Authors: Krish Didwania, Pratinav Seth, Aditya Kasliwal, Amit Agarwal
Abstract: Agriculture, vital for global sustenance, necessitates innovative solutions due to a lack of organized domain experts, particularly in developing countries where many farmers are impoverished and cannot afford expert consulting. Initiatives like Farmers Helpline play a crucial role in such countries, yet challenges such as high operational costs persist. Automating query resolution can alleviate the burden on traditional call centers, providing farmers with immediate and contextually relevant information. The integration of Agriculture and Artificial Intelligence (AI) offers a transformative opportunity to empower farmers and bridge information gaps. Language models like transformers, the rising stars of AI, possess remarkable language understanding capabilities, making them ideal for addressing information gaps in agriculture. This work explores and demonstrates the transformative potential of Large Language Models (LLMs) in automating query resolution for agricultural farmers, leveraging their expertise in deciphering natural language and understanding context. Using a subset of a vast dataset of real-world farmer queries collected in India, our study focuses on approximately 4 million queries from the state of Tamil Nadu, spanning various sectors, seasonal crops, and query types.
Authors: Eric Pasewark, Kyle Montgomery, Kefei Duan, Dawn Song, Chenguang Wang
Abstract: We present a new method for large language models to solve compositional tasks. Although they have shown strong performance on traditional language understanding tasks, large language models struggle to solve compositional tasks, where the solution depends on solving smaller instances of the same problem. We propose a natural approach to solve compositional tasks recursively. Our method, Re-Tuning, tunes models to break down a problem into subproblems, solve those subproblems, and combine the results. We show that our method significantly improves model performance on three representative compositional tasks: integer addition, dynamic programming, and parity. Compared to state-of-the-art methods that keep intermediate steps towards solving the problems, Re-Tuning achieves significantly higher accuracy and is more GPU memory efficient.
Authors: AbdelRahim Elmadany, Ife Adebara, Muhammad Abdul-Mageed
Abstract: We address a notable gap in Natural Language Processing (NLP) by introducing a collection of resources designed to improve Machine Translation (MT) for low-resource languages, with a specific focus on African languages. First, We introduce two language models (LMs), Cheetah-1.2B and Cheetah-3.7B, with 1.2 billion and 3.7 billion parameters respectively. Next, we finetune the aforementioned models to create toucan, an Afrocentric machine translation model designed to support 156 African language pairs. To evaluate Toucan, we carefully develop an extensive machine translation benchmark, dubbed AfroLingu-MT, tailored for evaluating machine translation. Toucan significantly outperforms other models, showcasing its remarkable performance on MT for African languages. Finally, we train a new model, spBLEU-1K, to enhance translation evaluation metrics, covering 1K languages, including 614 African languages. This work aims to advance the field of NLP, fostering cross-cultural understanding and knowledge exchange, particularly in regions with limited language resources such as Africa. The GitHub repository for the Toucan project is available at https://github.com/UBC-NLP/Toucan.
Authors: Chengjie Zhou, Bobo Li, Hao Fei, Fei Li, Chong Teng, Donghong Ji
Abstract: Structured Sentiment Analysis (SSA) was cast as a problem of bi-lexical dependency graph parsing by prior studies. Multiple formulations have been proposed to construct the graph, which share several intrinsic drawbacks: (1) The internal structures of spans are neglected, thus only the boundary tokens of spans are used for relation prediction and span recognition, thus hindering the model's expressiveness; (2) Long spans occupy a significant proportion in the SSA datasets, which further exacerbates the problem of internal structure neglect. In this paper, we treat the SSA task as a dependency parsing task on partially-observed dependency trees, regarding flat spans without determined tree annotations as latent subtrees to consider internal structures of spans. We propose a two-stage parsing method and leverage TreeCRFs with a novel constrained inside algorithm to model latent structures explicitly, which also takes advantages of joint scoring graph arcs and headed spans for global optimization and inference. Results of extensive experiments on five benchmark datasets reveal that our method performs significantly better than all previous bi-lexical methods, achieving new state-of-the-art.
Authors: Ivan Rodkin, Yuri Kuratov, Aydar Bulatov, Mikhail Burtsev
Abstract: This paper addresses the challenge of creating a neural architecture for very long sequences that requires constant time for processing new information at each time step. Our approach, Associative Recurrent Memory Transformer (ARMT), is based on transformer self-attention for local context and segment-level recurrence for storage of task specific information distributed over a long context. We demonstrate that ARMT outperfors existing alternatives in associative retrieval tasks and sets a new performance record in the recent BABILong multi-task long-context benchmark by answering single-fact questions over 50 million tokens with an accuracy of 79.9%. The source code for training and evaluation is available on github.
Authors: Jon Sporring, Ken Friis Larsen
Abstract: AI-supported programming has taken giant leaps with tools such as Meta's Llama and openAI's chatGPT. These are examples of stochastic sources of programs and have already greatly influenced how we produce code and teach programming. If we consider input to such models as a stochastic source, a natural question is, what is the relation between the input and the output distributions, between the chatGPT prompt and the resulting program? In this paper, we will show how the relation between random Python programs generated from chatGPT can be described geometrically and topologically using Tree-edit distances between the program's syntax trees and without explicit modeling of the underlying space. A popular approach to studying high-dimensional samples in a metric space is to use low-dimensional embedding using, e.g., multidimensional scaling. Such methods imply errors depending on the data and dimension of the embedding space. In this article, we propose to restrict such projection methods to purely visualization purposes and instead use geometric summary statistics, methods from spatial point statistics, and topological data analysis to characterize the configurations of random programs that do not rely on embedding approximations. To demonstrate their usefulness, we compare two publicly available models: ChatGPT-4 and TinyLlama, on a simple problem related to image processing. Application areas include understanding how questions should be asked to obtain useful programs; measuring how consistently a given large language model answers; and comparing the different large language models as a programming assistant. Finally, we speculate that our approach may in the future give new insights into the structure of programming languages.
Authors: Mihir Parmar, Hanieh Deilamsalehy, Franck Dernoncourt, Seunghyun Yoon, Ryan A. Rossi, Trung Bui
Abstract: Extractive summarization plays a pivotal role in natural language processing due to its wide-range applications in summarizing diverse content efficiently, while also being faithful to the original content. Despite significant advancement achieved in extractive summarization by Large Language Models (LLMs), these summaries frequently exhibit incoherence. An important aspect of the coherent summary is its readability for intended users. Although there have been many datasets and benchmarks proposed for creating coherent extractive summaries, none of them currently incorporate user intent to improve coherence in extractive summarization. Motivated by this, we propose a systematically created human-annotated dataset consisting of coherent summaries for five publicly available datasets and natural language user feedback, offering valuable insights into how to improve coherence in extractive summaries. We utilize this dataset for aligning LLMs through supervised fine-tuning with natural language human feedback to enhance the coherence of their generated summaries. Preliminary experiments with Falcon-40B and Llama-2-13B show significant performance improvements (~10% Rouge-L) in terms of producing coherent summaries. We further utilize human feedback to benchmark results over instruction-tuned models such as FLAN-T5 which resulted in several interesting findings. Data and source code are available at https://github.com/Mihir3009/Extract-AI.
Authors: Marcos M. Jos\'e, Fl\'avio N. Ca\c{c}\~ao, Maria F. Ribeiro, Rafael M. Cheang, Paulo Pirozelli, Fabio G. Cozman
Abstract: This paper proposes a novel architecture to generate multi-hop answers to open domain questions that require information from texts and tables, using the Open Table-and-Text Question Answering dataset for validation and training. One of the most common ways to generate answers in this setting is to retrieve information sequentially, where a selected piece of data helps searching for the next piece. As different models can have distinct behaviors when called in this sequential information search, a challenge is how to select models at each step. Our architecture employs reinforcement learning to choose between different state-of-the-art tools sequentially until, in the end, a desired answer is generated. This system achieved an F1-score of 19.03, comparable to iterative systems in the literature.
Authors: Ekin Ozince, Yi\u{g}it Ihlamur
Abstract: This study explores the application of large language models (LLMs) in venture capital (VC) decision-making, focusing on predicting startup success based on founder characteristics. We utilize LLM prompting techniques, like chain-of-thought, to generate features from limited data, then extract insights through statistics and machine learning. Our results reveal potential relationships between certain founder characteristics and success, as well as demonstrate the effectiveness of these characteristics in prediction. This framework for integrating ML techniques and LLMs has vast potential for improving startup success prediction, with important implications for VC firms seeking to optimize their investment strategies.
Authors: Zekun Li, Xianjun Yang, Kyuri Choi, Wanrong Zhu, Ryan Hsieh, HyeonJung Kim, Jin Hyuk Lim, Sungyoung Ji, Byungju Lee, Xifeng Yan, Linda Ruth Petzold, Stephen D. Wilson, Woosang Lim, William Yang Wang
Abstract: The rapid advancement of Large Language Models (LLMs) and Large Multimodal Models (LMMs) has heightened the demand for AI-based scientific assistants capable of understanding scientific articles and figures. Despite progress, there remains a significant gap in evaluating models' comprehension of professional, graduate-level, and even PhD-level scientific content. Current datasets and benchmarks primarily focus on relatively simple scientific tasks and figures, lacking comprehensive assessments across diverse advanced scientific disciplines. To bridge this gap, we collected a multimodal, multidisciplinary dataset from open-access scientific articles published in Nature Communications journals. This dataset spans 72 scientific disciplines, ensuring both diversity and quality. We created benchmarks with various tasks and settings to comprehensively evaluate LMMs' capabilities in understanding scientific figures and content. Our evaluation revealed that these tasks are highly challenging: many open-source models struggled significantly, and even GPT-4V and GPT-4o faced difficulties. We also explored using our dataset as training resources by constructing visual instruction-following data, enabling the 7B LLaVA model to achieve performance comparable to GPT-4V/o on our benchmark. Additionally, we investigated the use of our interleaved article texts and figure images for pre-training LMMs, resulting in improvements on the material generation task. The source dataset, including articles, figures, constructed benchmarks, and visual instruction-following data, is open-sourced.
Authors: Muhammad Abdul-Mageed, Amr Keleg, AbdelRahim Elmadany, Chiyu Zhang, Injy Hamed, Walid Magdy, Houda Bouamor, Nizar Habash
Abstract: We describe the findings of the fifth Nuanced Arabic Dialect Identification Shared Task (NADI 2024). NADI's objective is to help advance SoTA Arabic NLP by providing guidance, datasets, modeling opportunities, and standardized evaluation conditions that allow researchers to collaboratively compete on pre-specified tasks. NADI 2024 targeted both dialect identification cast as a multi-label task (Subtask~1), identification of the Arabic level of dialectness (Subtask~2), and dialect-to-MSA machine translation (Subtask~3). A total of 51 unique teams registered for the shared task, of whom 12 teams have participated (with 76 valid submissions during the test phase). Among these, three teams participated in Subtask~1, three in Subtask~2, and eight in Subtask~3. The winning teams achieved 50.57 F\textsubscript{1} on Subtask~1, 0.1403 RMSE for Subtask~2, and 20.44 BLEU in Subtask~3, respectively. Results show that Arabic dialect processing tasks such as dialect identification and machine translation remain challenging. We describe the methods employed by the participating teams and briefly offer an outlook for NADI.
Authors: Ethan Mendes, Yang Chen, James Hays, Sauvik Das, Wei Xu, Alan Ritter
Abstract: Vision Language Models (VLMs) are rapidly advancing in their capability to answer information-seeking questions. As these models are widely deployed in consumer applications, they could lead to new privacy risks due to emergent abilities to identify people in photos, geolocate images, etc. As we demonstrate, somewhat surprisingly, current open-source and proprietary VLMs are very capable image geolocators, making widespread geolocation with VLMs an immediate privacy risk, rather than merely a theoretical future concern. As a first step to address this challenge, we develop a new benchmark, GPTGeoChat, to test the ability of VLMs to moderate geolocation dialogues with users. We collect a set of 1,000 image geolocation conversations between in-house annotators and GPT-4v, which are annotated with the granularity of location information revealed at each turn. Using this new dataset, we evaluate the ability of various VLMs to moderate GPT-4v geolocation conversations by determining when too much location information has been revealed. We find that custom fine-tuned models perform on par with prompted API-based models when identifying leaked location information at the country or city level; however, fine-tuning on supervised data appears to be needed to accurately moderate finer granularities, such as the name of a restaurant or building.
Authors: Zhichao Xu, Ashim Gupta, Tao Li, Oliver Bentham, Vivek Srikumar
Abstract: Large language models (LLMs) are increasingly deployed in real-world scenarios with the help of recent model compression techniques. Such momentum towards local deployment means the use of compressed LLMs will widely impact a large population. However, prior analysis works often prioritize on preserving perplexity which is a direct analogy to training loss. The impact of compression method on other critical aspects of model behavior, particularly safety, still calls for a systematic assessment. To this end, we investigate the impact of model compression on four dimensions: 1) degeneration harm, i.e., bias and toxicity in generation; 2) representational harm, i.e., biases in discriminative tasks; 3) dialect bias; 4) language modeling and downstream task performance. We cover a wide spectrum of LLM compression techniques, including structured pruning, un/semi-structured ones, and quantization. Our analyses reveal that compression can lead to unexpected consequences. Although compression may unintentionally remedy LLMs' degeneration harm, it can still exacerbate on the representational harm axis. Moreover, there is a divergent impact on different protected groups as the compression rate grows. Finally, different compression methods have drastically different safety impacts, e.g., quantization mostly preserves bias while pruning degrades quickly. Our findings underscore the importance of integrating safety assessments into the development of compressed LLMs to ensure their reliability across real-world applications. Our full results are available here: \url{https://github.com/zhichaoxu-shufe/Beyond-Perplexity-Compression-Safety-Eval}
URLs: https://github.com/zhichaoxu-shufe/Beyond-Perplexity-Compression-Safety-Eval
Authors: Yuchen Fan, Xin Zhong, Chengsi Wang, Gaoche Wu, Bowen Zhou
Abstract: Summarization is a fundamental task in natural language processing (NLP) and since large language models (LLMs), such as GPT-4 and Claude, come out, increasing attention has been paid to long-form summarization whose input sequences are much longer, indicating more information contained. The current evaluation metrics either use similarity-based metrics like ROUGE and BERTScore which rely on similarity and fail to consider informativeness or LLM-based metrics, lacking quantitative analysis of information richness and are rather subjective. In this paper, we propose a new evaluation metric called EVA-Score using Atomic Fact Chain Generation and Document-level Relation Extraction together to automatically calculate the informativeness and give a definite number as an information score. Experiment results show that our metric shows a state-of-the-art correlation with humans. We also re-evaluate the performance of LLMs on long-form summarization comprehensively from the information aspect, forecasting future ways to use LLMs for long-form summarization.
Authors: Cheng Wang, Xinyang Lu, See-Kiong Ng, Bryan Kian Hsiang Low
Abstract: The rapid evolution of large language models (LLMs) represents a substantial leap forward in natural language understanding and generation. However, alongside these advancements come significant challenges related to the accountability and transparency of LLM responses. Reliable source attribution is essential to adhering to stringent legal and regulatory standards, including those set forth by the General Data Protection Regulation. Despite the well-established methods in source attribution within the computer vision domain, the application of robust attribution frameworks to natural language processing remains underexplored. To bridge this gap, we propose a novel and versatile TRansformer-based Attribution framework using Contrastive Embeddings called TRACE that, in particular, exploits contrastive learning for source attribution. We perform an extensive empirical evaluation to demonstrate the performance and efficiency of TRACE in various settings and show that TRACE significantly improves the ability to attribute sources accurately, making it a valuable tool for enhancing the reliability and trustworthiness of LLMs.
Authors: Yana Veitsman
Abstract: Research in the NLP sphere of the Turkic counterparts of Central Asian languages, namely Kazakh, Uzbek, Kyrgyz, and Turkmen, comes with the typical challenges of low-resource languages, like data scarcity and a general lack of linguistic resources. However, in the recent years research has greatly advanced via collection of language-specific datasets and development of downstream task technologies. Aiming to summarize this research up until May 2024, this paper also seeks to identify potential areas of future work. To achieve this, the paper gives a broad, high-level overview of the linguistic properties of the languages, the current coverage and performance of already developed technology, application of transfer learning techniques from higher-resource languages, and availability of labeled and unlabeled data for each language. Providing a summary of the current state of affairs, we hope that further research will be facilitated with the considerations we provide in the current paper.
Authors: Ting Wu, Xuefeng Li, Pengfei Liu
Abstract: Self-improvement through post-training methods such as iterative preference learning has been acclaimed for enhancing the problem-solving capabilities (e.g., mathematical reasoning) of Large Language Models (LLMs) without human intervention. However, as exploration deepens, it becomes crucial to assess whether these improvements genuinely signify progress in solving more challenging problems or if they could lead to unintended regressions. To address this, we propose a comprehensive evaluative framework that goes beyond the superficial pass@1 metric to scrutinize the underlying enhancements of post-training paradigms for self-improvement. Through rigorous experimentation and analysis across diverse problem-solving tasks, the empirical results point out the phenomenon of \emph{self-improvement reversal}, where models showing improved performance across benchmarks will paradoxically exhibit declines in broader, essential capabilities, like output diversity and out-of-distribution (OOD) generalization. These findings indicate that current self-improvement practices through post-training are inadequate for equipping models to tackle more complex problems. Furthermore, they underscore the necessity of our critical evaluation metrics in discerning the \emph{progress or regress} dichotomy for self-improving LLMs.
Authors: Bojana Ba\v{s}aragin, Adela Ljaji\'c, Darija Medvecki, Lorenzo Cassano, Milo\v{s} Ko\v{s}prdi\'c, Nikola Milo\v{s}evi\'c
Abstract: Large language models (LLMs) have recently become the leading source of answers for users' questions online. Despite their ability to offer eloquent answers, their accuracy and reliability can pose a significant challenge. This is especially true for sensitive domains such as biomedicine, where there is a higher need for factually correct answers. This paper introduces a biomedical retrieval-augmented generation (RAG) system designed to enhance the reliability of generated responses. The system is based on a fine-tuned LLM for the referenced question-answering, where retrieved relevant abstracts from PubMed are passed to LLM's context as input through a prompt. Its output is an answer based on PubMed abstracts, where each statement is referenced accordingly, allowing the users to verify the answer. Our retrieval system achieves an absolute improvement of 23% compared to the PubMed search engine. Based on the manual evaluation on a small sample, our fine-tuned LLM component achieves comparable results to GPT-4 Turbo in referencing relevant abstracts. We make the dataset used to fine-tune the models and the fine-tuned models based on Mistral-7B-instruct-v0.1 and v0.2 publicly available.
Authors: Esther Ploeger, Wessel Poelman, Andreas Holck H{\o}eg-Petersen, Anders Schlichtkrull, Miryam de Lhoneux, Johannes Bjerva
Abstract: Beyond individual languages, multilingual natural language processing (NLP) research increasingly aims to develop models that perform well across languages generally. However, evaluating these systems on all the world's languages is practically infeasible. To attain generalizability, representative language sampling is essential. Previous work argues that generalizable multilingual evaluation sets should contain languages with diverse typological properties. However, 'typologically diverse' language samples have been found to vary considerably in this regard, and popular sampling methods are flawed and inconsistent. We present a language sampling framework for selecting highly typologically diverse languages given a sampling frame, informed by language typology. We compare sampling methods with a range of metrics and find that our systematic methods consistently retrieve more typologically diverse language selections than previous methods in NLP. Moreover, we provide evidence that this affects generalizability in multilingual model evaluation, emphasizing the importance of diverse language sampling in NLP evaluation.
Authors: Ting-Yu Yen, Yun-Da Tsai, Keng-Te Liao, Shou-De Lin
Abstract: Converting different modalities into general text, serving as input prompts for large language models (LLMs), is a common method to align multimodal models when there is limited pairwise data. This text-centric approach leverages the unique properties of text as a modality space, transforming diverse inputs into a unified textual representation. This enables downstream models to effectively interpret various modal inputs. This study assesses the quality and robustness of multimodal representations in the presence of missing entries, noise, or absent modalities, revealing that current text-centric alignment methods compromise downstream robustness. To address this issue, we propose a new text-centric approach that achieves superior robustness compared to previous methods across various modalities in different settings. Our findings highlight the potential of this approach to enhance the robustness and adaptability of multimodal representations, offering a promising solution for dynamic and real-world applications.
Authors: Jingshen Zhang, Xinying Qiu, Teng Shen, Wenyu Wang, Kailin Zhang, Wenhe Feng
Abstract: Cross-lingual word alignment plays a crucial role in various natural language processing tasks, particularly for low-resource languages. Recent study proposes a BiLSTM-based encoder-decoder model that outperforms pre-trained language models in low-resource settings. However, their model only considers the similarity of word embedding spaces and does not explicitly model the differences between word embeddings. To address this limitation, we propose incorporating contrastive learning into the BiLSTM-based encoder-decoder framework. Our approach introduces a multi-view negative sampling strategy to learn the differences between word pairs in the shared cross-lingual embedding space. We evaluate our model on five bilingual aligned datasets spanning four ASEAN languages: Lao, Vietnamese, Thai, and Indonesian. Experimental results demonstrate that integrating contrastive learning consistently improves word alignment accuracy across all datasets, confirming the effectiveness of the proposed method in low-resource scenarios. We will release our data set and code to support future research on ASEAN or more low-resource word alignment.
Authors: Siqi He, Wei Zhao
Abstract: Computational and human perception are often considered separate approaches for studying sound changes over time; few works have touched on the intersection of both. To fill this research gap, we provide a pioneering review contrasting computational with human perception from the perspectives of methods and tasks. Overall, computational approaches rely on computer-driven models to perceive historical sound changes on etymological datasets, while human approaches use listener-driven models to perceive ongoing sound changes on recording corpora. Despite their differences, both approaches complement each other on phonetic and acoustic levels, showing the potential to achieve a more comprehensive perception of sound change. Moreover, we call for a comparative study on the datasets used by both approaches to investigate the influence of historical sound changes on ongoing changes. Lastly, we discuss the applications of sound change in computational linguistics, and point out that perceiving sound change alone is insufficient, as many processes of language change are complex, with entangled changes at syntactic, semantic, and phonetic levels.
Authors: Ergun Bi\c{c}ici
Abstract: We present a new parser performance prediction (PPP) model using machine translation performance prediction system (MTPPS), statistically independent of any language or parser, relying only on extrinsic and novel features based on textual, link structural, and bracketing tree structural information. This new system, MTPPS-PPP, can predict the performance of any parser in any language and can be useful for estimating the grammatical difficulty when understanding a given text, for setting expectations from parsing output, for parser selection for a specific domain, and for parser combination systems. We obtain SoA results in PPP of bracketing $F_1$ with better results over textual features and similar performance with previous results that use parser and linguistic label specific information. Our results show the contribution of different types of features as well as rankings of individual features in different experimental settings (cased vs. uncased), in different learning tasks (in-domain vs. out-of-domain), with different training sets, with different learning algorithms, and with different dimensionality reduction techniques. We achieve $0.0678$ MAE and $0.85$ RAE in setting +Link, which corresponds to about $7.4\%$ error when predicting the bracketing $F_1$ score for the Charniak and Johnson parser on the WSJ23 test set. MTPPS-PPP system can predict without parsing using only the text, without a supervised parser using only an unsupervised parser, without any parser or language dependent information, without using a reference parser output, and can be used to predict the performance of any parser in any language.
Authors: Ergun Bi\c{c}ici
Abstract: We use referential translation machines (RTMs) to identify the similarity between an attribute and two words in English by casting the task as machine translation performance prediction (MTPP) between the words and the attribute word and the distance between their similarities for Task 10 with stacked RTM models. RTMs are also used to predict the intensity of the structure and content in tweets in English, Arabic, and Spanish in Task 1 where MTPP is between the tweets and the set of words for the emotion selected from WordNet affect emotion lists. Stacked RTM models obtain encouraging results in both.
Authors: Jalil Nourmohammadi Khiarak, Ammar Ahmadi, Taher Ak-bari Saeed, Meysam Asgari-Chenaghlu, To\u{g}rul Atabay, Mohammad Reza Baghban Karimi, Ismail Ceferli, Farzad Hasanvand, Seyed Mahboub Mousavi, Morteza Noshad
Abstract: This paper introduces a pioneering English-Azerbaijani (Arabic Script) parallel corpus, designed to bridge the technological gap in language learning and machine translation (MT) for under-resourced languages. Consisting of 548,000 parallel sentences and approximately 9 million words per language, this dataset is derived from diverse sources such as news articles and holy texts, aiming to enhance natural language processing (NLP) applications and language education technology. This corpus marks a significant step forward in the realm of linguistic resources, particularly for Turkic languages, which have lagged in the neural machine translation (NMT) revolution. By presenting the first comprehensive case study for the English-Azerbaijani (Arabic Script) language pair, this work underscores the transformative potential of NMT in low-resource contexts. The development and utilization of this corpus not only facilitate the advancement of machine translation systems tailored for specific linguistic needs but also promote inclusive language learning through technology. The findings demonstrate the corpus's effectiveness in training deep learning MT systems and underscore its role as an essential asset for researchers and educators aiming to foster bilingual education and multilingual communication. This research covers the way for future explorations into NMT applications for languages lacking substantial digital resources, thereby enhancing global language education frameworks. The Python package of our code is available at https://pypi.org/project/chevir-kartalol/, and we also have a website accessible at https://translate.kartalol.com/.
URLs: https://pypi.org/project/chevir-kartalol/,, https://translate.kartalol.com/.
Authors: Weimin Lyu, Zexin Bi, Fusheng Wang, Chao Chen
Abstract: The advent of clinical language models integrated into electronic health records (EHR) for clinical decision support has marked a significant advancement, leveraging the depth of clinical notes for improved decision-making. Despite their success, the potential vulnerabilities of these models remain largely unexplored. This paper delves into the realm of backdoor attacks on clinical language models, introducing an innovative attention-based backdoor attack method, BadCLM (Bad Clinical Language Models). This technique clandestinely embeds a backdoor within the models, causing them to produce incorrect predictions when a pre-defined trigger is present in inputs, while functioning accurately otherwise. We demonstrate the efficacy of BadCLM through an in-hospital mortality prediction task with MIMIC III dataset, showcasing its potential to compromise model integrity. Our findings illuminate a significant security risk in clinical decision support systems and pave the way for future endeavors in fortifying clinical language models against such vulnerabilities.
Authors: Cheng-Han Chiang, Wei-Chih Chen, Chun-Yi Kuan, Chienchou Yang, Hung-yi Lee
Abstract: Using large language models (LLMs) for automatic evaluation has become an important evaluation method in NLP research. However, it is unclear whether these LLM-based evaluators can be applied in real-world classrooms to assess student assignments. This empirical report shares how we use GPT-4 as an automatic assignment evaluator in a university course with 1,028 students. Based on student responses, we find that LLM-based assignment evaluators are generally acceptable to students when students have free access to these LLM-based evaluators. However, students also noted that the LLM sometimes fails to adhere to the evaluation instructions. Additionally, we observe that students can easily manipulate the LLM-based evaluator to output specific strings, allowing them to achieve high scores without meeting the assignment rubric. Based on student feedback and our experience, we provide several recommendations for integrating LLM-based evaluators into future classrooms.
Authors: Soheil Shapouri, Saber Soleymani
Abstract: Emotion mining has become a crucial tool for understanding human emotions during disasters, leveraging the extensive data generated on social media platforms. This paper aims to summarize existing research on emotion mining within disaster contexts, highlighting both significant discoveries and persistent issues. On the one hand, emotion mining techniques have achieved acceptable accuracy enabling applications such as rapid damage assessment and mental health surveillance. On the other hand, with many studies adopting data-driven approaches, several methodological issues remain. These include arbitrary emotion classification, ignoring biases inherent in data collection from social media, such as the overrepresentation of individuals from higher socioeconomic status on Twitter, and the lack of application of theoretical frameworks like cross-cultural comparisons. These problems can be summarized as a notable lack of theory-driven research and ignoring insights from social and behavioral sciences. This paper underscores the need for interdisciplinary collaboration between computer scientists and social scientists to develop more robust and theoretically grounded approaches in emotion mining. By addressing these gaps, we aim to enhance the effectiveness and reliability of emotion mining methodologies, ultimately contributing to improved disaster preparedness, response, and recovery. Keywords: emotion mining, sentiment analysis, natural disasters, psychology, technological disasters
Authors: Jianlong Chen, Wei Xu, Zhicheng Ding, Jinxin Xu, Hao Yan, Xinyu Zhang
Abstract: Prompt recovery, a crucial task in natural language processing, entails the reconstruction of prompts or instructions that language models use to convert input text into a specific output. Although pivotal, the design and effectiveness of prompts represent a challenging and relatively untapped field within NLP research. This paper delves into an exhaustive investigation of prompt recovery methodologies, employing a spectrum of pre-trained language models and strategies. Our study is a comparative analysis aimed at gauging the efficacy of various models on a benchmark dataset, with the goal of pinpointing the most proficient approach for prompt recovery. Through meticulous experimentation and detailed analysis, we elucidate the outstanding performance of the Gemma-2b-it + Phi2 model + Pretrain. This model surpasses its counterparts, showcasing its exceptional capability in accurately reconstructing prompts for text transformation tasks. Our findings offer a significant contribution to the existing knowledge on prompt recovery, shedding light on the intricacies of prompt design and offering insightful perspectives for future innovations in text rewriting and the broader field of natural language processing.
Authors: Yubo Zhang, Shudi Hou, Mingyu Derek Ma, Wei Wang, Muhao Chen, Jieyu Zhao
Abstract: Large language models (LLMs) are increasingly applied to clinical decision-making. However, their potential to exhibit bias poses significant risks to clinical equity. Currently, there is a lack of benchmarks that systematically evaluate such clinical bias in LLMs. While in downstream tasks, some biases of LLMs can be avoided such as by instructing the model to answer "I'm not sure...", the internal bias hidden within the model still lacks deep studies. We introduce CLIMB (shorthand for A Benchmark of Clinical Bias in Large Language Models), a pioneering comprehensive benchmark to evaluate both intrinsic (within LLMs) and extrinsic (on downstream tasks) bias in LLMs for clinical decision tasks. Notably, for intrinsic bias, we introduce a novel metric, AssocMAD, to assess the disparities of LLMs across multiple demographic groups. Additionally, we leverage counterfactual intervention to evaluate extrinsic bias in a task of clinical diagnosis prediction. Our experiments across popular and medically adapted LLMs, particularly from the Mistral and LLaMA families, unveil prevalent behaviors with both intrinsic and extrinsic bias. This work underscores the critical need to mitigate clinical bias and sets a new standard for future evaluations of LLMs' clinical bias.
Authors: Zhiwen You, HaeJin Lee, Shubhanshu Mishra, Sullam Jeoung, Apratim Mishra, Jinseok Kim, Jana Diesner
Abstract: Name-based gender prediction has traditionally categorized individuals as either female or male based on their names, using a binary classification system. That binary approach can be problematic in the cases of gender-neutral names that do not align with any one gender, among other reasons. Relying solely on binary gender categories without recognizing gender-neutral names can reduce the inclusiveness of gender prediction tasks. We introduce an additional gender category, i.e., "neutral", to study and address potential gender biases in Large Language Models (LLMs). We evaluate the performance of several foundational and large language models in predicting gender based on first names only. Additionally, we investigate the impact of adding birth years to enhance the accuracy of gender prediction, accounting for shifting associations between names and genders over time. Our findings indicate that most LLMs identify male and female names with high accuracy (over 80%) but struggle with gender-neutral names (under 40%), and the accuracy of gender prediction is higher for English-based first names than non-English names. The experimental results show that incorporating the birth year does not improve the overall accuracy of gender prediction, especially for names with evolving gender associations. We recommend using caution when applying LLMs for gender identification in downstream tasks, particularly when dealing with non-binary gender labels.
Authors: Junjie Wu, Lemao Liu, Wei Bi, Dit-Yan Yeung
Abstract: Targeted adversarial attacks are widely used to evaluate the robustness of neural machine translation systems. Unfortunately, this paper first identifies a critical issue in the existing settings of NMT targeted adversarial attacks, where their attacking results are largely overestimated. To this end, this paper presents a new setting for NMT targeted adversarial attacks that could lead to reliable attacking results. Under the new setting, it then proposes a Targeted Word Gradient adversarial Attack (TWGA) method to craft adversarial examples. Experimental results demonstrate that our proposed setting could provide faithful attacking results for targeted adversarial attacks on NMT systems, and the proposed TWGA method can effectively attack such victim NMT systems. In-depth analyses on a large-scale dataset further illustrate some valuable findings. 1 Our code and data are available at https://github.com/wujunjie1998/TWGA.
Authors: Leonidas Zotos, Hedderik van Rijn, Malvina Nissim
Abstract: Estimating the difficulty of multiple-choice questions would be great help for educators who must spend substantial time creating and piloting stimuli for their tests, and for learners who want to practice. Supervised approaches to difficulty estimation have yielded to date mixed results. In this contribution we leverage an aspect of generative large models which might be seen as a weakness when answering questions, namely their uncertainty, and exploit it towards exploring correlations between two different metrics of uncertainty, and the actual student response distribution. While we observe some present but weak correlations, we also discover that the models' behaviour is different in the case of correct vs wrong answers, and that correlations differ substantially according to the different question types which are included in our fine-grained, previously unused dataset of 451 questions from a Biopsychology course. In discussing our findings, we also suggest potential avenues to further leverage model uncertainty as an additional proxy for item difficulty.
Authors: Zirun Guo, Tao Jin, Zhou Zhao
Abstract: The development of multimodal models has significantly advanced multimodal sentiment analysis and emotion recognition. However, in real-world applications, the presence of various missing modality cases often leads to a degradation in the model's performance. In this work, we propose a novel multimodal Transformer framework using prompt learning to address the issue of missing modalities. Our method introduces three types of prompts: generative prompts, missing-signal prompts, and missing-type prompts. These prompts enable the generation of missing modality features and facilitate the learning of intra- and inter-modality information. Through prompt learning, we achieve a substantial reduction in the number of trainable parameters. Our proposed method outperforms other methods significantly across all evaluation metrics. Extensive experiments and ablation studies are conducted to demonstrate the effectiveness and robustness of our method, showcasing its ability to effectively handle missing modalities.
Authors: Abhinav Joshi, Shounak Paul, Akshat Sharma, Pawan Goyal, Saptarshi Ghosh, Ashutosh Modi
Abstract: Legal systems worldwide are inundated with exponential growth in cases and documents. There is an imminent need to develop NLP and ML techniques for automatically processing and understanding legal documents to streamline the legal system. However, evaluating and comparing various NLP models designed specifically for the legal domain is challenging. This paper addresses this challenge by proposing IL-TUR: Benchmark for Indian Legal Text Understanding and Reasoning. IL-TUR contains monolingual (English, Hindi) and multi-lingual (9 Indian languages) domain-specific tasks that address different aspects of the legal system from the point of view of understanding and reasoning over Indian legal documents. We present baseline models (including LLM-based) for each task, outlining the gap between models and the ground truth. To foster further research in the legal domain, we create a leaderboard (available at: https://exploration-lab.github.io/IL-TUR/) where the research community can upload and compare legal text understanding systems.
Authors: Abhinav Joshi, Romit Mohanty, Mounika Kanakanti, Andesha Mangla, Sudeep Choudhary, Monali Barbate, Ashutosh Modi
Abstract: Indian Sign Language has limited resources for developing machine learning and data-driven approaches for automated language processing. Though text/audio-based language processing techniques have shown colossal research interest and tremendous improvements in the last few years, Sign Languages still need to catch up due to the need for more resources. To bridge this gap, in this work, we propose iSign: a benchmark for Indian Sign Language (ISL) Processing. We make three primary contributions to this work. First, we release one of the largest ISL-English datasets with more than 118K video-sentence/phrase pairs. To the best of our knowledge, it is the largest sign language dataset available for ISL. Second, we propose multiple NLP-specific tasks (including SignVideo2Text, SignPose2Text, Text2Pose, Word Prediction, and Sign Semantics) and benchmark them with the baseline models for easier access to the research community. Third, we provide detailed insights into the proposed benchmarks with a few linguistic insights into the workings of ISL. We streamline the evaluation of Sign Language processing, addressing the gaps in the NLP research community for Sign Languages. We release the dataset, tasks, and models via the following website: https://exploration-lab.github.io/iSign/
Authors: Weizhi Tang, Vaishak Belle
Abstract: Temporal reasoning (TR) is a critical component of artificial intelligence, encompassing understanding and processing temporal information and relationships between events. To discover and study the TR ability in Large Language Models (LLMs), various datasets have been constructed in different ways for evaluating various aspects of TR ability. Our work proposes a novel approach to design and develop a pipeline for constructing datasets to evaluate the TR ability of LLMs by leveraging random directed graph generation, LTL formula, and the NuSMV model checker. Based on the pipeline, we have also constructed a dataset as a benchmark, namely LTLBench, consisting of 2,000 TR challenges and evaluated six LLMs with it. Furthermore, we have conducted additional experiments to discover the impact of increasing the number of events and formula operators on the complexity of TR problems and the performance of LLMs. We have demonstrated that although LLMs exhibit some promise in handling TR challenges, they still struggle with complex TR. We expect this work can offer insights into TR ability in LLMs while also providing a valuable tool for future TR evaluations.
Authors: Elisei Rykov, Konstantin Zaytsev, Ivan Anisimov, Alexandr Voronin
Abstract: This paper presents a solution for the Multilingual Text Detoxification task in the PAN-2024 competition of the SmurfCat team. Using data augmentation through machine translation and a special filtering procedure, we collected an additional multilingual parallel dataset for text detoxification. Using the obtained data, we fine-tuned several multilingual sequence-to-sequence models, such as mT0 and Aya, on a text detoxification task. We applied the ORPO alignment technique to the final model. Our final model has only 3.7 billion parameters and achieves state-of-the-art results for the Ukrainian language and near state-of-the-art results for other languages. In the competition, our team achieved first place in the automated evaluation with a score of 0.52 and second place in the final human evaluation with a score of 0.74.
Authors: Jiaxin Ge, Xueying Jia, Vijay Viswanathan, Hongyin Luo, Graham Neubig
Abstract: One of the most reliable ways to create deployable models for specialized tasks is to obtain an adequate amount of high-quality task-specific data. However, for specialized tasks, often such datasets do not exist. Existing methods address this by creating such data from large language models (LLMs) and then distilling such knowledge into smaller models. However, these methods are limited by the quality of the LLMs output, and tend to generate repetitive or incorrect data. In this work, we present Retrieval Based Distillation (ReBase), a method that first retrieves data from rich online sources and then transforms them into domain-specific data. This method greatly enhances data diversity. Moreover, ReBase generates Chain-of-Thought reasoning and distills the reasoning capacity of LLMs. We test our method on 4 benchmarks and results show that our method significantly improves performance by up to 7.8% on SQuAD, 1.37% on MNLI, and 1.94% on BigBench-Hard.
Authors: Wenxin Zhou
Abstract: In this paper, we present our system for the BioNNE English track, which aims to extract 8 types of biomedical nested named entities from biomedical text. We use a large language model (Mixtral 8x7B instruct) and ScispaCy NER model to identify entities in an article and build custom heuristics based on unified medical language system (UMLS) semantic types to categorize the entities. We discuss the results and limitations of our system and propose future improvements. Our system achieved an F1 score of 0.39 on the BioNNE validation set and 0.348 on the test set.
Authors: Simran Arora, Aman Timalsina, Aaryan Singhal, Benjamin Spector, Sabri Eyuboglu, Xinyi Zhao, Ashish Rao, Atri Rudra, Christopher R\'e
Abstract: Recurrent large language models that compete with Transformers in language modeling perplexity are emerging at a rapid rate (e.g., Mamba, RWKV). Excitingly, these architectures use a constant amount of memory during inference. However, due to the limited memory, recurrent LMs cannot recall and use all the information in long contexts leading to brittle in-context learning (ICL) quality. A key challenge for efficient LMs is selecting what information to store versus discard. In this work, we observe the order in which information is shown to the LM impacts the selection difficulty. To formalize this, we show that the hardness of information recall reduces to the hardness of a problem called set disjointness (SD), a quintessential problem in communication complexity that requires a streaming algorithm (e.g., recurrent model) to decide whether inputted sets are disjoint. We empirically and theoretically show that the recurrent memory required to solve SD changes with set order, i.e., whether the smaller set appears first in-context. Our analysis suggests, to mitigate the reliance on data order, we can put information in the right order in-context or process prompts non-causally. Towards that end, we propose: (1) JRT-Prompt, where context gets repeated multiple times in the prompt, effectively showing the model all data orders. This gives $11.0 \pm 1.3$ points of improvement, averaged across $16$ recurrent LMs and the $6$ ICL tasks, with $11.9\times$ higher throughput than FlashAttention-2 for generation prefill (length $32$k, batch size $16$, NVidia H100). We then propose (2) JRT-RNN, which uses non-causal prefix-linear-attention to process prompts and provides $99\%$ of Transformer quality at $360$M params., $30$B tokens and $96\%$ at $1.3$B params., $50$B tokens on average across the tasks, with $19.2\times$ higher throughput for prefill than FA2.
Authors: Hugo Pitorro, Pavlo Vasylenko, Marcos Treviso, Andr\'e F. T. Martins
Abstract: Transformers are the current architecture of choice for NLP, but their attention layers do not scale well to long contexts. Recent works propose to replace attention with linear recurrent layers -- this is the case for state space models, which enjoy efficient training and inference. However, it remains unclear whether these models are competitive with transformers in machine translation (MT). In this paper, we provide a rigorous and comprehensive experimental comparison between transformers and linear recurrent models for MT. Concretely, we experiment with RetNet, Mamba, and hybrid versions of Mamba which incorporate attention mechanisms. Our findings demonstrate that Mamba is highly competitive with transformers on sentence and paragraph-level datasets, where in the latter both models benefit from shifting the training distribution towards longer sequences. Further analysis show that integrating attention into Mamba improves translation quality, robustness to sequence length extrapolation, and the ability to recall named entities.
Authors: Nikhil Sharma, Kenton Murray, Ziang Xiao
Abstract: With Retrieval Augmented Generation (RAG), Large Language Models (LLMs) are playing a pivotal role in information search and are being adopted globally. Although the multilingual capability of LLMs offers new opportunities to bridge the language barrier, do these capabilities translate into real-life scenarios where linguistic divide and knowledge conflicts between multilingual sources are known occurrences? In this paper, we studied LLM's linguistic preference in a RAG-based information search setting. We found that LLMs displayed systemic bias towards information in the same language as the query language in both information retrieval and answer generation. Furthermore, in scenarios where there is little information in the language of the query, LLMs prefer documents in high-resource languages, reinforcing the dominant views. Such bias exists for both factual and opinion-based queries. Our results highlight the linguistic divide within multilingual LLMs in information search systems. The seemingly beneficial multilingual capability of LLMs may backfire on information parity by reinforcing language-specific information cocoons or filter bubbles further marginalizing low-resource views.
Authors: Tianyi Tang, Yiwen Hu, Bingqian Li, Wenyang Luo, Zijing Qin, Haoxiang Sun, Jiapeng Wang, Shiyi Xu, Xiaoxue Cheng, Geyang Guo, Han Peng, Bowen Zheng, Yiru Tang, Yingqian Min, Yushuo Chen, Jie Chen, Yuanqian Zhao, Luran Ding, Yuhao Wang, Zican Dong, Chunxuan Xia, Junyi Li, Kun Zhou, Wayne Xin Zhao, Ji-Rong Wen
Abstract: To facilitate the research on large language models (LLMs), this paper presents a comprehensive and unified library, LLMBox, to ease the development, use, and evaluation of LLMs. This library is featured with three main merits: (1) a unified data interface that supports the flexible implementation of various training strategies, (2) a comprehensive evaluation that covers extensive tasks, datasets, and models, and (3) more practical consideration, especially on user-friendliness and efficiency. With our library, users can easily reproduce existing methods, train new models, and conduct comprehensive performance comparisons. To rigorously test LLMBox, we conduct extensive experiments in a diverse coverage of evaluation settings, and experimental results demonstrate the effectiveness and efficiency of our library in supporting various implementations related to LLMs. The detailed introduction and usage guidance can be found at https://github.com/RUCAIBox/LLMBox.
Authors: Francisco Zanartu, Yulia Otmakhova, John Cook, Lea Frermann
Abstract: Misinformation about climate change causes numerous negative impacts, necessitating corrective responses. Psychological research has offered various strategies for reducing the influence of climate misinformation, such as the fact-myth-fallacy-fact-structure. However, practically implementing corrective interventions at scale represents a challenge. Automatic detection and correction of misinformation offers a solution to the misinformation problem. This study documents the development of large language models that accept as input a climate myth and produce a debunking that adheres to the fact-myth-fallacy-fact (``truth sandwich'') structure, by incorporating contrarian claim classification and fallacy detection into an LLM prompting framework. We combine open (Mixtral, Palm2) and proprietary (GPT-4) LLMs with prompting strategies of varying complexity. Experiments reveal promising performance of GPT-4 and Mixtral if combined with structured prompts. We identify specific challenges of debunking generation and human evaluation, and map out avenues for future work. We release a dataset of high-quality truth-sandwich debunkings, source code and a demo of the debunking system.
Authors: Xintong Li, Jinya Jiang, Ria Dharmani, Jayanth Srinivasa, Gaowen Liu, Jingbo Shang
Abstract: We study open-world multi-label text classification under extremely weak supervision (XWS), where the user only provides a brief description for classification objectives without any labels or ground-truth label space. Similar single-label XWS settings have been explored recently, however, these methods cannot be easily adapted for multi-label. We observe that (1) most documents have a dominant class covering the majority of content and (2) long-tail labels would appear in some documents as a dominant class. Therefore, we first utilize the user description to prompt a large language model (LLM) for dominant keyphrases of a subset of raw documents, and then construct a (initial) label space via clustering. We further apply a zero-shot multi-label classifier to locate the documents with small top predicted scores, so we can revisit their dominant keyphrases for more long-tail labels. We iterate this process to discover a comprehensive label space and construct a multi-label classifier as a novel method, X-MLClass. X-MLClass exhibits a remarkable increase in ground-truth label space coverage on various datasets, for example, a 40% improvement on the AAPD dataset over topic modeling and keyword extraction methods. Moreover, X-MLClass achieves the best end-to-end multi-label classification accuracy.
Authors: Surya Agustian, Muhammad Irfan Syah, Nurul Fatiara, Rahmad Abdillah
Abstract: The stakeholders' needs in sentiment analysis for various issues, whether positive or negative, are speed and accuracy. One new challenge in sentiment analysis tasks is the limited training data, which often leads to suboptimal machine learning models and poor performance on test data. This paper discusses the problem of text classification based on limited training data (300 to 600 samples) into three classes: positive, negative, and neutral. A benchmark dataset is provided for training and testing data on the issue of Kaesang Pangarep's appointment as Chairman of PSI. External data for aggregation and augmentation purposes are provided, consisting of two datasets: the topic of Covid Vaccination sentiment and an open topic. The official score used is the F1-score, which balances precision and recall among the three classes, positive, negative, and neutral. A baseline score is provided as a reference for researchers for unoptimized classification methods. The optimized score is provided as a reference for the target score to be achieved by any proposed method. Both scoring (baseline and optimized) use the SVM method, which is widely reported as the state-of-the-art in conventional machine learning methods. The F1-scores achieved by the baseline and optimized methods are 40.83% and 51.28%, respectively.
Authors: Hao Sun, Yong Jiang, Bo Wang, Yingyan Hou, Yan Zhang, Pengjun Xie, Fei Huang
Abstract: In-context learning (ICL) has been instrumental in adapting Large Language Models (LLMs) to downstream tasks using correct input-output examples. Recent advances have attempted to improve model performance through principles derived from mistakes, yet these approaches suffer from lack of customization and inadequate error coverage. To address these limitations, we propose Retrieved In-Context Principles (RICP), a novel teacher-student framework. In RICP, the teacher model analyzes mistakes from the student model to generate reasons and insights for preventing similar mistakes. These mistakes are clustered based on their underlying reasons for developing task-level principles, enhancing the error coverage of principles. During inference, the most relevant mistakes for each question are retrieved to create question-level principles, improving the customization of the provided guidance. RICP is orthogonal to existing prompting methods and does not require intervention from the teacher model during inference. Experimental results across seven reasoning benchmarks reveal that RICP effectively enhances performance when applied to various prompting strategies.
Authors: Bowen Shen, Zheng Lin, Daren Zha, Wei Liu, Jian Luan, Bin Wang, Weiping Wang
Abstract: Structured pruning fundamentally reduces computational and memory overheads of large language models (LLMs) and offers a feasible solution for end-side LLM deployment. Structurally pruned models remain dense and high-precision, highly compatible with further tuning and compression. However, as the coarse-grained structured pruning poses large damage to the highly interconnected model, achieving a high compression ratio for scaled-up LLMs remains a challenge. In this paper, we introduce a task-agnostic structured pruning approach coupled with a compact Transformer architecture design. The proposed approach, named TransAct, reduces transitional activations inside multi-head attention (MHA) and multi-layer perceptron (MLP) modules, while preserving the inter-module activations that are sensitive to perturbations. Hence, the LLM is pruned into an intra-module low-rank architecture, significantly reducing weights, KV Cache and attention computation. TransAct is implemented on the LLaMA model and evaluated on downstream benchmarks. Results verify the optimality of our approach at high compression with respect to both efficiency and performance. Further, ablation studies reveal the strength of activation-guided iterative pruning and provide experimental analysis on the redundancy of MHA and MLP modules.
Authors: Yutong Wu, Di Huang, Wenxuan Shi, Wei Wang, Lingzhe Gao, Shihao Liu, Ziyuan Nan, Kaizhao Yuan, Rui Zhang, Xishan Zhang, Zidong Du, Qi Guo, Yewen Pu, Dawei Yin, Xing Hu, Yunji Chen
Abstract: Recent advancements in open-source code large language models (LLMs) have demonstrated remarkable coding abilities by fine-tuning on the data generated from powerful closed-source LLMs such as GPT-3.5 and GPT-4 for instruction tuning. This paper explores how to further improve an instruction-tuned code LLM by generating data from itself rather than querying closed-source LLMs. Our key observation is the misalignment between the translation of formal and informal languages: translating formal language (i.e., code) to informal language (i.e., natural language) is more straightforward than the reverse. Based on this observation, we propose INVERSE-INSTRUCT, which summarizes instructions from code snippets instead of the reverse. Specifically, given an instruction tuning corpus for code and the resulting instruction-tuned code LLM, we ask the code LLM to generate additional high-quality instructions for the original corpus through code summarization and self-evaluation. Then, we fine-tune the base LLM on the combination of the original corpus and the self-generated one, which yields a stronger instruction-tuned LLM. We present a series of code LLMs named InverseCoder, which surpasses the performance of the original code LLMs on a wide range of benchmarks, including Python text-to-code generation, multilingual coding, and data-science code generation.
Authors: Chenxu Yang, Zheng Lin, Chong Tian, Liang Pang, Lanrui Wang, Zhengyang Tong, Qirong Ho, Yanan Cao, Weiping Wang
Abstract: Grounding external knowledge can enhance the factuality of responses in dialogue generation. However, excessive emphasis on it might result in the lack of engaging and diverse expressions. Through the introduction of randomness in sampling, current approaches can increase the diversity. Nevertheless, such sampling method could undermine the factuality in dialogue generation. In this study, to discover a solution for advancing creativity without relying on questionable randomness and to subtly reconcile the factuality and diversity within the source-grounded paradigm, a novel method named DoGe is proposed. DoGe can dynamically alternate between the utilization of internal parameter knowledge and external source knowledge based on the model's factual confidence. Extensive experiments on three widely-used datasets show that DoGe can not only enhance response diversity but also maintain factuality, and it significantly surpasses other various decoding strategy baselines.
Authors: Jinpeng Hu, Tengteng Dong, Hui Ma, Peng Zou, Xiao Sun, Meng Wang
Abstract: Mental health has attracted substantial attention in recent years and LLM can be an effective technology for alleviating this problem owing to its capability in text understanding and dialogue. However, existing research in this domain often suffers from limitations, such as training on datasets lacking crucial prior knowledge and evidence, and the absence of comprehensive evaluation methods. In this paper, we propose a specialized psychological large language model (LLM), named PsycoLLM, trained on a proposed high-quality psychological dataset, including single-turn QA, multi-turn dialogues enriched with prior knowledge and knowledge-based QA. Additionally, to compare the performance of PsycoLLM with other LLMs, we develop a comprehensive psychological benchmark based on authoritative psychological counseling examinations in China, which includes assessments of professional ethics, theoretical proficiency, and case analysis. The experimental results on the benchmark illustrates the effectiveness of PsycoLLM, which demonstrates superior performance compared to other LLMs.
Authors: Seungju Kim, Meounggun Jo
Abstract: Large Language Models (LLMs) have shown promise in Automated Essay Scoring (AES), but their zero-shot and few-shot performance often falls short compared to state-of-the-art models and human raters. However, fine-tuning LLMs for each specific task is impractical due to the variety of essay prompts and rubrics used in real-world educational contexts. This study proposes a novel approach combining LLMs and Comparative Judgment (CJ) for AES, using zero-shot prompting to choose between two essays. We demonstrate that a CJ method surpasses traditional rubric-based scoring in essay scoring using LLMs.
Authors: Daniela N. Rim, Heeyoul Choi
Abstract: This work explores the capability of conversational chatbots powered by large language models (LLMs), to understand and characterize predicate symmetry, a cognitive linguistic function traditionally believed to be an inherent human trait. Leveraging in-context learning (ICL), a paradigm shift enabling chatbots to learn new tasks from prompts without re-training, we assess the symmetrical reasoning of five chatbots: ChatGPT 4, Huggingface chat AI, Microsoft's Copilot AI, LLaMA through Perplexity, and Gemini Advanced. Using the Symmetry Inference Sentence (SIS) dataset by Tanchip et al. (2020), we compare chatbot responses against human evaluations to gauge their understanding of predicate symmetry. Experiment results reveal varied performance among chatbots, with some approaching human-like reasoning capabilities. Gemini, for example, reaches a correlation of 0.85 with human scores, while providing a sounding justification for each symmetry evaluation. This study underscores the potential and limitations of LLMs in mirroring complex cognitive processes as symmetrical reasoning.
Authors: Shangrui Nie, Michael Fromm, Charles Welch, Rebekka G\"orge, Akbar Karimi, Joan Plepi, Nazia Afsan Mowmita, Nicolas Flores-Herr, Mehdi Ali, Lucie Flek
Abstract: While preliminary findings indicate that multilingual LLMs exhibit reduced bias compared to monolingual ones, a comprehensive understanding of the effect of multilingual training on bias mitigation, is lacking. This study addresses this gap by systematically training six LLMs of identical size (2.6B parameters) and architecture: five monolingual models (English, German, French, Italian, and Spanish) and one multilingual model trained on an equal distribution of data across these languages, all using publicly available data. To ensure robust evaluation, standard bias benchmarks were automatically translated into the five target languages and verified for both translation quality and bias preservation by human annotators. Our results consistently demonstrate that multilingual training effectively mitigates bias. Moreover, we observe that multilingual models achieve not only lower bias but also superior prediction accuracy when compared to monolingual models with the same amount of training data, model architecture, and size.
Authors: Weiming Li, Manni Duan, Dong An, Yan Shao
Abstract: Large language models (LLMs) demonstrate extraordinary abilities in a wide range of natural language processing (NLP) tasks. In this paper, we show that, beyond text understanding capability, LLMs are capable of processing text layouts that are denoted by spatial markers. They are able to answer questions that require explicit spatial perceiving and reasoning, while a drastic performance drop is observed when the spatial markers from the original data are excluded. We perform a series of experiments with the GPT-3.5, Baichuan2, Llama2 and ChatGLM3 models on various types of layout-sensitive datasets for further analysis. The experimental results reveal that the layout understanding ability of LLMs is mainly introduced by the coding data for pretraining, which is further enhanced at the instruction-tuning stage. In addition, layout understanding can be enhanced by integrating low-cost, auto-generated data approached by a novel text game. Finally, we show that layout understanding ability is beneficial for building efficient visual question-answering (VQA) systems.
Authors: Alex Nguyen, Dheeraj Mekala, Chengyu Dong, Jingbo Shang
Abstract: Self-consistency (Wang et al., 2023) suggests that the most consistent answer obtained through large language models (LLMs) is more likely to be correct. In this paper, we challenge this argument and propose a nuanced correction. Our observations indicate that consistent answers derived through more computation i.e. longer reasoning texts, rather than simply the most consistent answer across all outputs, are more likely to be correct. This is predominantly because we demonstrate that LLMs can autonomously produce chain-of-thought (CoT) style reasoning with no custom prompts merely while generating longer responses, which lead to consistent predictions that are more accurate. In the zero-shot setting, by sampling Mixtral-8x7B model multiple times and considering longer responses, we achieve 86% of its self-consistency performance obtained through zero-shot CoT prompting on the GSM8K and MultiArith datasets. Finally, we demonstrate that the probability of LLMs generating a longer response is quite low, highlighting the need for decoding strategies conditioned on output length.
Authors: Atin Sakkeer Hussain, Anu Thomas
Abstract: Domain-specific Entity Recognition holds significant importance in legal contexts, serving as a fundamental task that supports various applications such as question-answering systems, text summarization, machine translation, sentiment analysis, and information retrieval specifically within case law documents. Recent advancements have highlighted the efficacy of Large Language Models in natural language processing tasks, demonstrating their capability to accurately detect and classify domain-specific facts (entities) from specialized texts like clinical and financial documents. This research investigates the application of Large Language Models in identifying domain-specific entities (e.g., courts, petitioner, judge, lawyer, respondents, FIR nos.) within case law documents, with a specific focus on their aptitude for handling domain-specific language complexity and contextual variations. The study evaluates the performance of state-of-the-art Large Language Model architectures, including Large Language Model Meta AI 3, Mistral, and Gemma, in the context of extracting judicial facts tailored to Indian judicial texts. Mistral and Gemma emerged as the top-performing models, showcasing balanced precision and recall crucial for accurate entity identification. These findings confirm the value of Large Language Models in judicial documents and demonstrate how they can facilitate and quicken scientific research by producing precise, organised data outputs that are appropriate for in-depth examination.
Authors: Nandini Mundra, Aditya Nanda Kishore, Raj Dabre, Ratish Puduppully, Anoop Kunchukuttan, Mitesh M. Khapra
Abstract: Language Models (LMs) excel in natural language processing tasks for English but show reduced performance in most other languages. This problem is commonly tackled by continually pre-training and fine-tuning these models for said languages. A significant issue in this process is the limited vocabulary coverage in the original model's tokenizer, leading to inadequate representation of new languages and necessitating an expansion of the tokenizer. The initialization of the embeddings corresponding to new vocabulary items presents a further challenge. Current strategies require cross-lingual embeddings and lack a solid theoretical foundation as well as comparisons with strong baselines. In this paper, we first establish theoretically that initializing within the convex hull of existing embeddings is a good initialization, followed by a novel but simple approach, Constrained Word2Vec (CW2V), which does not require cross-lingual embeddings. Our study evaluates different initialization methods for expanding RoBERTa and LLaMA 2 across four languages and five tasks. The results show that CW2V performs equally well or even better than more advanced techniques. Additionally, simpler approaches like multivariate initialization perform on par with these advanced methods indicating that efficient large-scale multilingual continued pretraining can be achieved even with simpler initialization methods.
Authors: Yanxu Zhu, Jinlin Xiao, Yuhang Wang, Jitao Sang
Abstract: Recent studies have demonstrated that large language models (LLMs) are susceptible to being misled by false premise questions (FPQs), leading to errors in factual knowledge, know as factuality hallucination. Existing benchmarks that assess this vulnerability primarily rely on manual construction, resulting in limited scale and lack of scalability. In this work, we introduce an automated, scalable pipeline to create FPQs based on knowledge graphs (KGs). The first step is modifying true triplets extracted from KGs to create false premises. Subsequently, utilizing the state-of-the-art capabilities of GPTs, we generate semantically rich FPQs. Based on the proposed method, we present a comprehensive benchmark, the Knowledge Graph-based False Premise Questions (KG-FPQ), which contains approximately 178k FPQs across three knowledge domains, at six levels of confusability, and in two task formats. Using KG-FPQ, we conduct extensive evaluations on several representative LLMs and provide valuable insights. The KG-FPQ dataset and code are available at~https://github.com/yanxuzhu/KG-FPQ.
Authors: Sanjeet Singh, Shreya Gupta, Niralee Gupta, Naimish Sharma, Lokesh Srivastava, Vibhu Agarwal, Ashutosh Modi
Abstract: The consequences of a healthcare data breach can be devastating for the patients, providers, and payers. The average financial impact of a data breach in recent months has been estimated to be close to USD 10 million. This is especially significant for healthcare organizations in India that are managing rapid digitization while still establishing data governance procedures that align with the letter and spirit of the law. Computer-based systems for de-identification of personal information are vulnerable to data drift, often rendering them ineffective in cross-institution settings. Therefore, a rigorous assessment of existing de-identification against local health datasets is imperative to support the safe adoption of digital health initiatives in India. Using a small set of de-identified patient discharge summaries provided by an Indian healthcare institution, in this paper, we report the nominal performance of de-identification algorithms (based on language models) trained on publicly available non-Indian datasets, pointing towards a lack of cross-institutional generalization. Similarly, experimentation with off-the-shelf de-identification systems reveals potential risks associated with the approach. To overcome data scarcity, we explore generating synthetic clinical reports (using publicly available and Indian summaries) by performing in-context learning over Large Language Models (LLMs). Our experiments demonstrate the use of generated reports as an effective strategy for creating high-performing de-identification systems with good generalization capabilities.
Authors: Anum Afzal, Alexander Kowsik, Rajna Fani, Florian Matthes
Abstract: Large Language Models have found application in various mundane and repetitive tasks including Human Resource (HR) support. We worked with the domain experts of SAP SE to develop an HR support chatbot as an efficient and effective tool for addressing employee inquiries. We inserted a human-in-the-loop in various parts of the development cycles such as dataset collection, prompt optimization, and evaluation of generated output. By enhancing the LLM-driven chatbot's response quality and exploring alternative retrieval methods, we have created an efficient, scalable, and flexible tool for HR professionals to address employee inquiries effectively. Our experiments and evaluation conclude that GPT-4 outperforms other models and can overcome inconsistencies in data through internal reasoning capabilities. Additionally, through expert analysis, we infer that reference-free evaluation metrics such as G-Eval and Prometheus demonstrate reliability closely aligned with that of human evaluation.
Authors: Yinquan Lu, Wenhao Zhu, Lei Li, Yu Qiao, Fei Yuan
Abstract: Large Language Models~(LLMs) demonstrate remarkable translation capabilities in high-resource language tasks, yet their performance in low-resource languages is hindered by insufficient multilingual data during pre-training. To address this, we dedicate 35,000 A100-SXM4-80GB GPU hours in conducting extensive multilingual continual pre-training on the LLaMA series models, enabling translation support across more than 100 languages. Through a comprehensive analysis of training strategies, such as vocabulary expansion and data augmentation, we develop LLaMAX. Remarkably, without sacrificing its generalization ability, LLaMAX achieves significantly higher translation performance compared to existing open-source LLMs~(by more than 10 spBLEU points) and performs on-par with specialized translation model~(M2M-100-12B) on the Flores-101 benchmark. Extensive experiments indicate that LLaMAX can serve as a robust multilingual foundation model. The code~\footnote{\url{https://github.com/CONE-MT/LLaMAX/.}} and models~\footnote{\url{https://huggingface.co/LLaMAX/.}} are publicly available.
URLs: https://github.com/CONE-MT/LLaMAX/., https://huggingface.co/LLaMAX/.
Authors: Chani Jung (KAIST), Dongkwan Kim (KAIST), Jiho Jin (KAIST), Jiseon Kim (KAIST), Yeon Seonwoo (Amazon), Yejin Choi (Allen Institute for AI, University of Washington), Alice Oh (KAIST), Hyunwoo Kim (Allen Institute for AI)
Abstract: While humans naturally develop theory of mind (ToM), the capability to understand other people's mental states and beliefs, state-of-the-art large language models (LLMs) underperform on simple ToM benchmarks. We posit that we can extend our understanding of LLMs' ToM abilities by evaluating key human ToM precursors -- perception inference and perception-to-belief inference -- in LLMs. We introduce two datasets, Percept-ToMi and Percept-FANToM, to evaluate these precursory inferences for ToM in LLMs by annotating characters' perceptions on ToMi and FANToM, respectively. Our evaluation of eight state-of-the-art LLMs reveals that the models generally perform well in perception inference while exhibiting limited capability in perception-to-belief inference (e.g., lack of inhibitory control). Based on these results, we present PercepToM, a novel ToM method leveraging LLMs' strong perception inference capability while supplementing their limited perception-to-belief inference. Experimental results demonstrate that PercepToM significantly enhances LLM's performance, especially in false belief scenarios.
Authors: Tommaso Mario Buonocore, Simone Rancati, Enea Parimbelli
Abstract: The development of domain-specific language models has significantly advanced natural language processing applications in various specialized fields, particularly in biomedicine. However, the focus has largely been on English-language models, leaving a gap for less-resourced languages such as Italian. This paper introduces Igea, the first decoder-only language model designed explicitly for biomedical text generation in Italian. Built on the Minerva model and continually pretrained on a diverse corpus of Italian medical texts, Igea is available in three model sizes: 350 million, 1 billion, and 3 billion parameters. The models aim to balance computational efficiency and performance, addressing the challenges of managing the peculiarities of medical terminology in Italian. We evaluate Igea using a mix of in-domain biomedical corpora and general-purpose benchmarks, highlighting its efficacy and retention of general knowledge even after the domain-specific training. This paper discusses the model's development and evaluation, providing a foundation for future advancements in Italian biomedical NLP.
Authors: Ping Yu, Jing Xu, Jason Weston, Ilia Kulikov
Abstract: Large language models (LLMs) can spend extra compute during inference to generate intermediate thoughts, which helps to produce better final responses. Since Chain-of-Thought (Wei et al., 2022), many such System 2 techniques have been proposed such as Rephrase and Respond (Deng et al., 2023a), System 2 Attention (Weston and Sukhbaatar, 2023) and Branch-Solve-Merge (Saha et al., 2023). In this work we investigate self-supervised methods to ``compile'' (distill) higher quality outputs from System 2 techniques back into LLM generations without intermediate reasoning token sequences, as this reasoning has been distilled into System 1. We show that several such techniques can be successfully distilled, resulting in improved results compared to the original System 1 performance, and with less inference cost than System 2. We posit that such System 2 distillation will be an important feature of future continually learning AI systems, enabling them to focus System 2 capabilities on the reasoning tasks that they cannot yet do well.
Authors: Miao Zheng, Hao Liang, Fan Yang, Haoze Sun, Tianpeng Li, Lingchu Xiong, Yan Zhang, Yozhen Wu, Kun Li, Yanjun Sheng, Mingan Lin, Tao Zhang, Guosheng Dong, Yujing Qiao, Kun Fang, Weipeng Chen, Bin Cui, Wentao Zhang, Zenan Zhou
Abstract: In recent years, the rise of Large Language Models (LLMs) has spurred a growing demand for plug-and-play AI systems. Among the various AI techniques, prompt engineering stands out as particularly significant. However, users often face challenges in writing prompts due to the steep learning curve and significant time investment, and existing automatic prompt engineering (APE) models can be difficult to use. To address this issue, we propose PAS, an LLM-based plug-and-play APE system. PAS utilizes LLMs trained on high-quality, automatically generated prompt complementary datasets, resulting in exceptional performance. In comprehensive benchmarks, PAS achieves state-of-the-art (SoTA) results compared to previous APE models, with an average improvement of 6.09 points. Moreover, PAS is highly efficient, achieving SoTA performance with only 9000 data points. Additionally, PAS can autonomously generate prompt augmentation data without requiring additional human labor. Its flexibility also allows it to be compatible with all existing LLMs and applicable to a wide range of tasks. PAS excels in human evaluations, underscoring its suitability as a plug-in for users. This combination of high performance, efficiency, and flexibility makes PAS a valuable system for enhancing the usability and effectiveness of LLMs through improved prompt engineering.
Authors: Nikit Srivastava, Mengshi Ma, Daniel Vollmers, Hamada Zahera, Diego Moussallem, Axel-Cyrille Ngonga Ngomo
Abstract: Knowledge Graph Question Answering (KGQA) simplifies querying vast amounts of knowledge stored in a graph-based model using natural language. However, the research has largely concentrated on English, putting non-English speakers at a disadvantage. Meanwhile, existing multilingual KGQA systems face challenges in achieving performance comparable to English systems, highlighting the difficulty of generating SPARQL queries from diverse languages. In this research, we propose a simplified approach to enhance multilingual KGQA systems by incorporating linguistic context and entity information directly into the processing pipeline of a language model. Unlike existing methods that rely on separate encoders for integrating auxiliary information, our strategy leverages a single, pretrained multilingual transformer-based language model to manage both the primary input and the auxiliary data. Our methodology significantly improves the language model's ability to accurately convert a natural language query into a relevant SPARQL query. It demonstrates promising results on the most recent QALD datasets, namely QALD-9-Plus and QALD-10. Furthermore, we introduce and evaluate our approach on Chinese and Japanese, thereby expanding the language diversity of the existing datasets.
Authors: Alan Wu, Ye Yuan, Ming Zhang
Abstract: Visually impaired people are a large group who can only use braille for reading and writing. However, the lack of special educational resources is the bottleneck for educating them. Educational equity is a reflection of the level of social civilization, cultural equality, and individual dignity. Facilitating and improving lifelong learning channels for the visually impaired is of great significance. Their written braille homework or exam papers cannot be understood by sighted teachers, because of the lack of a highly accurate braille translation system, especially in Chinese which has tone marks. braille writers often omit tone marks to save space, leading to confusion when braille with the same consonants and vowels is translated into Chinese. Previous algorithms were insufficient in extracting contextual information, resulting in low accuracy of braille translations into Chinese. This project informatively fine-tuned the mT5 model with an Encoder-decoder architecture for braille to Chinese character conversion. This research created a training set of braille and corresponding Chinese text from the Leipzig Corpora. This project significantly reduced the confusion in braille, achieving $62.4$ and $62.3$ BLEU scores in the validation and test sets, with a curriculum learning fine-tuning method. By incorporating the braille recognition algorithm, this project is the first publicly available braille translation system and can benefit lots of visually impaired students and families who are preparing for the Chinese College Test and help to propel their college dreams in the future. There is a demo on our homepage\footnote{\url{https://vision-braille.com/}}.
Authors: Afra Amini, Tim Vieira, Ryan Cotterell
Abstract: Best-of-N (BoN) is a popular and effective algorithm for aligning language models to human preferences. The algorithm works as follows: at inference time, N samples are drawn from the language model, and the sample with the highest reward, as judged by a reward model, is returned as the output. Despite its effectiveness, BoN is computationally expensive; it reduces sampling throughput by a factor of N. To make BoN more efficient at inference time, one strategy is to fine-tune the language model to mimic what BoN does during inference. To achieve this, we derive the distribution induced by the BoN algorithm. We then propose to fine-tune the language model to minimize backward KL divergence to the BoN distribution. Our approach is analogous to mean-field variational inference and, thus, we term it variational BoN (vBoN). To the extent this fine-tuning is successful and we end up with a good approximation, we have reduced the inference cost by a factor of N. Our experiments on a controlled generation task suggest that while variational BoN is not as effective as BoN in aligning language models, it is close to BoN performance as vBoN appears more often on the Pareto frontier of reward and KL divergence compared to models trained with KL-constrained RL objective.
Authors: Maor Ivgi, Ori Yoran, Jonathan Berant, Mor Geva
Abstract: Large language models (LLMs) often exhibit undesirable behaviors, such as hallucinations and sequence repetitions. We propose to view these behaviors as fallbacks that models exhibit under uncertainty, and investigate the connection between them. We categorize fallback behaviors -- sequence repetitions, degenerate text, and hallucinations -- and extensively analyze them in models from the same family that differ by the amount of pretraining tokens, parameter count, or the inclusion of instruction-following training. Our experiments reveal a clear and consistent ordering of fallback behaviors, across all these axes: the more advanced an LLM is (i.e., trained on more tokens, has more parameters, or instruction-tuned), its fallback behavior shifts from sequence repetitions, to degenerate text, and then to hallucinations. Moreover, the same ordering is observed throughout a single generation, even for the best-performing models; as uncertainty increases, models shift from generating hallucinations to producing degenerate text and then sequence repetitions. Lastly, we demonstrate that while common decoding techniques, such as random sampling, might alleviate some unwanted behaviors like sequence repetitions, they increase harder-to-detect hallucinations.
Authors: Jinliang Lu, Ziliang Pang, Min Xiao, Yaochen Zhu, Rui Xia, Jiajun Zhang
Abstract: The remarkable success of Large Language Models (LLMs) has ushered natural language processing (NLP) research into a new era. Despite their diverse capabilities, LLMs trained on different corpora exhibit varying strengths and weaknesses, leading to challenges in maximizing their overall efficiency and versatility. To address these challenges, recent studies have explored collaborative strategies for LLMs. This paper provides a comprehensive overview of this emerging research area, highlighting the motivation behind such collaborations. Specifically, we categorize collaborative strategies into three primary approaches: Merging, Ensemble, and Cooperation. Merging involves integrating multiple LLMs in the parameter space. Ensemble combines the outputs of various LLMs. Cooperation} leverages different LLMs to allow full play to their diverse capabilities for specific tasks. We provide in-depth introductions to these methods from different perspectives and discuss their potential applications. Additionally, we outline future research directions, hoping this work will catalyze further studies on LLM collaborations and paving the way for advanced NLP applications.
Authors: Kenya Andrews, Lamogha Chiazor
Abstract: Injustice occurs when someone experiences unfair treatment or their rights are violated and is often due to the presence of implicit biases and prejudice such as stereotypes. The automated identification of injustice in text has received little attention, due in part to the fact that underlying implicit biases or stereotypes are rarely explicitly stated and that instances often occur unconsciously due to the pervasive nature of prejudice in society. Here, we describe a novel framework that combines the use of a fine-tuned BERT-based bias detection model, two stereotype detection models, and a lexicon-based approach to show that epistemological biases (i.e., words, which presupposes, entails, asserts, hedges, or boosts text to erode or assert a person's capacity as a knower) can assist with the automatic detection of injustice in text. The news media has many instances of injustice (i.e. discriminatory narratives), thus it is our use case here. We conduct and discuss an empirical qualitative research study which shows how the framework can be applied to detect injustices, even at higher volumes of data.
Authors: Yadong Zhang, Shaoguang Mao, Wenshan Wu, Yan Xia, Tao Ge, Man Lan, Furu Wei
Abstract: This paper introduces BI-Directional DEliberation Reasoning (BIDDER), a novel reasoning approach to enhance the decision rationality of language models. Traditional reasoning methods typically rely on historical information and employ uni-directional (left-to-right) reasoning strategy. This lack of bi-directional deliberation reasoning results in limited awareness of potential future outcomes and insufficient integration of historical context, leading to suboptimal decisions. BIDDER addresses this gap by incorporating principles of rational decision-making, specifically managing uncertainty and predicting expected utility. Our approach involves three key processes: Inferring hidden states to represent uncertain information in the decision-making process from historical data; Using these hidden states to predict future potential states and potential outcomes; Integrating historical information (past contexts) and long-term outcomes (future contexts) to inform reasoning. By leveraging bi-directional reasoning, BIDDER ensures thorough exploration of both past and future contexts, leading to more informed and rational decisions. We tested BIDDER's effectiveness in two well-defined scenarios: Poker (Limit Texas Hold'em) and Negotiation. Our experiments demonstrate that BIDDER significantly improves the decision-making capabilities of LLMs and LLM agents.
Authors: Ethan Chern, Jiadi Su, Yan Ma, Pengfei Liu
Abstract: Previous open-source large multimodal models (LMMs) have faced several limitations: (1) they often lack native integration, requiring adapters to align visual representations with pre-trained large language models (LLMs); (2) many are restricted to single-modal generation; (3) while some support multimodal generation, they rely on separate diffusion models for visual modeling and generation. To mitigate these limitations, we present Anole, an open, autoregressive, native large multimodal model for interleaved image-text generation. We build Anole from Meta AI's Chameleon, adopting an innovative fine-tuning strategy that is both data-efficient and parameter-efficient. Anole demonstrates high-quality, coherent multimodal generation capabilities. We have open-sourced our model, training framework, and instruction tuning data.
Authors: Lukas Netz, Jan Reimar, Bernhard Rumpe
Abstract: We present and evaluate a method called grammar masking, which is used to guide large language models (LLMs) toward producing syntactically correct models for a given context-free grammar. Prompt engineering methods such as few-shot learning or priming can be used to improve the chances of an LLM producing correct syntax, but the more complex the grammar, the more time-consuming and less promising these methods become. Previous work is focused primarily on the usage of either language model training or prompt engineering. In this work, a method is presented that restricts the output to a given grammar using constrained decoding to ensure the output adheres to a valid syntax. We use several DSLs built with MontiCore and task multiple LLMs to produce models with and without constrained decoding. A corresponding parser is used to confirm the syntactic correctness of each model. We show that grammar masking can dramatically improve the modeling capabilities of several LLMs, reducing the need for well-refined prompting while increasing the chance of producing correct models.
Authors: Zesen Liu, Tianshuo Cong, Xinlei He, Qi Li
Abstract: Large Language Models (LLMs) excel in various applications, including text generation and complex tasks. However, the misuse of LLMs raises concerns about the authenticity and ethical implications of the content they produce, such as deepfake news, academic fraud, and copyright infringement. Watermarking techniques, which embed identifiable markers in machine-generated text, offer a promising solution to these issues by allowing for content verification and origin tracing. Unfortunately, the robustness of current LLM watermarking schemes under potential watermark removal attacks has not been comprehensively explored. In this paper, to fill this gap, we first systematically comb the mainstream watermarking schemes and removal attacks on machine-generated texts, and then we categorize them into pre-text (before text generation) and post-text (after text generation) classes so that we can conduct diversified analyses. In our experiments, we evaluate eight watermarks (five pre-text, three post-text) and twelve attacks (two pre-text, ten post-text) across 87 scenarios. Evaluation results indicate that (1) KGW and Exponential watermarks offer high text quality and watermark retention but remain vulnerable to most attacks; (2) Post-text attacks are found to be more efficient and practical than pre-text attacks; (3) Pre-text watermarks are generally more imperceptible, as they do not alter text fluency, unlike post-text watermarks; (4) Additionally, combined attack methods can significantly increase effectiveness, highlighting the need for more robust watermarking solutions. Our study underscores the vulnerabilities of current techniques and the necessity for developing more resilient schemes.
Authors: Zhaorun Chen, Yichao Du, Zichen Wen, Yiyang Zhou, Chenhang Cui, Zhenzhen Weng, Haoqin Tu, Chaoqi Wang, Zhengwei Tong, Qinglan Huang, Canyu Chen, Qinghao Ye, Zhihong Zhu, Yuqing Zhang, Jiawei Zhou, Zhuokai Zhao, Rafael Rafailov, Chelsea Finn, Huaxiu Yao
Abstract: While text-to-image models like DALLE-3 and Stable Diffusion are rapidly proliferating, they often encounter challenges such as hallucination, bias, and the production of unsafe, low-quality output. To effectively address these issues, it is crucial to align these models with desired behaviors based on feedback from a multimodal judge. Despite their significance, current multimodal judges frequently undergo inadequate evaluation of their capabilities and limitations, potentially leading to misalignment and unsafe fine-tuning outcomes. To address this issue, we introduce MJ-Bench, a novel benchmark which incorporates a comprehensive preference dataset to evaluate multimodal judges in providing feedback for image generation models across four key perspectives: alignment, safety, image quality, and bias. Specifically, we evaluate a large variety of multimodal judges including smaller-sized CLIP-based scoring models, open-source VLMs (e.g. LLaVA family), and close-source VLMs (e.g. GPT-4o, Claude 3) on each decomposed subcategory of our preference dataset. Experiments reveal that close-source VLMs generally provide better feedback, with GPT-4o outperforming other judges in average. Compared with open-source VLMs, smaller-sized scoring models can provide better feedback regarding text-image alignment and image quality, while VLMs provide more accurate feedback regarding safety and generation bias due to their stronger reasoning capabilities. Further studies in feedback scale reveal that VLM judges can generally provide more accurate and stable feedback in natural language (Likert-scale) than numerical scales. Notably, human evaluations on end-to-end fine-tuned models using separate feedback from these multimodal judges provide similar conclusions, further confirming the effectiveness of MJ-Bench. All data, code, models are available at https://huggingface.co/MJ-Bench.
Authors: Lucas Saldyt, Subbarao Kambhampati
Abstract: Important tasks such as reasoning and planning are fundamentally algorithmic, meaning that solving them robustly requires acquiring true reasoning or planning algorithms, rather than shortcuts. Large Language Models lack true algorithmic ability primarily because of the limitations of neural network optimization algorithms, their optimization data and optimization objective, but also due to architectural inexpressivity. To solve this, our paper proposes augmenting LLMs with a library of fundamental operations and sophisticated differentiable programs, so that common algorithms do not need to be learned from scratch. We add memory, registers, basic operations, and adaptive recurrence to a transformer architecture built on LLaMA3. Then, we define a method for directly compiling algorithms into a differentiable starting library, which is used natively and propagates gradients for optimization. In this preliminary study, we explore the feasability of augmenting LLaMA3 with a differentiable computer, for instance by fine-tuning small transformers on simple algorithmic tasks with variable computational depth.
Authors: Tiancheng Zhao, Qianqian Zhang, Kyusong Lee, Peng Liu, Lu Zhang, Chunxin Fang, Jiajia Liao, Kelei Jiang, Yibo Ma, Ruochen Xu
Abstract: We introduce OmChat, a model designed to excel in handling long contexts and video understanding tasks. OmChat's new architecture standardizes how different visual inputs are processed, making it more efficient and adaptable. It uses a dynamic vision encoding process to effectively handle images of various resolutions, capturing fine details across a range of image qualities. OmChat utilizes an active progressive multimodal pretraining strategy, which gradually increases the model's capacity for long contexts and enhances its overall abilities. By selecting high-quality data during training, OmChat learns from the most relevant and informative data points. With support for a context length of up to 512K, OmChat demonstrates promising performance in tasks involving multiple images and videos, outperforming most open-source models in these benchmarks. Additionally, OmChat proposes a prompting strategy for unifying complex multimodal inputs including single image text, multi-image text and videos, and achieving competitive performance on single-image benchmarks. To further evaluate the model's capabilities, we proposed a benchmark dataset named Temporal Visual Needle in a Haystack. This dataset assesses OmChat's ability to comprehend temporal visual details within long videos. Our analysis highlights several key factors contributing to OmChat's success: support for any-aspect high image resolution, the active progressive pretraining strategy, and high-quality supervised fine-tuning datasets. This report provides a detailed overview of OmChat's capabilities and the strategies that enhance its performance in visual understanding.
Authors: Yijia Xiao, Edward Sun, Tianyu Liu, Wei Wang
Abstract: We propose LogicVista, an evaluation benchmark that assesses the integrated logical reasoning capabilities of multimodal large language models (MLLMs) in Visual contexts. Recent advancements in MLLMs have demonstrated various fascinating abilities, from crafting poetry based on an image to performing mathematical reasoning. However, there is still a lack of systematic evaluation of MLLMs' proficiency in logical reasoning tasks, which are essential for activities like navigation and puzzle-solving. Thus we evaluate general logical cognition abilities across 5 logical reasoning tasks encompassing 9 different capabilities, using a sample of 448 multiple-choice questions. Each question is annotated with the correct answer and the human-written reasoning behind the selection, enabling both open-ended and multiple-choice evaluation. A total of 8 MLLMs are comprehensively evaluated using LogicVista. Code and Data Available at https://github.com/Yijia-Xiao/LogicVista.
Authors: Ulin Nuha, Chih-Hsueh Lin
Abstract: Several machine learning schemes have attempted to perform the detection of spam messages. However, those schemes mostly require a huge amount of labeled data. The existing techniques addressing the lack of data availability have issues with effectiveness and robustness. Therefore, this paper proposes a conditional semi-supervised data augmentation (CSSDA) for a spam detection model lacking the availability of data. The main architecture of CSSDA comprises feature extraction and enhanced generative network. Here, we exploit unlabeled data for data augmentation to extend training data. The enhanced generative in our proposed scheme produces latent variables as fake samples from unlabeled data through a conditional scheme. Latent variables can come from labeled and unlabeled data as the input for the final classifier in our spam detection model. The experimental results indicate that our proposed CSSDA achieves excellent results compared to several related methods both exploiting unlabeled data and not. In the experiment stage with various amounts of unlabeled data, CSSDA is the only robust model that obtains a balanced accuracy of about 85% when the availability of labeled data is large. We also conduct several ablation studies to investigate our proposed scheme in detail. The result also shows that several ablation studies strengthen our proposed innovations. These experiments indicate that unlabeled data has a significant contribution to data augmentation using the conditional semi-supervised scheme for spam detection.
Authors: Shaowen Wang, Linxi Yu, Jian Li
Abstract: Fine-tuning large-scale pretrained models is prohibitively expensive in terms of computational and memory costs. LoRA, as one of the most popular Parameter-Efficient Fine-Tuning (PEFT) methods, offers a cost-effective alternative by fine-tuning an auxiliary low-rank model that has significantly fewer parameters. Although LoRA reduces the computational and memory requirements significantly at each iteration, extensive empirical evidence indicates that it converges at a considerably slower rate compared to full fine-tuning, ultimately leading to increased overall compute and often worse test performance. In our paper, we perform an in-depth investigation of the initialization method of LoRA and show that careful initialization (without any change of the architecture and the training algorithm) can significantly enhance both efficiency and performance. In particular, we introduce a novel initialization method, LoRA-GA (Low Rank Adaptation with Gradient Approximation), which aligns the gradients of low-rank matrix product with those of full fine-tuning at the first step. Our extensive experiments demonstrate that LoRA-GA achieves a convergence rate comparable to that of full fine-tuning (hence being significantly faster than vanilla LoRA as well as various recent improvements) while simultaneously attaining comparable or even better performance. For example, on the subset of the GLUE dataset with T5-Base, LoRA-GA outperforms LoRA by 5.69% on average. On larger models such as Llama 2-7B, LoRA-GA shows performance improvements of 0.34, 11.52%, and 5.05% on MT-bench, GSM8K, and Human-eval, respectively. Additionally, we observe up to 2-4 times convergence speed improvement compared to vanilla LoRA, validating its effectiveness in accelerating convergence and enhancing model performance. Code is available at https://github.com/Outsider565/LoRA-GA.
Authors: Kai Shen, Lingfei Wu, Siliang Tang, Fangli Xu, Bo Long, Yueting Zhuang, Jian Pei
Abstract: The visual question generation (VQG) task aims to generate human-like questions from an image and potentially other side information (e.g. answer type). Previous works on VQG fall in two aspects: i) They suffer from one image to many questions mapping problem, which leads to the failure of generating referential and meaningful questions from an image. ii) They fail to model complex implicit relations among the visual objects in an image and also overlook potential interactions between the side information and image. To address these limitations, we first propose a novel learning paradigm to generate visual questions with answer-awareness and region-reference. Concretely, we aim to ask the right visual questions with Double Hints - textual answers and visual regions of interests, which could effectively mitigate the existing one-to-many mapping issue. Particularly, we develop a simple methodology to self-learn the visual hints without introducing any additional human annotations. Furthermore, to capture these sophisticated relationships, we propose a new double-hints guided Graph-to-Sequence learning framework, which first models them as a dynamic graph and learns the implicit topology end-to-end, and then utilizes a graph-to-sequence model to generate the questions with double hints. Experimental results demonstrate the priority of our proposed method.
Authors: Peng Xia, Kangyu Zhu, Haoran Li, Hongtu Zhu, Yun Li, Gang Li, Linjun Zhang, Huaxiu Yao
Abstract: The recent emergence of Medical Large Vision Language Models (Med-LVLMs) has enhanced medical diagnosis. However, current Med-LVLMs frequently encounter factual issues, often generating responses that do not align with established medical facts. Retrieval-Augmented Generation (RAG), which utilizes external knowledge, can improve the factual accuracy of these models but introduces two major challenges. First, limited retrieved contexts might not cover all necessary information, while excessive retrieval can introduce irrelevant and inaccurate references, interfering with the model's generation. Second, in cases where the model originally responds correctly, applying RAG can lead to an over-reliance on retrieved contexts, resulting in incorrect answers. To address these issues, we propose RULE, which consists of two components. First, we introduce a provably effective strategy for controlling factuality risk through the calibrated selection of the number of retrieved contexts. Second, based on samples where over-reliance on retrieved contexts led to errors, we curate a preference dataset to fine-tune the model, balancing its dependence on inherent knowledge and retrieved contexts for generation. We demonstrate the effectiveness of RULE on three medical VQA datasets, achieving an average improvement of 20.8% in factual accuracy. We publicly release our benchmark and code in https://github.com/richard-peng-xia/RULE.
Authors: Kuei-Chun Kao, Ruochen Wang, Cho-Jui Hsieh
Abstract: Large Language Models (LLMs) have demonstrated remarkable performance in solving math problems, a hallmark of human intelligence. Despite high success rates on current benchmarks; however, these often feature simple problems with only one or two unknowns, which do not sufficiently challenge their reasoning capacities. This paper introduces a novel benchmark, BeyondX, designed to address these limitations by incorporating problems with multiple unknowns. Recognizing the challenges in proposing multi-unknown problems from scratch, we developed BeyondX using an innovative automated pipeline that progressively increases complexity by expanding the number of unknowns in simpler problems. Empirical study on BeyondX reveals that the performance of existing LLMs, even those fine-tuned specifically on math tasks, significantly decreases as the number of unknowns increases - with a performance drop of up to 70\% observed in GPT-4. To tackle these challenges, we propose the Formulate-and-Solve strategy, a generalized prompting approach that effectively handles problems with an arbitrary number of unknowns. Our findings reveal that this strategy not only enhances LLM performance on the BeyondX benchmark but also provides deeper insights into the computational limits of LLMs when faced with more complex mathematical challenges.
Authors: Sander Noels, S\'ebastien Viaene, Tijl De Bie
Abstract: This paper addresses a long-standing problem in the field of accounting: mapping company-specific ledger accounts to a standardized chart of accounts. We propose a novel solution, TopoLedgerBERT, a unique sentence embedding method devised specifically for ledger account mapping. This model integrates hierarchical information from the charts of accounts into the sentence embedding process, aiming to accurately capture both the semantic similarity and the hierarchical structure of the ledger accounts. In addition, we introduce a data augmentation strategy that enriches the training data and, as a result, increases the performance of our proposed model. Compared to benchmark methods, TopoLedgerBERT demonstrates superior performance in terms of accuracy and mean reciprocal rank.
Authors: Ben Fitzgerald
Abstract: Models like Delphi have been able to label ethical dilemmas as moral or immoral with astonishing accuracy. This paper challenges accuracy as a holistic metric for ethics modeling by identifying issues with translating moral dilemmas into text-based input. It demonstrates these issues with contrast sets that substantially reduce the performance of classifiers trained on the dataset Moral Stories. Ultimately, we obtain concrete estimates for how much specific forms of data misrepresentation harm classifier accuracy. Specifically, label-changing tweaks to the descriptive content of a situation (as small as 3-5 words) can reduce classifier accuracy to as low as 51%, almost half the initial accuracy of 99.8%. Associating situations with a misleading social norm lowers accuracy to 98.8%, while adding textual bias (i.e. an implication that a situation already fits a certain label) lowers accuracy to 77%. These results suggest not only that many ethics models have substantially overfit, but that several precautions are required to ensure that input accurately captures a moral dilemma. This paper recommends re-examining the structure of a social norm, training models to ask for context with defeasible reasoning, and filtering input for textual bias. Doing so not only gives us the first concrete estimates of the average cost to accuracy of misrepresenting ethics data, but gives researchers practical tips for considering these estimates in research.
Authors: Yan Wang, Yawen Zeng, Jingsheng Zheng, Xiaofen Xing, Jin Xu, Xiangmin Xu
Abstract: Multimodal large language models (MLLMs) are flourishing, but mainly focus on images with less attention than videos, especially in sub-fields such as prompt engineering, video chain-of-thought (CoT), and instruction tuning on videos. Therefore, we try to explore the collection of CoT datasets in videos to lead to video OpenQA and improve the reasoning ability of MLLMs. Unfortunately, making such video CoT datasets is not an easy task. Given that human annotation is too cumbersome and expensive, while machine-generated is not reliable due to the hallucination issue, we develop an automatic annotation tool that combines machine and human experts, under the active learning paradigm. Active learning is an interactive strategy between the model and human experts, in this way, the workload of human labeling can be reduced and the quality of the dataset can be guaranteed. With the help of the automatic annotation tool, we strive to contribute three datasets, namely VideoCoT, TopicQA, TopicCoT. Furthermore, we propose a simple but effective benchmark based on the collected datasets, which exploits CoT to maximize the complex reasoning capabilities of MLLMs. Extensive experiments demonstrate the effectiveness our solution.
Authors: Haorui He, Zengqiang Shang, Chaoren Wang, Xuyuan Li, Yicheng Gu, Hua Hua, Liwei Liu, Chen Yang, Jiaqi Li, Peiyang Shi, Yuancheng Wang, Kai Chen, Pengyuan Zhang, Zhizheng Wu
Abstract: Recently, speech generation models have made significant progress by using large-scale training data. However, the research community struggle to produce highly spontaneous and human-like speech due to the lack of large-scale, diverse, and spontaneous speech data. This paper presents \textit{Emilia}, the first multilingual speech generation dataset from in-the-wild speech data, and Emilia-Pipe, the first open-source preprocessing pipeline designed to transform in-the-wild speech data into high-quality training data with annotations for speech generation. Emilia starts with over 101k hours of speech in six languages and features diverse speech with varied speaking styles. To facilitate the scale-up of Emilia, the open-source pipeline Emilia-Pipe can process one hour of raw speech data ready for model training in a few mins, which enables the research community to collaborate on large-scale speech generation research. Experimental results validate the effectiveness of Emilia. Demos are available at: https://emilia-dataset.github.io/Emilia-Demo-Page/.
Authors: Lai-Man Po, Yuyang Liu, Haoxuan Wu, Tianqi Zhang, Wing-Yin Yu, Zeyu Jiang, Kun Li
Abstract: This paper introduces Standard Basis LoRA (SBoRA), a novel parameter-efficient fine-tuning approach for Large Language Models that builds upon the pioneering works of Low-Rank Adaptation (LoRA) and Orthogonal Adaptation. SBoRA further reduces the computational and memory requirements of LoRA while enhancing learning performance. By leveraging orthogonal standard basis vectors to initialize one of the low-rank matrices, either A or B, SBoRA enables regional weight updates and memory-efficient fine-tuning. This approach gives rise to two variants, SBoRA-FA and SBoRA-FB, where only one of the matrices is updated, resulting in a sparse update matrix with a majority of zero rows or columns. Consequently, the majority of the fine-tuned model's weights remain unchanged from the pre-trained weights. This characteristic of SBoRA, wherein regional weight updates occur, is reminiscent of the modular organization of the human brain, which efficiently adapts to new tasks. Our empirical results demonstrate the superiority of SBoRA-FA over LoRA in various fine-tuning tasks, including commonsense reasoning and arithmetic reasoning. Furthermore, we evaluate the effectiveness of QSBoRA on quantized LLaMA models of varying scales, highlighting its potential for efficient adaptation to new tasks. Code is available at https://github.com/CityUHK-AI/SBoRA
Authors: Vishnu S. Pendyala, Madhulika Dutta
Abstract: Misinformation is still a major societal problem and the arrival of Large Language Models (LLMs) only added to it. This paper analyzes synthetic, false, and genuine information in the form of text from spectral analysis, visualization, and explainability perspectives to find the answer to why the problem is still unsolved despite multiple years of research and a plethora of solutions in the literature. Various embedding techniques on multiple datasets are used to represent information for the purpose. The diverse spectral and non-spectral methods used on these embeddings include t-distributed Stochastic Neighbor Embedding (t-SNE), Principal Component Analysis (PCA), and Variational Autoencoders (VAEs). Classification is done using multiple machine learning algorithms. Local Interpretable Model-Agnostic Explanations (LIME), SHapley Additive exPlanations (SHAP), and Integrated Gradients are used for the explanation of the classification. The analysis and the explanations generated show that misinformation is quite closely intertwined with genuine information and the machine learning algorithms are not as effective in separating the two despite the claims in the literature.
Authors: Dongxu Zhang, Varun Gangal, Barrett Martin Lattimer, Yi Yang
Abstract: Detecting hallucinations in large language model (LLM) outputs is pivotal, yet traditional fine-tuning for this classification task is impeded by the expensive and quickly outdated annotation process, especially across numerous vertical domains and in the face of rapid LLM advancements. In this study, we introduce an approach that automatically generates both faithful and hallucinated outputs by rewriting system responses. Experimental findings demonstrate that a T5-base model, fine-tuned on our generated dataset, surpasses state-of-the-art zero-shot detectors and existing synthetic generation methods in both accuracy and latency, indicating efficacy of our approach.
Authors: Mingchen Li, Xuechen Zhang, Yixiao Huang, Samet Oymak
Abstract: The transformer architecture has catalyzed revolutionary advances in language modeling. However, recent architectural recipes, such as state-space models, have bridged the performance gap. Motivated by this, we examine the benefits of Convolution-Augmented Transformer (CAT) for recall, copying, and length generalization tasks. CAT incorporates convolutional filters in the K/Q/V embeddings of an attention layer. Through CAT, we show that the locality of the convolution synergizes with the global view of the attention. Unlike comparable architectures, such as Mamba or transformer, CAT can provably solve the associative recall (AR) and copying tasks using a single layer while also enjoying guaranteed length generalization. We also establish computational tradeoffs between convolution and attention by characterizing how convolution can mitigate the need for full attention by summarizing the context window and creating salient summary tokens to attend. Evaluations on real datasets corroborate our findings and demonstrate that CAT and its variations indeed enhance the language modeling performance.
Authors: Xiaoxiao Miao, Ruijie Tao, Chang Zeng, Xin Wang
Abstract: Privacy-preserving voice protection approaches primarily suppress privacy-related information derived from paralinguistic attributes while preserving the linguistic content. Existing solutions focus on single-speaker scenarios. However, they lack practicality for real-world applications, i.e., multi-speaker scenarios. In this paper, we present an initial attempt to provide a multi-speaker anonymization benchmark by defining the task and evaluation protocol, proposing benchmarking solutions, and discussing the privacy leakage of overlapping conversations. Specifically, ideal multi-speaker anonymization should preserve the number of speakers and the turn-taking structure of the conversation, ensuring accurate context conveyance while maintaining privacy. To achieve that, a cascaded system uses speaker diarization to aggregate the speech of each speaker and speaker anonymization to conceal speaker privacy and preserve speech content. Additionally, we propose two conversation-level speaker vector anonymization methods to improve the utility further. Both methods aim to make the original and corresponding pseudo-speaker identities of each speaker unlinkable while preserving or even improving the distinguishability among pseudo-speakers in a conversation. The first method minimizes the differential similarity across speaker pairs in the original and anonymized conversations to maintain original speaker relationships in the anonymized version. The other method minimizes the aggregated similarity across anonymized speakers to achieve better differentiation between speakers. Experiments conducted on both non-overlap simulated and real-world datasets demonstrate the effectiveness of the multi-speaker anonymization system with the proposed speaker anonymizers. Additionally, we analyzed overlapping speech regarding privacy leakage and provide potential solutions.
Authors: Ken Nishida, Kojiro Machi, Kazuma Onishi, Katsuhiko Hayashi, Hidetaka Kamigaito
Abstract: The extreme multi-label classification~(XMC) task involves learning a classifier that can predict from a large label set the most relevant subset of labels for a data instance. While deep neural networks~(DNNs) have demonstrated remarkable success in XMC problems, the task is still challenging because it must deal with a large number of output labels, which make the DNN training computationally expensive. This paper addresses the issue by exploring the use of random circular vectors, where each vector component is represented as a complex amplitude. In our framework, we can develop an output layer and loss function of DNNs for XMC by representing the final output layer as a fully connected layer that directly predicts a low-dimensional circular vector encoding a set of labels for a data instance. We conducted experiments on synthetic datasets to verify that circular vectors have better label encoding capacity and retrieval ability than normal real-valued vectors. Then, we conducted experiments on actual XMC datasets and found that these appealing properties of circular vectors contribute to significant improvements in task performance compared with a previous model using random real-valued vectors, while reducing the size of the output layers by up to 99%.
Authors: Harshit Joshi, Shicheng Liu, James Chen, Robert Weigle, Monica S. Lam
Abstract: Programming LLM-based knowledge and task assistants that faithfully conform to developer-provided policies is challenging. These agents must retrieve and provide consistent, accurate, and relevant information to address user's queries and needs. Yet such agents generate unfounded responses ("hallucinate"). Traditional dialogue trees can only handle a limited number of conversation flows, making them inherently brittle. To this end, we present KITA - a programmable framework for creating task-oriented conversational agents that are designed to handle complex user interactions. Unlike LLMs, KITA provides reliable grounded responses, with controllable agent policies through its expressive specification, KITA Worksheet. In contrast to dialog trees, it is resilient to diverse user queries, helpful with knowledge sources, and offers ease of programming policies through its declarative paradigm. Through a real-user study involving 62 participants, we show that KITA beats the GPT-4 with function calling baseline by 26.1, 22.5, and 52.4 points on execution accuracy, dialogue act accuracy, and goal completion rate, respectively. We also release 22 real-user conversations with KITA manually corrected to ensure accuracy.
Authors: Sara Hooker
Abstract: At face value, this essay is about understanding a fairly esoteric governance tool called compute thresholds. However, in order to grapple with whether these thresholds will achieve anything, we must first understand how they came to be. This requires engaging with a decades-old debate at the heart of computer science progress, namely, is bigger always better? Hence, this essay may be of interest not only to policymakers and the wider public but also to computer scientists interested in understanding the role of compute in unlocking breakthroughs. Does a certain inflection point of compute result in changes to the risk profile of a model? This discussion is increasingly urgent given the wide adoption of governance approaches that suggest greater compute equates with higher propensity for harm. Several leading frontier AI companies have released responsible scaling policies. Both the White House Executive Orders on AI Safety (EO) and the EU AI Act encode the use of FLOP or floating-point operations as a way to identify more powerful systems. What is striking about the choice of compute thresholds to-date is that no models currently deployed in the wild fulfill the current criteria set by the EO. This implies that the emphasis is often not on auditing the risks and harms incurred by currently deployed models - but rather is based upon the belief that future levels of compute will introduce unforeseen new risks. A key conclusion of this essay is that compute thresholds as currently implemented are shortsighted and likely to fail to mitigate risk. Governance that is overly reliant on compute fails to understand that the relationship between compute and risk is highly uncertain and rapidly changing. It also overestimates our ability to predict what abilities emerge at different scales. This essay ends with recommendations for a better way forward.
Authors: Jiaqi Chen, Bingqian Lin, Xinmin Liu, Xiaodan Liang, Kwan-Yee K. Wong
Abstract: LLM-based agents have demonstrated impressive zero-shot performance in the vision-language navigation (VLN) task. However, these zero-shot methods focus only on solving high-level task planning by selecting nodes in predefined navigation graphs for movements, overlooking low-level control in realistic navigation scenarios. To bridge this gap, we propose AO-Planner, a novel affordances-oriented planning framework for continuous VLN task. Our AO-Planner integrates various foundation models to achieve affordances-oriented motion planning and action decision-making, both performed in a zero-shot manner. Specifically, we employ a visual affordances prompting (VAP) approach, where visible ground is segmented utilizing SAM to provide navigational affordances, based on which the LLM selects potential next waypoints and generates low-level path planning towards selected waypoints. We further introduce a high-level agent, PathAgent, to identify the most probable pixel-based path and convert it into 3D coordinates to fulfill low-level motion. Experimental results on the challenging R2R-CE benchmark demonstrate that AO-Planner achieves state-of-the-art zero-shot performance (5.5% improvement in SPL). Our method establishes an effective connection between LLM and 3D world to circumvent the difficulty of directly predicting world coordinates, presenting novel prospects for employing foundation models in low-level motion control.
Authors: Nikhil Abhyankar, Vivek Gupta, Dan Roth, Chandan K. Reddy
Abstract: Tabular reasoning involves interpreting unstructured queries against structured tables, requiring a synthesis of textual understanding and symbolic reasoning. Existing methods rely on either of the approaches and are constrained by their respective limitations. Textual reasoning excels in semantic interpretation unlike symbolic reasoning (SQL logic), but falls short in mathematical reasoning where SQL excels. In this paper, we introduce a novel algorithm H-STAR, comprising table extraction and adaptive reasoning, integrating both symbolic and semantic (text-based) approaches. To enhance evidence extraction, H-STAR employs a multi-view approach, incorporating step-by-step row and column retrieval. It also adapts reasoning strategies based on question types, utilizing symbolic reasoning for quantitative and logical tasks, and semantic reasoning for direct lookup and complex lexical queries. Our extensive experiments demonstrate that H-STAR significantly outperforms state-of-the-art methods across three tabular question-answering (QA) and fact-verification datasets, underscoring its effectiveness and efficiency.
Authors: Yibo Miao, Yifan Zhu, Yinpeng Dong, Lijia Yu, Jun Zhu, Xiao-Shan Gao
Abstract: The recent development of Sora leads to a new era in text-to-video (T2V) generation. Along with this comes the rising concern about its security risks. The generated videos may contain illegal or unethical content, and there is a lack of comprehensive quantitative understanding of their safety, posing a challenge to their reliability and practical deployment. Previous evaluations primarily focus on the quality of video generation. While some evaluations of text-to-image models have considered safety, they cover fewer aspects and do not address the unique temporal risk inherent in video generation. To bridge this research gap, we introduce T2VSafetyBench, a new benchmark designed for conducting safety-critical assessments of text-to-video models. We define 12 critical aspects of video generation safety and construct a malicious prompt dataset using LLMs and jailbreaking prompt attacks. Based on our evaluation results, we draw several important findings, including: 1) no single model excels in all aspects, with different models showing various strengths; 2) the correlation between GPT-4 assessments and manual reviews is generally high; 3) there is a trade-off between the usability and safety of text-to-video generative models. This indicates that as the field of video generation rapidly advances, safety risks are set to surge, highlighting the urgency of prioritizing video safety. We hope that T2VSafetyBench can provide insights for better understanding the safety of video generation in the era of generative AI.
Authors: Qi Jia, Baoyu Fan, Cong Xu, Lu Liu, Liang Jin, Guoguang Du, Zhenhua Guo, Yaqian Zhao, Xuanjing Huang, Rengang Li
Abstract: Existing video multi-modal sentiment analysis mainly focuses on the sentiment expression of people within the video, yet often neglects the induced sentiment of viewers while watching the videos. Induced sentiment of viewers is essential for inferring the public response to videos, has broad application in analyzing public societal sentiment, effectiveness of advertising and other areas. The micro videos and the related comments provide a rich application scenario for viewers induced sentiment analysis. In light of this, we introduces a novel research task, Multi-modal Sentiment Analysis for Comment Response of Video Induced(MSA-CRVI), aims to inferring opinions and emotions according to the comments response to micro video. Meanwhile, we manually annotate a dataset named Comment Sentiment toward to Micro Video (CSMV) to support this research. It is the largest video multi-modal sentiment dataset in terms of scale and video duration to our knowledge, containing 107,267 comments and 8,210 micro videos with a video duration of 68.83 hours. To infer the induced sentiment of comment should leverage the video content, so we propose the Video Content-aware Comment Sentiment Analysis (VC-CSA) method as baseline to address the challenges inherent in this new task. Extensive experiments demonstrate that our method is showing significant improvements over other established baselines.
Authors: Shihan Dou, Haoxiang Jia, Shenxi Wu, Huiyuan Zheng, Weikang Zhou, Muling Wu, Mingxu Chai, Jessica Fan, Caishuang Huang, Yunbo Tao, Yan Liu, Enyu Zhou, Ming Zhang, Yuhao Zhou, Yueming Wu, Rui Zheng, Ming Wen, Rongxiang Weng, Jingang Wang, Xunliang Cai, Tao Gui, Xipeng Qiu, Qi Zhang, Xuanjing Huang
Abstract: The increasing development of large language models (LLMs) in code generation has drawn significant attention among researchers. To enhance LLM-based code generation ability, current efforts are predominantly directed towards collecting high-quality datasets and leveraging diverse training technologies. However, there is a notable lack of comprehensive studies examining the limitations and boundaries of these existing methods. To bridge this gap, we conducted an extensive empirical study evaluating the performance of three leading closed-source LLMs and four popular open-source LLMs on three commonly used benchmarks. Our investigation, which evaluated the length, cyclomatic complexity and API number of the generated code, revealed that these LLMs face challenges in generating successful code for more complex problems, and tend to produce code that is shorter yet more complicated as compared to canonical solutions. Additionally, we developed a taxonomy of bugs for incorrect codes that includes three categories and 12 sub-categories, and analyze the root cause for common bug types. Furthermore, to better understand the performance of LLMs in real-world projects, we manually created a real-world benchmark comprising 140 code generation tasks. Our analysis highlights distinct differences in bug distributions between actual scenarios and existing benchmarks. Finally, we propose a novel training-free iterative method that introduces self-critique, enabling LLMs to critique and correct their generated code based on bug types and compiler feedback. Experimental results demonstrate that our approach can significantly mitigate bugs and increase the passing rate by 29.2% after two iterations, indicating substantial potential for LLMs to handle more complex problems.
Authors: Jin Peng Zhou, Christian K. Belardi, Ruihan Wu, Travis Zhang, Carla P. Gomes, Wen Sun, Kilian Q. Weinberger
Abstract: Large language models (LLMs) currently dominate the field of natural language processing (NLP), representing the state-of-the-art across a diverse array of tasks. Developing a model of this nature, from training to inference, requires making numerous decisions which define a combinatorial search problem. For example, selecting the optimal pre-trained LLM, prompt, or hyperparameters to attain the best performance for a task often requires evaluating multiple candidates on an entire test set. This exhaustive evaluation can be time-consuming and costly, as both inference and metric computation with LLMs are resource-intensive. In this paper, we address the challenge of identifying the best method within a limited budget for evaluating methods on test examples. By leveraging the well-studied multi-armed bandit framework, which sequentially selects the next method-example pair to evaluate, our approach, combining multi-armed bandit algorithms with low-rank factorization, significantly reduces the required resources. Experiments show that our algorithms can identify the top-performing method using only 5-15\% of the typically needed resources, resulting in an 85-95\% reduction in cost.
Authors: Antonia Karamolegkou, Phillip Rust, Yong Cao, Ruixiang Cui, Anders S{\o}gaard, Daniel Hershcovich
Abstract: Large vision-language models (VLMs) can assist visually impaired people by describing images from their daily lives. Current evaluation datasets may not reflect diverse cultural user backgrounds or the situational context of this use case. To address this problem, we create a survey to determine caption preferences and propose a culture-centric evaluation benchmark by filtering VizWiz, an existing dataset with images taken by people who are blind. We then evaluate several VLMs, investigating their reliability as visual assistants in a culturally diverse setting. While our results for state-of-the-art models are promising, we identify challenges such as hallucination and misalignment of automatic evaluation metrics with human judgment. We make our survey, data, code, and model outputs publicly available.
Authors: Xuweiyi Chen, Ziqiao Ma, Xuejun Zhang, Sihan Xu, Shengyi Qian, Jianing Yang, David F. Fouhey, Joyce Chai
Abstract: Large vision language models (LVLMs) often suffer from object hallucination, producing objects not present in the given images. While current benchmarks for object hallucination primarily concentrate on the presence of a single object class rather than individual entities, this work systematically investigates multi-object hallucination, examining how models misperceive (e.g., invent nonexistent objects or become distracted) when tasked with focusing on multiple objects simultaneously. We introduce Recognition-based Object Probing Evaluation (ROPE), an automated evaluation protocol that considers the distribution of object classes within a single image during testing and uses visual referring prompts to eliminate ambiguity. With comprehensive empirical studies and analysis of potential factors leading to multi-object hallucination, we found that (1) LVLMs suffer more hallucinations when focusing on multiple objects compared to a single object. (2) The tested object class distribution affects hallucination behaviors, indicating that LVLMs may follow shortcuts and spurious correlations.(3) Hallucinatory behaviors are influenced by data-specific factors, salience and frequency, and model intrinsic behaviors. We hope to enable LVLMs to recognize and reason about multiple objects that often occur in realistic visual scenes, provide insights, and quantify our progress towards mitigating the issues.
Authors: Dominik Schlechtweg, Nina Tahmasebi, Simon Hengchen, Haim Dubossarsky, Barbara McGillivray
Abstract: Word meaning is notoriously difficult to capture, both synchronically and diachronically. In this paper, we describe the creation of the largest resource of graded contextualized, diachronic word meaning annotation in four different languages, based on 100,000 human semantic proximity judgments. We thoroughly describe the multi-round incremental annotation process, the choice for a clustering algorithm to group usages into senses, and possible - diachronic and synchronic - uses for this dataset.
Authors: Jiaang Li, Yova Kementchedjhieva, Constanza Fierro, Anders S{\o}gaard
Abstract: Large-scale pretrained language models (LMs) are said to ``lack the ability to connect utterances to the world'' (Bender and Koller, 2020), because they do not have ``mental models of the world' '(Mitchell and Krakauer, 2023). If so, one would expect LM representations to be unrelated to representations induced by vision models. We present an empirical evaluation across four families of LMs (BERT, GPT-2, OPT and LLaMA-2) and three vision model architectures (ResNet, SegFormer, and MAE). Our experiments show that LMs partially converge towards representations isomorphic to those of vision models, subject to dispersion, polysemy and frequency. This has important implications for both multi-modal processing and the LM understanding debate (Mitchell and Krakauer, 2023).
Authors: Vittoria Dentella, Elliot Murphy, Gary Marcus, Evelina Leivada
Abstract: Large Language Models (LLMs) are recruited in applications that span from clinical assistance and legal support to question answering and education. Their success in specialized tasks has led to the claim that they possess human-like linguistic capabilities related to compositional understanding and reasoning. Yet, reverse-engineering is bound by Moravec's Paradox, according to which easy skills are hard. We systematically assess 7 state-of-the-art models on a novel benchmark. Models answered a series of comprehension questions, each prompted multiple times in two settings, permitting one-word or open-length replies. Each question targets a short text featuring high-frequency linguistic constructions. To establish a baseline for achieving human-like performance, we tested 400 humans on the same prompts. Based on a dataset of n=26,680 datapoints, we discovered that LLMs perform at chance accuracy and waver considerably in their answers. Quantitatively, the tested models are outperformed by humans, and qualitatively their answers showcase distinctly non-human errors in language understanding. We interpret this evidence as suggesting that, despite their usefulness in various tasks, current AI models fall short of understanding language in a way that matches humans, and we argue that this may be due to their lack of a compositional operator for regulating grammatical and semantic information.
Authors: Amir Hossein Kargaran, Nafiseh Nikeghbal, Abbas Heydarnoori, Hinrich Sch\"utze
Abstract: Menu system design for user interfaces is a challenging task involving many design options and various human factors. For example, one crucial factor that designers need to consider is the semantic and systematic relation of menu commands. However, capturing these relations can be challenging due to limited available resources. Large language models can be helpful in this regard, using their pre-training knowledge to design and refine menu systems. In this paper, we propose MenuCraft, an AI-assisted designer for menu design that enables collaboration between the designer and a dialogue system to design menus. MenuCraft offers an interactive language-based menu design tool that simplifies the menu design process and enables easy customization of design options. MenuCraft supports a variety of interactions through dialog that allows performing in-context learning.
Authors: Sunkyoung Kim, Dayeon Ki, Yireun Kim, Jinsik Lee
Abstract: Multilingual large language models (MLLMs) have demonstrated significant cross-lingual capabilities through in-context learning. Existing approaches typically construct monolingual few-shot examples, either in the source or target language. However, translating entire in-context examples into the target language might compromise contextual integrity and be costly in the case of long-context passages. To address this, we introduce Cross-lingual QA, a cross-lingual prompting method that translates only the question and answer parts, thus reducing translation costs. Experiments on four typologically diverse multilingual benchmarks show that Cross-lingual QA prompting effectively stimulates models to elicit their cross-lingual knowledge, outperforming prior monolingual few-shot prompting approaches. Furthermore, we show that prompting open-source MLLMs with cross-lingual few-shot examples enhances performance as the model scale increases.
Authors: Zhengping Zhou, Lezhi Li, Xinxi Chen, Andy Li
Abstract: ChatGPT is phenomenal. However, it is prohibitively expensive to train and refine such giant models. Fortunately, small language models are flourishing and becoming more and more competent. We call them "mini-giants". We argue that open source community like Kaggle and mini-giants will win-win in many ways, technically, ethically and socially. In this article, we present a brief yet rich background, discuss how to attain small language models, present a comparative study of small language models and a brief discussion of evaluation methods, discuss the application scenarios where small language models are most needed in the real world, and conclude with discussion and outlook.
Authors: Johannes Schneider, Bernd Schenk, Christina Niklaus
Abstract: Grading exams is an important, labor-intensive, subjective, repetitive, and frequently challenging task. The feasibility of autograding textual responses has greatly increased thanks to the availability of large language models (LLMs) such as ChatGPT and the substantial influx of data brought about by digitalization. However, entrusting AI models with decision-making roles raises ethical considerations, mainly stemming from potential biases and issues related to generating false information. Thus, in this manuscript, we provide an evaluation of a large language model for the purpose of autograding, while also highlighting how LLMs can support educators in validating their grading procedures. Our evaluation is targeted towards automatic short textual answers grading (ASAG), spanning various languages and examinations from two distinct courses. Our findings suggest that while "out-of-the-box" LLMs provide a valuable tool to provide a complementary perspective, their readiness for independent automated grading remains a work in progress, necessitating human oversight.
Authors: Aniruddha Deb, Neeva Oza, Sarthak Singla, Dinesh Khandelwal, Dinesh Garg, Parag Singla
Abstract: While forward reasoning (i.e., find the answer given the question) has been explored extensively in recent literature, backward reasoning is relatively unexplored. We examine the backward reasoning capabilities of LLMs on Math Word Problems (MWPs): given a mathematical question and its answer, with some details omitted from the question, can LLMs effectively retrieve the missing information? On modifying three benchmark datasets for this task, to evaluate this task: GSM8k, SVAMP, and MultiArith, we find a significant drop in the accuracy of models on this task compared to forward reasoning across SOTA LLMs (GPT4, GPT3.5, PaLM-2, and LLaMa). Motivated by the fact backward reasoning can be seen as the ''inverse'' of forward reasoning, we propose variations of three different forward reasoning strategies to improve performance. Rephrase reformulates the given problem into a forward reasoning problem, PAL-Tools combines the idea of Program-Aided LLMs to produce a set of equations that can be solved by an external solver, and Check your Work exploits the availability of natural verifier of high accuracy in the forward direction, interleaving solving and verification steps. Finally, realizing that each of our base methods correctly solves a different set of problems, we propose a novel Bayesian formulation for creating an ensemble over the base methods to further boost the accuracy. Extensive experimentation demonstrates successive improvement in the performance of LLMs on the backward reasoning task, using our strategies, with our ensemble-based method resulting in significant performance gains compared to the SOTA forward reasoning strategies we adapt.
Authors: Michael Lan, Phillip Torr, Fazl Barez
Abstract: While transformer models exhibit strong capabilities on linguistic tasks, their complex architectures make them difficult to interpret. Recent work has aimed to reverse engineer transformer models into human-readable representations called circuits that implement algorithmic functions. We extend this research by analyzing and comparing circuits for similar sequence continuation tasks, which include increasing sequences of Arabic numerals, number words, and months. By applying circuit interpretability analysis, we identify a key sub-circuit in both GPT-2 Small and Llama-2-7B responsible for detecting sequence members and for predicting the next member in a sequence. Our analysis reveals that semantically related sequences rely on shared circuit subgraphs with analogous roles. Additionally, we show that this sub-circuit has effects on various math-related prompts, such as on intervaled circuits, Spanish number word and months continuation, and natural language word problems. Overall, documenting shared computational structures enables better model behavior predictions, identification of errors, and safer editing procedures. This mechanistic understanding of transformers is a critical step towards building more robust, aligned, and interpretable language models.
Authors: Gantavya Bhatt, Yifang Chen, Arnav M. Das, Jifan Zhang, Sang T. Truong, Stephen Mussmann, Yinglun Zhu, Jeffrey Bilmes, Simon S. Du, Kevin Jamieson, Jordan T. Ash, Robert D. Nowak
Abstract: Supervised finetuning (SFT) on instruction datasets has played a crucial role in achieving the remarkable zero-shot generalization capabilities observed in modern large language models (LLMs). However, the annotation efforts required to produce high quality responses for instructions are becoming prohibitively expensive, especially as the number of tasks spanned by instruction datasets continues to increase. Active learning is effective in identifying useful subsets of samples to annotate from an unlabeled pool, but its high computational cost remains a barrier to its widespread applicability in the context of LLMs. To mitigate the annotation cost of SFT and circumvent the computational bottlenecks of active learning, we propose using experimental design. Experimental design techniques select the most informative samples to label, and typically maximize some notion of uncertainty and/or diversity. In our work, we implement a framework that evaluates several existing and novel experimental design techniques and find that these methods consistently yield significant gains in label efficiency with little computational overhead. On generative tasks, our methods achieve the same generalization performance with only $50\%$ of annotation cost required by random sampling.
Authors: Zabir Al Nazi, Wei Peng
Abstract: The deployment of large language models (LLMs) within the healthcare sector has sparked both enthusiasm and apprehension. These models exhibit the remarkable capability to provide proficient responses to free-text queries, demonstrating a nuanced understanding of professional medical knowledge. This comprehensive survey delves into the functionalities of existing LLMs designed for healthcare applications, elucidating the trajectory of their development, starting from traditional Pretrained Language Models (PLMs) to the present state of LLMs in healthcare sector. First, we explore the potential of LLMs to amplify the efficiency and effectiveness of diverse healthcare applications, particularly focusing on clinical language understanding tasks. These tasks encompass a wide spectrum, ranging from named entity recognition and relation extraction to natural language inference, multi-modal medical applications, document classification, and question-answering. Additionally, we conduct an extensive comparison of the most recent state-of-the-art LLMs in the healthcare domain, while also assessing the utilization of various open-source LLMs and highlighting their significance in healthcare applications. Furthermore, we present the essential performance metrics employed to evaluate LLMs in the biomedical domain, shedding light on their effectiveness and limitations. Finally, we summarize the prominent challenges and constraints faced by large language models in the healthcare sector, offering a holistic perspective on their potential benefits and shortcomings. This review provides a comprehensive exploration of the current landscape of LLMs in healthcare, addressing their role in transforming medical applications and the areas that warrant further research and development.
Authors: Shengchao Liu, Xiaoming Liu, Yichen Wang, Zehua Cheng, Chengzhengxu Li, Zhaohan Zhang, Yu Lan, Chao Shen
Abstract: The burgeoning generative capabilities of large language models (LLMs) have raised growing concerns about abuse, demanding automatic machine-generated text detectors. DetectGPT, a zero-shot metric-based detector, first introduces perturbation and shows great performance improvement. However, in DetectGPT, the random perturbation strategy could introduce noise, and logit regression depends on the threshold, harming the generalizability and applicability of individual or small-batch inputs. Hence, we propose a novel fine-tuned detector, Pecola, bridging metric-based and fine-tuned methods by contrastive learning on selective perturbation. Selective strategy retains important tokens during perturbation and weights for multi-pair contrastive learning. The experiments show that Pecola outperforms the state-of-the-art (SOTA) by 1.20% in accuracy on average on four public datasets. And we further analyze the effectiveness, robustness, and generalization of the method.
Authors: Xinyu Li, Zachary C. Lipton, Liu Leqi
Abstract: Reinforcement Learning from Human Feedback (RLHF) is commonly used to fine-tune large language models to better align with human preferences. However, the underlying premise of algorithms developed under this framework can be problematic when user preferences encoded in human feedback are diverse. In this work, we aim to address this problem by developing methods for building personalized language models. We first formally introduce the task of learning from personalized human feedback and explain why vanilla RLHF can be ineffective in this context. We then propose a general Personalized-RLHF (P-RLHF) framework, including a user model that maps user information to user representations and can flexibly encode our assumptions on user preferences. We develop new learning objectives to perform personalized Direct Preference Optimization that jointly learns a user model and a personalized language model. We demonstrate the efficacy of our proposed method through (1) a synthetic task where we fine-tune a GPT-J 6B model to align with users with conflicting preferences on generation length; and (2) an instruction following task where we fine-tune a Tulu-7B model to generate responses for users with diverse preferences on the style of responses. In both cases, our learned models can generate personalized responses that are better aligned with the preferences of individual users.
Authors: Yoonjeong Park, Hyunjin Kim, Chanyeol Choi, Junseong Kim, Jy-yong Sohn
Abstract: Chain-of-thought (CoT) prompting is a simple and effective method for improving the reasoning capabilities of Large Language Models (LLMs). The basic idea of CoT is to let LLMs break down their thought processes step-by-step by putting exemplars in the input prompt. However, the densely structured prompt exemplars of CoT may cause the cognitive overload of LLMs. Inspired by human cognition, we introduce COT-SEP, a method that strategically employs separators at the end of each exemplar in CoT prompting. These separators are designed to help the LLMs understand their thought processes better while reasoning. Interestingly, it turns out that COT-SEP significantly improves the LLMs' performances on complex reasoning tasks (e.g., GSM8K, AQuA, CSQA), compared with the vanilla CoT, which does not use separators. We also study the effects of the type and the location of separators tested on multiple LLMs, including GPT-3.5-Turbo, GPT-4, and LLaMA-2 7B.
Authors: Zongxia Li, Ishani Mondal, Yijun Liang, Huy Nghiem, Jordan Lee Boyd-Graber
Abstract: Question answering (QA) can only make progress if we know if an answer is correct, but for many of the most challenging and interesting QA examples, current efficient answer correctness (AC) metrics do not align with human judgments, particularly verbose, free-form answers from large language models (LLMs). There are two challenges: a lack of diverse evaluation data and that models are too big and non-transparent; LLM-based scorers correlate better with humans, but this expensive task has only been tested on limited QA datasets. We rectify these issues by providing guidelines and datasets for evaluating machine QA adopted from human QA community. We also propose an efficient, low-resource, and interpretable QA evaluation method more stable than an exact match and neural methods.
Authors: Badr AlKhamissi, Muhammad ElNokrashy, Mai AlKhamissi, Mona Diab
Abstract: The intricate relationship between language and culture has long been a subject of exploration within the realm of linguistic anthropology. Large Language Models (LLMs), promoted as repositories of collective human knowledge, raise a pivotal question: do these models genuinely encapsulate the diverse knowledge adopted by different cultures? Our study reveals that these models demonstrate greater cultural alignment along two dimensions -- firstly, when prompted with the dominant language of a specific culture, and secondly, when pretrained with a refined mixture of languages employed by that culture. We quantify cultural alignment by simulating sociological surveys, comparing model responses to those of actual survey participants as references. Specifically, we replicate a survey conducted in various regions of Egypt and the United States through prompting LLMs with different pretraining data mixtures in both Arabic and English with the personas of the real respondents and the survey questions. Further analysis reveals that misalignment becomes more pronounced for underrepresented personas and for culturally sensitive topics, such as those probing social values. Finally, we introduce Anthropological Prompting, a novel method leveraging anthropological reasoning to enhance cultural alignment. Our study emphasizes the necessity for a more balanced multilingual pretraining dataset to better represent the diversity of human experience and the plurality of different cultures with many implications on the topic of cross-lingual transfer.
Authors: Chen Xu, Tian Lan, Changlong Yu, Wei Wang, Jun Gao, Yu Ji, Qunxi Dong, Kun Qian, Piji Li, Wei Bi, Bin Hu
Abstract: Constrained decoding approaches aim to control the meaning or style of text generated by a Pre-trained Language Model (PLM) using specific target words during inference. However, these methods often guide plausible continuations by greedily selecting targets, which, while completing the task, may disrupt the natural patterns of human language generation. In this work, we propose a novel decoding framework, DECIDER, which enables us to program rules on how we complete tasks to control a PLM. Differing from previous work, our framework transforms the encouragement of target words into the encouragement of all words that satisfy the rule. Specifically, DECIDER is a dual system where a PLM is equipped with a First-OrderLogic (FOL) reasoner to express and evaluate the rules, and a decision function to merge the outputs from both systems to steer the generation. Experiments on CommonGen and PersonaChat demonstrate that DECIDER can effectively follow given rules to achieve generation tasks in a more human-like manner.
Authors: Gregor Bachmann, Vaishnavh Nagarajan
Abstract: Can a mere next-token predictor faithfully model human intelligence? We crystallize this emerging concern and correct popular misconceptions surrounding it, and advocate a simple multi-token objective. As a starting point, we argue that the two often-conflated phases of next-token prediction -- autoregressive inference and teacher-forced training -- must be treated distinctly. The popular criticism that errors can compound during autoregressive inference, crucially assumes that teacher-forcing has learned an accurate next-token predictor. This assumption sidesteps a more deep-rooted problem we expose: in certain classes of tasks, teacher-forcing can simply fail to learn an accurate next-token predictor in the first place. We describe a general mechanism of how teacher-forcing can fail, and design a minimal planning task where both the Transformer and the Mamba architecture empirically fail in that manner -- remarkably, despite the task being straightforward to learn. Finally, we provide preliminary evidence that this failure can be resolved using a simple modification that predicts multiple tokens in advance. We hope this finding can ground future debates and inspire explorations beyond the next-token prediction paradigm. We make our code available under https://github.com/gregorbachmann/Next-Token-Failures
Authors: H S V N S Kowndinya Renduchintala, Sumit Bhatia, Ganesh Ramakrishnan
Abstract: Instruction Tuning involves finetuning a language model on a collection of instruction-formatted datasets in order to enhance the generalizability of the model to unseen tasks. Studies have shown the importance of balancing different task proportions during finetuning, but finding the right balance remains challenging. Unfortunately, there's currently no systematic method beyond manual tuning or relying on practitioners' intuition. In this paper, we introduce SMART (Submodular data Mixture strAtegy for instRuction Tuning) - a novel data mixture strategy which makes use of a submodular function to assign importance scores to tasks which are then used to determine the mixture weights. Given a fine-tuning budget, SMART redistributes the budget among tasks and selects non-redundant samples from each task. Experimental results demonstrate that SMART significantly outperforms traditional methods such as examples proportional mixing and equal mixing. Furthermore, SMART facilitates the creation of data mixtures based on a few representative subsets of tasks alone and through task pruning analysis, we reveal that in a limited budget setting, allocating budget among a subset of representative tasks yields superior performance compared to distributing the budget among all tasks. The code for reproducing our results is open-sourced at https://github.com/kowndinya-renduchintala/SMART.
Authors: Fahim Faisal, Orevaoghene Ahia, Aarohi Srivastava, Kabir Ahuja, David Chiang, Yulia Tsvetkov, Antonios Anastasopoulos
Abstract: Language technologies should be judged on their usefulness in real-world use cases. An often overlooked aspect in natural language processing (NLP) research and evaluation is language variation in the form of non-standard dialects or language varieties (hereafter, varieties). Most NLP benchmarks are limited to standard language varieties. To fill this gap, we propose DIALECTBENCH, the first-ever large-scale benchmark for NLP on varieties, which aggregates an extensive set of task-varied variety datasets (10 text-level tasks covering 281 varieties). This allows for a comprehensive evaluation of NLP system performance on different language varieties. We provide substantial evidence of performance disparities between standard and non-standard language varieties, and we also identify language clusters with large performance divergence across tasks. We believe DIALECTBENCH provides a comprehensive view of the current state of NLP for language varieties and one step towards advancing it further. Code/data: https://github.com/ffaisal93/DialectBench
Authors: Cheng Jiayang, Lin Qiu, Chunkit Chan, Xin Liu, Yangqiu Song, Zheng Zhang
Abstract: Narrative reasoning relies on the understanding of eventualities in story contexts, which requires a wealth of background world knowledge. To help machines leverage such knowledge, existing solutions can be categorized into two groups. Some focus on implicitly modeling eventuality knowledge by pretraining language models (LMs) with eventuality-aware objectives. However, this approach breaks down knowledge structures and lacks interpretability. Others explicitly collect world knowledge of eventualities into structured eventuality-centric knowledge graphs (KGs). However, existing research on leveraging these knowledge sources for free-texts is limited. In this work, we propose an initial comprehensive framework called EventGround, which aims to tackle the problem of grounding free-texts to eventuality-centric KGs for contextualized narrative reasoning. We identify two critical problems in this direction: the event representation and sparsity problems. We provide simple yet effective parsing and partial information extraction methods to tackle these problems. Experimental results demonstrate that our approach consistently outperforms baseline models when combined with graph neural network (GNN) or large language model (LLM) based graph reasoning models. Our framework, incorporating grounded knowledge, achieves state-of-the-art performance while providing interpretable evidence.
Authors: Samuel Cahyawijaya, Holy Lovenia, Fajri Koto, Rifki Afina Putri, Emmanuel Dave, Jhonson Lee, Nuur Shadieq, Wawan Cenggoro, Salsabil Maulana Akbar, Muhammad Ihza Mahendra, Dea Annisayanti Putri, Bryan Wilie, Genta Indra Winata, Alham Fikri Aji, Ayu Purwarianti, Pascale Fung
Abstract: Large language models (LLMs) show remarkable human-like capability in various domains and languages. However, a notable quality gap arises in low-resource languages, e.g., Indonesian indigenous languages, rendering them ineffective and inefficient in such linguistic contexts. To bridge this quality gap, we introduce Cendol, a collection of Indonesian LLMs encompassing both decoder-only and encoder-decoder architectures across a range of model sizes. We highlight Cendol's effectiveness across a diverse array of tasks, attaining 20% improvement, and demonstrate its capability to generalize to unseen tasks and indigenous languages of Indonesia. Furthermore, Cendol models showcase improved human favorability despite their limitations in capturing indigenous knowledge and cultural values in Indonesia. In addition, we discuss the shortcomings of parameter-efficient tunings, such as LoRA, for language adaptation. Alternatively, we propose the usage of vocabulary adaptation to enhance efficiency. Lastly, we evaluate the safety of Cendol and showcase that safety in pre-training in one language such as English is transferable to low-resource languages, such as Indonesian, even without RLHF and safety fine-tuning.
Authors: Xiaoshu Chen, Sihang Zhou, Ke Liang, Xinwang Liu
Abstract: Chain of thought finetuning (cot-finetuning) aims to endow small language models (SLM) with reasoning ability to improve their performance towards specific tasks by allowing them to imitate the reasoning procedure of large language models (LLM) beyond simply predicting the answers. Most existing cot-finetuning methods adopt a pre-thinking mechanism, allowing the SLM to generate a rationale before providing an answer. This mechanism enables SLM to analyze and think about complex questions, but it also makes answer correctness highly sensitive to minor errors in rationale. Therefore, we propose a robust post-thinking mechanism to generate answers before rationale. Thanks to this answer-first setting, 1) the answer can escape from the adverse effects caused by minor errors in the rationale; 2) the rationale serves as an error amplifier to the answer, which makes the SLM focus on learning hard samples; 3) the inferring efficiency can also benefit from the setting since users can stop the generation right after answers are outputted when inference is conducted. However, although the post-thinking mechanism brings many advantages and improves the overall performance of SLM on specific tasks, it may lose the ability to think about the questions and decompose complex questions into simple sub-questions compared to pre-thinking mechanism. Therefore, a plug-and-play adaptive-thinking mechanism is proposed with the aid of the soft prompt tuning to integrate the merits of the pre-thinking mechanism and post-thinking mechanism, in which a perception module is introduced to adaptively prompt SLM answer or think first based on perceiving the complexity of the questions. Extensive experiments are conducted across 12 reasoning tasks and 2 representative language models to demonstrate the effectiveness of the proposed mechanism.
Authors: Zhili Liu, Yunhao Gou, Kai Chen, Lanqing Hong, Jiahui Gao, Fei Mi, Yu Zhang, Zhenguo Li, Xin Jiang, Qun Liu, James T. Kwok
Abstract: As the capabilities of large language models (LLMs) have expanded dramatically, aligning these models with human values presents a significant challenge. Traditional alignment strategies rely heavily on human intervention, such as Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF), or on the self-alignment capacities of LLMs, which usually require a strong LLM's emergent ability to improve its original bad answer. To address these challenges, we propose a novel self-alignment method that utilizes a Chain of Thought (CoT) approach, termed AlignCoT. This method encompasses stages of Question Analysis, Answer Guidance, and Safe Answer production. It is designed to enable LLMs to generate high-quality, safe responses throughout various stages of their development. Furthermore, we introduce the Mixture of insighTful Experts (MoTE) architecture, which applies mixture of experts to enhance each component of the AlignCoT process, markedly increasing alignment efficiency. The MoTE approach not only outperforms existing methods in aligning LLMs with human values but also highlights the benefits of using self-generated data, revealing the dual benefits of improved alignment and training efficiency.
Authors: Christian Arnold, Andreas K\"upfer
Abstract: When studying political communication, combining the information from text, audio, and video signals promises to reflect the richness of human communication more comprehensively than confining it to individual modalities alone. However, its heterogeneity, connectedness, and interaction are challenging to address when modeling such multimodal data. We argue that aligning the respective modalities can be an essential step in entirely using the potential of multimodal data because it informs the model with human understanding. Taking care of the data-generating process of multimodal data, our framework proposes four principles to organize alignment and, thus, address the challenges of multimodal data. We illustrate the utility of these principles by analyzing how German MPs address members of the far-right AfD in their speeches and predicting the tone of video advertising in the context of the 2020 US presidential race. Our paper offers important insights to all keen to analyze multimodal data effectively.
Authors: Abhishek Divekar, Greg Durrett
Abstract: It is often desirable to distill the capabilities of large language models (LLMs) into smaller student models due to compute and memory constraints. One way to do this for classification tasks is via dataset synthesis, which can be accomplished by generating examples of each label from the LLM. Prior approaches to synthesis use few-shot prompting, which relies on the LLM's parametric knowledge to generate usable examples. However, this leads to issues of repetition, bias towards popular entities, and stylistic differences from human text. In this work, we propose Synthesize by Retrieval and Refinement (SynthesizRR), which uses retrieval augmentation to introduce variety into the dataset synthesis process: as retrieved passages vary, the LLM is seeded with different content to generate its examples. We empirically study the synthesis of six datasets, covering topic classification, sentiment analysis, tone detection, and humor, requiring complex synthesis strategies. We find that SynthesizRR greatly improves lexical and semantic diversity, similarity to human-written text, and distillation performance, when compared to 32-shot prompting and four prior approaches. We release our extensive codebase at https://github.com/amazon-science/synthesizrr
Authors: Mireia Hernandez Caralt, Clarence Boon Liang Ng, Marek Rei
Abstract: Electronic Health Records (EHR) serve as a valuable source of patient information, offering insights into medical histories, treatments, and outcomes. Previous research has developed systems for detecting applicable ICD codes that should be assigned while writing a given EHR document, mainly focusing on discharge summaries written at the end of a hospital stay. In this work, we investigate the potential of predicting these codes for the whole patient stay at different time points during their stay, even before they are officially assigned by clinicians. The development of methods to predict diagnoses and treatments earlier in advance could open opportunities for predictive medicine, such as identifying disease risks sooner, suggesting treatments, and optimizing resource allocation. Our experiments show that predictions regarding final ICD codes can be made already two days after admission and we propose a custom model that improves performance on this early prediction task.
Authors: Xingyuan Pan, Luyang Huang, Liyan Kang, Zhicheng Liu, Yu Lu, Shanbo Cheng
Abstract: Large Language Models (LLMs) have demonstrated remarkable abilities in general scenarios. Instruction finetuning empowers them to align with humans in various tasks. Nevertheless, the Diversity and Quality of the instruction data remain two main challenges for instruction finetuning. With regard to this, in this paper, we propose a novel gradient-based method to automatically select high-quality and diverse instruction finetuning data for machine translation. Our key innovation centers around analyzing how individual training examples influence the model during training. Specifically, we select training examples that exert beneficial influences on the model as high-quality ones by means of Influence Function plus a small high-quality seed dataset. Moreover, to enhance the diversity of the training data we maximize the variety of influences they have on the model by clustering on their gradients and resampling. Extensive experiments on WMT22 and FLORES translation tasks demonstrate the superiority of our methods, and in-depth analysis further validates their effectiveness and generalization.
Authors: Fanfan Wang, Heqing Ma, Jianfei Yu, Rui Xia, Erik Cambria
Abstract: The ability to understand emotions is an essential component of human-like artificial intelligence, as emotions greatly influence human cognition, decision making, and social interactions. In addition to emotion recognition in conversations, the task of identifying the potential causes behind an individual's emotional state in conversations, is of great importance in many application scenarios. We organize SemEval-2024 Task 3, named Multimodal Emotion Cause Analysis in Conversations, which aims at extracting all pairs of emotions and their corresponding causes from conversations. Under different modality settings, it consists of two subtasks: Textual Emotion-Cause Pair Extraction in Conversations (TECPE) and Multimodal Emotion-Cause Pair Extraction in Conversations (MECPE). The shared task has attracted 143 registrations and 216 successful submissions. In this paper, we introduce the task, dataset and evaluation settings, summarize the systems of the top teams, and discuss the findings of the participants.
Authors: Yu Meng, Mengzhou Xia, Danqi Chen
Abstract: Direct Preference Optimization (DPO) is a widely used offline preference optimization algorithm that reparameterizes reward functions in reinforcement learning from human feedback (RLHF) to enhance simplicity and training stability. In this work, we propose SimPO, a simpler yet more effective approach. The effectiveness of SimPO is attributed to a key design: using the average log probability of a sequence as the implicit reward. This reward formulation better aligns with model generation and eliminates the need for a reference model, making it more compute and memory efficient. Additionally, we introduce a target reward margin to the Bradley-Terry objective to encourage a larger margin between the winning and losing responses, further enhancing the algorithm's performance. We compare SimPO to DPO and its latest variants across various state-of-the-art training setups, including both base and instruction-tuned models like Mistral and Llama3. We evaluated on extensive instruction-following benchmarks, including AlpacaEval 2, MT-Bench, and the recent challenging Arena-Hard benchmark. Our results demonstrate that SimPO consistently and significantly outperforms existing approaches without substantially increasing response length. Specifically, SimPO outperforms DPO by up to 6.4 points on AlpacaEval 2 and by up to 7.5 points on Arena-Hard. Our top-performing model, built on Llama3-8B-Instruct, achieves a remarkable 53.7 length-controlled win rate on AlpacaEval 2 -- surpassing Claude 3 Opus on the leaderboard, and a 36.5 win rate on Arena-Hard -- making it the strongest 8B open-source model.
Authors: Zixuan Dong, Baoyun Peng, Yufei Wang, Jia Fu, Xiaodong Wang, Yongxue Shan, Xin Zhou
Abstract: While large language models (LLMs) have shown remarkable capabilities in natural language processing, they struggle with complex, multi-step reasoning tasks involving knowledge graphs (KGs). Existing approaches that integrate LLMs and KGs either underutilize the reasoning abilities of LLMs or suffer from prohibitive computational costs due to tight coupling. To address these limitations, we propose a novel collaborative framework named EffiQA that can strike a balance between performance and efficiency via an iterative paradigm. EffiQA consists of three stages: global planning, efficient KG exploration, and self-reflection. Specifically, EffiQA leverages the commonsense capability of LLMs to explore potential reasoning pathways through global planning. Then, it offloads semantic pruning to a small plug-in model for efficient KG exploration. Finally, the exploration results are fed to LLMs for self-reflection to further improve the global planning and efficient KG exploration. Empirical evidence on multiple KBQA benchmarks shows EffiQA's effectiveness, achieving an optimal balance between reasoning accuracy and computational costs. We hope the proposed new framework will pave the way for efficient, knowledge-intensive querying by redefining the integration of LLMs and KGs, fostering future research on knowledge-based question answering.
Authors: Zi-Yi Dou, Cheng-Fu Yang, Xueqing Wu, Kai-Wei Chang, Nanyun Peng
Abstract: Finetuning language agents with reasoning-action trajectories is effective, but obtaining these trajectories from human annotations or stronger models is costly and sometimes impractical. In this paper, we investigate the use of self-training in language agents, which can generate supervision from the agent itself, offering a promising alternative without relying on human or stronger model demonstrations. Self-training, however, requires high-quality model-generated samples, which are hard to obtain for challenging language agent tasks. To address this, we present Reflection-Reinforced Self-Training (Re-ReST), which uses a \textit{reflector} to refine low-quality generated samples during self-training. The reflector takes the agent's output and feedback from an external environment (e.g., unit test results in code generation) to produce improved samples. This technique enhances the quality of inferior samples and efficiently enriches the self-training dataset with higher-quality samples. We conduct extensive experiments on open-source language agents across tasks, including multi-hop question answering, sequential decision-making, code generation, visual question answering, and text-to-image generation. The results demonstrate the effectiveness of self-training and Re-ReST in language agent tasks, with self-training improving baselines by 7.6\% on HotpotQA and 28.4\% on AlfWorld, and Re-ReST further boosting performance by 2.0\% and 14.1\%, respectively. Our studies also confirm the efficiency of using a reflector to generate high-quality samples for self-training. Moreover, we demonstrate a method to employ reflection during inference without ground-truth feedback, addressing the limitation of previous reflection work. Our code is released at https://github.com/PlusLabNLP/Re-ReST.
Authors: Jinghan Zhang, Xiting Wang, Yiqiao Jin, Changyu Chen, Xinhao Zhang, Kunpeng Liu
Abstract: The reward model for Reinforcement Learning from Human Feedback (RLHF) has proven effective in fine-tuning Large Language Models (LLMs). Notably, collecting human feedback for RLHF can be resource-intensive and lead to scalability issues for LLMs and complex tasks. Our proposed framework Proto-RM leverages prototypical networks to enhance reward models under limited human feedback. By enabling stable and reliable structural learning from fewer samples, Proto-RM significantly enhances LLMs' adaptability and accuracy in interpreting human preferences. Extensive experiments on various datasets demonstrate that Proto-RM significantly improves the performance of reward models and LLMs in human feedback tasks, achieving comparable and usually better results than traditional methods, while requiring significantly less data. in data-limited scenarios. This research offers a promising direction for enhancing the efficiency of reward models and optimizing the fine-tuning of language models under restricted feedback conditions.
Authors: Holy Lovenia, Rahmad Mahendra, Salsabil Maulana Akbar, Lester James V. Miranda, Jennifer Santoso, Elyanah Aco, Akhdan Fadhilah, Jonibek Mansurov, Joseph Marvin Imperial, Onno P. Kampman, Joel Ruben Antony Moniz, Muhammad Ravi Shulthan Habibi, Frederikus Hudi, Railey Montalan, Ryan Ignatius, Joanito Agili Lopo, William Nixon, B\"orje F. Karlsson, James Jaya, Ryandito Diandaru, Yuze Gao, Patrick Amadeus, Bin Wang, Jan Christian Blaise Cruz, Chenxi Whitehouse, Ivan Halim Parmonangan, Maria Khelli, Wenyu Zhang, Lucky Susanto, Reynard Adha Ryanda, Sonny Lazuardi Hermawan, Dan John Velasco, Muhammad Dehan Al Kautsar, Willy Fitra Hendria, Yasmin Moslem, Noah Flynn, Muhammad Farid Adilazuarda, Haochen Li, Johanes Lee, R. Damanhuri, Shuo Sun, Muhammad Reza Qorib, Amirbek Djanibekov, Wei Qi Leong, Quyet V. Do, Niklas Muennighoff, Tanrada Pansuwan, Ilham Firdausi Putra, Yan Xu, Ngee Chia Tai, Ayu Purwarianti, Sebastian Ruder, William Tjhi, Peerat Limkonchotiwat, Alham Fikri Aji, Sedrick Keh, Genta Indra Winata, Ruochen Zhang, Fajri Koto, Zheng-Xin Yong, Samuel Cahyawijaya
Abstract: Southeast Asia (SEA) is a region rich in linguistic diversity and cultural variety, with over 1,300 indigenous languages and a population of 671 million people. However, prevailing AI models suffer from a significant lack of representation of texts, images, and audio datasets from SEA, compromising the quality of AI models for SEA languages. Evaluating models for SEA languages is challenging due to the scarcity of high-quality datasets, compounded by the dominance of English training data, raising concerns about potential cultural misrepresentation. To address these challenges, we introduce SEACrowd, a collaborative initiative that consolidates a comprehensive resource hub that fills the resource gap by providing standardized corpora in nearly 1,000 SEA languages across three modalities. Through our SEACrowd benchmarks, we assess the quality of AI models on 36 indigenous languages across 13 tasks, offering valuable insights into the current AI landscape in SEA. Furthermore, we propose strategies to facilitate greater AI advancements, maximizing potential utility and resource equity for the future of AI in SEA.
Authors: Mert Yazan, Suzan Verberne, Frederik Situmeang
Abstract: Post-training quantization reduces the computational demand of Large Language Models (LLMs) but can weaken some of their capabilities. Since LLM abilities emerge with scale, smaller LLMs are more sensitive to quantization. In this paper, we explore how quantization affects smaller LLMs' ability to perform retrieval-augmented generation (RAG), specifically in longer contexts. We chose personalization for evaluation because it is a challenging domain to perform using RAG as it requires long-context reasoning over multiple documents. We compare the original FP16 and the quantized INT4 performance of multiple 7B and 8B LLMs on two tasks while progressively increasing the number of retrieved documents to test how quantized models fare against longer contexts. To better understand the effect of retrieval, we evaluate three retrieval models in our experiments. Our findings reveal that if a 7B LLM performs the task well, quantization does not impair its performance and long-context reasoning capabilities. We conclude that it is possible to utilize RAG with quantized smaller LLMs.
Authors: Adane Nega Tarekegn
Abstract: The news landscape is continuously evolving, with an ever-increasing volume of information from around the world. Automated event detection within this vast data repository is essential for monitoring, identifying, and categorizing significant news occurrences across diverse platforms. This paper presents an event detection framework that leverages Large Language Models (LLMs) combined with clustering analysis to detect news events from the Global Database of Events, Language, and Tone (GDELT). The framework enhances event clustering through both pre-event detection tasks (keyword extraction and text embedding) and post-event detection tasks (event summarization and topic labelling). We also evaluate the impact of various textual embeddings on the quality of clustering outcomes, ensuring robust news categorization. Additionally, we introduce a novel Cluster Stability Assessment Index (CSAI) to assess the validity and robustness of clustering results. CSAI utilizes multiple feature vectors to provide a new way of measuring clustering quality. Our experiments indicate that the use of LLM embedding in the event detection framework has significantly improved the results, demonstrating greater robustness in terms of CSAI scores. Moreover, post-event detection tasks generate meaningful insights, facilitating effective interpretation of event clustering results. Overall, our experimental results indicate that the proposed framework offers valuable insights and could enhance the accuracy in news analysis and reporting.
Authors: Zhuang Li, Yuncheng Hua, Thuy-Trang Vu, Haolan Zhan, Lizhen Qu, Gholamreza Haffari
Abstract: Recent studies have shown that maintaining a consistent response style by human experts and enhancing data quality in training sets can significantly improve the performance of fine-tuned Large Language Models (LLMs) while reducing the number of training examples needed. However, the precise definition of style and the relationship between style, data quality, and LLM performance remains unclear. This research decomposes response style into presentation and composition styles and finds that, among training data of similar quality, those with higher style consistency lead to better LLM performance. Inspired by this, we introduce Style Consistency-Aware Response Ranking (SCAR), which automatically prioritizes instruction-response pairs in the training set based on their response stylistic consistency. By selecting the most style-consistent examples, ranging from the top 25% to 0.7% of the full dataset, the fine-tuned LLMs can match or even surpass the performance of models trained on the entire dataset in coding and open-ended question-answering benchmarks. Code and data are available at https://github.com/zhuang-li/SCAR .
Authors: Bofei Gao, Zefan Cai, Runxin Xu, Peiyi Wang, Ce Zheng, Runji Lin, Keming Lu, Dayiheng Liu, Chang Zhou, Wen Xiao, Junjie Hu, Tianyu Liu, Baobao Chang
Abstract: Mathematical verfier achieves success in mathematical reasoning tasks by validating the correctness of solutions. However, existing verifiers are trained with binary classification labels, which are not informative enough for the model to accurately assess the solutions. To mitigate the aforementioned insufficiency of binary labels, we introduce step-wise natural language feedbacks as rationale labels (i.e., the correctness of the current step and the explanations). In this paper, we propose \textbf{Math-Minos}, a natural language feedback enhanced verifier by constructing automatically-generated training data and a two-stage training paradigm for effective training and efficient inference. Our experiments reveal that a small set (30k) of natural language feedbacks can significantly boost the performance of the verifier by the accuracy of 1.6\% (86.6\% $\rightarrow$ 88.2\%) on GSM8K and 0.8\% (37.8\% $\rightarrow$ 38.6\%) on MATH. We have released our code and data for further exploration.
Authors: Chia-Yu Hung, Navonil Majumder, Ambuj Mehrish, Soujanya Poria
Abstract: The widespread applicability and increasing omnipresence of LLMs have instigated a need to align LLM responses to user and stakeholder preferences. Many preference optimization approaches have been proposed that fine-tune LLM parameters to achieve good alignment. However, such parameter tuning is known to interfere with model performance on many tasks. Moreover, keeping up with shifting user preferences is tricky in such a situation. Decoding-time alignment with reward model guidance solves these issues at the cost of increased inference time. However, most of such methods fail to strike the right balance between exploration and exploitation of reward -- often due to the conflated formulation of these two aspects - to give well-aligned responses. To remedy this we decouple these two aspects and implement them in an evolutionary fashion: exploration is enforced by decoding from mutated instructions and exploitation is represented as the periodic replacement of poorly-rewarded generations with well-rewarded ones. Empirical evidences indicate that this strategy outperforms many preference optimization and decode-time alignment approaches on two widely accepted alignment benchmarks AlpacaEval 2 and MT-Bench. Our implementation will be available at: https://darwin-alignment.github.io.
Authors: Xingjian Hu, Baole Wei, Liangcai Gao, Jun Wang
Abstract: Text line detection is a key task in historical document analysis facing many challenges of arbitrary-shaped text lines, dense texts, and text lines with high aspect ratios, etc. In this paper, we propose a general framework for historical document text detection (SegHist), enabling existing segmentation-based text detection methods to effectively address the challenges, especially text lines with high aspect ratios. Integrating the SegHist framework with the commonly used method DB++, we develop DB-SegHist. This approach achieves SOTA on the CHDAC, MTHv2, and competitive results on HDRC datasets, with a significant improvement of 1.19% on the most challenging CHDAC dataset which features more text lines with high aspect ratios. Moreover, our method attains SOTA on rotated MTHv2 and rotated HDRC, demonstrating its rotational robustness. The code is available at https://github.com/LumionHXJ/SegHist.
Authors: Khai Duy Doan, Abdul Waheed, Muhammad Abdul-Mageed
Abstract: Zero-shot multi-speaker text-to-speech (ZS-TTS) systems have advanced for English, however, it still lags behind due to insufficient resources. We address this gap for Arabic, a language of more than 450 million native speakers, by first adapting a sizeable existing dataset to suit the needs of speech synthesis. Additionally, we employ a set of Arabic dialect identification models to explore the impact of pre-defined dialect labels on improving the ZS-TTS model in a multi-dialect setting. Subsequently, we fine-tune the XTTS\footnote{https://docs.coqui.ai/en/latest/models/xtts.html}\footnote{https://medium.com/machine-learns/xtts-v2-new-version-of-the-open-source-text-to-speech-model-af73914db81f}\footnote{https://medium.com/@erogol/xtts-v1-techincal-notes-eb83ff05bdc} model, an open-source architecture. We then evaluate our models on a dataset comprising 31 unseen speakers and an in-house dialectal dataset. Our automated and human evaluation results show convincing performance while capable of generating dialectal speech. Our study highlights significant potential for improvements in this emerging area of research in Arabic.
URLs: https://docs.coqui.ai/en/latest/models/xtts.html, https://medium.com/machine-learns/xtts-v2-new-version-of-the-open-source-text-to-speech-model-af73914db81f, https://medium.com/@erogol/xtts-v1-techincal-notes-eb83ff05bdc
Authors: Buu Phan, Marton Havasi, Matthew Muckley, Karen Ullrich
Abstract: State-of-the-art language models are autoregressive and operate on subword units known as tokens. Specifically, one must encode the conditioning string into a list of tokens before passing to the language models for next-token prediction. We show that popular encoding schemes, such as maximum prefix encoding (MPE) and byte-pair-encoding (BPE), induce a sampling bias that cannot be mitigated with more training or data. To counter this universal problem, for each encoding scheme above, we propose a novel algorithm to obtain unbiased estimates from any language model trained on tokenized data. Our methods do not require finetuning the model, and the complexity, defined as the number of model runs, scales linearly with the sequence length in the case of MPE. As a result, we show that one can simulate token-free behavior from a tokenized language model. We empirically verify the correctness of our method through a Markov-chain setup, where it accurately recovers the transition probabilities, as opposed to the conventional method of directly prompting tokens into the language model.
Authors: Yusheng Liao, Shuyang Jiang, Yanfeng Wang, Yu Wang
Abstract: Large language models (LLMs) have shown substantial progress in natural language understanding and generation, proving valuable especially in the medical field. Despite advancements, challenges persist due to the complexity and diversity inherent in medical tasks, which can be categorized as knowledge-intensive tasks and alignment-required tasks. Previous approaches either ignore the latter task or focus on a minority of tasks and hence lose generalization. To address these drawbacks, we propose a progressive fine-tuning pipeline. This pipeline employs a Knowledge Aggregator and a Noise aggregator to encode diverse knowledge in the first stage and filter out detrimental information. In the second stage, we drop the Noise Aggregator to avoid the interference of suboptimal representation and leverage an additional alignment module optimized towards an orthogonal direction to the knowledge space to mitigate knowledge forgetting. Based on this two-stage paradigm, we proposed a Medical LLM through decoupling Clinical Alignment and Knowledge Aggregation (MedCare), which is designed to achieve state-of-the-art (SOTA) performance on over 20 medical tasks, as well as SOTA results on specific medical alignment tasks. Various model sizes of MedCare (1.8B, 7B, 14B) all demonstrate significant improvements over existing models with similar model sizes.
Authors: Aditya Kalyanpur, Kailash Saravanakumar, Victor Barres, CJ McFate, Lori Moon, Nati Seifu, Maksim Eremeev, Jose Barrera, Eric Brown, David Ferrucci
Abstract: The advent of Large Language Models (LLMs) and Generative AI has revolutionized natural language applications across various domains. However, high-stakes decision-making tasks in fields such as medical, legal and finance require a level of precision, comprehensiveness, and logical consistency that pure LLM or Retrieval-Augmented-Generation (RAG) approaches often fail to deliver. At Elemental Cognition (EC), we have developed a neuro-symbolic AI platform to tackle these problems. The platform integrates fine-tuned LLMs for knowledge extraction and alignment with a robust symbolic reasoning engine for logical inference, planning and interactive constraint solving. We describe Cora, a Collaborative Research Assistant built on this platform, that is designed to perform complex research and discovery tasks in high-stakes domains. This paper discusses the multi-step inference challenges inherent in such domains, critiques the limitations of existing LLM-based methods, and demonstrates how Cora's neuro-symbolic approach effectively addresses these issues. We provide an overview of the system architecture, key algorithms for knowledge extraction and formal reasoning, and present preliminary evaluation results that highlight Cora's superior performance compared to well-known LLM and RAG baselines.
Authors: Aakanksha, Arash Ahmadian, Beyza Ermis, Seraphina Goldfarb-Tarrant, Julia Kreutzer, Marzieh Fadaee, Sara Hooker
Abstract: A key concern with the concept of "alignment" is the implicit question of "alignment to what?". AI systems are increasingly used across the world, yet safety alignment is often focused on homogeneous monolingual settings. Additionally, preference training and safety measures often overfit to harms common in Western-centric datasets. Here, we explore the viability of different alignment approaches when balancing dual objectives: addressing and optimizing for a non-homogeneous set of languages and cultural preferences while minimizing both global and local harms. We collect the first set of human annotated red-teaming prompts in different languages distinguishing between global and local harm, which serve as a laboratory for understanding the reliability of alignment techniques when faced with preference distributions that are non-stationary across geographies and languages. While this setting is seldom covered by the literature to date, which primarily centers on English harm mitigation, it captures real-world interactions with AI systems around the world. We establish a new precedent for state-of-the-art alignment techniques across 6 languages with minimal degradation in general performance. Our work provides important insights into cross-lingual transfer and novel optimization approaches to safeguard AI systems designed to serve global populations.
Authors: Longchao Da, Tiejin Chen, Lu Cheng, Hua Wei
Abstract: The Large language models (LLMs) have showcased superior capabilities in sophisticated tasks across various domains, stemming from basic question-answer (QA), they are nowadays used as decision assistants or explainers for unfamiliar content. However, they are not always correct due to the data sparsity in specific domain corpus, or the model's hallucination problems. Given this, how much should we trust the responses from LLMs? This paper presents a novel way to evaluate the uncertainty that captures the directional instability, by constructing a directional graph from entailment probabilities, and we innovatively conduct Random Walk Laplacian given the asymmetric property of a constructed directed graph, then the uncertainty is aggregated by the derived eigenvalues from the Laplacian process. We also provide a way to incorporate the existing work's semantics uncertainty with our proposed layer. Besides, this paper identifies the vagueness issues in the raw response set and proposes an augmentation approach to mitigate such a problem, we conducted extensive empirical experiments and demonstrated the superiority of our proposed solutions.
Authors: Raj Sanjay Shah, Khushi Bhardwaj, Sashank Varma
Abstract: Recent studies show evidence for emergent cognitive abilities in Large Pre-trained Language Models (PLMs). The increasing cognitive alignment of these models has made them candidates for cognitive science theories. Prior research into the emergent cognitive abilities of PLMs has largely been path-independent to model training, i.e., has focused on the final model weights and not the intermediate steps. However, building plausible models of human cognition using PLMs would benefit from considering the developmental alignment of their performance during training to the trajectories of children's thinking. Guided by psychometric tests of human intelligence, we choose four sets of tasks to investigate the alignment of ten popular families of PLMs and evaluate their available intermediate and final training steps. These tasks are Numerical ability, Linguistic abilities, Conceptual understanding, and Fluid reasoning. We find a striking regularity: regardless of model size, the developmental trajectories of PLMs consistently exhibit a window of maximal alignment to human cognitive development. Before that window, training appears to endow "blank slate" models with the requisite structure to be poised to rapidly learn from experience. After that window, training appears to serve the engineering goal of reducing loss but not the scientific goal of increasing alignment with human cognition.
Authors: Taeho Hwang, Soyeong Jeong, Sukmin Cho, SeungYoon Han, Jong C. Park
Abstract: Recent advancements in Large Language Models (LLMs) have significantly improved their performance across various Natural Language Processing (NLP) tasks. However, LLMs still struggle with generating non-factual responses due to limitations in their parametric memory. Retrieval-Augmented Generation (RAG) systems address this issue by incorporating external knowledge with a retrieval module. Despite their successes, however, current RAG systems face challenges with retrieval failures and the limited ability of LLMs to filter out irrelevant information. Therefore, in this work, we propose DSLR (Document Refinement with Sentence-Level Re-ranking and Reconstruction), an unsupervised framework that decomposes retrieved documents into sentences, filters out irrelevant sentences, and reconstructs them again into coherent passages. We experimentally validate DSLR on multiple open-domain QA datasets and the results demonstrate that DSLR significantly enhances the RAG performance over conventional fixed-size passage. Furthermore, our DSLR enhances performance in specific, yet realistic scenarios without the need for additional training, providing an effective and efficient solution for refining retrieved documents in RAG systems.
Authors: Polina Tsvilodub, Michael Franke, Fausto Carcassi
Abstract: To what extent can LLMs be used as part of a cognitive model of language generation? In this paper, we approach this question by exploring a neuro-symbolic implementation of an algorithmic cognitive model of referential expression generation by Dale & Reiter (1995). The symbolic task analysis implements the generation as an iterative procedure that scaffolds symbolic and gpt-3.5-turbo-based modules. We compare this implementation to an ablated model and a one-shot LLM-only baseline on the A3DS dataset (Tsvilodub & Franke, 2023). We find that our hybrid approach is cognitively plausible and performs well in complex contexts, while allowing for more open-ended modeling of language generation in a larger domain.
Authors: Lukas Christ, Shahin Amiriparian, Alexander Kathan, Niklas M\"uller, Andreas K\"onig, Bj\"orn W. Schuller
Abstract: Humor is a substantial element of human social behavior, affect, and cognition. Its automatic understanding can facilitate a more naturalistic human-AI interaction. Current methods of humor detection have been exclusively based on staged data, making them inadequate for "real-world" applications. We contribute to addressing this deficiency by introducing the novel Passau-Spontaneous Football Coach Humor (Passau-SFCH) dataset, comprising about 11 hours of recordings. The Passau-SFCH dataset is annotated for the presence of humor and its dimensions (sentiment and direction) as proposed in Martin's Humor Style Questionnaire. We conduct a series of experiments employing pretrained Transformers, convolutional neural networks, and expert-designed features. The performance of each modality (text, audio, video) for spontaneous humor recognition is analyzed and their complementarity is investigated. Our findings suggest that for the automatic analysis of humor and its sentiment, facial expressions are most promising, while humor direction can be best modeled via text-based features. Further, we experiment with different multimodal approaches to humor recognition, including decision-level fusion and MulT, a multimodal Transformer approach. In this context, we propose a novel multimodal architecture that yields the best overall results. Finally, we make our code publicly available at https://www.github.com/lc0197/passau-sfch. The Passau-SFCH dataset is available upon request.
Authors: Subba Reddy Oota, Zijiao Chen, Manish Gupta, Raju S. Bapi, Gael Jobard, Frederic Alexandre, Xavier Hinaut
Abstract: Can we obtain insights about the brain using AI models? How is the information in deep learning models related to brain recordings? Can we improve AI models with the help of brain recordings? Such questions can be tackled by studying brain recordings like functional magnetic resonance imaging (fMRI). As a first step, the neuroscience community has contributed several large cognitive neuroscience datasets related to passive reading/listening/viewing of concept words, narratives, pictures, and movies. Encoding and decoding models using these datasets have also been proposed in the past two decades. These models serve as additional tools for basic cognitive science and neuroscience research. Encoding models aim at generating fMRI brain representations given a stimulus automatically. They have several practical applications in evaluating and diagnosing neurological conditions and thus may also help design therapies for brain damage. Decoding models solve the inverse problem of reconstructing the stimuli given the fMRI. They are useful for designing brain-machine or brain-computer interfaces. Inspired by the effectiveness of deep learning models for natural language processing, computer vision, and speech, several neural encoding and decoding models have been recently proposed. In this survey, we will first discuss popular representations of language, vision and speech stimuli, and present a summary of neuroscience datasets. Further, we will review popular deep learning based encoding and decoding architectures and note their benefits and limitations. Finally, we will conclude with a summary and discussion about future trends. Given the large amount of recently published work in the computational cognitive neuroscience (CCN) community, we believe that this survey enables an entry point for DNN researchers to diversify into CCN research.
Authors: Hye Sun Yun, Mehdi Arjmand, Phillip Sherlock, Michael K. Paasche-Orlow, James W. Griffith, Timothy Bickmore
Abstract: Standardized, validated questionnaires are vital tools in research and healthcare, offering dependable self-report data. Prior work has revealed that virtual agent-administered questionnaires are almost equivalent to self-administered ones in an electronic form. Despite being an engaging method, repeated use of virtual agent-administered questionnaires in longitudinal or pre-post studies can induce respondent fatigue, impacting data quality via response biases and decreased response rates. We propose using large language models (LLMs) to generate diverse questionnaire versions while retaining good psychometric properties. In a longitudinal study, participants interacted with our agent system and responded daily for two weeks to one of the following questionnaires: a standardized depression questionnaire, question variants generated by LLMs, or question variants accompanied by LLM-generated small talk. The responses were compared to a validated depression questionnaire. Psychometric testing revealed consistent covariation between the external criterion and focal measure administered across the three conditions, demonstrating the reliability and validity of the LLM-generated variants. Participants found that the variants were significantly less repetitive than repeated administrations of the same standardized questionnaire. Our findings highlight the potential of LLM-generated variants to invigorate agent-administered questionnaires and foster engagement and interest, without compromising their validity.
Authors: Jitang Li, Jinzheng Li
Abstract: With the development in cognitive science and Large Language Models (LLMs), increasing connections have come to light between these two distinct fields. Building upon these connections, we propose a conjecture suggesting the existence of a duality between LLMs and Tulving's theory of memory. We identify a potential correspondence between Tulving's synergistic ecphory model (SEM) of retrieval and the emergent abilities observed in LLMs, serving as supporting evidence for our conjecture. Furthermore, we speculate that consciousness may be considered a form of emergent ability based on this duality. We also discuss how other theories of consciousness intersect with our research.
Authors: Yuling Shi, Hongyu Zhang, Chengcheng Wan, Xiaodong Gu
Abstract: Large language models have catalyzed an unprecedented wave in code generation. While achieving significant advances, they blur the distinctions between machine- and human-authored source code, causing integrity and authenticity issues of software artifacts. Previous methods such as DetectGPT have proven effective in discerning machine-generated texts, but they do not identify and harness the unique patterns of machine-generated code. Thus, its applicability falters when applied to code. In this paper, we carefully study the specific patterns that characterize machine- and human-authored code. Through a rigorous analysis of code attributes such as lexical diversity, conciseness, and naturalness, we expose unique patterns inherent to each source. We particularly notice that the syntactic segmentation of code is a critical factor in identifying its provenance. Based on our findings, we propose DetectCodeGPT, a novel method for detecting machine-generated code, which improves DetectGPT by capturing the distinct stylized patterns of code. Diverging from conventional techniques that depend on external LLMs for perturbations, DetectCodeGPT perturbs the code corpus by strategically inserting spaces and newlines, ensuring both efficacy and efficiency. Experiment results show that our approach significantly outperforms state-of-the-art techniques in detecting machine-generated code.
Authors: Chenxi Liu, Sun Yang, Qianxiong Xu, Zhishuai Li, Cheng Long, Ziyue Li, Rui Zhao
Abstract: Traffic prediction, an essential component for intelligent transportation systems, endeavours to use historical data to foresee future traffic features at specific locations. Although existing traffic prediction models often emphasize developing complex neural network structures, their accuracy has not improved. Recently, large language models have shown outstanding capabilities in time series analysis. Differing from existing models, LLMs progress mainly through parameter expansion and extensive pretraining while maintaining their fundamental structures. Motivated by these developments, we propose a Spatial-Temporal Large Language Model (ST-LLM) for traffic prediction. In the ST-LLM, we define timesteps at each location as tokens and design a spatial-temporal embedding to learn the spatial location and global temporal patterns of these tokens. Additionally, we integrate these embeddings by a fusion convolution to each token for a unified spatial-temporal representation. Furthermore, we innovate a partially frozen attention strategy to adapt the LLM to capture global spatial-temporal dependencies for traffic prediction. Comprehensive experiments on real traffic datasets offer evidence that ST-LLM is a powerful spatial-temporal learner that outperforms state-of-the-art models. Notably, the ST-LLM also exhibits robust performance in both few-shot and zero-shot prediction scenarios. The code is publicly available at https://github.com/ChenxiLiu-HNU/ST-LLM.
Authors: Zongyu Wu, Hongcheng Gao, Yueze Wang, Xiang Zhang, Suhang Wang
Abstract: Text-to-Image (T2I) models have shown great performance in generating images based on textual prompts. However, these models are vulnerable to unsafe input to generate unsafe content like sexual, harassment and illegal-activity images. Existing studies based on image checker, model fine-tuning and embedding blocking are impractical in real-world applications. Hence, we propose the first universal prompt optimizer for safe T2I (POSI) generation in black-box scenario. We first construct a dataset consisting of toxic-clean prompt pairs by GPT-3.5 Turbo. To guide the optimizer to have the ability of converting toxic prompt to clean prompt while preserving semantic information, we design a novel reward function measuring toxicity and text alignment of generated images and train the optimizer through Proximal Policy Optimization. Experiments show that our approach can effectively reduce the likelihood of various T2I models in generating inappropriate images, with no significant impact on text alignment. It is also flexible to be combined with methods to achieve better performance. Our code is available at https://github.com/wzongyu/POSI.
Authors: Pengzhou Cheng, Yidong Ding, Tianjie Ju, Zongru Wu, Wei Du, Ping Yi, Zhuosheng Zhang, Gongshen Liu
Abstract: Large language models (LLMs) have raised concerns about potential security threats despite performing significantly in Natural Language Processing (NLP). Backdoor attacks initially verified that LLM is doing substantial harm at all stages, but the cost and robustness have been criticized. Attacking LLMs is inherently risky in security review, while prohibitively expensive. Besides, the continuous iteration of LLMs will degrade the robustness of backdoors. In this paper, we propose TrojanRAG, which employs a joint backdoor attack in the Retrieval-Augmented Generation, thereby manipulating LLMs in universal attack scenarios. Specifically, the adversary constructs elaborate target contexts and trigger sets. Multiple pairs of backdoor shortcuts are orthogonally optimized by contrastive learning, thus constraining the triggering conditions to a parameter subspace to improve the matching. To improve the recall of the RAG for the target contexts, we introduce a knowledge graph to construct structured data to achieve hard matching at a fine-grained level. Moreover, we normalize the backdoor scenarios in LLMs to analyze the real harm caused by backdoors from both attackers' and users' perspectives and further verify whether the context is a favorable tool for jailbreaking models. Extensive experimental results on truthfulness, language understanding, and harmfulness show that TrojanRAG exhibits versatility threats while maintaining retrieval capabilities on normal queries.
Authors: Jingyang Ou, Shen Nie, Kaiwen Xue, Fengqi Zhu, Jiacheng Sun, Zhenguo Li, Chongxuan Li
Abstract: Discrete diffusion models with absorbing processes have shown promise in language modeling. The key quantities to be estimated are the ratios between the marginal probabilities of two transitive states at all timesteps, called the concrete score. In this paper, we reveal that the concrete score in absorbing diffusion can be expressed as conditional probabilities of clean data, multiplied by a time-dependent scalar in an analytic form. Motivated by this finding, we propose reparameterized absorbing discrete diffusion (RADD), a dedicated diffusion model without time-condition that characterizes the time-independent conditional probabilities. Besides its simplicity, RADD can reduce the number of function evaluations (NFEs) by caching the output of the time-independent network when the noisy sample remains unchanged in a sampling interval. Empirically, RADD is up to 3.5 times faster while achieving similar performance with the strongest baseline. Built upon the new perspective of conditional distributions, we further unify absorbing discrete diffusion and any-order autoregressive models (AO-ARMs), showing that the upper bound on the negative log-likelihood for the diffusion model can be interpreted as an expected negative log-likelihood for AO-ARMs. Further, our RADD models achieve SOTA performance among diffusion models on 5 zero-shot language modeling benchmarks (measured by perplexity) at the GPT-2 scale. Our code is available at https://github.com/ML-GSAI/RADD.
Authors: Andy Zou, Long Phan, Justin Wang, Derek Duenas, Maxwell Lin, Maksym Andriushchenko, Rowan Wang, Zico Kolter, Matt Fredrikson, Dan Hendrycks
Abstract: AI systems can take harmful actions and are highly vulnerable to adversarial attacks. We present an approach, inspired by recent advances in representation engineering, that interrupts the models as they respond with harmful outputs with "circuit breakers." Existing techniques aimed at improving alignment, such as refusal training, are often bypassed. Techniques such as adversarial training try to plug these holes by countering specific attacks. As an alternative to refusal training and adversarial training, circuit-breaking directly controls the representations that are responsible for harmful outputs in the first place. Our technique can be applied to both text-only and multimodal language models to prevent the generation of harmful outputs without sacrificing utility -- even in the presence of powerful unseen attacks. Notably, while adversarial robustness in standalone image recognition remains an open challenge, circuit breakers allow the larger multimodal system to reliably withstand image "hijacks" that aim to produce harmful content. Finally, we extend our approach to AI agents, demonstrating considerable reductions in the rate of harmful actions when they are under attack. Our approach represents a significant step forward in the development of reliable safeguards to harmful behavior and adversarial attacks.
Authors: Yiqing Shen, Zan Chen, Michail Mamalakis, Luhan He, Haiyang Xia, Tianbin Li, Yanzhou Su, Junjun He, Yu Guang Wang
Abstract: The parallels between protein sequences and natural language in their sequential structures have inspired the application of large language models (LLMs) to protein understanding. Despite the success of LLMs in NLP, their effectiveness in comprehending protein sequences remains an open question, largely due to the absence of datasets linking protein sequences to descriptive text. Researchers have then attempted to adapt LLMs for protein understanding by integrating a protein sequence encoder with a pre-trained LLM. However, this adaptation raises a fundamental question: "Can LLMs, originally designed for NLP, effectively comprehend protein sequences as a form of language?" Current datasets fall short in addressing this question due to the lack of a direct correlation between protein sequences and corresponding text descriptions, limiting the ability to train and evaluate LLMs for protein understanding effectively. To bridge this gap, we introduce ProteinLMDataset, a dataset specifically designed for further self-supervised pretraining and supervised fine-tuning (SFT) of LLMs to enhance their capability for protein sequence comprehension. Specifically, ProteinLMDataset includes 17.46 billion tokens for pretraining and 893,000 instructions for SFT. Additionally, we present ProteinLMBench, the first benchmark dataset consisting of 944 manually verified multiple-choice questions for assessing the protein understanding capabilities of LLMs. ProteinLMBench incorporates protein-related details and sequences in multiple languages, establishing a new standard for evaluating LLMs' abilities in protein comprehension. The large language model InternLM2-7B, pretrained and fine-tuned on the ProteinLMDataset, outperforms GPT-4 on ProteinLMBench, achieving the highest accuracy score.
Authors: Shwai He, Guoheng Sun, Zheyu Shen, Ang Li
Abstract: Scaling Transformer-based large language models (LLMs) has demonstrated promising performance across various tasks. However, this scaling also introduces redundant structures, posing challenges for real-world deployment. Despite some recognition of redundancy in LLMs, the variability of redundancy across different structures, such as MLP and Attention layers, is under-explored. In this work, we investigate the varying redundancy across different modules within Transformers, including Blocks, MLP, and Attention layers, using a similarity-based metric. This metric operates on the premise that redundant structures produce outputs highly similar to their inputs. Surprisingly, while attention layers are essential for transformers and distinguish them from other mainstream architectures, we found that a large proportion of attention layers exhibit excessively high similarity and can be safely pruned without degrading performance, leading to reduced memory and computation costs. Additionally, we further propose a method that jointly drops Attention and MLP layers, achieving improved performance and dropping ratios. Extensive experiments demonstrate the effectiveness of our methods, e.g., Llama-3-70B maintains comparable performance even after pruning half of the attention layers. Our findings provide valuable insights for future network architecture design. The code will be released at: \url{https://github.com/Shwai-He/LLM-Drop}.
Authors: Yuheng Zhang, Dian Yu, Baolin Peng, Linfeng Song, Ye Tian, Mingyue Huo, Nan Jiang, Haitao Mi, Dong Yu
Abstract: Reinforcement Learning with Human Feedback (RLHF) has achieved great success in aligning large language models (LLMs) with human preferences. Prevalent RLHF approaches are reward-based, following the Bradley-Terry (BT) model assumption, which may not fully capture the complexity of human preferences. In this paper, we explore RLHF under a general preference framework and approach it from a game-theoretic perspective. Specifically, we formulate the problem as a two-player game and propose a novel algorithm, iterative Nash policy optimization (INPO). The key idea is to let the policy play against itself via no-regret learning, thereby approximating the Nash policy. Unlike previous methods, INPO bypasses the need for estimating the expected win rate for individual responses, which typically incurs high computational or annotation costs. Instead, we introduce a new loss objective that is directly minimized over a preference dataset. We provide theoretical analysis for our approach and demonstrate its effectiveness through experiments on various representative benchmarks. With an LLaMA-3-8B-based SFT model, INPO achieves a 41.5% length-controlled win rate on AlpacaEval 2.0 and a 38.3% win rate on Arena-Hard, showing substantial improvement over the state-of-the-art iterative algorithm [Dong et al., 2024] under the BT model assumption. Additionally, our ablation study highlights the benefits of incorporating KL regularization for response length control.
Authors: Joan Nwatu, Oana Ignat, Rada Mihalcea
Abstract: Unequal representation across cultures and socioeconomic groups in AI is a significant and challenging problem, often leading to uneven model performance. As a step toward addressing this issue, we formulate translated non-English, geographic, and socioeconomic integrated prompts and evaluate their impact on VL model performance for data from different countries and income groups. Our findings show that geographic and socioeconomic integrated prompts improve VL performance on lower-income data and favor the retrieval of topic appearances commonly found in data from low-income households. From our analyses, we identify and highlight contexts where these strategies yield the most improvements. Our model analysis code is publicly available at https://github.com/Anniejoan/Uplifting-Lower-income-data .
URLs: https://github.com/Anniejoan/Uplifting-Lower-income-data
Authors: Hannes Kunstmann, Joseph Ollier, Joel Persson, Florian von Wangenheim
Abstract: Large language models (LLMs) present an enormous evolution in the strategic potential of conversational recommender systems (CRS). Yet to date, research has predominantly focused upon technical frameworks to implement LLM-driven CRS, rather than end-user evaluations or strategic implications for firms, particularly from the perspective of a small to medium enterprises (SME) that makeup the bedrock of the global economy. In the current paper, we detail the design of an LLM-driven CRS in an SME setting, and its subsequent performance in the field using both objective system metrics and subjective user evaluations. While doing so, we additionally outline a short-form revised ResQue model for evaluating LLM-driven CRS, enabling replicability in a rapidly evolving field. Our results reveal good system performance from a user experience perspective (85.5% recommendation accuracy) but underscore latency, cost, and quality issues challenging business viability. Notably, with a median cost of $0.04 per interaction and a latency of 5.7s, cost-effectiveness and response time emerge as crucial areas for achieving a more user-friendly and economically viable LLM-driven CRS for SME settings. One major driver of these costs is the use of an advanced LLM as a ranker within the retrieval-augmented generation (RAG) technique. Our results additionally indicate that relying solely on approaches such as Prompt-based learning with ChatGPT as the underlying LLM makes it challenging to achieve satisfying quality in a production environment. Strategic considerations for SMEs deploying an LLM-driven CRS are outlined, particularly considering trade-offs in the current technical landscape.