Authors: Mariana Yukari Noguti, Edduardo Vellasques, Luiz Eduardo Soares Oliveira
Abstract: Recent advances in language modelling has significantly decreased the need of labelled data in text classification tasks. Transformer-based models, pre-trained on unlabeled data, can outmatch the performance of models trained from scratch for each task. However, the amount of labelled data need to fine-tune such type of model is still considerably high for domains requiring expert-level annotators, like the legal domain. This paper investigates the best strategies for optimizing the use of a small labeled dataset and large amounts of unlabeled data and perform a classification task in the legal area with 50 predefined topics. More specifically, we use the records of demands to a Brazilian Public Prosecutor's Office aiming to assign the descriptions in one of the subjects, which currently demands deep legal knowledge for manual filling. The task of optimizing the performance of classifiers in this scenario is especially challenging, given the low amount of resources available regarding the Portuguese language, especially in the legal domain. Our results demonstrate that classic supervised models such as logistic regression and SVM and the ensembles random forest and gradient boosting achieve better performance along with embeddings extracted with word2vec when compared to BERT language model. The latter demonstrates superior performance in association with the architecture of the model itself as a classifier, having surpassed all previous models in that regard. The best result was obtained with Unsupervised Data Augmentation (UDA), which jointly uses BERT, data augmentation, and strategies of semi-supervised learning, with an accuracy of 80.7% in the aforementioned task.
Authors: Xichen Tang
Abstract: Using computerized verifiable formal languages like Lean 4 to prove mathematical theorems has a significant impact on mathematical formalization. Lean 4 offers prominent potential for advancing mathematical reasoning. However, existing efforts are limited to mathematical formalization languages in substantial online corpora and are dedicated to keeping pace with rapidly evolving languages. To bridge the gap between the traditional and computerized proof, my approach to formalizing theorem proving involves generating formal steps and complete proofs using Large Language Models (LLMs) based on Natural Language (NL) proofs. The method is to introduce the basic structure and tactics in general, determine how AI can assist the mathematical formalization process to improve its performance, and give examples of solving problems in Lean 4 comparing to NL, mainly in IMO, and a sample theorem proving in abstract algebra.
Authors: Francisco Valentini, Viviana Cotik, Dami\'an Furman, Ivan Bercovich, Edgar Altszyler, Juan Manuel P\'erez
Abstract: Information retrieval (IR) is the task of finding relevant documents in response to a user query. Although Spanish is the second most spoken native language, current IR benchmarks lack Spanish data, hindering the development of information access tools for Spanish speakers. We introduce MessIRve, a large-scale Spanish IR dataset with around 730 thousand queries from Google's autocomplete API and relevant documents sourced from Wikipedia. MessIRve's queries reflect diverse Spanish-speaking regions, unlike other datasets that are translated from English or do not consider dialectal variations. The large size of the dataset allows it to cover a wide variety of topics, unlike smaller datasets. We provide a comprehensive description of the dataset, comparisons with existing datasets, and baseline evaluations of prominent IR models. Our contributions aim to advance Spanish IR research and improve information access for Spanish speakers.
Authors: Lukas Garbas, Max Ploner, Alan Akbik
Abstract: Classification tasks in NLP are typically addressed by selecting a pre-trained language model (PLM) from a model hub, and fine-tuning it for the task at hand. However, given the very large number of PLMs that are currently available, a practical challenge is to determine which of them will perform best for a specific downstream task. With this paper, we introduce TransformerRanker, a lightweight library that efficiently ranks PLMs for classification tasks without the need for computationally costly fine-tuning. Our library implements current approaches for transferability estimation (LogME, H-Score, kNN), in combination with layer aggregation options, which we empirically showed to yield state-of-the-art rankings of PLMs (Garbas et al., 2024). We designed the interface to be lightweight and easy to use, allowing users to directly connect to the HuggingFace Transformers and Dataset libraries. Users need only select a downstream classification task and a list of PLMs to create a ranking of likely best-suited PLMs for their task. We make TransformerRanker available as a pip-installable open-source library https://github.com/flairNLP/transformer-ranker.
Authors: Leanne Nortje, Dan Oneata, Herman Kamper
Abstract: Given an image query, visually prompted keyword localisation (VPKL) aims to find occurrences of the depicted word in a speech collection. This can be useful when transcriptions are not available for a low-resource language (e.g. if it is unwritten). Previous work showed that VPKL can be performed with a visually grounded speech model trained on paired images and unlabelled speech. But all experiments were done on English. Moreover, transcriptions were used to get positive and negative pairs for the contrastive loss. This paper introduces a few-shot learning scheme to mine pairs automatically without transcriptions. On English, this results in only a small drop in performance. We also - for the first time - consider VPKL on a real low-resource language, Yoruba. While scores are reasonable, here we see a bigger drop in performance compared to using ground truth pairs because the mining is less accurate in Yoruba.
Authors: Christina Walker, Joan C. Timoneda
Abstract: Extant work shows that generative AI models such as GPT-3.5 and 4 perpetuate social stereotypes and biases. One concerning but less explored source of bias is ideology. Do GPT models take ideological stances on politically sensitive topics? In this article, we provide an original approach to identifying ideological bias in generative models, showing that bias can stem from both the training data and the filtering algorithm. We leverage linguistic variation in countries with contrasting political attitudes to evaluate bias in average GPT responses to sensitive political topics in those languages. First, we find that GPT output is more conservative in languages that map well onto conservative societies (i.e., Polish), and more liberal in languages used uniquely in liberal societies (i.e., Swedish). This result provides strong evidence of training data bias in GPT models. Second, differences across languages observed in GPT-3.5 persist in GPT-4, even though GPT-4 is significantly more liberal due to OpenAI's filtering policy. Our main takeaway is that generative model training must focus on high-quality, curated datasets to reduce bias, even if it entails a compromise in training data size. Filtering responses after training only introduces new biases and does not remove the underlying training biases.
Authors: Joymallya Chakraborty, Wei Xia, Anirban Majumder, Dan Ma, Walid Chaabene, Naveed Janvekar
Abstract: Large language models (LLMs) have demonstrated remarkable capabilities in natural language processing tasks. However, their practical application in high-stake domains, such as fraud and abuse detection, remains an area that requires further exploration. The existing applications often narrowly focus on specific tasks like toxicity or hate speech detection. In this paper, we present a comprehensive benchmark suite designed to assess the performance of LLMs in identifying and mitigating fraudulent and abusive language across various real-world scenarios. Our benchmark encompasses a diverse set of tasks, including detecting spam emails, hate speech, misogynistic language, and more. We evaluated several state-of-the-art LLMs, including models from Anthropic, Mistral AI, and the AI21 family, to provide a comprehensive assessment of their capabilities in this critical domain. The results indicate that while LLMs exhibit proficient baseline performance in individual fraud and abuse detection tasks, their performance varies considerably across tasks, particularly struggling with tasks that demand nuanced pragmatic reasoning, such as identifying diverse forms of misogynistic language. These findings have important implications for the responsible development and deployment of LLMs in high-risk applications. Our benchmark suite can serve as a tool for researchers and practitioners to systematically evaluate LLMs for multi-task fraud detection and drive the creation of more robust, trustworthy, and ethically-aligned systems for fraud and abuse detection.
Authors: Yujian Gan, Changling Li, Jinxia Xie, Luou Wen, Matthew Purver, Massimo Poesio
Abstract: We introduce ClarQ-LLM, an evaluation framework consisting of bilingual English-Chinese conversation tasks, conversational agents and evaluation metrics, designed to serve as a strong benchmark for assessing agents' ability to ask clarification questions in task-oriented dialogues. The benchmark includes 31 different task types, each with 10 unique dialogue scenarios between information seeker and provider agents. The scenarios require the seeker to ask questions to resolve uncertainty and gather necessary information to complete tasks. Unlike traditional benchmarks that evaluate agents based on fixed dialogue content, ClarQ-LLM includes a provider conversational agent to replicate the original human provider in the benchmark. This allows both current and future seeker agents to test their ability to complete information gathering tasks through dialogue by directly interacting with our provider agent. In tests, LLAMA3.1 405B seeker agent managed a maximum success rate of only 60.05\%, showing that ClarQ-LLM presents a strong challenge for future research.
Authors: Shervin Ghasemlou, Ashish Katiyar, Aparajita Saraf, Seungwhan Moon, Mangesh Pujari, Pinar Donmez, Babak Damavandi, Anuj Kumar
Abstract: In this paper, we investigate the problem of "generation supervision" in large language models, and present a novel bicameral architecture to separate supervision signals from their core capability, helpfulness. Doppelg\"anger, a new module parallel to the underlying language model, supervises the generation of each token, and learns to concurrently predict the supervision score(s) of the sequences up to and including each token. In this work, we present the theoretical findings, and leave the report on experimental results to a forthcoming publication.
Authors: Sung-Lin Yeh, Hao Tang
Abstract: Representing speech with discrete units has been widely used in speech codec and speech generation. However, there are several unverified claims about self-supervised discrete units, such as disentangling phonetic and speaker information with k-means, or assuming information loss after k-means. In this work, we take an information-theoretic perspective to answer how much information is present (information completeness) and how much information is accessible (information accessibility), before and after residual vector quantization. We show a lower bound for information completeness and estimate completeness on discretized HuBERT representations after residual vector quantization. We find that speaker information is sufficiently present in HuBERT discrete units, and that phonetic information is sufficiently present in the residual, showing that vector quantization does not achieve disentanglement. Our results offer a comprehensive assessment on the choice of discrete units, and suggest that a lot more information in the residual should be mined rather than discarded.
Authors: Neha Prakriya, Jui-Nan Yen, Cho-Jui Hsieh, Jason Cong
Abstract: Large Language Model (LLM) pretraining traditionally relies on autoregressive language modeling on randomly sampled data blocks from web-scale datasets. We take inspiration from human learning techniques like spaced repetition to hypothesize that random data sampling for LLMs leads to high training cost and low quality models which tend to forget data. In order to effectively commit web-scale information to long-term memory, we propose the LFR (Learn, Focus, and Review) pedagogy, a new dynamic training paradigm which focuses and repeatedly reviews complex data blocks at systematic intervals based on the model's learning pace and progress. LFR records the model perplexities for different data blocks and frequently revisits blocks with higher perplexity which are more likely to be forgotten. We pretrain the GPT-2 models (124M - 1.5B) from scratch on the OpenWebText dataset using LFR. We test on downstream tasks from the language modeling, question answering, translation, and problem solving domains to achieve consistently lower perplexity and higher accuracy than the baseline OpenAI models, while obtaining a 20x pretraining speed-up.
Authors: Yining Chen, Jianqiang Li, Changwei Song, Qing Zhao, Yongsheng Tong, Guanghui Fu
Abstract: Suicide is a pressing global issue, demanding urgent and effective preventive interventions. Among the various strategies in place, psychological support hotlines had proved as a potent intervention method. Approximately two million people in China attempt suicide annually, with many individuals making multiple attempts. Prompt identification and intervention for high-risk individuals are crucial to preventing tragedies. With the rapid advancement of artificial intelligence (AI), especially the development of large-scale language models (LLMs), new technological tools have been introduced to the field of mental health. This study included 1284 subjects, and was designed to validate whether deep learning models and LLMs, using audio and transcribed text from support hotlines, can effectively predict suicide risk. We proposed a simple LLM-based pipeline that first summarizes transcribed text from approximately one hour of speech to extract key features, and then predict suicidial bahaviours in the future. We compared our LLM-based method with the traditional manual scale approach in a clinical setting and with five advanced deep learning models. Surprisingly, the proposed simple LLM pipeline achieved strong performance on a test set of 46 subjects, with an F1 score of 76\% when combined with manual scale rating. This is 7\% higher than the best speech-based deep learning models and represents a 27.82\% point improvement in F1 score compared to using the manual scale apporach alone. Our study explores new applications of LLMs and demonstrates their potential for future use in suicide prevention efforts.
Authors: Georgios Chochlakis, Niyantha Maruthu Pandiyan, Kristina Lerman, Shrikanth Narayanan
Abstract: In-Context Learning (ICL) in Large Language Models (LLM) has emerged as the dominant technique for performing natural language tasks, as it does not require updating the model parameters with gradient-based methods. ICL promises to "adapt" the LLM to perform the present task at a competitive or state-of-the-art level at a fraction of the computational cost. ICL can be augmented by incorporating the reasoning process to arrive at the final label explicitly in the prompt, a technique called Chain-of-Thought (CoT) prompting. However, recent work has found that ICL relies mostly on the retrieval of task priors and less so on "learning" to perform tasks, especially for complex subjective domains like emotion and morality, where priors ossify posterior predictions. In this work, we examine whether "enabling" reasoning also creates the same behavior in LLMs, wherein the format of CoT retrieves reasoning priors that remain relatively unchanged despite the evidence in the prompt. We find that, surprisingly, CoT indeed suffers from the same posterior collapse as ICL for larger language models. Code is avalaible at https://github.com/gchochla/cot-priors.
Authors: Sandeep Kumar, Tirthankar Ghosal, Vinayak Goyal, Asif Ekbal
Abstract: "An idea is nothing more nor less than a new combination of old elements" (Young, J.W.). The widespread adoption of Large Language Models (LLMs) and publicly available ChatGPT have marked a significant turning point in the integration of Artificial Intelligence (AI) into people's everyday lives. This study explores the capability of LLMs in generating novel research ideas based on information from research papers. We conduct a thorough examination of 4 LLMs in five domains (e.g., Chemistry, Computer, Economics, Medical, and Physics). We found that the future research ideas generated by Claude-2 and GPT-4 are more aligned with the author's perspective than GPT-3.5 and Gemini. We also found that Claude-2 generates more diverse future research ideas than GPT-4, GPT-3.5, and Gemini 1.0. We further performed a human evaluation of the novelty, relevancy, and feasibility of the generated future research ideas. This investigation offers insights into the evolving role of LLMs in idea generation, highlighting both its capability and limitations. Our work contributes to the ongoing efforts in evaluating and utilizing language models for generating future research ideas. We make our datasets and codes publicly available.
Authors: Yoonji Nam, TaeWoong Seo, Gyeongcheol Shin, Sangji Lee, JaeEun Im
Abstract: To mitigate the difficulties of university freshmen in adapting to university life, we developed NOVI, a chatbot system based on GPT-4o. This system utilizes post and comment data from SKKU 'Everytime', a university community site. Developed using LangChain, NOVI's performance has been evaluated with a BLEU score, Perplexity score, ROUGE-1 score, ROUGE-2 score, ROUGE-L score and METEOR score. This approach is not only limited to help university freshmen but is also expected to help various people adapting to new environments with different data. This research explores the development and potential application of new educational technology tools, contributing to easier social adaptation for beginners and settling a foundation for future advancement in LLM studies.
Authors: Kohei Tsuji, Tatsuya Hiraoka, Yuchang Cheng, Tomoya Iwakura
Abstract: Many datasets of natural language processing (NLP) sometimes include annotation errors. Researchers have attempted to develop methods to reduce the adverse effect of errors in datasets automatically. However, an existing method is time-consuming because it requires many trained models to detect errors. We propose a novel method to reduce the time of error detection. Specifically, we use a tokenization technique called subword regularization to create pseudo-multiple models which are used to detect errors. Our proposed method, SubRegWeigh, can perform annotation weighting four to five times faster than the existing method. Additionally, SubRegWeigh improved performance in both document classification and named entity recognition tasks. In experiments with pseudo-incorrect labels, pseudo-incorrect labels were adequately detected.
Authors: Sakshi Deo Shukla, Pavel Denisov, Tugtekin Turan
Abstract: Recent advancements in speech-based topic segmentation have highlighted the potential of pretrained speech encoders to capture semantic representations directly from speech. Traditionally, topic segmentation has relied on a pipeline approach in which transcripts of the automatic speech recognition systems are generated, followed by text-based segmentation algorithms. In this paper, we introduce an end-to-end scheme that bypasses this conventional two-step process by directly employing semantic speech encoders for segmentation. Focused on the broadcasted news domain, which poses unique challenges due to the diversity of speakers and topics within single recordings, we address the challenge of accessing topic change points efficiently in an end-to-end manner. Furthermore, we propose a new benchmark for spoken news topic segmentation by utilizing a dataset featuring approximately 1000 hours of publicly available recordings across six European languages and including an evaluation set in Hindi to test the model's cross-domain performance in a cross-lingual, zero-shot scenario. This setup reflects real-world diversity and the need for models adapting to various linguistic settings. Our results demonstrate that while the traditional pipeline approach achieves a state-of-the-art $P_k$ score of 0.2431 for English, our end-to-end model delivers a competitive $P_k$ score of 0.2564. When trained multilingually, these scores further improve to 0.1988 and 0.2370, respectively. To support further research, we release our model along with data preparation scripts, facilitating open research on multilingual spoken news topic segmentation.
Authors: Linfeng Zhang, Changyue Hu, Zhiyu Quan
Abstract: As the body of academic literature continues to grow, researchers face increasing difficulties in effectively searching for relevant resources. Existing databases and search engines often fall short of providing a comprehensive and contextually relevant collection of academic literature. To address this issue, we propose a novel framework that leverages Natural Language Processing (NLP) techniques. This framework automates the retrieval, summarization, and clustering of academic literature within a specific research domain. To demonstrate the effectiveness of our approach, we introduce CyLit, an NLP-powered repository specifically designed for the cyber risk literature. CyLit empowers researchers by providing access to context-specific resources and enabling the tracking of trends in the dynamic and rapidly evolving field of cyber risk. Through the automatic processing of large volumes of data, our NLP-powered solution significantly enhances the efficiency and specificity of academic literature searches. We compare the literature categorization results of CyLit to those presented in survey papers or generated by ChatGPT, highlighting the distinctive insights this tool provides into cyber risk research literature. Using NLP techniques, we aim to revolutionize the way researchers discover, analyze, and utilize academic resources, ultimately fostering advancements in various domains of knowledge.
Authors: Jihyun Lee, Gary Geunbae Lee
Abstract: Traditional dialogue state tracking approaches heavily rely on extensive training data and handcrafted features, limiting their scalability and adaptability to new domains. In this paper, we propose a novel method that leverages inference and in-context learning with ChatGPT for domain transfer in dialogue state tracking, without any parameter updates. By guiding ChatGPT's chain of thought, we enable it to retrieve relevant examples and generalize knowledge to accurately infer dialogue states, solely through inference. Experimental results on the MultiWOZ dataset demonstrate competitive performance and promising generalization across domains. Our parameter-free approach offers a scalable and adaptable solution, opening new research directions in domain transfer learning.
Authors: Jihyun Lee, Solee Im, Wonjun Lee, Gary Geunbae Lee
Abstract: Dialogue State Tracking (DST) is a key part of task-oriented dialogue systems, identifying important information in conversations. However, its accuracy drops significantly in spoken dialogue environments due to named entity errors from Automatic Speech Recognition (ASR) systems. We introduce a simple yet effective data augmentation method that targets those entities to improve the robustness of DST model. Our novel method can control the placement of errors using keyword-highlighted prompts while introducing phonetically similar errors. As a result, our method generated sufficient error patterns on keywords, leading to improved accuracy in noised and low-accuracy ASR environments.
Authors: Nicholas Pochinkov, Angelo Benoit, Lovkush Agarwal, Zainab Ali Majid, Lucile Ter-Minassian
Abstract: Generative large language models (LLMs) excel in natural language processing tasks, yet their inner workings remain underexplored beyond token-level predictions. This study investigates the degree to which these models decide the content of a paragraph at its onset, shedding light on their contextual understanding. By examining the information encoded in single-token activations, specifically the "\textbackslash n\textbackslash n" double newline token, we demonstrate that patching these activations can transfer significant information about the context of the following paragraph, providing further insights into the model's capacity to plan ahead.
Authors: Zi Yang
Abstract: We argue that there are two major distinct capabilities in long context understanding: retrieval and holistic understanding. Understanding and further improving LLMs' long context capabilities would not be possible without knowing the tasks' focus categories. We aim to automatically identify retrieval focused and holistic understanding focused problems from suites of benchmarks and quantitatively measure the difficulty within each focus. In this paper, we present the Dolce framework, which parameterizes each problem by $\lambda$ (complexity) and $k$ (redundancy) and assigns to one of five predefined focus categories. We propose to sample short contexts from the full context and estimate the probability an LLM solves the problem using the sampled spans. To find the $\lambda$ and $k$ for each problem, we further propose a mixture model of a non-parametric background noise component and a parametric/non-parametric hybrid oracle component, where we derive the probability functions parameterized by $\lambda$ and $k$ for both the correct-or-wrong (COW) scenario and the partial-point-in-grading (PIG) scenario. Our proposed methods can identify 0% to 67% of the problems are retrieval focused and 0% to 90% of the problems are holistic understanding focused across 44 existing long context evaluation tasks.
Authors: Lennart Keller, Goran Glava\v{s}
Abstract: Recent advancements in multilingual speech encoding as well as transcription raise the question of the most effective approach to semantic speech classification. Concretely, can (1) end-to-end (E2E) classifiers obtained by fine-tuning state-of-the-art multilingual speech encoders (MSEs) match or surpass the performance of (2) cascading (CA), where speech is first transcribed into text and classification is delegated to a text-based classifier. To answer this, we first construct SpeechTaxi, an 80-hour multilingual dataset for semantic speech classification of Bible verses, covering 28 diverse languages. We then leverage SpeechTaxi to conduct a wide range of experiments comparing E2E and CA in monolingual semantic speech classification as well as in cross-lingual transfer. We find that E2E based on MSEs outperforms CA in monolingual setups, i.e., when trained on in-language data. However, MSEs seem to have poor cross-lingual transfer abilities, with E2E substantially lagging CA both in (1) zero-shot transfer to languages unseen in training and (2) multilingual training, i.e., joint training on multiple languages. Finally, we devise a novel CA approach based on transcription to Romanized text as a language-agnostic intermediate representation and show that it represents a robust solution for languages without native ASR support. Our SpeechTaxi dataset is publicly available at: https://huggingface.co/ datasets/LennartKeller/SpeechTaxi/.
URLs: https://huggingface.co/
Authors: Masato Kikuchi, Masatsugu Ono, Toshioki Soga, Tetsu Tanabe, Tadachika Ozono
Abstract: WordNet is one of the largest handcrafted concept dictionaries visualizing word connections through semantic relationships. It is widely used as a word sense inventory in natural language processing tasks. However, WordNet's fine-grained senses have been criticized for limiting its usability. In this paper, we semantically match sense definitions from Cambridge dictionaries and WordNet and develop new coarse-grained sense inventories. We verify the effectiveness of our inventories by comparing their semantic coherences with that of Coarse Sense Inventory. The advantages of the proposed inventories include their low dependency on large-scale resources, better aggregation of closely related senses, CEFR-level assignments, and ease of expansion and improvement.
Authors: Yi-Cheng Wang, Li-Ting Pai, Bi-Cheng Yan, Hsin-Wei Wang, Chi-Han Lin, Berlin Chen
Abstract: End-to-end (E2E) automatic speech recognition (ASR) models have become standard practice for various commercial applications. However, in real-world scenarios, the long-tailed nature of word distribution often leads E2E ASR models to perform well on common words but fall short in recognizing uncommon ones. Recently, the notion of a contextual adapter (CA) was proposed to infuse external knowledge represented by a context word list into E2E ASR models. Although CA can improve recognition performance on rare words, two crucial data imbalance problems remain. First, when using low-frequency words as context words during training, since these words rarely occur in the utterance, CA becomes prone to overfit on attending to the
Authors: Juhwan Choi, YoungBin Kim
Abstract: Large language models (LLMs) have become a dominant approach in natural language processing, yet their internal knowledge structures remain largely unexplored. In this paper, we analyze the internal knowledge structures of LLMs using historical medal tallies from the Olympic Games. We task the models with providing the medal counts for each team and identifying which teams achieved specific rankings. Our results reveal that while state-of-the-art LLMs perform remarkably well in reporting medal counts for individual teams, they struggle significantly with questions about specific rankings. This suggests that the internal knowledge structures of LLMs are fundamentally different from those of humans, who can easily infer rankings from known medal counts. To support further research, we publicly release our code, dataset, and model outputs.
Authors: Jan Elfes
Abstract: Given the profound impact of narratives across various societal levels, from personal identities to international politics, it is crucial to understand their distribution and development over time. This is particularly important in online spaces. On the Web, narratives can spread rapidly and intensify societal divides and conflicts. While many qualitative approaches exist, quantifying narratives remains a significant challenge. Computational narrative analysis lacks frameworks that are both comprehensive and generalizable. To address this gap, we introduce a numerical narrative representation grounded in structuralist linguistic theory. Chiefly, Greimas' Actantial Model represents a narrative through a constellation of six functional character roles. These so-called actants are genre-agnostic, making the model highly generalizable. We extract the actants using an open-source LLM and integrate them into a Narrative-Structured Text Embedding that captures both the semantics and narrative structure of a text. We demonstrate the analytical insights of the method on the example of 5000 full-text news articles from Al Jazeera and The Washington Post on the Israel-Palestine conflict. Our method successfully distinguishes articles that cover the same topics but differ in narrative structure.
Authors: Victor Bocharov, Romaric Besan\c{c}on, Ga\"el de Chalendar, Olivier Ferret, Nasredine Semmar
Abstract: In this article, we describe the architecture of the LIMA (Libre Multilingual Analyzer) framework and its recent evolution with the addition of new text analysis modules based on deep neural networks. We extended the functionality of LIMA in terms of the number of supported languages while preserving existing configurable architecture and the availability of previously developed rule-based and statistical analysis components. Models were trained for more than 60 languages on the Universal Dependencies 2.5 corpora, WikiNer corpora, and CoNLL-03 dataset. Universal Dependencies allowed us to increase the number of supported languages and to generate models that could be integrated into other platforms. This integration of ubiquitous Deep Learning Natural Language Processing models and the use of standard annotated collections using Universal Dependencies can be viewed as a new path of interoperability, through the normalization of models and data, that are complementary to a more standard technical interoperability, implemented in LIMA through services available in Docker containers on Docker Hub.
Authors: Vivi Nastase, Chunyang Jiang, Giuseppe Samo, Paola Merlo
Abstract: In this paper, our goal is to investigate to what degree multilingual pretrained language models capture cross-linguistically valid abstract linguistic representations. We take the approach of developing curated synthetic data on a large scale, with specific properties, and using them to study sentence representations built using pretrained language models. We use a new multiple-choice task and datasets, Blackbird Language Matrices (BLMs), to focus on a specific grammatical structural phenomenon -- subject-verb agreement across a variety of sentence structures -- in several languages. Finding a solution to this task requires a system detecting complex linguistic patterns and paradigms in text representations. Using a two-level architecture that solves the problem in two steps -- detect syntactic objects and their properties in individual sentences, and find patterns across an input sequence of sentences -- we show that despite having been trained on multilingual texts in a consistent manner, multilingual pretrained language models have language-specific differences, and syntactic structure is not shared, even across closely related languages.
Authors: Sacha Muller, Ant\'onio Loison, Bilel Omrani, Gautier Viaud
Abstract: Retrieval-Augmented Generation (RAG) has emerged as a common paradigm to use Large Language Models (LLMs) alongside private and up-to-date knowledge bases. In this work, we address the challenges of using LLM-as-a-Judge when evaluating grounded answers generated by RAG systems. To assess the calibration and discrimination capabilities of judge models, we identify 7 generator failure modes and introduce GroUSE (Grounded QA Unitary Scoring of Evaluators), a meta-evaluation benchmark of 144 unit tests. This benchmark reveals that existing automated RAG evaluation frameworks often overlook important failure modes, even when using GPT-4 as a judge. To improve on the current design of automated RAG evaluation frameworks, we propose a novel pipeline and find that while closed models perform well on GroUSE, state-of-the-art open-source judges do not generalize to our proposed criteria, despite strong correlation with GPT-4's judgement. Our findings suggest that correlation with GPT-4 is an incomplete proxy for the practical performance of judge models and should be supplemented with evaluations on unit tests for precise failure mode detection. We further show that finetuning Llama-3 on GPT-4's reasoning traces significantly boosts its evaluation capabilities, improving upon both correlation with GPT-4's evaluations and calibration on reference situations.
Authors: Yetao Wu, Yihong Wang, Teng Chen, Chenxi Liu, Ningyuan Xi, Qingqing Gu, Hongyang Lei, Zhonglin Jiang, Yong Chen, Luo Ji
Abstract: Hallucinations is a major challenge for large language models (LLMs), prevents adoption in diverse fields. Uncertainty estimation could be used for alleviating the damages of hallucinations. The skeptical emotion of human could be useful for enhancing the ability of self estimation. Inspirited by this observation, we proposed a new approach called Skepticism Modeling (SM). This approach is formalized by combining the information of token and logits for self estimation. We construct the doubt emotion aware data, perform continual pre-training, and then fine-tune the LLMs, improve their ability of self estimation. Experimental results demonstrate this new approach effectively enhances a model's ability to estimate their uncertainty, and validate its generalization ability of other tasks by out-of-domain experiments.
Authors: Vivi Nastase, Giuseppe Samo, Chunyang Jiang, Paola Merlo
Abstract: We investigate to what degree existing LLMs encode abstract linguistic information in Italian in a multi-task setting. We exploit curated synthetic data on a large scale -- several Blackbird Language Matrices (BLMs) problems in Italian -- and use them to study how sentence representations built using pre-trained language models encode specific syntactic and semantic information. We use a two-level architecture to model separately a compression of the sentence embeddings into a representation that contains relevant information for a task, and a BLM task. We then investigate whether we can obtain compressed sentence representations that encode syntactic and semantic information relevant to several BLM tasks. While we expected that the sentence structure -- in terms of sequence of phrases/chunks -- and chunk properties could be shared across tasks, performance and error analysis show that the clues for the different tasks are encoded in different manners in the sentence embeddings, suggesting that abstract linguistic notions such as constituents or thematic roles does not seem to be present in the pretrained sentence embeddings.
Authors: Ningyuan Xi, Yetao Wu, Kun Fan, Teng Chen, Qingqing Gu, Peng Yu, Jinxian Qu, Chenxi Liu, Zhonglin Jiang, Yong Chen, Luo Ji
Abstract: Large Language Models (LLM) often needs to be Continual Pre-Trained (CPT) to obtain the unfamiliar language skill or adapt into new domains. The huge training cost of CPT often asks for cautious choice of key hyper-parameters such as the mixture ratio of extra language or domain corpus. However, there is no systematic study which bridge the gap between the optimal mixture ratio and the actual model performance, and the gap between experimental scaling law and the actual deployment in the full model size. In this paper, we perform CPT on Llama-3 8B and 70B to enhance its Chinese ability. We study the optimal correlation between the Additional Language Mixture Ratio (ALMR) and the Learning Rate (LR) on the 8B size which directly indicate the optimal experimental set up. By thorough choice of hyper-parameter, and subsequent fine-tuning, the model capability is improved not only on the Chinese-related benchmark, but also some specific domains including math, coding and emotional intelligence. We deploy the final 70B version of LLM on an real-life chat system which obtain satisfying performance.
Authors: Kyudan Jung, Nam-Joon Kim, Hyongon Ryu, Sieun Hyeon, Seung-jun Lee, Hyeok-jae Lee
Abstract: LaTeX is highly suited to creating documents with special formatting, particularly in the fields of science, technology, mathematics, and computer science. Despite the increasing use of mathematical expressions in LaTeX format with language models, there are no evaluation metrics for evaluating them. In this study, we propose TeXBLEU, an evaluation metric tailored for mathematical expressions in LaTeX format, based on the n-gram-based BLEU metric that is widely used for translation tasks. The proposed TeXBLEU includes a predefined tokenizer trained on the arXiv paper dataset and a finetuned embedding model. It also considers the positional embedding of tokens. Simultaneously, TeXBLEU compares tokens based on n-grams and computes the score using exponentiation of a logarithmic sum, similar to the original BLEU. Experimental results show that TeXBLEU outperformed traditional evaluation metrics such as BLEU, Rouge, CER, and WER when compared to human evaluation data on the test dataset of the MathBridge dataset, which contains 1,000 data points. The average correlation coefficient with human evaluation was 0.71, which is an improvement of 87% compared with BLEU, which had the highest correlation with human evaluation data among the existing metrics. The code is available at https://github.com/KyuDan1/TeXBLEU.
Authors: Qingkai Fang, Shoutao Guo, Yan Zhou, Zhengrui Ma, Shaolei Zhang, Yang Feng
Abstract: Models like GPT-4o enable real-time interaction with large language models (LLMs) through speech, significantly enhancing user experience compared to traditional text-based interaction. However, there is still a lack of exploration on how to build speech interaction models based on open-source LLMs. To address this, we propose LLaMA-Omni, a novel model architecture designed for low-latency and high-quality speech interaction with LLMs. LLaMA-Omni integrates a pretrained speech encoder, a speech adaptor, an LLM, and a streaming speech decoder. It eliminates the need for speech transcription, and can simultaneously generate text and speech responses directly from speech instructions with extremely low latency. We build our model based on the latest Llama-3.1-8B-Instruct model. To align the model with speech interaction scenarios, we construct a dataset named InstructS2S-200K, which includes 200K speech instructions and corresponding speech responses. Experimental results show that compared to previous speech-language models, LLaMA-Omni provides better responses in both content and style, with a response latency as low as 226ms. Additionally, training LLaMA-Omni takes less than 3 days on just 4 GPUs, paving the way for the efficient development of speech-language models in the future.
Authors: Zihan Liao, Jun Wang, Hang Yu, Lingxiao Wei, Jianguo Li, Jun Wang, Wei Zhang
Abstract: In the realm of Large Language Models (LLMs), the ability to process long contexts is increasingly crucial for tasks such as multi-round dialogues, code generation, and document summarization. This paper addresses the challenges of enhancing the long-context performance, reducing computational complexity, and leveraging pretrained models collectively termed the "impossible triangle." We introduce E2LLM (Encoder Elongated Large Language Models), a novel approach that effectively navigates this paradox. The method involves splitting long contexts into chunks, compressing each into embedding vectors via a pretrained text encoder, and utilizing an adapter to align these representations with a decoder-only LLM. Two training objectives, focusing on reconstruction of the encoder output and long-context instruction fine-tuning, are employed to facilitate the understanding of soft prompts by the LLM. Experimental results demonstrate that E2LLM achieves superior performance in long-context scenarios while balancing efficiency, performance, and compatibility with pretrained models. Our framework thus represents a significant advancement in the field, contributing to effective long-text modeling.
Authors: Bruce W. Lee, Inkit Padhi, Karthikeyan Natesan Ramamurthy, Erik Miehling, Pierre Dognin, Manish Nagireddy, Amit Dhurandhar
Abstract: LLMs have shown remarkable capabilities, but precisely controlling their response behavior remains challenging. Existing activation steering methods alter LLM behavior indiscriminately, limiting their practical applicability in settings where selective responses are essential, such as content moderation or domain-specific assistants. In this paper, we propose Conditional Activation Steering (CAST), which analyzes LLM activation patterns during inference to selectively apply or withhold activation steering based on the input context. Our method is based on the observation that different categories of prompts activate distinct patterns in the model's hidden states. Using CAST, one can systematically control LLM behavior with rules like "if input is about hate speech or adult content, then refuse" or "if input is not about legal advice, then refuse." This allows for selective modification of responses to specific content while maintaining normal responses to other content, all without requiring weight optimization. We release an open-source implementation of our framework.
Authors: Lars-Peter Meyer, Johannes Frey, Felix Brei, Natanael Arndt
Abstract: The integration of Large Language Models (LLMs) with Knowledge Graphs (KGs) offers significant synergistic potential for knowledge-driven applications. One possible integration is the interpretation and generation of formal languages, such as those used in the Semantic Web, with SPARQL being a core technology for accessing KGs. In this paper, we focus on measuring out-of-the box capabilities of LLMs to work with SPARQL and more specifically with SPARQL SELECT queries applying a quantitative approach. We implemented various benchmarking tasks in the LLM-KG-Bench framework for automated execution and evaluation with several LLMs. The tasks assess capabilities along the dimensions of syntax, semantic read, semantic create, and the role of knowledge graph prompt inclusion. With this new benchmarking tasks, we evaluated a selection of GPT, Gemini, and Claude models. Our findings indicate that working with SPARQL SELECT queries is still challenging for LLMs and heavily depends on the specific LLM as well as the complexity of the task. While fixing basic syntax errors seems to pose no problems for the best of the current LLMs evaluated, creating semantically correct SPARQL SELECT queries is difficult in several cases.
Authors: Zahra Khanjani, Tolulope Ale, Jianwu Wang, Lavon Davis, Christine Mallinson, Vandana P. Janeja
Abstract: Several types of spoofed audio, such as mimicry, replay attacks, and deepfakes, have created societal challenges to information integrity. Recently, researchers have worked with sociolinguistics experts to label spoofed audio samples with Expert Defined Linguistic Features (EDLFs) that can be discerned by the human ear: pitch, pause, word-initial and word-final release bursts of consonant stops, audible intake or outtake of breath, and overall audio quality. It is established that there is an improvement in several deepfake detection algorithms when they augmented the traditional and common features of audio data with these EDLFs. In this paper, using a hybrid dataset comprised of multiple types of spoofed audio augmented with sociolinguistic annotations, we investigate causal discovery and inferences between the discernible linguistic features and the label in the audio clips, comparing the findings of the causal models with the expert ground truth validation labeling process. Our findings suggest that the causal models indicate the utility of incorporating linguistic features to help discern spoofed audio, as well as the overall need and opportunity to incorporate human knowledge into models and techniques for strengthening AI models. The causal discovery and inference can be used as a foundation of training humans to discern spoofed audio as well as automating EDLFs labeling for the purpose of performance improvement of the common AI-based spoofed audio detectors.
Authors: Yuan Tian, Jonathan K. Kummerfeld, Toby Jia-Jun Li, Tianyi Zhang
Abstract: Though recent advances in machine learning have led to significant improvements in natural language interfaces for databases, the accuracy and reliability of these systems remain limited, especially in high-stakes domains. This paper introduces SQLucid, a novel user interface that bridges the gap between non-expert users and complex database querying processes. SQLucid addresses existing limitations by integrating visual correspondence, intermediate query results, and editable step-by-step SQL explanations in natural language to facilitate user understanding and engagement. This unique blend of features empowers users to understand and refine SQL queries easily and precisely. Two user studies and one quantitative experiment were conducted to validate SQLucid's effectiveness, showing significant improvement in task completion accuracy and user confidence compared to existing interfaces. Our code is available at https://github.com/magic-YuanTian/SQLucid.
Authors: Wanli Qian, Chenfeng Gao, Anup Sathya, Ryo Suzuki, Ken Nakagaki
Abstract: This paper introduces text-to-shape-display, a novel approach to generating dynamic shape changes in pin-based shape displays through natural language commands. By leveraging large language models (LLMs) and AI-chaining, our approach allows users to author shape-changing behaviors on demand through text prompts without programming. We describe the foundational aspects necessary for such a system, including the identification of key generative elements (primitive, animation, and interaction) and design requirements to enhance user interaction, based on formative exploration and iterative design processes. Based on these insights, we develop SHAPE-IT, an LLM-based authoring tool for a 24 x 24 shape display, which translates the user's textual command into executable code and allows for quick exploration through a web-based control interface. We evaluate the effectiveness of SHAPE-IT in two ways: 1) performance evaluation and 2) user evaluation (N= 10). The study conclusions highlight the ability to facilitate rapid ideation of a wide range of shape-changing behaviors with AI. However, the findings also expose accuracy-related challenges and limitations, prompting further exploration into refining the framework for leveraging AI to better suit the unique requirements of shape-changing systems.
Authors: Jaeseong Lee, seung-won hwang, Aurick Qiao, Daniel F Campos, Zhewei Yao, Yuxiong He
Abstract: Mixture-of-experts (MoEs) have been adopted for reducing inference costs by sparsely activating experts in Large language models (LLMs). Despite this reduction, the massive number of experts in MoEs still makes them expensive to serve. In this paper, we study how to address this, by pruning MoEs. Among pruning methodologies, unstructured pruning has been known to achieve the highest performance for a given pruning ratio, compared to structured pruning, since the latter imposes constraints on the sparsification structure. This is intuitive, as the solution space of unstructured pruning subsumes that of structured pruning. However, our counterintuitive finding reveals that expert pruning, a form of structured pruning, can actually precede unstructured pruning to outperform unstructured-only pruning. As existing expert pruning, requiring $O(\frac{k^n}{\sqrt{n}})$ forward passes for $n$ experts, cannot scale for recent MoEs, we propose a scalable alternative with $O(1)$ complexity, yet outperforming the more expensive methods. The key idea is leveraging a latent structure between experts, based on behavior similarity, such that the greedy decision of whether to prune closely captures the joint pruning effect. Ours is highly effective -- for Snowflake Arctic, a 480B-sized MoE with 128 experts, our method needs only one H100 and two hours to achieve nearly no loss in performance with 40% sparsity, even in generative tasks such as GSM8K, where state-of-the-art unstructured pruning fails to. The code will be made publicly available.
Authors: Arvind Krishna Sridhar, Yinyi Guo, Erik Visser
Abstract: The Audio Question Answering task includes audio event classification, audio captioning, and open ended reasoning. Recently, Audio Question Answering has garnered attention due to the advent of Large Audio Language Models. Current literature focuses on constructing LALMs by integrating audio encoders with text only Large Language Models through a projection module. While Large Audio Language Models excel in general audio understanding, they are limited in temporal reasoning which may hinder their commercial applications and on device deployment. This paper addresses these challenges and limitations in audio temporal reasoning. First, we introduce a data augmentation technique for generating reliable audio temporal questions and answers using an LLM. Second, we propose a continued finetuning curriculum learning strategy to specialize in temporal reasoning without compromising performance on finetuned tasks. Finally, we develop a reliable and transparent automated metric, assisted by an LLM, to measure the correlation between Large Audio Language Model responses and ground truth data intelligently. We demonstrate the effectiveness of our proposed techniques using SOTA LALMs on public audio benchmark datasets.
Authors: Weicong Qin, Yi Xu, Weijie Yu, Chenglei Shen, Xiao Zhang, Ming He, Jianping Fan, Jun Xu
Abstract: Sequence recommendation (SeqRec) aims to predict the next item a user will interact with by understanding user intentions and leveraging collaborative filtering information. Large language models (LLMs) have shown great promise in recommendation tasks through prompt-based, fixed reflection libraries, and fine-tuning techniques. However, these methods face challenges, including lack of supervision, inability to optimize reflection sources, inflexibility to diverse user needs, and high computational costs. Despite promising results, current studies primarily focus on reflections of users' explicit preferences (e.g., item titles) while neglecting implicit preferences (e.g., brands) and collaborative filtering information. This oversight hinders the capture of preference shifts and dynamic user behaviors. Additionally, existing approaches lack mechanisms for reflection evaluation and iteration, often leading to suboptimal recommendations. To address these issues, we propose the Mixture of REflectors (MoRE) framework, designed to model and learn dynamic user preferences in SeqRec. Specifically, MoRE introduces three reflectors for generating LLM-based reflections on explicit preferences, implicit preferences, and collaborative signals. Each reflector incorporates a self-improving strategy, termed refining-and-iteration, to evaluate and iteratively update reflections. Furthermore, a meta-reflector employs a contextual bandit algorithm to select the most suitable expert and corresponding reflections for each user's recommendation, effectively capturing dynamic preferences. Extensive experiments on three real-world datasets demonstrate that MoRE consistently outperforms state-of-the-art methods, requiring less training time and GPU memory compared to other LLM-based approaches in SeqRec.
Authors: Wei Liu, Yang Bai, Chengcheng Han, Rongxiang Weng, Jun Xu, Xuezhi Cao, Jingang Wang, Xunliang Cai
Abstract: Direct Preference Optimization (DPO) is widely utilized in the Reinforcement Learning from Human Feedback (RLHF) phase to align Large Language Models (LLMs) with human preferences, thereby enhancing both their harmlessness and efficacy. However, it has been observed that DPO tends to over-optimize for verbosity, which can detrimentally affect both performance and user experience. In this paper, we conduct an in-depth theoretical analysis of DPO's optimization objective and reveal a strong correlation between its implicit reward and data length. This correlation misguides the optimization direction, resulting in length sensitivity during the DPO training and leading to verbosity. To address this issue, we propose a length-desensitization improvement method for DPO, termed LD-DPO. The proposed method aims to desensitize DPO to data length by decoupling explicit length preference, which is relatively insignificant, from the other implicit preferences, thereby enabling more effective learning of the intrinsic preferences. We utilized two settings (Base and Instruct) of Llama2-13B, Llama3-8B, and Qwen2-7B for experimental validation on various benchmarks including MT-Bench and AlpacaEval 2. The experimental results indicate that LD-DPO consistently outperforms DPO and other baseline methods, achieving more concise responses with a 10-40\% reduction in length compared to DPO. We conducted in-depth experimental analyses to demonstrate that LD-DPO can indeed achieve length desensitization and align the model more closely with human-real preferences.
Authors: Hossein Hajipour, Lea Sch\"onherr, Thorsten Holz, Mario Fritz
Abstract: Large language models (LLMs) have shown great potential for automatic code generation and form the basis for various tools such as GitHub Copilot. However, recent studies highlight that many LLM-generated code contains serious security vulnerabilities. While previous work tries to address this by training models that generate secure code, these attempts remain constrained by limited access to training data and labor-intensive data preparation. In this paper, we introduce HexaCoder, a novel approach to enhance the ability of LLMs to generate secure codes by automatically synthesizing secure codes, which reduces the effort of finding suitable training data. HexaCoder comprises two key components: an oracle-guided data synthesis pipeline and a two-step process for secure code generation. The data synthesis pipeline generates pairs of vulnerable and fixed codes for specific Common Weakness Enumeration (CWE) types by utilizing a state-of-the-art LLM for repairing vulnerable code. A security oracle identifies vulnerabilities, and a state-of-the-art LLM repairs them by extending and/or editing the codes, creating data pairs for fine-tuning using the Low-Rank Adaptation (LoRA) method. Each example of our fine-tuning dataset includes the necessary security-related libraries and code that form the basis of our novel two-step generation approach. This allows the model to integrate security-relevant libraries before generating the main code, significantly reducing the number of generated vulnerable codes by up to 85% compared to the baseline methods. We perform extensive evaluations on three different benchmarks for four LLMs, demonstrating that HexaCoder not only improves the security of the generated code but also maintains a high level of functional correctness.
Authors: Wenyu Zhang, Shuo Sun, Bin Wang, Xunlong Zou, Zhuohan Liu, Yingxu He, Geyu Lin, Nancy F. Chen, Ai Ti Aw
Abstract: The rapid advancements in large language models (LLMs) have significantly enhanced natural language processing capabilities, facilitating the development of AudioLLMs that process and understand speech and audio inputs alongside text. Existing AudioLLMs typically combine a pre-trained audio encoder with a pre-trained LLM, which are subsequently finetuned on specific audio tasks. However, the pre-trained audio encoder has constrained capacity to capture features for new tasks and datasets. To address this, we propose to incorporate mixtures of `weak' encoders (MoWE) into the AudioLLM framework. MoWE supplements a base encoder with a pool of relatively light weight encoders, selectively activated based on the audio input to enhance feature extraction without significantly increasing model size. Our empirical results demonstrate that MoWE effectively improves multi-task performance, broadening the applicability of AudioLLMs to more diverse audio tasks.
Authors: Taejin Park, Ivan Medennikov, Kunal Dhawan, Weiqing Wang, He Huang, Nithin Rao Koluguri, Krishna C. Puvvada, Jagadeesh Balam, Boris Ginsburg
Abstract: We propose Sortformer, a novel neural model for speaker diarization, trained with unconventional objectives compared to existing end-to-end diarization models. The permutation problem in speaker diarization has long been regarded as a critical challenge. Most prior end-to-end diarization systems employ permutation invariant loss (PIL), which optimizes for the permutation that yields the lowest error. In contrast, we introduce Sort Loss, which enables a diarization model to autonomously resolve permutation, with or without PIL. We demonstrate that combining Sort Loss and PIL achieves performance competitive with state-of-the-art end-to-end diarization models trained exclusively with PIL. Crucially, we present a streamlined multispeaker ASR architecture that leverages Sortformer as a speaker supervision model, embedding speaker label estimation within the ASR encoder state using a sinusoidal kernel function. This approach resolves the speaker permutation problem through sorted objectives, effectively bridging speaker-label timestamps and speaker tokens. In our experiments, we show that the proposed multispeaker ASR architecture, enhanced with speaker supervision, improves performance via adapter techniques. Code and trained models will be made publicly available via the NVIDIA NeMo framework
Authors: Hiroki Furuta, Kuang-Huei Lee, Shixiang Shane Gu, Yutaka Matsuo, Aleksandra Faust, Heiga Zen, Izzeddin Gur
Abstract: Many algorithms for aligning LLMs with human preferences assume that human preferences are binary and deterministic. However, it is reasonable to think that they can vary with different individuals, and thus should be distributional to reflect the fine-grained relationship between the responses. In this work, we introduce the distributional soft preference labels and improve Direct Preference Optimization (DPO) with a weighted geometric average of the LLM output likelihood in the loss function. In doing so, the scale of learning loss is adjusted based on the soft labels, and the loss with equally preferred responses would be close to zero. This simple modification can be easily applied to any DPO family and helps the models escape from the over-optimization and objective mismatch prior works suffer from. In our experiments, we simulate the soft preference labels with AI feedback from LLMs and demonstrate that geometric averaging consistently improves performance on standard benchmarks for alignment research. In particular, we observe more preferable responses than binary labels and significant improvements with data where modestly-confident labels are in the majority.
Authors: Daniel Fern\'andez-Gonz\'alez
Abstract: Intelligent voice assistants, such as Apple Siri and Amazon Alexa, are widely used nowadays. These task-oriented dialogue systems require a semantic parsing module in order to process user utterances and understand the action to be performed. This semantic parsing component was initially implemented by rule-based or statistical slot-filling approaches for processing simple queries; however, the appearance of more complex utterances demanded the application of shift-reduce parsers or sequence-to-sequence models. Although shift-reduce approaches were initially considered the most promising option, the emergence of sequence-to-sequence neural systems has propelled them to the forefront as the highest-performing method for this particular task. In this article, we advance the research on shift-reduce semantic parsing for task-oriented dialogue. We implement novel shift-reduce parsers that rely on Stack-Transformers. This framework allows to adequately model transition systems on the Transformer neural architecture, notably boosting shift-reduce parsing performance. Furthermore, our approach goes beyond the conventional top-down algorithm: we incorporate alternative bottom-up and in-order transition systems derived from constituency parsing into the realm of task-oriented parsing. We extensively test our approach on multiple domains from the Facebook TOP benchmark, improving over existing shift-reduce parsers and state-of-the-art sequence-to-sequence models in both high-resource and low-resource settings. We also empirically prove that the in-order algorithm substantially outperforms the commonly-used top-down strategy. Through the creation of innovative transition systems and harnessing the capabilities of a robust neural architecture, our study showcases the superiority of shift-reduce parsers over leading sequence-to-sequence methods on the main benchmark.
Authors: Mehedi Hasan Bijoy, Nahid Hossain, Salekul Islam, Swakkhar Shatabda
Abstract: Spelling error correction is the task of identifying and rectifying misspelled words in texts. It is a potential and active research topic in Natural Language Processing because of numerous applications in human language understanding. The phonetically or visually similar yet semantically distinct characters make it an arduous task in any language. Earlier efforts on spelling error correction in Bangla and resource-scarce Indic languages focused on rule-based, statistical, and machine learning-based methods which we found rather inefficient. In particular, machine learning-based approaches, which exhibit superior performance to rule-based and statistical methods, are ineffective as they correct each character regardless of its appropriateness. In this paper, we propose a novel detector-purificator-corrector framework, DPCSpell based on denoising transformers by addressing previous issues. In addition to that, we present a method for large-scale corpus creation from scratch which in turn resolves the resource limitation problem of any left-to-right scripted language. The empirical outcomes demonstrate the effectiveness of our approach, which outperforms previous state-of-the-art methods by attaining an exact match (EM) score of 94.78%, a precision score of 0.9487, a recall score of 0.9478, an f1 score of 0.948, an f0.5 score of 0.9483, and a modified accuracy (MA) score of 95.16% for Bangla spelling error correction. The models and corpus are publicly available at https://tinyurl.com/DPCSpell.
Authors: Mengjie Qian, Rao Ma, Adian Liusie, Erfan Loweimi, Kate M. Knill, Mark J. F. Gales
Abstract: Multimodal Video Search by Examples (MVSE) investigates using video clips as the query term for information retrieval, rather than the more traditional text query. This enables far richer search modalities such as images, speaker, content, topic, and emotion. A key element for this process is highly rapid and flexible search to support large archives, which in MVSE is facilitated by representing video attributes with embeddings. This work aims to compensate for any performance loss from this rapid archive search by examining reranking approaches. In particular, zero-shot reranking methods using large language models (LLMs) are investigated as these are applicable to any video archive audio content. Performance is evaluated for topic-based retrieval on a publicly available video archive, the BBC Rewind corpus. Results demonstrate that reranking significantly improves retrieval ranking without requiring any task-specific in-domain training data. Furthermore, three sources of information (ASR transcriptions, automatic summaries and synopses) as input for LLM reranking were compared. To gain a deeper understanding and further insights into the performance differences and limitations of these text sources, we employ a fact-checking approach to analyse the information consistency among them.
Authors: Matthew Renze, Erhan Guven
Abstract: In this paper, we introduce Concise Chain-of-Thought (CCoT) prompting. We compared standard CoT and CCoT prompts to see how conciseness impacts response length and correct-answer accuracy. We evaluated this using GPT-3.5 and GPT-4 with a multiple-choice question-and-answer (MCQA) benchmark. CCoT reduced average response length by 48.70% for both GPT-3.5 and GPT-4 while having a negligible impact on problem-solving performance. However, on math problems, GPT-3.5 with CCoT incurs a performance penalty of 27.69%. Overall, CCoT leads to an average per-token cost reduction of 22.67%.
Authors: Raymond Douglas, Andis Draguns, Tom\'a\v{s} Gaven\v{c}iak
Abstract: The broad capabilities of Language Models (LMs) can be limited by their sensitivity to distractor tasks: LMs can infer secondary tasks from the prompt in addition to the intended one, leading to unwanted outputs. For example, prompt injection attacks can cause models to deviate from explicit directives. In some 'inverse scaling' cases, this unwanted behaviour actually worsens as models scale up to at least 540B parameters. We present a theoretical framework that interprets LMs as a product of experts that combine multiple data generation processes. Based on this framework, we demonstrate prior-aware decoding (PAD) - a simple contrastive inference method to reduce the influence of distractor tasks. We apply PAD to eleven models, across four datasets, and find improvements in 41 out of 44 task-model combinations, with a median increase in task completion proportion of 40%. The results suggest a promising direction for further development towards more reliable language models.
Authors: Ga\"el Gendron, Bao Trung Nguyen, Alex Yuxuan Peng, Michael Witbrock, Gillian Dobbie
Abstract: Despite impressive performance on language modelling and complex reasoning tasks, Large Language Models (LLMs) fall short on the same tasks in uncommon settings or with distribution shifts, exhibiting a lack of generalisation ability. By contrast, systems such as causal models, that learn abstract variables and causal relationships, can demonstrate increased robustness against changes in the distribution. One reason for this success is the existence and use of Independent Causal Mechanisms (ICMs) representing high-level concepts that only sparsely interact. In this work, we apply two concepts from causality to learn ICMs within LLMs. We develop a new LLM architecture composed of multiple sparsely interacting language modelling modules. We show that such causal constraints can improve out-of-distribution performance on abstract and causal reasoning tasks. We also investigate the level of independence and domain specialisation and show that LLMs rely on pre-trained partially domain-invariant mechanisms resilient to fine-tuning.
Authors: Lin Ning, Luyang Liu, Jiaxing Wu, Neo Wu, Devora Berlowitz, Sushant Prakash, Bradley Green, Shawn O'Banion, Jun Xie
Abstract: Large language models (LLMs) have achieved remarkable success across various domains, but effectively incorporating complex and potentially noisy user timeline data into LLMs remains a challenge. Current approaches often involve translating user timelines into text descriptions before feeding them to LLMs, which can be inefficient and may not fully capture the nuances of user behavior. Inspired by how LLMs are effectively integrated with images through direct embeddings, we propose User-LLM, a novel framework that leverages user embeddings to directly contextualize LLMs with user history interactions. These embeddings, generated by a user encoder pretrained using self-supervised learning on diverse user interactions, capture latent user behaviors and interests as well as their evolution over time. We integrate these user embeddings with LLMs through cross-attention, enabling LLMs to dynamically adapt their responses based on the context of a user's past actions and preferences. Our approach achieves significant efficiency gains by representing user timelines directly as embeddings, leading to substantial inference speedups of up to 78.1X. Comprehensive experiments on MovieLens, Amazon Review, and Google Local Review datasets demonstrate that User-LLM outperforms text-prompt-based contextualization on tasks requiring deep user understanding, with improvements of up to 16.33%, particularly excelling on long sequences that capture subtle shifts in user behavior. Furthermore, the incorporation of Perceiver layers streamlines the integration between user encoders and LLMs, yielding additional computational savings.
Authors: Md Abrar Jahin, Md Sakib Hossain Shovon, M. F. Mridha, Md Rashedul Islam, Yutaka Watanobe
Abstract: Sentiment analysis is crucial for understanding public opinion and consumer behavior. Existing models face challenges with linguistic diversity, generalizability, and explainability. We propose TRABSA, a hybrid framework integrating transformer-based architectures, attention mechanisms, and BiLSTM networks to address this. Leveraging RoBERTa-trained on 124M tweets, we bridge gaps in sentiment analysis benchmarks, ensuring state-of-the-art accuracy. Augmenting datasets with tweets from 32 countries and US states, we compare six word-embedding techniques and three lexicon-based labeling techniques, selecting the best for optimal sentiment analysis. TRABSA outperforms traditional ML and deep learning models with 94% accuracy and significant precision, recall, and F1-score gains. Evaluation across diverse datasets demonstrates consistent superiority and generalizability. SHAP and LIME analyses enhance interpretability, improving confidence in predictions. Our study facilitates pandemic resource management, aiding resource planning, policy formation, and vaccination tactics.
Authors: Pu Li, Xiaoyan Yu, Hao Peng, Yantuan Xian, Linqin Wang, Li Sun, Jingyun Zhang, Philip S. Yu
Abstract: Social Event Detection (SED) aims to identify significant events from social streams, and has a wide application ranging from public opinion analysis to risk management. In recent years, Graph Neural Network (GNN) based solutions have achieved state-of-the-art performance. However, GNN-based methods often struggle with missing and noisy edges between messages, affecting the quality of learned message embedding. Moreover, these methods statically initialize node embedding before training, which, in turn, limits the ability to learn from message texts and relations simultaneously. In this paper, we approach social event detection from a new perspective based on Pre-trained Language Models (PLMs), and present RPLM_SED (Relational prompt-based Pre-trained Language Models for Social Event Detection). We first propose a new pairwise message modeling strategy to construct social messages into message pairs with multi-relational sequences. Secondly, a new multi-relational prompt-based pairwise message learning mechanism is proposed to learn more comprehensive message representation from message pairs with multi-relational prompts using PLMs. Thirdly, we design a new clustering constraint to optimize the encoding process by enhancing intra-cluster compactness and inter-cluster dispersion, making the message representation more distinguishable. We evaluate the RPLM_SED on three real-world datasets, demonstrating that the RPLM_SED model achieves state-of-the-art performance in offline, online, low-resource, and long-tail distribution scenarios for social event detection tasks.
Authors: Yilong Xu, Jinhua Gao, Xiaoming Yu, Baolong Bi, Huawei Shen, Xueqi Cheng
Abstract: Large Language Models (LLMs) can enhance the credibility and verifiability by generating text with citations. However, existing tasks and evaluation methods are predominantly limited to sentence-level statement, neglecting the significance of positional fine-grained citations that can appear anywhere within sentences. To facilitate further exploration of the fine-grained citation generation, we propose ALiiCE, the first automatic evaluation framework for this task. Our framework first parses the sentence claim into atomic claims via dependency analysis and then calculates citation quality at the atomic claim level. ALiiCE introduces three novel metrics for positional fined-grained citation quality assessment, including positional fine-grained citation recall and precision, and coefficient of variation of citation positions. We evaluate the positional fine-grained citation generation performance of several LLMs on two long-form QA datasets. Our experiments and analyses demonstrate the effectiveness and reasonableness of ALiiCE. The results also indicate that existing LLMs still struggle to provide positional fine-grained citations.
Authors: An Yang, Baosong Yang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Zhou, Chengpeng Li, Chengyuan Li, Dayiheng Liu, Fei Huang, Guanting Dong, Haoran Wei, Huan Lin, Jialong Tang, Jialin Wang, Jian Yang, Jianhong Tu, Jianwei Zhang, Jianxin Ma, Jianxin Yang, Jin Xu, Jingren Zhou, Jinze Bai, Jinzheng He, Junyang Lin, Kai Dang, Keming Lu, Keqin Chen, Kexin Yang, Mei Li, Mingfeng Xue, Na Ni, Pei Zhang, Peng Wang, Ru Peng, Rui Men, Ruize Gao, Runji Lin, Shijie Wang, Shuai Bai, Sinan Tan, Tianhang Zhu, Tianhao Li, Tianyu Liu, Wenbin Ge, Xiaodong Deng, Xiaohuan Zhou, Xingzhang Ren, Xinyu Zhang, Xipin Wei, Xuancheng Ren, Xuejing Liu, Yang Fan, Yang Yao, Yichang Zhang, Yu Wan, Yunfei Chu, Yuqiong Liu, Zeyu Cui, Zhenru Zhang, Zhifang Guo, Zhihao Fan
Abstract: This report introduces the Qwen2 series, the latest addition to our large language models and large multimodal models. We release a comprehensive suite of foundational and instruction-tuned language models, encompassing a parameter range from 0.5 to 72 billion, featuring dense models and a Mixture-of-Experts model. Qwen2 surpasses most prior open-weight models, including its predecessor Qwen1.5, and exhibits competitive performance relative to proprietary models across diverse benchmarks on language understanding, generation, multilingual proficiency, coding, mathematics, and reasoning. The flagship model, Qwen2-72B, showcases remarkable performance: 84.2 on MMLU, 37.9 on GPQA, 64.6 on HumanEval, 89.5 on GSM8K, and 82.4 on BBH as a base language model. The instruction-tuned variant, Qwen2-72B-Instruct, attains 9.1 on MT-Bench, 48.1 on Arena-Hard, and 35.7 on LiveCodeBench. Moreover, Qwen2 demonstrates robust multilingual capabilities, proficient in approximately 30 languages, spanning English, Chinese, Spanish, French, German, Arabic, Russian, Korean, Japanese, Thai, Vietnamese, and more, underscoring its versatility and global reach. To foster community innovation and accessibility, we have made the Qwen2 model weights openly available on Hugging Face and ModelScope, and the supplementary materials including example code on GitHub. These platforms also include resources for quantization, fine-tuning, and deployment, facilitating a wide range of applications and research endeavors.
Authors: Jayr Pereira, Andre Assumpcao, Roberto Lotufo
Abstract: Evaluating the quality of text generated by large language models (LLMs) remains a significant challenge. Traditional metrics often fail to align well with human judgments, particularly in tasks requiring creativity and nuance. In this paper, we propose \textsc{Check-Eval}, a novel evaluation framework leveraging LLMs to assess the quality of generated text through a checklist-based approach. \textsc{Check-Eval} can be employed as both a reference-free and reference-dependent evaluation method, providing a structured and interpretable assessment of text quality. The framework consists of two main stages: checklist generation and checklist evaluation. We validate \textsc{Check-Eval} on two benchmark datasets: Portuguese Legal Semantic Textual Similarity and \textsc{SummEval}. Our results demonstrate that \textsc{Check-Eval} achieves higher correlations with human judgments compared to existing metrics, such as \textsc{G-Eval} and \textsc{GPTScore}, underscoring its potential as a more reliable and effective evaluation framework for natural language generation tasks. The code for our experiments is available at \url{https://anonymous.4open.science/r/check-eval-0DB4}
Authors: Chuhan Wu, Ruiming Tang
Abstract: Guided by the belief of the scaling law, large language models (LLMs) have achieved impressive performance in recent years. However, scaling law only gives a qualitative estimation of loss, which is influenced by various factors such as model architectures, data distributions, tokenizers, and computation precision. Thus, estimating the real performance of LLMs with different training settings rather than loss may be quite useful in practical development. In this article, we present an empirical equation named "Performance Law" to directly predict the MMLU score of an LLM, which is a widely used metric to indicate the general capability of LLMs in real-world conversations and applications. Based on only a few key hyperparameters of the LLM architecture and the size of training data, we obtain a quite accurate MMLU prediction of various LLMs with diverse sizes and architectures developed by different organizations in different years. Performance law can be used to guide the choice of LLM architecture and the effective allocation of computational resources without extensive experiments.
Authors: Abu Ubaida Akash, Ahmed Fahmy, Amine Trabelsi
Abstract: Stance detection (SD) assesses a text's position towards a target, typically labeled as "favor," "against," or "neutral." We introduce Open-Target Stance Detection (OTSD), where targets are neither seen during training nor provided as input. Evaluating Large Language Models (LLMs) like GPT-3.5, GPT-4o, Llama 3, and Mistral, we compare their performance with the Target-Stance Extraction (TSE) approach, which has the advantage of using predefined targets. LLMs perform better than TSE in target generation when the real target is explicitly and not explicitly mentioned in the text. For stance detection, LLMs perform better in explicit scenarios but fail in non-explicit ones.
Authors: Ike Ebubechukwu, Johane Takeuchi, Antonello Ceravola, Frank Joublin
Abstract: As dialogue systems and chatbots increasingly integrate into everyday interactions, the need for efficient and accurate evaluation methods becomes paramount. This study explores the comparative performance of human and AI assessments across a range of dialogue scenarios, focusing on seven key performance indicators (KPIs): Coherence, Innovation, Concreteness, Goal Contribution, Commonsense Contradiction, Incorrect Fact, and Redundancy. Utilizing the GPT-4o API, we generated a diverse dataset of conversations and conducted a two-part experimental analysis. In Experiment 1, we evaluated multi-party conversations on Coherence, Innovation, Concreteness, and Goal Contribution, revealing that GPT models align closely with human judgments. Notably, both human and AI evaluators exhibited a tendency towards binary judgment rather than linear scaling, highlighting a shared challenge in these assessments. Experiment 2 extended the work of Finch et al. (2023) by focusing on dyadic dialogues and assessing Commonsense Contradiction, Incorrect Fact, and Redundancy. The results indicate that while GPT-4o demonstrates strong performance in maintaining factual accuracy and commonsense reasoning, it still struggles with reducing redundancy and self-contradiction. Our findings underscore the potential of GPT models to closely replicate human evaluation in dialogue systems, while also pointing to areas for improvement. This research offers valuable insights for advancing the development and implementation of more refined dialogue evaluation methodologies, contributing to the evolution of more effective and human-like AI communication tools.
Authors: Yuhao Wu, Ming Shan Hee, Zhiqing Hu, Roy Ka-Wei Lee
Abstract: The abilities of long-context language models (LMs) are often evaluated using the "Needle-in-a-Haystack" (NIAH) test, which comprises tasks designed to assess a model's ability to identify specific information ("needle") within large text sequences ("haystack"). While these benchmarks measure how well models understand long-context input sequences, they do not effectively gauge the quality of long-form text generation--a critical aspect for applications such as design proposals and creative writing. To address this gap, we have introduced a new long-form text evaluation benchmark, Spinning the Golden Thread (SGT), which tests models' ability to identify specific events within generated long text sequences. In this benchmark, we prompt long-context LMs to create long-form text that must include particular events or constraints and evaluate their ability to incorporate these elements. We evaluated ten long-context LMs across four distinct scenarios, three types of prompt instructions, and two different generation-length settings (16K and 32K). Although these models perform well on NIAH benchmarks, none demonstrated satisfactory performance on the Spinning the Golden Thread, raising concerns about their ability to generate coherent long-form text that follows instructions. Additionally, as the length of the generated text increases, all models exhibit a significant drop in performance.
Authors: Xiang Yue, Tianyu Zheng, Yuansheng Ni, Yubo Wang, Kai Zhang, Shengbang Tong, Yuxuan Sun, Botao Yu, Ge Zhang, Huan Sun, Yu Su, Wenhu Chen, Graham Neubig
Abstract: This paper introduces MMMU-Pro, a robust version of the Massive Multi-discipline Multimodal Understanding and Reasoning (MMMU) benchmark. MMMU-Pro rigorously assesses multimodal models' true understanding and reasoning capabilities through a three-step process based on MMMU: (1) filtering out questions answerable by text-only models, (2) augmenting candidate options, and (3) introducing a vision-only input setting where questions are embedded within images. This setting challenges AI to truly "see" and "read" simultaneously, testing a fundamental human cognitive skill of seamlessly integrating visual and textual information. Results show that model performance is substantially lower on MMMU-Pro than on MMMU, ranging from 16.8% to 26.9% across models. We explore the impact of OCR prompts and Chain of Thought (CoT) reasoning, finding that OCR prompts have minimal effect while CoT generally improves performance. MMMU-Pro provides a more rigorous evaluation tool, closely mimicking real-world scenarios and offering valuable directions for future research in multimodal AI.
Authors: Jiajie Zhang, Yushi Bai, Xin Lv, Wanjun Gu, Danqing Liu, Minhao Zou, Shulin Cao, Lei Hou, Yuxiao Dong, Ling Feng, Juanzi Li
Abstract: Though current long-context large language models (LLMs) have demonstrated impressive capacities in answering user questions based on extensive text, the lack of citations in their responses makes user verification difficult, leading to concerns about their trustworthiness due to their potential hallucinations. In this work, we aim to enable long-context LLMs to generate responses with fine-grained sentence-level citations, improving their faithfulness and verifiability. We first introduce LongBench-Cite, an automated benchmark for assessing current LLMs' performance in Long-Context Question Answering with Citations (LQAC), revealing considerable room for improvement. To this end, we propose CoF (Coarse to Fine), a novel pipeline that utilizes off-the-shelf LLMs to automatically generate long-context QA instances with precise sentence-level citations, and leverage this pipeline to construct LongCite-45k, a large-scale SFT dataset for LQAC. Finally, we train LongCite-8B and LongCite-9B using the LongCite-45k dataset, successfully enabling their generation of accurate responses and fine-grained sentence-level citations in a single output. The evaluation results on LongBench-Cite show that our trained models achieve state-of-the-art citation quality, surpassing advanced proprietary models including GPT-4o.
Authors: Inacio Vieira, Will Allred, S\'eamus Lankford, Sheila Castilho, Andy Way
Abstract: Decoder-only LLMs have shown impressive performance in MT due to their ability to learn from extensive datasets and generate high-quality translations. However, LLMs often struggle with the nuances and style required for organisation-specific translation. In this study, we explore the effectiveness of fine-tuning Large Language Models (LLMs), particularly Llama 3 8B Instruct, leveraging translation memories (TMs), as a valuable resource to enhance accuracy and efficiency. We investigate the impact of fine-tuning the Llama 3 model using TMs from a specific organisation in the software sector. Our experiments cover five translation directions across languages of varying resource levels (English to Brazilian Portuguese, Czech, German, Finnish, and Korean). We analyse diverse sizes of training datasets (1k to 207k segments) to evaluate their influence on translation quality. We fine-tune separate models for each training set and evaluate their performance based on automatic metrics, BLEU, chrF++, TER, and COMET. Our findings reveal improvement in translation performance with larger datasets across all metrics. On average, BLEU and COMET scores increase by 13 and 25 points, respectively, on the largest training set against the baseline model. Notably, there is a performance deterioration in comparison with the baseline model when fine-tuning on only 1k and 2k examples; however, we observe a substantial improvement as the training dataset size increases. The study highlights the potential of integrating TMs with LLMs to create bespoke translation models tailored to the specific needs of businesses, thus enhancing translation quality and reducing turn-around times. This approach offers a valuable insight for organisations seeking to leverage TMs and LLMs for optimal translation outcomes, especially in narrower domains.
Authors: Arnon Turetzky, Yossi Adi
Abstract: Speech tokenization serves as the foundation of speech language model (LM), enabling them to perform various tasks such as spoken language modeling, text-to-speech, speech-to-text, etc. Most speech tokenizers are trained independently of the LM training process, relying on separate acoustic models and quantization methods. Following such an approach may create a mismatch between the tokenization process and its usage afterward. In this study, we propose a novel approach to training a speech tokenizer by leveraging objectives from pre-trained textual LMs. We advocate for the integration of this objective into the process of learning discrete speech representations. Our aim is to transform features from a pre-trained speech model into a new feature space that enables better clustering for speech LMs. We empirically investigate the impact of various model design choices, including speech vocabulary size and text LM size. Our results demonstrate the proposed tokenization method outperforms the evaluated baselines considering both spoken language modeling and speech-to-text. More importantly, unlike prior work, the proposed method allows the utilization of a single pre-trained LM for processing both speech and text inputs, setting it apart from conventional tokenization approaches.
Authors: Hong Xingyun Hong, Shao Yan Shao, Wang Zhilin Wang, Duan Manni Duan, Jin Xiongnan
Abstract: The development of LLMs has greatly enhanced the intelligence and fluency of question answering, while the emergence of retrieval enhancement has enabled models to better utilize external information. However, the presence of noise and errors in retrieved information poses challenges to the robustness of LLMs. In this work, to evaluate the model's performance under multiple interferences, we first construct a dataset based on machine reading comprehension datasets simulating various scenarios, including critical information absence, noise, and conflicts. To address the issue of model accuracy decline caused by noisy external information, we propose a data augmentation-based fine-tuning method to enhance LLM's robustness against noise. Additionally, contrastive learning approach is utilized to preserve the model's discrimination capability of external information. We have conducted experiments on both existing LLMs and our approach, the results are evaluated by GPT-4, which indicates that our proposed methods improve model robustness while strengthening the model's discrimination capability.
Authors: Hongjin Qian, Peitian Zhang, Zheng Liu, Kelong Mao, Zhicheng Dou
Abstract: Retrieval-Augmented Generation (RAG) leverages retrieval tools to access external databases, thereby enhancing the generation quality of large language models (LLMs) through optimized context. However, the existing retrieval methods are constrained inherently, as they can only perform relevance matching between explicitly stated queries and well-formed knowledge, but unable to handle tasks involving ambiguous information needs or unstructured knowledge. Consequently, existing RAG systems are primarily effective for straightforward question-answering tasks. In this work, we propose MemoRAG, a novel retrieval-augmented generation paradigm empowered by long-term memory. MemoRAG adopts a dual-system architecture. On the one hand, it employs a light but long-range LLM to form the global memory of database. Once a task is presented, it generates draft answers, cluing the retrieval tools to locate useful information within the database. On the other hand, it leverages an expensive but expressive LLM, which generates the ultimate answer based on the retrieved information. Building on this general framework, we further optimize MemoRAG's performance by enhancing its cluing mechanism and memorization capacity. In our experiment, MemoRAG achieves superior performance across a variety of evaluation tasks, including both complex ones where conventional RAG fails and straightforward ones where RAG is commonly applied.
Authors: Tianlong Li, Wenhao Liu, Changze Lv, Yufei Gu, Jianhan Xu, Cenyuan Zhang, Muling Wu, Xiaoqing Zheng, Xuanjing Huang
Abstract: Spiking Neural Networks (SNNs) have emerged as a promising alternative to conventional Artificial Neural Networks (ANNs), demonstrating comparable performance in both visual and linguistic tasks while offering the advantage of improved energy efficiency. Despite these advancements, the integration of linguistic and visual features into a unified representation through spike trains poses a significant challenge, and the application of SNNs to multimodal scenarios remains largely unexplored. This paper presents SpikeCLIP, a novel framework designed to bridge the modality gap in spike-based computation. Our approach employs a two-step recipe: an ``alignment pre-training'' to align features across modalities, followed by a ``dual-loss fine-tuning'' to refine the model's performance. Extensive experiments reveal that SNNs achieve results on par with ANNs while substantially reducing energy consumption across various datasets commonly used for multimodal model evaluation. Furthermore, SpikeCLIP maintains robust image classification capabilities, even when dealing with classes that fall outside predefined categories. This study marks a significant advancement in the development of energy-efficient and biologically plausible multimodal learning systems.
Authors: Aaron Haag, Vlad Argatu, Oliver Lohse
Abstract: Large Language Models (LLMs) have achieved impressive performance in text understanding and have become an essential tool for building smart assistants. Originally focusing on text, they have been enhanced with multimodal capabilities in recent works that successfully built visual instruction following assistants. As far as the graph modality goes, however, no such assistants have yet been developed. Graph structures are complex in that they represent relation between different features and are permutation invariant. Moreover, representing them in purely textual form does not always lead to good LLM performance even for finetuned models. As a result, there is a need to develop a new method to integrate graphs in LLMs for general graph understanding. This work explores the integration of the graph modality in LLM for general graph instruction following tasks. It aims at producing a deep learning model that enhances an underlying LLM with graph embeddings and trains it to understand them and to produce, given an instruction, an answer grounded in the graph representation. The approach performs significantly better than a graph to text approach and remains consistent even for larger graphs.
Authors: Chaoyou Fu, Haojia Lin, Zuwei Long, Yunhang Shen, Meng Zhao, Yifan Zhang, Shaoqi Dong, Xiong Wang, Di Yin, Long Ma, Xiawu Zheng, Ran He, Rongrong Ji, Yunsheng Wu, Caifeng Shan, Xing Sun
Abstract: The remarkable multimodal capabilities and interactive experience of GPT-4o underscore their necessity in practical applications, yet open-source models rarely excel in both areas. In this paper, we introduce VITA, the first-ever open-source Multimodal Large Language Model (MLLM) adept at simultaneous processing and analysis of Video, Image, Text, and Audio modalities, and meanwhile has an advanced multimodal interactive experience. Starting from Mixtral 8x7B as a language foundation, we expand its Chinese vocabulary followed by bilingual instruction tuning. We further endow the language model with visual and audio capabilities through two-stage multi-task learning of multimodal alignment and instruction tuning. VITA demonstrates robust foundational capabilities of multilingual, vision, and audio understanding, as evidenced by its strong performance across a range of both unimodal and multimodal benchmarks. Beyond foundational capabilities, we have made considerable progress in enhancing the natural multimodal human-computer interaction experience. VITA is the first step for the open-source community to explore the seamless integration of multimodal understanding and interaction. While there is still lots of work to be done on VITA to get close to close-source counterparts, we hope that its role as a pioneer can serve as a cornerstone for subsequent research. Project Page: https://vita-home.github.io.
Authors: Boci Peng, Yun Zhu, Yongchao Liu, Xiaohe Bo, Haizhou Shi, Chuntao Hong, Yan Zhang, Siliang Tang
Abstract: Recently, Retrieval-Augmented Generation (RAG) has achieved remarkable success in addressing the challenges of Large Language Models (LLMs) without necessitating retraining. By referencing an external knowledge base, RAG refines LLM outputs, effectively mitigating issues such as ``hallucination'', lack of domain-specific knowledge, and outdated information. However, the complex structure of relationships among different entities in databases presents challenges for RAG systems. In response, GraphRAG leverages structural information across entities to enable more precise and comprehensive retrieval, capturing relational knowledge and facilitating more accurate, context-aware responses. Given the novelty and potential of GraphRAG, a systematic review of current technologies is imperative. This paper provides the first comprehensive overview of GraphRAG methodologies. We formalize the GraphRAG workflow, encompassing Graph-Based Indexing, Graph-Guided Retrieval, and Graph-Enhanced Generation. We then outline the core technologies and training methods at each stage. Additionally, we examine downstream tasks, application domains, evaluation methodologies, and industrial use cases of GraphRAG. Finally, we explore future research directions to inspire further inquiries and advance progress in the field. In order to track recent progress in this field, we set up a repository at \url{https://github.com/pengboci/GraphRAG-Survey}.
Authors: Yang Cao
Abstract: The rapid advancement in large language models (LLMs) comes with a significant increase in their parameter size, presenting challenges for adaptation and fine-tuning. Parameter-efficient fine-tuning (PEFT) methods are widely used to adapt LLMs for downstream tasks efficiently. In this paper, we propose Singular Values and Orthonormal Regularized Singular Vectors Adaptation, or SORSA, a novel PEFT method. We introduce a method to analyze the variation of the parameters by performing singular value decomposition (SVD) and discuss and analyze SORSA's superiority in minimizing the alteration in the SVD aspect. Each SORSA adapter consists of two main parts: trainable principal singular weights $W_p = U_p \Sigma_p V^\top_p$, and frozen residual weights $W_r = U_r \Sigma_r V^\top_r$. These parts are initialized by performing SVD on pre-trained weights. Moreover, we implement and analyze an orthonormal regularizer, which could effectively transfer the scaling information into $\Sigma_p$ and ultimately allows the training process to be more efficient. SORSA adapters could be merged during inference, thus eliminating any inference latency. After all, SORSA shows a faster convergence than PiSSA and LoRA in our experiments. On the MATH benchmark, Llama 2 7B adapted using SORSA achieved 10.36% accuracy, outperforming LoRA (5.50%), Full FT (7.22%), and PiSSA (7.44%). On the GSM-8K benchmark, SORSA achieved 56.03% accuracy, surpassing LoRA (42.30%), Full FT (49.05%), and PiSSA (53.07%). We conclude that SORSA offers a new perspective on parameter-efficient fine-tuning, demonstrating remarkable performance. The code is available at https://github.com/Gunale0926/SORSA.
Authors: Hemanth Kandula, Damianos Karakos, Haoling Qiu, Benjamin Rozonoyer, Ian Soboroff, Lee Tarlin, Bonan Min
Abstract: Frequently, users of an Information Retrieval (IR) system start with an overarching information need (a.k.a., an analytic task) and proceed to define finer-grained queries covering various important aspects (i.e., sub-topics) of that analytic task. We present a novel, interactive system called $\textit{QueryBuilder}$, which allows a novice, English-speaking user to create queries with a small amount of effort, through efficient exploration of an English development corpus in order to rapidly develop cross-lingual information retrieval queries corresponding to the user's information needs. QueryBuilder performs near real-time retrieval of documents based on user-entered search terms; the user looks through the retrieved documents and marks sentences as relevant to the information needed. The marked sentences are used by the system as additional information in query formation and refinement: query terms (and, optionally, event features, which capture event $'triggers'$ (indicator terms) and agent/patient roles) are appropriately weighted, and a neural-based system, which better captures textual meaning, retrieves other relevant content. The process of retrieval and marking is repeated as many times as desired, giving rise to increasingly refined queries in each iteration. The final product is a fine-grained query used in Cross-Lingual Information Retrieval (CLIR). Our experiments using analytic tasks and requests from the IARPA BETTER IR datasets show that with a small amount of effort (at most 10 minutes per sub-topic), novice users can form $\textit{useful}$ fine-grained queries including in languages they don't understand. QueryBuilder also provides beneficial capabilities to the traditional corpus exploration and query formation process. A demonstration video is released at https://vimeo.com/734795835
Authors: Hongbo Wang, Junyu Lu, Yan Han, Kai Ma, Liang Yang, Hongfei Lin
Abstract: Patronizing and Condescending Language (PCL) is a form of discriminatory toxic speech targeting vulnerable groups, threatening both online and offline safety. While toxic speech research has mainly focused on overt toxicity, such as hate speech, microaggressions in the form of PCL remain underexplored. Additionally, dominant groups' discriminatory facial expressions and attitudes toward vulnerable communities can be more impactful than verbal cues, yet these frame features are often overlooked. In this paper, we introduce the PCLMM dataset, the first Chinese multimodal dataset for PCL, consisting of 715 annotated videos from Bilibili, with high-quality PCL facial frame spans. We also propose the MultiPCL detector, featuring a facial expression detection module for PCL recognition, demonstrating the effectiveness of modality complementarity in this challenging task. Our work makes an important contribution to advancing microaggression detection within the domain of toxic speech.