do not support arbitrarily interleaved images and text for both inputs and outputs, and they only cover a limited number of domains and use cases. Also, current works predominantly use similarity-based metrics which fall short in assessing the quality in open-ended scenarios. To this end, we introduce InterleavedBench, the first benchmark carefully curated for the evaluation of interleaved text-and-image generation. InterleavedBench features a rich array of tasks to cover diverse real-world use cases. In addition, we present InterleavedEval, a strong reference-free metric powered by GPT-4o to deliver accurate and explainable evaluation. We carefully define five essential evaluation aspects for InterleavedEval, including text quality, perceptual quality, image coherence, text-image coherence, and helpfulness, to ensure a comprehensive and fine-grained assessment. Through extensive experiments and rigorous human evaluation, we show that our benchmark and metric can effectively evaluate the existing models with a strong correlation with human judgments surpassing previous reference-based metrics. We also provide substantial findings and insights to foster future research in interleaved generation and its evaluation.
Authors: Shashank Kirtania, Priyanshu Gupta, Arjun Radhakirshna
Abstract: In this paper we examine the limitations of Large Language Models (LLMs) for complex reasoning tasks. Although recent works have started to employ formal languages as an intermediate representation for reasoning tasks, they often face challenges in accurately generating and refining these formal specifications to ensure correctness. To address these issues, this paper proposes Logic-LM++, an improvement on Logic-LM . It uses the ability of LLMs to do pairwise comparisons, allowing the evaluation of the refinements suggested by the LLM. The paper demonstrates that Logic-LM++ outperforms Logic-LM and other contemporary techniques across natural language reasoning tasks on three datasets, FOLIO, ProofWriter and AR-LSAT, with an average improvement of 18.5% on standard prompting, 12.3% on chain of thought prompting and 5% on Logic-LM.
Authors: Hannah Brown, Leon Lin, Kenji Kawaguchi, Michael Shieh
Abstract: We introduce a defense against adversarial attacks on LLMs utilizing self-evaluation. Our method requires no model fine-tuning, instead using pre-trained models to evaluate the inputs and outputs of a generator model, significantly reducing the cost of implementation in comparison to other, finetuning-based methods. Our method can significantly reduce the attack success rate of attacks on both open and closed-source LLMs, beyond the reductions demonstrated by Llama-Guard2 and commonly used content moderation APIs. We present an analysis of the effectiveness of our method, including attempts to attack the evaluator in various settings, demonstrating that it is also more resilient to attacks than existing methods. Code and data will be made available at https://github.com/Linlt-leon/self-eval.
Authors: Siwei Wu, Kang Zhu, Yu Bai, Yiming Liang, Yizhi Li, Haoning Wu, J. H. Liu, Ruibo Liu, Xingwei Qu, Xuxin Cheng, Ge Zhang, Wenhao Huang, Chenghua Lin
Abstract: Given the remarkable success that large visual language models (LVLMs) have achieved in image perception tasks, the endeavor to make LVLMs perceive the world like humans is drawing increasing attention. Current multi-modal benchmarks primarily focus on facts or specific topic-related knowledge contained within individual images. However, they often overlook the associative relations between multiple images, which require the identification and analysis of similarities among entities or content present in different images. Therefore, we propose the multi-image relation association task and a meticulously curated Multi-granularity Multi-image Relational Association (MMRA) benchmark, comprising 1,024 samples. In order to systematically and comprehensively evaluate current LVLMs, we establish an associational relation system among images that contain 11 subtasks (e.g, UsageSimilarity, SubEvent) at two granularity levels (i.e., image and entity) according to the relations in ConceptNet. Our experiments reveal that on the MMRA benchmark, current multi-image LVLMs exhibit distinct advantages and disadvantages across various subtasks. Notably, fine-grained, entity-level multi-image perception tasks pose a greater challenge for LVLMs compared to image-level tasks. Moreover, LVLMs perform poorly on spatial-related tasks, indicating that LVLMs still have limited spatial awareness. Additionally, our findings indicate that while LVLMs demonstrate a strong capability to perceive image details, enhancing their ability to associate information across multiple images hinges on improving the reasoning capabilities of their language model component. Moreover, we explored the ability of LVLMs to perceive image sequences within the context of our multi-image association task. Our experiments show that the majority of current LVLMs do not adequately model image sequences during the pre-training process.
Authors: Rubing Chen, Xulu Zhang, Jiaxin Wu, Wenqi Fan, Xiao-Yong Wei, Qing Li
Abstract: This paper addresses the need for improved precision in existing Retrieval-Augmented Generation (RAG) methods that primarily focus on enhancing recall. We propose a multi-layer knowledge pyramid approach within the RAG framework to achieve a better balance between precision and recall. The knowledge pyramid consists of three layers: Ontologies, Knowledge Graphs (KGs), and chunk-based raw text. We employ cross-layer augmentation techniques for comprehensive knowledge coverage and dynamic updates of the Ontology schema and instances. To ensure compactness, we utilize cross-layer filtering methods for knowledge condensation in KGs. Our approach, named PolyRAG, follows a waterfall model for retrieval, starting from the top of the pyramid and progressing down until a confident answer is obtained. We introduce two benchmarks for domain-specific knowledge retrieval, one in the academic domain and the other in the financial domain. The effectiveness of the methods has been validated through comprehensive experiments by outperforming 19 SOTA methods. An encouraging observation is that the proposed method has augmented the GPT-4, providing 395\% F1 gain by improving its performance from 0.1636 to 0.8109.
Authors: Hongjun An, Yifan Chen, Xiaozhen Qiao, Zhe Sun, Xuelong Li
Abstract: Current large language models (LLMs) primarily utilize next-token prediction method for inference, which significantly impedes their processing speed. In this paper, we introduce a novel inference methodology termed next-sentence prediction, aimed at enhancing the inference efficiency of LLMs. We present Sentence Variational Autoencoder (SentenceVAE), a tiny model consisting of a Sentence Encoder and a Sentence Decoder. The Sentence Encoder can effectively condense the information within a sentence into a singular token, while the Sentence Decoder can reconstruct this compressed token back into sentence. By integrating SentenceVAE into the input and output layers of LLMs, we develop Sentence-level LLMs (SLLMs) that employ a sentence-by-sentence inference method. In addition, the SentenceVAE module of SLLMs can maintain the integrity of the original semantic content by segmenting the context into sentences, thereby improving accuracy while boosting inference speed. Moreover, compared to previous LLMs, SLLMs process fewer tokens over equivalent context length, significantly reducing memory demands for self-attention computation and facilitating the handling of longer context. Extensive experiments on Wanjuan dataset have reveal that the proposed method can accelerate inference speed by 204~365%, reduce perplexity (PPL) to 46~75% of its original metric, and decrease memory overhead by 86~91% for the equivalent context length, compared to the token-by-token method.
Authors: Yulei Qin, Yuncheng Yang, Pengcheng Guo, Gang Li, Hang Shao, Yuchen Shi, Zihan Xu, Yun Gu, Ke Li, Xing Sun
Abstract: Instruction tuning plays a critical role in aligning large language models (LLMs) with human preference. Despite the vast amount of open instruction datasets, naively training a LLM on all existing instructions may not be optimal and practical. To pinpoint the most beneficial datapoints, data assessment and selection methods have been proposed in the fields of natural language processing (NLP) and deep learning. However, under the context of instruction tuning, there still exists a gap in knowledge on what kind of data evaluation metrics can be employed and how they can be integrated into the selection mechanism. To bridge this gap, we present a comprehensive review on existing literature of data assessment and selection especially for instruction tuning of LLMs. We systematically categorize all applicable methods into quality-based, diversity-based, and importance-based ones where a unified, fine-grained taxonomy is structured. For each category, representative methods are elaborated to describe the landscape of relevant research. In addition, comparison between latest methods is conducted on their officially reported results to provide in-depth discussions on their limitations. Finally, we summarize the open challenges and propose the promosing avenues for future studies. All related contents are available at https://github.com/yuleiqin/fantastic-data-engineering.
URLs: https://github.com/yuleiqin/fantastic-data-engineering.