Authors: Tong Wang, Ninad Kulkarni, Yanjun Qi
Abstract: Assessing the factual consistency of automatically generated texts in relation to source context is crucial for developing reliable natural language generation applications. Recent literature proposes AlignScore which uses a unified alignment model to evaluate factual consistency and substantially outperforms previous methods across many benchmark tasks. In this paper, we take a closer look of datasets used in AlignScore and uncover an unexpected finding: utilizing a smaller number of data points can actually improve performance. We process the original AlignScore training dataset to remove noise, augment with robustness-enhanced samples, and utilize a subset comprising 10\% of the data to train an improved factual consistency evaluation model, we call LIM-RA (Less Is More for Robust AlignScore). LIM-RA demonstrates superior performance, consistently outperforming AlignScore and other strong baselines like ChatGPT across four benchmarks (two utilizing traditional natural language generation datasets and two focused on large language model outputs). Our experiments show that LIM-RA achieves the highest score on 24 of the 33 test datasets, while staying competitive on the rest, establishing the new state-of-the-art benchmarks.
Authors: Jane Dwivedi-Yu, Raaz Dwivedi, Timo Schick
Abstract: The accurate evaluation of differential treatment in language models to specific groups is critical to ensuring a positive and safe user experience. An ideal evaluation should have the properties of being robust, extendable to new groups or attributes, and being able to capture biases that appear in typical usage (rather than just extreme, rare cases). Relatedly, bias evaluation should surface not only egregious biases but also ones that are subtle and commonplace, such as a likelihood for talking about appearances with regard to women. We present FairPair, an evaluation framework for assessing differential treatment that occurs during ordinary usage. FairPair operates through counterfactual pairs, but crucially, the paired continuations are grounded in the same demographic group, which ensures equivalent comparison. Additionally, unlike prior work, our method factors in the inherent variability that comes from the generation process itself by measuring the sampling variability. We present an evaluation of several commonly used generative models and a qualitative analysis that indicates a preference for discussing family and hobbies with regard to women.
Authors: Jeongrok Yu, Seong Ug Kim, Jacob Choi, Jinho D. Choi
Abstract: Bias is a disproportionate prejudice in favor of one side against another. Due to the success of transformer-based Masked Language Models (MLMs) and their impact on many NLP tasks, a systematic evaluation of bias in these models is needed more than ever. While many studies have evaluated gender bias in English MLMs, only a few works have been conducted for the task in other languages. This paper proposes a multilingual approach to estimate gender bias in MLMs from 5 languages: Chinese, English, German, Portuguese, and Spanish. Unlike previous work, our approach does not depend on parallel corpora coupled with English to detect gender bias in other languages using multilingual lexicons. Moreover, a novel model-based method is presented to generate sentence pairs for a more robust analysis of gender bias, compared to the traditional lexicon-based method. For each language, both the lexicon-based and model-based methods are applied to create two datasets respectively, which are used to evaluate gender bias in an MLM specifically trained for that language using one existing and 3 new scoring metrics. Our results show that the previous approach is data-sensitive and not stable as it does not remove contextual dependencies irrelevant to gender. In fact, the results often flip when different scoring metrics are used on the same dataset, suggesting that gender bias should be studied on a large dataset using multiple evaluation metrics for best practice.
Authors: Omid Ghahroodi, Marzia Nouri, Mohammad Vali Sanian, Alireza Sahebi, Doratossadat Dastgheib, Ehsaneddin Asgari, Mahdieh Soleymani Baghshah, Mohammad Hossein Rohban
Abstract: Evaluating Large Language Models (LLMs) is challenging due to their generative nature, necessitating precise evaluation methodologies. Additionally, non-English LLM evaluation lags behind English, resulting in the absence or weakness of LLMs for many languages. In response to this necessity, we introduce Khayyam Challenge (also known as PersianMMLU), a meticulously curated collection comprising 20,192 four-choice questions sourced from 38 diverse tasks extracted from Persian examinations, spanning a wide spectrum of subjects, complexities, and ages. The primary objective of the Khayyam Challenge is to facilitate the rigorous evaluation of LLMs that support the Persian language. Distinctive features of the Khayyam Challenge are (i) its comprehensive coverage of various topics, including literary comprehension, mathematics, sciences, logic, intelligence testing, etc., aimed at assessing different facets of LLMs such as language comprehension, reasoning, and information retrieval across various educational stages, from lower primary school to upper secondary school (ii) its inclusion of rich metadata such as human response rates, difficulty levels, and descriptive answers (iii) its utilization of new data to avoid data contamination issues prevalent in existing frameworks (iv) its use of original, non-translated data tailored for Persian speakers, ensuring the framework is free from translation challenges and errors while encompassing cultural nuances (v) its inherent scalability for future data updates and evaluations without requiring special human effort. Previous works lacked an evaluation framework that combined all of these features into a single comprehensive benchmark. Furthermore, we evaluate a wide range of existing LLMs that support the Persian language, with statistical analyses and interpretations of their outputs.
Authors: Cheng-Ping Hsieh, Simeng Sun, Samuel Kriman, Shantanu Acharya, Dima Rekesh, Fei Jia, Boris Ginsburg
Abstract: The needle-in-a-haystack (NIAH) test, which examines the ability to retrieve a piece of information (the "needle") from long distractor texts (the "haystack"), has been widely adopted to evaluate long-context language models (LMs). However, this simple retrieval-based test is indicative of only a superficial form of long-context understanding. To provide a more comprehensive evaluation of long-context LMs, we create a new synthetic benchmark RULER with flexible configurations for customized sequence length and task complexity. RULER expands upon the vanilla NIAH test to encompass variations with diverse types and quantities of needles. Moreover, RULER introduces new task categories multi-hop tracing and aggregation to test behaviors beyond searching from context. We evaluate ten long-context LMs with 13 representative tasks in RULER. Despite achieving nearly perfect accuracy in the vanilla NIAH test, all models exhibit large performance drops as the context length increases. While these models all claim context sizes of 32K tokens or greater, only four models (GPT-4, Command-R, Yi-34B, and Mixtral) can maintain satisfactory performance at the length of 32K. Our analysis of Yi-34B, which supports context length of 200K, reveals large room for improvement as we increase input length and task complexity. We open source RULER to spur comprehensive evaluation of long-context LMs.
Authors: Nikhita Vedula, Giuseppe Castellucci, Eugene Agichtein, Oleg Rokhlenko, Shervin Malmasi
Abstract: Conversational Task Assistants (CTAs) guide users in performing a multitude of activities, such as making recipes. However, ensuring that interactions remain engaging, interesting, and enjoyable for CTA users is not trivial, especially for time-consuming or challenging tasks. Grounded in psychological theories of human interest, we propose to engage users with contextual and interesting statements or facts during interactions with a multi-modal CTA, to reduce fatigue and task abandonment before a task is complete. To operationalize this idea, we train a high-performing classifier (82% F1-score) to automatically identify relevant and interesting facts for users. We use it to create an annotated dataset of task-specific interesting facts for the domain of cooking. Finally, we design and validate a dialogue policy to incorporate the identified relevant and interesting facts into a conversation, to improve user engagement and task completion. Live testing on a leading multi-modal voice assistant shows that 66% of the presented facts were received positively, leading to a 40% gain in the user satisfaction rating, and a 37% increase in conversation length. These findings emphasize that strategically incorporating interesting facts into the CTA experience can promote real-world user participation for guided task interactions.
Authors: Yu Ying Chiu, Liwei Jiang, Maria Antoniak, Chan Young Park, Shuyue Stella Li, Mehar Bhatia, Sahithya Ravi, Yulia Tsvetkov, Vered Shwartz, Yejin Choi
Abstract: Frontier large language models (LLMs) are developed by researchers and practitioners with skewed cultural backgrounds and on datasets with skewed sources. However, LLMs' (lack of) multicultural knowledge cannot be effectively assessed with current methods for developing benchmarks. Existing multicultural evaluations primarily rely on expensive and restricted human annotations or potentially outdated internet resources. Thus, they struggle to capture the intricacy, dynamics, and diversity of cultural norms. LLM-generated benchmarks are promising, yet risk propagating the same biases they are meant to measure. To synergize the creativity and expert cultural knowledge of human annotators and the scalability and standardizability of LLM-based automation, we introduce CulturalTeaming, an interactive red-teaming system that leverages human-AI collaboration to build truly challenging evaluation dataset for assessing the multicultural knowledge of LLMs, while improving annotators' capabilities and experiences. Our study reveals that CulturalTeaming's various modes of AI assistance support annotators in creating cultural questions, that modern LLMs fail at, in a gamified manner. Importantly, the increased level of AI assistance (e.g., LLM-generated revision hints) empowers users to create more difficult questions with enhanced perceived creativity of themselves, shedding light on the promises of involving heavier AI assistance in modern evaluation dataset creation procedures. Through a series of 1-hour workshop sessions, we gather CULTURALBENCH-V0.1, a compact yet high-quality evaluation dataset with users' red-teaming attempts, that different families of modern LLMs perform with accuracy ranging from 37.7% to 72.2%, revealing a notable gap in LLMs' multicultural proficiency.
Authors: Anna Wegmann, Tijs van den Broek, Dong Nguyen
Abstract: Best practices for high conflict conversations like counseling or customer support almost always include recommendations to paraphrase the previous speaker. Although paraphrase classification has received widespread attention in NLP, paraphrases are usually considered independent from context, and common models and datasets are not applicable to dialog settings. In this work, we investigate paraphrases in dialog (e.g., Speaker 1: "That book is mine." becomes Speaker 2: "That book is yours."). We provide an operationalization of context-dependent paraphrases, and develop a training for crowd-workers to classify paraphrases in dialog. We introduce a dataset with utterance pairs from NPR and CNN news interviews annotated for context-dependent paraphrases. To enable analyses on label variation, the dataset contains 5,581 annotations on 600 utterance pairs. We present promising results with in-context learning and with token classification models for automatic paraphrase detection in dialog.
Authors: Shashi Kant Gupta, Aditya Basu, Bradley Taylor, Anai Kothari, Hrituraj Singh
Abstract: Retrieving information from EHR systems is essential for answering specific questions about patient journeys and improving the delivery of clinical care. Despite this fact, most EHR systems still rely on keyword-based searches. With the advent of generative large language models (LLMs), retrieving information can lead to better search and summarization capabilities. Such retrievers can also feed Retrieval-augmented generation (RAG) pipelines to answer any query. However, the task of retrieving information from EHR real-world clinical data contained within EHR systems in order to solve several downstream use cases is challenging due to the difficulty in creating query-document support pairs. We provide a blueprint for creating such datasets in an affordable manner using large language models. Our method results in a retriever that is 30-50 F-1 points better than propriety counterparts such as Ada and Mistral for oncology data elements. We further compare our model, called Onco-Retriever, against fine-tuned PubMedBERT model as well. We conduct an extensive manual evaluation on real-world EHR data along with latency analysis of the different models and provide a path forward for healthcare organizations to build domain-specific retrievers.
Authors: Longwei Zou, Qingyang Wang, Han Zhao, Jiangang Kong, Yi Yang, Yangdong Deng
Abstract: The fast-growing large scale language models are delivering unprecedented performance on almost all natural language processing tasks. However, the effectiveness of large language models are reliant on an exponentially increasing number of parameters. The overwhelming computation complexity incurs a high inference latency that negatively affects user experience. Existing methods to improve inference efficiency, such as tensor parallelism and quantization, target to reduce per-layer computing latency, yet overlook the cumulative latency due to the number of layers. Recent works on reducing the cumulative latency through layer removing, however, lead to significant performance drop. Motivated by the similarity of inputs among adjacent layers, we propose to identify quasi-independent layers, which can be concurrently computed to significantly decrease inference latency. We also introduce a bypassing technique to mitigate the effect of information loss. Empirical experiments of the proposed approach on the LLaMA models confirm that Concurrent Computation of Quasi-Independent Layers (CQIL) can reduce latency by up to 48.3% on the LLaMA-33B model, while maintaining a close level of performance.
Authors: Murong Yue, Wijdane Mifdal, Yixuan Zhang, Jennifer Suh, Ziyu Yao
Abstract: Mathematical modeling (MM) is considered a fundamental skill for students in STEM disciplines. Practicing the MM skill is often the most effective when students can engage in group discussion and collaborative problem-solving. However, due to unevenly distributed teachers and educational resources needed to monitor such group activities, students do not always receive equal opportunities for this practice. Excitingly, large language models (LLMs) have recently demonstrated strong capability in both modeling mathematical problems and simulating characters with different traits and properties. Drawing inspiration from the advancement of LLMs, in this work, we present MATHVC, the very first LLM-powered virtual classroom containing multiple LLM-simulated student characters, with whom a human student can practice their MM skill. To encourage each LLM character's behaviors to be aligned with their specified math-relevant properties (termed "characteristics alignment") and the overall conversational procedure to be close to an authentic student MM discussion (termed "conversational procedural alignment"), we proposed three innovations: integrating MM domain knowledge into the simulation, defining a symbolic schema as the ground for character simulation, and designing a meta planner at the platform level to drive the conversational procedure. Through experiments and ablation studies, we confirmed the effectiveness of our simulation approach and showed the promise for MATHVC to benefit real-life students in the future.
Authors: Xincan Feng, Akifumi Yoshimoto
Abstract: Recent advancements in Natural Language Processing (NLP) have seen Large-scale Language Models (LLMs) excel at producing high-quality text for various purposes. Notably, in Text-To-Speech (TTS) systems, the integration of BERT for semantic token generation has underscored the importance of semantic content in producing coherent speech outputs. Despite this, the specific utility of LLMs in enhancing TTS synthesis remains considerably limited. This research introduces an innovative approach, Llama-VITS, which enhances TTS synthesis by enriching the semantic content of text using LLM. Llama-VITS integrates semantic embeddings from Llama2 with the VITS model, a leading end-to-end TTS framework. By leveraging Llama2 for the primary speech synthesis process, our experiments demonstrate that Llama-VITS matches the naturalness of the original VITS (ORI-VITS) and those incorporate BERT (BERT-VITS), on the LJSpeech dataset, a substantial collection of neutral, clear speech. Moreover, our method significantly enhances emotive expressiveness on the EmoV_DB_bea_sem dataset, a curated selection of emotionally consistent speech from the EmoV_DB dataset, highlighting its potential to generate emotive speech.
Authors: Xiaokang Zhang, Zijun Yao, Jing Zhang, Kaifeng Yun, Jifan Yu, Juanzi Li, Jie Tang
Abstract: Detecting non-factual content is a longstanding goal to increase the trustworthiness of large language models (LLMs) generations. Current factuality probes, trained using humanannotated labels, exhibit limited transferability to out-of-distribution content, while online selfconsistency checking imposes extensive computation burden due to the necessity of generating multiple outputs. This paper proposes PINOSE, which trains a probing model on offline self-consistency checking results, thereby circumventing the need for human-annotated data and achieving transferability across diverse data distributions. As the consistency check process is offline, PINOSE reduces the computational burden of generating multiple responses by online consistency verification. Additionally, it examines various aspects of internal states prior to response decoding, contributing to more effective detection of factual inaccuracies. Experiment results on both factuality detection and question answering benchmarks show that PINOSE achieves surpassing results than existing factuality detection methods. Our code and datasets are publicly available on this anonymized repository.
Authors: Jianxiang Xiang, Zhenhua Liu, Haodong Liu, Yin Bai, Jia Cheng, Wenliang Chen
Abstract: In real-life conversations, the content is diverse, and there exists the one-to-many problem that requires diverse generation. Previous studies attempted to introduce discrete or Gaussian-based continuous latent variables to address the one-to-many problem, but the diversity is limited. Recently, diffusion models have made breakthroughs in computer vision, and some attempts have been made in natural language processing. In this paper, we propose DiffusionDialog, a novel approach to enhance the diversity of dialogue generation with the help of diffusion model. In our approach, we introduce continuous latent variables into the diffusion model. The problem of using latent variables in the dialog task is how to build both an effective prior of the latent space and an inferring process to obtain the proper latent given the context. By combining the encoder and latent-based diffusion model, we encode the response's latent representation in a continuous space as the prior, instead of fixed Gaussian distribution or simply discrete ones. We then infer the latent by denoising step by step with the diffusion model. The experimental results show that our model greatly enhances the diversity of dialog responses while maintaining coherence. Furthermore, in further analysis, we find that our diffusion model achieves high inference efficiency, which is the main challenge of applying diffusion models in natural language processing.
Authors: Zhengyuan Liu, Stella Xin Yin, Geyu Lin, Nancy F. Chen
Abstract: Intelligent Tutoring Systems (ITSs) can provide personalized and self-paced learning experience. The emergence of large language models (LLMs) further enables better human-machine interaction, and facilitates the development of conversational ITSs in various disciplines such as math and language learning. In dialogic teaching, recognizing and adapting to individual characteristics can significantly enhance student engagement and learning efficiency. However, characterizing and simulating student's persona remain challenging in training and evaluating conversational ITSs. In this work, we propose a framework to construct profiles of different student groups by refining and integrating both cognitive and noncognitive aspects, and leverage LLMs for personality-aware student simulation in a language learning scenario. We further enhance the framework with multi-aspect validation, and conduct extensive analysis from both teacher and student perspectives. Our experimental results show that state-of-the-art LLMs can produce diverse student responses according to the given language ability and personality traits, and trigger teacher's adaptive scaffolding strategies.
Authors: Ruotong Pan, Boxi Cao, Hongyu Lin, Xianpei Han, Jia Zheng, Sirui Wang, Xunliang Cai, Le Sun
Abstract: The rapid development of large language models has led to the widespread adoption of Retrieval-Augmented Generation (RAG), which integrates external knowledge to alleviate knowledge bottlenecks and mitigate hallucinations. However, the existing RAG paradigm inevitably suffers from the impact of flawed information introduced during the retrieval phrase, thereby diminishing the reliability and correctness of the generated outcomes. In this paper, we propose Credibility-aware Generation (CAG), a universally applicable framework designed to mitigate the impact of flawed information in RAG. At its core, CAG aims to equip models with the ability to discern and process information based on its credibility. To this end, we propose an innovative data transformation framework that generates data based on credibility, thereby effectively endowing models with the capability of CAG. Furthermore, to accurately evaluate the models' capabilities of CAG, we construct a comprehensive benchmark covering three critical real-world scenarios. Experimental results demonstrate that our model can effectively understand and utilize credibility for generation, significantly outperform other models with retrieval augmentation, and exhibit resilience against the disruption caused by noisy documents, thereby maintaining robust performance. Moreover, our model supports customized credibility, offering a wide range of potential applications.
Authors: Mingrui Fu, Weijiang Li
Abstract: The purpose of emotion-cause pair extraction is to extract the pair of emotion clauses and cause clauses. On the one hand, the existing methods do not take fully into account the relationship between the emotion extraction of two auxiliary tasks. On the other hand, the existing two-stage model has the problem of error propagation. In addition, existing models do not adequately address the emotion and cause-induced locational imbalance of samples. To solve these problems, an end-to-end multitasking model (MM-ECPE) based on shared interaction between GRU, knowledge graph and transformer modules is proposed. Furthermore, based on MM-ECPE, in order to use the encoder layer to better solve the problem of imbalanced distribution of clause distances between clauses and emotion clauses, we propose a novel encoding based on BERT, sentiment lexicon, and position-aware interaction module layer of emotion motif pair retrieval model (MM-ECPE(BERT)). The model first fully models the interaction between different tasks through the multi-level sharing module, and mines the shared information between emotion-cause pair extraction and the emotion extraction and cause extraction. Second, to solve the imbalanced distribution of emotion clauses and cause clauses problem, suitable labels are screened out according to the knowledge graph path length and task-specific features are constructed so that the model can focus on extracting pairs with corresponding emotion-cause relationships. Experimental results on the ECPE benchmark dataset show that the proposed model achieves good performance, especially on position-imbalanced samples.
Authors: Li Zhou, Taelin Karidi, Nicolas Garneau, Yong Cao, Wanlong Liu, Wenyu Chen, Daniel Hershcovich
Abstract: Recent studies have highlighted the presence of cultural biases in Large Language Models (LLMs), yet often lack a robust methodology to dissect these phenomena comprehensively. Our work aims to bridge this gap by delving into the Food domain, a universally relevant yet culturally diverse aspect of human life. We introduce FmLAMA, a multilingual dataset centered on food-related cultural facts and variations in food practices. We analyze LLMs across various architectures and configurations, evaluating their performance in both monolingual and multilingual settings. By leveraging templates in six different languages, we investigate how LLMs interact with language-specific and cultural knowledge. Our findings reveal that (1) LLMs demonstrate a pronounced bias towards food knowledge prevalent in the United States; (2) Incorporating relevant cultural context significantly improves LLMs' ability to access cultural knowledge; (3) The efficacy of LLMs in capturing cultural nuances is highly dependent on the interplay between the probing language, the specific model architecture, and the cultural context in question. This research underscores the complexity of integrating cultural understanding into LLMs and emphasizes the importance of culturally diverse datasets to mitigate biases and enhance model performance across different cultural domains.
Authors: Miriam Ansch\"utz, Edoardo Mosca, Georg Groh
Abstract: Text simplification seeks to improve readability while retaining the original content and meaning. Our study investigates whether pre-trained classifiers also maintain such coherence by comparing their predictions on both original and simplified inputs. We conduct experiments using 11 pre-trained models, including BERT and OpenAI's GPT 3.5, across six datasets spanning three languages. Additionally, we conduct a detailed analysis of the correlation between prediction change rates and simplification types/strengths. Our findings reveal alarming inconsistencies across all languages and models. If not promptly addressed, simplified inputs can be easily exploited to craft zero-iteration model-agnostic adversarial attacks with success rates of up to 50%
Authors: Jinghong Chen, Weizhe Lin, Jingbiao Mei, Bill Byrne
Abstract: The Directed Acyclic Transformer is a fast non-autoregressive (NAR) model that performs well in Neural Machine Translation. Two issues prevent its application to general Natural Language Generation (NLG) tasks: frequent Out-Of-Vocabulary (OOV) errors and the inability to faithfully generate entity names. We introduce Control-DAG, a constrained decoding algorithm for our Directed Acyclic T5 (DA-T5) model which offers lexical, vocabulary and length control. We show that Control-DAG significantly enhances DA-T5 on the Schema Guided Dialogue and the DART datasets, establishing strong NAR results for Task-Oriented Dialogue and Data-to-Text NLG.
Authors: Thomas Merth, Qichen Fu, Mohammad Rastegari, Mahyar Najibi
Abstract: Despite the successes of large language models (LLMs), they exhibit significant drawbacks, particularly when processing long contexts. Their inference cost scales quadratically with respect to sequence length, making it expensive for deployment in some real-world text processing applications, such as retrieval-augmented generation (RAG). Additionally, LLMs also exhibit the "distraction phenomenon," where irrelevant context in the prompt degrades output quality. To address these drawbacks, we propose a novel RAG prompting methodology, superposition prompting, which can be directly applied to pre-trained transformer-based LLMs without the need for fine-tuning. At a high level, superposition prompting allows the LLM to process input documents in parallel prompt paths, discarding paths once they are deemed irrelevant. We demonstrate the capability of our method to simultaneously enhance time efficiency across a variety of question-answering benchmarks using multiple pre-trained LLMs. Furthermore, our technique significantly improves accuracy when the retrieved context is large relative the context the model was trained on. For example, our approach facilitates an 93x reduction in compute time while improving accuracy by 43\% on the NaturalQuestions-Open dataset with the MPT-7B instruction-tuned model over naive RAG.
Authors: Shuzhou Yuan, Michael F\"arber
Abstract: Pretrained Language Models (PLMs) benefit from external knowledge stored in graph structures for various downstream tasks. However, bridging the modality gap between graph structures and text remains a significant challenge. Traditional methods like linearizing graphs for PLMs lose vital graph connectivity, whereas Graph Neural Networks (GNNs) require cumbersome processes for integration into PLMs. In this work, we propose a novel graph-guided self-attention mechanism, GraSAME. GraSAME seamlessly incorporates token-level structural information into PLMs without necessitating additional alignment or concatenation efforts. As an end-to-end, lightweight multimodal module, GraSAME follows a multi-task learning strategy and effectively bridges the gap between graph and textual modalities, facilitating dynamic interactions between GNNs and PLMs. Our experiments on the graph-to-text generation task demonstrate that GraSAME outperforms baseline models and achieves results comparable to state-of-the-art (SOTA) models on WebNLG datasets. Furthermore, compared to SOTA models, GraSAME eliminates the need for extra pre-training tasks to adjust graph inputs and reduces the number of trainable parameters by over 100 million.
Authors: Shishir G. Patil, Tianjun Zhang, Vivian Fang, Noppapon C., Roy Huang, Aaron Hao, Martin Casado, Joseph E. Gonzalez, Raluca Ada Popa, Ion Stoica
Abstract: Large Language Models (LLMs) are evolving beyond their classical role of providing information within dialogue systems to actively engaging with tools and performing actions on real-world applications and services. Today, humans verify the correctness and appropriateness of the LLM-generated outputs (e.g., code, functions, or actions) before putting them into real-world execution. This poses significant challenges as code comprehension is well known to be notoriously difficult. In this paper, we study how humans can efficiently collaborate with, delegate to, and supervise autonomous LLMs in the future. We argue that in many cases, "post-facto validation" - verifying the correctness of a proposed action after seeing the output - is much easier than the aforementioned "pre-facto validation" setting. The core concept behind enabling a post-facto validation system is the integration of an intuitive undo feature, and establishing a damage confinement for the LLM-generated actions as effective strategies to mitigate the associated risks. Using this, a human can now either revert the effect of an LLM-generated output or be confident that the potential risk is bounded. We believe this is critical to unlock the potential for LLM agents to interact with applications and services with limited (post-facto) human involvement. We describe the design and implementation of our open-source runtime for executing LLM actions, Gorilla Execution Engine (GoEX), and present open research questions towards realizing the goal of LLMs and applications interacting with each other with minimal human supervision. We release GoEX at https://github.com/ShishirPatil/gorilla/.
Authors: Rahul Mehta, Andrew Hoblitzell, Jack O'Keefe, Hyeju Jang, Vasudeva Varma
Abstract: This paper presents our winning solution for the SemEval-2024 Task 6 competition. We propose a meta-regressor framework of large language models (LLMs) for model evaluation and integration that achieves the highest scores on the leader board. Our approach leverages uncertainty signals present in a diverse basket of LLMs to detect hallucinations more robustly.
Authors: Yijin Liu, Fandong Meng, Jie Zhou
Abstract: Recently, dynamic computation methods have shown notable acceleration for Large Language Models (LLMs) by skipping several layers of computations through elaborate heuristics or additional predictors. However, in the decoding process of existing approaches, different samples are assigned different computational budgets, which cannot guarantee a stable and precise acceleration effect. Furthermore, existing approaches generally skip multiple contiguous layers at the bottom or top of the layers, leading to a drastic change in the model's layer-wise representations, and thus a consequent performance degeneration. Therefore, we propose a Unified Layer Skipping strategy, which selects the number of layers to skip computation based solely on the target speedup ratio, and then skips the corresponding number of intermediate layer computations in a balanced manner. Since the Unified Layer Skipping strategy is independent of input samples, it naturally supports popular acceleration techniques such as batch decoding and KV caching, thus demonstrating more practicality for real-world applications. Experimental results on two common tasks, i.e., machine translation and text summarization, indicate that given a target speedup ratio, the Unified Layer Skipping strategy significantly enhances both the inference performance and the actual model throughput over existing dynamic approaches.
Authors: Martin Popel, Lucie Pol\'akov\'a, Michal Nov\'ak, Jind\v{r}ich Helcl, Jind\v{r}ich Libovick\'y, Pavel Stra\v{n}\'ak, Tom\'a\v{s} Kraba\v{c}, Jaroslava Hlav\'a\v{c}ov\'a, Mariia Anisimova, Tereza Chla\v{n}ov\'a
Abstract: We present Charles Translator, a machine translation system between Ukrainian and Czech, developed as part of a society-wide effort to mitigate the impact of the Russian-Ukrainian war on individuals and society. The system was developed in the spring of 2022 with the help of many language data providers in order to quickly meet the demand for such a service, which was not available at the time in the required quality. The translator was later implemented as an online web interface and as an Android app with speech input, both featuring Cyrillic-Latin script transliteration. The system translates directly, compared to other available systems that use English as a pivot, and thus take advantage of the typological similarity of the two languages. It uses the block back-translation method, which allows for efficient use of monolingual training data. The paper describes the development process, including data collection and implementation, evaluation, mentions several use cases, and outlines possibilities for the further development of the system for educational purposes.
Authors: Peipei Liu, Gaosheng Wang, Ying Tong, Jian Liang, Zhenquan Ding, Hongsong Zhu
Abstract: Few-shot named entity recognition can identify new types of named entities based on a few labeled examples. Previous methods employing token-level or span-level metric learning suffer from the computational burden and a large number of negative sample spans. In this paper, we propose the Hybrid Multi-stage Decoding for Few-shot NER with Entity-aware Contrastive Learning (MsFNER), which splits the general NER into two stages: entity-span detection and entity classification. There are 3 processes for introducing MsFNER: training, finetuning, and inference. In the training process, we train and get the best entity-span detection model and the entity classification model separately on the source domain using meta-learning, where we create a contrastive learning module to enhance entity representations for entity classification. During finetuning, we finetune the both models on the support dataset of target domain. In the inference process, for the unlabeled data, we first detect the entity-spans, then the entity-spans are jointly determined by the entity classification model and the KNN. We conduct experiments on the open FewNERD dataset and the results demonstrate the advance of MsFNER.
Authors: Maite Heredia, Julen Etxaniz, Muitze Zulaika, Xabier Saralegi, Jeremy Barnes, Aitor Soroa
Abstract: XNLI is a popular Natural Language Inference (NLI) benchmark widely used to evaluate cross-lingual Natural Language Understanding (NLU) capabilities across languages. In this paper, we expand XNLI to include Basque, a low-resource language that can greatly benefit from transfer-learning approaches. The new dataset, dubbed XNLIeu, has been developed by first machine-translating the English XNLI corpus into Basque, followed by a manual post-edition step. We have conducted a series of experiments using mono- and multilingual LLMs to assess a) the effect of professional post-edition on the MT system; b) the best cross-lingual strategy for NLI in Basque; and c) whether the choice of the best cross-lingual strategy is influenced by the fact that the dataset is built by translation. The results show that post-edition is necessary and that the translate-train cross-lingual strategy obtains better results overall, although the gain is lower when tested in a dataset that has been built natively from scratch. Our code and datasets are publicly available under open licenses.
Authors: Linan Yue, Qi Liu, Lili Zhao, Li Wang, Weibo Gao, Yanqing An
Abstract: With the development of legal intelligence, Criminal Court View Generation has attracted much attention as a crucial task of legal intelligence, which aims to generate concise and coherent texts that summarize case facts and provide explanations for verdicts. Existing researches explore the key information in case facts to yield the court views. Most of them employ a coarse-grained approach that partitions the facts into broad segments (e.g., verdict-related sentences) to make predictions. However, this approach fails to capture the complex details present in the case facts, such as various criminal elements and legal events. To this end, in this paper, we propose an Event Grounded Generation (EGG) method for criminal court view generation with cooperative (Large) Language Models, which introduces the fine-grained event information into the generation. Specifically, we first design a LLMs-based extraction method that can extract events in case facts without massive annotated events. Then, we incorporate the extracted events into court view generation by merging case facts and events. Besides, considering the computational burden posed by the use of LLMs in the extraction phase of EGG, we propose a LLMs-free EGG method that can eliminate the requirement for event extraction using LLMs in the inference phase. Extensive experimental results on a real-world dataset clearly validate the effectiveness of our proposed method.
Authors: Igor Tufanov, Karen Hambardzumyan, Javier Ferrando, Elena Voita
Abstract: We present the LM Transparency Tool (LM-TT), an open-source interactive toolkit for analyzing the internal workings of Transformer-based language models. Differently from previously existing tools that focus on isolated parts of the decision-making process, our framework is designed to make the entire prediction process transparent, and allows tracing back model behavior from the top-layer representation to very fine-grained parts of the model. Specifically, it (1) shows the important part of the whole input-to-output information flow, (2) allows attributing any changes done by a model block to individual attention heads and feed-forward neurons, (3) allows interpreting the functions of those heads or neurons. A crucial part of this pipeline is showing the importance of specific model components at each step. As a result, we are able to look at the roles of model components only in cases where they are important for a prediction. Since knowing which components should be inspected is key for analyzing large models where the number of these components is extremely high, we believe our tool will greatly support the interpretability community both in research settings and in practical applications.
Authors: Kuo-Yu Liao, Cheng-Shang Chang, Y. -W. Peter Hong
Abstract: Recent advances in Large Language Models (LLMs) have demonstrated the emergence of capabilities (learned skills) when the number of system parameters and the size of training data surpass certain thresholds. The exact mechanisms behind such phenomena are not fully understood and remain a topic of active research. Inspired by the skill-text bipartite graph model presented in [1] for modeling semantic language, we develop a mathematical theory to explain the emergence of learned skills, taking the learning (or training) process into account. Our approach models the learning process for skills in the skill-text bipartite graph as an iterative decoding process in Low-Density Parity Check (LDPC) codes and Irregular Repetition Slotted ALOHA (IRSA). Using density evolution analysis, we demonstrate the emergence of learned skills when the ratio of the size of training texts to the number of skills exceeds a certain threshold. Our analysis also yields a scaling law for testing errors relative to the size of training texts. Upon completion of the training, we propose a method for semantic compression and discuss its application in semantic communication.
Authors: Yunlong Feng, Yang Xu, Libo Qin, Yasheng Wang, Wanxiang Che
Abstract: Large-scale high-quality training data is important for improving the performance of models. After trained with data that has rationales (reasoning steps), models gain reasoning capability. However, the dataset with high-quality rationales is relatively scarce due to the high annotation cost. To address this issue, we propose \textit{Self-motivated Learning} framework. The framework motivates the model itself to automatically generate rationales on existing datasets. Based on the inherent rank from correctness across multiple rationales, the model learns to generate better rationales, leading to higher reasoning capability. Specifically, we train a reward model with the rank to evaluate the quality of rationales, and improve the performance of reasoning through reinforcement learning. Experiment results of Llama2 7B on multiple reasoning datasets show that our method significantly improves the reasoning ability of models, even outperforming text-davinci-002 in some datasets.
Authors: Jaya Caporusso, Damar Hoogland, Mojca Brglez, Boshko Koloski, Matthew Purver, Senja Pollak
Abstract: Dehumanisation involves the perception and or treatment of a social group's members as less than human. This phenomenon is rarely addressed with computational linguistic techniques. We adapt a recently proposed approach for English, making it easier to transfer to other languages and to evaluate, introducing a new sentiment resource, the use of zero-shot cross-lingual valence and arousal detection, and a new method for statistical significance testing. We then apply it to study attitudes to migration expressed in Slovene newspapers, to examine changes in the Slovene discourse on migration between the 2015-16 migration crisis following the war in Syria and the 2022-23 period following the war in Ukraine. We find that while this discourse became more negative and more intense over time, it is less dehumanising when specifically addressing Ukrainian migrants compared to others.
Authors: Elisa Sanchez-Bayona, Rodrigo Agerri
Abstract: Metaphors, although occasionally unperceived, are ubiquitous in our everyday language. Thus, it is crucial for Language Models to be able to grasp the underlying meaning of this kind of figurative language. In this work, we present Meta4XNLI, a novel parallel dataset for the tasks of metaphor detection and interpretation that contains metaphor annotations in both Spanish and English. We investigate language models' metaphor identification and understanding abilities through a series of monolingual and cross-lingual experiments by leveraging our proposed corpus. In order to comprehend how these non-literal expressions affect models' performance, we look over the results and perform an error analysis. Additionally, parallel data offers many potential opportunities to investigate metaphor transferability between these languages and the impact of translation on the development of multilingual annotated resources.
Authors: Alessandro Stolfo
Abstract: We present an empirical study of groundedness in long-form question answering (LFQA) by retrieval-augmented large language models (LLMs). In particular, we evaluate whether every generated sentence is grounded in the retrieved documents or the model's pre-training data. Across 3 datasets and 4 model families, our findings reveal that a significant fraction of generated sentences are consistently ungrounded, even when those sentences contain correct ground-truth answers. Additionally, we examine the impacts of factors such as model size, decoding strategy, and instruction tuning on groundedness. Our results show that while larger models tend to ground their outputs more effectively, a significant portion of correct answers remains compromised by hallucinations. This study provides novel insights into the groundedness challenges in LFQA and underscores the necessity for more robust mechanisms in LLMs to mitigate the generation of ungrounded content.
Authors: Mingyu Jin, Qinkai Yu, Jingyuan Huang, Qingcheng Zeng, Zhenting Wang, Wenyue Hua, Haiyan Zhao, Kai Mei, Yanda Meng, Kaize Ding, Fan Yang, Mengnan Du, Yongfeng Zhang
Abstract: This paper studies the phenomenon that different concepts are learned in different layers of large language models, i.e. more difficult concepts are fully acquired with deeper layers. We define the difficulty of concepts by the level of abstraction, and here it is crudely categorized by factual, emotional, and inferential. Each category contains a spectrum of tasks, arranged from simple to complex. For example, within the factual dimension, tasks range from lie detection to categorizing mathematical problems. We employ a probing technique to extract representations from different layers of the model and apply these to classification tasks. Our findings reveal that models tend to efficiently classify simpler tasks, indicating that these concepts are learned in shallower layers. Conversely, more complex tasks may only be discernible at deeper layers, if at all. This paper explores the implications of these findings for our understanding of model learning processes and internal representations. Our implementation is available at \url{https://github.com/Luckfort/CD}.
Authors: Jianzhi Liu, Hexiang Gu, Tianyu Zheng, Liuyu Xiang, Huijia Wu, Jie Fu, Zhaofeng He
Abstract: In the realm of mimicking human deliberation, large language models (LLMs) show promising performance, thereby amplifying the importance of this research area. Deliberation is influenced by both logic and personality. However, previous studies predominantly focused on the logic of LLMs, neglecting the exploration of personality aspects. In this work, we introduce Dynamic Personality Generation (DPG), a dynamic personality generation method based on Hypernetworks. Initially, we embed the Big Five personality theory into GPT-4 to form a personality assessment machine, enabling it to evaluate characters' personality traits from dialogues automatically. We propose a new metric to assess personality generation capability based on this evaluation method. Then, we use this personality assessment machine to evaluate dialogues in script data, resulting in a personality-dialogue dataset. Finally, we fine-tune DPG on the personality-dialogue dataset. Experiments prove that DPG's personality generation capability is stronger after fine-tuning on this dataset than traditional fine-tuning methods, surpassing prompt-based GPT-4.
Authors: Bowen Jin, Chulin Xie, Jiawei Zhang, Kashob Kumar Roy, Yu Zhang, Suhang Wang, Yu Meng, Jiawei Han
Abstract: Large language models (LLMs), while exhibiting exceptional performance, suffer from hallucinations, especially on knowledge-intensive tasks. Existing works propose to augment LLMs with individual text units retrieved from external knowledge corpora to alleviate the issue. However, in many domains, texts are interconnected (e.g., academic papers in a bibliographic graph are linked by citations and co-authorships) which form a (text-attributed) graph. The knowledge in such graphs is encoded not only in single texts/nodes but also in their associated connections. To facilitate the research of augmenting LLMs with graphs, we manually construct a Graph Reasoning Benchmark dataset called GRBench, containing 1,740 questions that can be answered with the knowledge from 10 domain graphs. Then, we propose a simple and effective framework called Graph Chain-of-thought (Graph-CoT) to augment LLMs with graphs by encouraging LLMs to reason on the graph iteratively. Each Graph-CoT iteration consists of three sub-steps: LLM reasoning, LLM-graph interaction, and graph execution. We conduct systematic experiments with three LLM backbones on GRBench, where Graph-CoT outperforms the baselines consistently. The code is available at https://github.com/PeterGriffinJin/Graph-CoT.
Authors: Yongqiang Ma, Lizhi Qin, Jiawei Liu, Yangyang Kang, Yue Zhang, Wei Lu, Xiaozhong Liu, Qikai Cheng
Abstract: Evaluating large language models (LLMs) is fundamental, particularly in the context of practical applications. Conventional evaluation methods, typically designed primarily for LLM development, yield numerical scores that ignore the user experience. Therefore, our study shifts the focus from model-centered to human-centered evaluation in the context of AI-powered writing assistance applications. Our proposed metric, termed ``Revision Distance,'' utilizes LLMs to suggest revision edits that mimic the human writing process. It is determined by counting the revision edits generated by LLMs. Benefiting from the generated revision edit details, our metric can provide a self-explained text evaluation result in a human-understandable manner beyond the context-independent score. Our results show that for the easy-writing task, ``Revision Distance'' is consistent with established metrics (ROUGE, Bert-score, and GPT-score), but offers more insightful, detailed feedback and better distinguishes between texts. Moreover, in the context of challenging academic writing tasks, our metric still delivers reliable evaluations where other metrics tend to struggle. Furthermore, our metric also holds significant potential for scenarios lacking reference texts.
Authors: Sara Kangaslahti, David Alvarez-Melis
Abstract: As large language models (LLMs) have gained popularity for a variety of use cases, making them adaptable and controllable has become increasingly important, especially for user-facing applications. While the existing literature on LLM adaptation primarily focuses on finding a model (or models) that optimizes a single predefined objective, here we focus on the challenging case where the model must dynamically adapt to diverse -- and often changing -- user preferences. For this, we leverage adaptation methods based on linear weight interpolation, casting them as continuous multi-domain interpolators that produce models with specific prescribed generation characteristics on-the-fly. Specifically, we use low-rank updates to fine-tune a base model to various different domains, yielding a set of anchor models with distinct generation profiles. Then, we use the weight updates of these anchor models to parametrize the entire (infinite) class of models contained within their convex hull. We empirically show that varying the interpolation weights yields predictable and consistent change in the model outputs with respect to all of the controlled attributes. We find that there is little entanglement between most attributes and identify and discuss the pairs of attributes for which this is not the case. Our results suggest that linearly interpolating between the weights of fine-tuned models facilitates predictable, fine-grained control of model outputs with respect to multiple stylistic characteristics simultaneously.
Authors: Jinwei Lu, Yuanfeng Song, Haodi Zhang, Chen Zhang, Raymond Chi-Wing Wong
Abstract: Text-to-Vis is an emerging task in the natural language processing (NLP) area that aims to automatically generate data visualizations from natural language questions (NLQs). Despite their progress, existing text-to-vis models often heavily rely on lexical matching between words in the questions and tokens in data schemas. This overreliance on lexical matching may lead to a diminished level of model robustness against input variations. In this study, we thoroughly examine the robustness of current text-to-vis models, an area that has not previously been explored. In particular, we construct the first robustness dataset nvBench-Rob, which contains diverse lexical and phrasal variations based on the original text-to-vis benchmark nvBench. Then, we found that the performance of existing text-to-vis models on this new dataset dramatically drops, implying that these methods exhibit inadequate robustness overall. Finally, we propose a novel framework based on Retrieval-Augmented Generation (RAG) technique, named GRED, specifically designed to address input perturbations in these two variants. The framework consists of three parts: NLQ-Retrieval Generator, Visualization Query-Retrieval Retuner and Annotation-based Debugger, which are used to tackle the challenges posed by natural language variants, programming style differences and data schema variants, respectively. Extensive experimental evaluations show that, compared to the state-of-the-art model RGVisNet in the Text-to-Vis field, RGDR performs better in terms of model robustness, with a 32% increase in accuracy on the proposed nvBench-Rob dataset.
Authors: Tsendsuren Munkhdalai, Manaal Faruqui, Siddharth Gopal
Abstract: This work introduces an efficient method to scale Transformer-based Large Language Models (LLMs) to infinitely long inputs with bounded memory and computation. A key component in our proposed approach is a new attention technique dubbed Infini-attention. The Infini-attention incorporates a compressive memory into the vanilla attention mechanism and builds in both masked local attention and long-term linear attention mechanisms in a single Transformer block. We demonstrate the effectiveness of our approach on long-context language modeling benchmarks, 1M sequence length passkey context block retrieval and 500K length book summarization tasks with 1B and 8B LLMs. Our approach introduces minimal bounded memory parameters and enables fast streaming inference for LLMs.
Authors: Xinfeng Li, Yuchen Yang, Jiangyi Deng, Chen Yan, Yanjiao Chen, Xiaoyu Ji, Wenyuan Xu
Abstract: Text-to-image (T2I) models, such as Stable Diffusion, have exhibited remarkable performance in generating high-quality images from text descriptions in recent years. However, text-to-image models may be tricked into generating not-safe-for-work (NSFW) content, particularly in sexual scenarios. Existing countermeasures mostly focus on filtering inappropriate inputs and outputs, or suppressing improper text embeddings, which can block explicit NSFW-related content (e.g., naked or sexy) but may still be vulnerable to adversarial prompts inputs that appear innocent but are ill-intended. In this paper, we present SafeGen, a framework to mitigate unsafe content generation by text-to-image models in a text-agnostic manner. The key idea is to eliminate unsafe visual representations from the model regardless of the text input. In this way, the text-to-image model is resistant to adversarial prompts since unsafe visual representations are obstructed from within. Extensive experiments conducted on four datasets demonstrate SafeGen's effectiveness in mitigating unsafe content generation while preserving the high-fidelity of benign images. SafeGen outperforms eight state-of-the-art baseline methods and achieves 99.1% sexual content removal performance. Furthermore, our constructed benchmark of adversarial prompts provides a basis for future development and evaluation of anti-NSFW-generation methods.
Authors: Leying Zhang, Yao Qian, Long Zhou, Shujie Liu, Dongmei Wang, Xiaofei Wang, Midia Yousefi, Yanmin Qian, Jinyu Li, Lei He, Sheng Zhao, Michael Zeng
Abstract: Recent advancements in zero-shot text-to-speech (TTS) modeling have led to significant strides in generating high-fidelity and diverse speech. However, dialogue generation, along with achieving human-like naturalness in speech, continues to be a challenge in the field. In this paper, we introduce CoVoMix: Conversational Voice Mixture Generation, a novel model for zero-shot, human-like, multi-speaker, multi-round dialogue speech generation. CoVoMix is capable of first converting dialogue text into multiple streams of discrete tokens, with each token stream representing semantic information for individual talkers. These token streams are then fed into a flow-matching based acoustic model to generate mixed mel-spectrograms. Finally, the speech waveforms are produced using a HiFi-GAN model. Furthermore, we devise a comprehensive set of metrics for measuring the effectiveness of dialogue modeling and generation. Our experimental results show that CoVoMix can generate dialogues that are not only human-like in their naturalness and coherence but also involve multiple talkers engaging in multiple rounds of conversation. These dialogues, generated within a single channel, are characterized by seamless speech transitions, including overlapping speech, and appropriate paralinguistic behaviors such as laughter. Audio samples are available at https://aka.ms/covomix.
URLs: https://aka.ms/covomix.
Authors: Yingbo Ma, Suraj Kolla, Zhenhong Hu, Dhruv Kaliraman, Victoria Nolan, Ziyuan Guan, Yuanfang Ren, Brooke Armfield, Tezcan Ozrazgat-Baslanti, Jeremy A. Balch, Tyler J. Loftus, Parisa Rashidi, Azra Bihorac, Benjamin Shickel
Abstract: Modern electronic health records (EHRs) hold immense promise in tracking personalized patient health trajectories through sequential deep learning, owing to their extensive breadth, scale, and temporal granularity. Nonetheless, how to effectively leverage multiple modalities from EHRs poses significant challenges, given its complex characteristics such as high dimensionality, multimodality, sparsity, varied recording frequencies, and temporal irregularities. To this end, this paper introduces a novel multimodal contrastive learning framework, specifically focusing on medical time series and clinical notes. To tackle the challenge of sparsity and irregular time intervals in medical time series, the framework integrates temporal cross-attention transformers with a dynamic embedding and tokenization scheme for learning multimodal feature representations. To harness the interconnected relationships between medical time series and clinical notes, the framework equips a global contrastive loss, aligning a patient's multimodal feature representations with the corresponding discharge summaries. Since discharge summaries uniquely pertain to individual patients and represent a holistic view of the patient's hospital stay, machine learning models are led to learn discriminative multimodal features via global contrasting. Extensive experiments with a real-world EHR dataset demonstrated that our framework outperformed state-of-the-art approaches on the exemplar task of predicting the occurrence of nine postoperative complications for more than 120,000 major inpatient surgeries using multimodal data from UF health system split among three hospitals (UF Health Gainesville, UF Health Jacksonville, and UF Health Jacksonville-North).
Authors: Jon Kleinberg, Sendhil Mullainathan
Abstract: Although current large language models are complex, the most basic specifications of the underlying language generation problem itself are simple to state: given a finite set of training samples from an unknown language, produce valid new strings from the language that don't already appear in the training data. Here we ask what we can conclude about language generation using only this specification, without further assumptions. In particular, suppose that an adversary enumerates the strings of an unknown target language L that is known only to come from one of a possibly infinite list of candidates. A computational agent is trying to learn to generate from this language; we say that the agent generates from L in the limit if after some finite point in the enumeration of L, the agent is able to produce new elements that come exclusively from L and that have not yet been presented by the adversary. Our main result is that there is an agent that is able to generate in the limit for every countable list of candidate languages. This contrasts dramatically with negative results due to Gold and Angluin in a well-studied model of language learning where the goal is to identify an unknown language from samples; the difference between these results suggests that identifying a language is a fundamentally different problem than generating from it.
Authors: Yuexi Chen, Zhicheng Liu
Abstract: Non-native English speakers (NNES) face challenges in digital workspace communication (e.g., emails, Slack messages), often inadvertently translating expressions from their native languages, which can lead to awkward or incorrect usage. Current AI-assisted writing tools are equipped with fluency enhancement and rewriting suggestions; however, NNES may struggle to grasp the subtleties among various expressions, making it challenging to choose the one that accurately reflects their intent. Such challenges are exacerbated in high-stake text-based communications, where the absence of non-verbal cues heightens the risk of misinterpretation. By leveraging the latest advancements in large language models (LLM) and word embeddings, we propose WordDecipher, an explainable AI-assisted writing tool to enhance digital workspace communication for NNES. WordDecipher not only identifies the perceived social intentions detected in users' writing, but also generates rewriting suggestions aligned with users' intended messages, either numerically or by inferring from users' writing in their native language. Then, WordDecipher provides an overview of nuances to help NNES make selections. Through a usage scenario, we demonstrate how WordDecipher can significantly enhance an NNES's ability to communicate her request, showcasing its potential to transform workspace communication for NNES.
Authors: Weihao Xia, Raoul de Charette, Cengiz \"Oztireli, Jing-Hao Xue
Abstract: We address prevailing challenges of the brain-powered research, departing from the observation that the literature hardly recover accurate spatial information and require subject-specific models. To address these challenges, we propose UMBRAE, a unified multimodal decoding of brain signals. First, to extract instance-level conceptual and spatial details from neural signals, we introduce an efficient universal brain encoder for multimodal-brain alignment and recover object descriptions at multiple levels of granularity from subsequent multimodal large language model (MLLM). Second, we introduce a cross-subject training strategy mapping subject-specific features to a common feature space. This allows a model to be trained on multiple subjects without extra resources, even yielding superior results compared to subject-specific models. Further, we demonstrate this supports weakly-supervised adaptation to new subjects, with only a fraction of the total training data. Experiments demonstrate that UMBRAE not only achieves superior results in the newly introduced tasks but also outperforms methods in well established tasks. To assess our method, we construct and share with the community a comprehensive brain understanding benchmark BrainHub. Our code and benchmark are available at https://weihaox.github.io/UMBRAE.
Authors: Yoichi Ishibashi, Sho Yokoi, Katsuhito Sudoh, Satoshi Nakamura
Abstract: In the field of natural language processing (NLP), continuous vector representations are crucial for capturing the semantic meanings of individual words. Yet, when it comes to the representations of sets of words, the conventional vector-based approaches often struggle with expressiveness and lack the essential set operations such as union, intersection, and complement. Inspired by quantum logic, we realize the representation of word sets and corresponding set operations within pre-trained word embedding spaces. By grounding our approach in the linear subspaces, we enable efficient computation of various set operations and facilitate the soft computation of membership functions within continuous spaces. Moreover, we allow for the computation of the F-score directly within word vectors, thereby establishing a direct link to the assessment of sentence similarity. In experiments with widely-used pre-trained embeddings and benchmarks, we show that our subspace-based set operations consistently outperform vector-based ones in both sentence similarity and set retrieval tasks.
Authors: Danial Kamali, Joseph Romain, Huiyi Liu, Wei Peng, Jingbo Meng, Parisa Kordjamshidi
Abstract: Nowadays, the spread of misinformation is a prominent problem in society. Our research focuses on aiding the automatic identification of misinformation by analyzing the persuasive strategies employed in textual documents. We introduce a novel annotation scheme encompassing common persuasive writing tactics to achieve our objective. Additionally, we provide a dataset on health misinformation, thoroughly annotated by experts utilizing our proposed scheme. Our contribution includes proposing a new task of annotating pieces of text with their persuasive writing strategy types. We evaluate fine-tuning and prompt-engineering techniques with pre-trained language models of the BERT family and the generative large language models of the GPT family using persuasive strategies as an additional source of information. We evaluate the effects of employing persuasive strategies as intermediate labels in the context of misinformation detection. Our results show that those strategies enhance accuracy and improve the explainability of misinformation detection models. The persuasive strategies can serve as valuable insights and explanations, enabling other models or even humans to make more informed decisions regarding the trustworthiness of the information.
Authors: Merlin Carl (Europa-Universit\"at Flensburg)
Abstract: The Diproche system is an automated proof checker for texts written in a controlled fragment of German, designed for didactical applications in classes introducing students to proofs for the first time. The first version of the system used a controlled natural language for which a Prolog formalization routine was written. In this paper, we explore the possibility of prompting large language models for autoformalization in the context of Diproche, with encouraging first results.
Authors: Merlin Carl (Europa-Universit\"at Flensburg)
Abstract: We describe two systems currently being developed that use large language models for the automatized correction of (i) exercises in translating back and forth between natural language and the languages of propositional logic and first-order predicate logic and (ii) exercises in writing simple arguments in natural language in non-mathematical scenarios.
Authors: Yunlong Feng, Bohan Li, Libo Qin, Xiao Xu, Wanxiang Che
Abstract: Cross-domain text classification aims to adapt models to a target domain that lacks labeled data. It leverages or reuses rich labeled data from the different but related source domain(s) and unlabeled data from the target domain. To this end, previous work focuses on either extracting domain-invariant features or task-agnostic features, ignoring domain-aware features that may be present in the target domain and could be useful for the downstream task. In this paper, we propose a two-stage framework for cross-domain text classification. In the first stage, we finetune the model with mask language modeling (MLM) and labeled data from the source domain. In the second stage, we further fine-tune the model with self-supervised distillation (SSD) and unlabeled data from the target domain. We evaluate its performance on a public cross-domain text classification benchmark and the experiment results show that our method achieves new state-of-the-art results for both single-source domain adaptations (94.17% $\uparrow$1.03%) and multi-source domain adaptations (95.09% $\uparrow$1.34%).
Authors: Yuxian Gu, Li Dong, Furu Wei, Minlie Huang
Abstract: Knowledge Distillation (KD) is a promising technique for reducing the high computational demand of large language models (LLMs). However, previous KD methods are primarily applied to white-box classification models or training small models to imitate black-box model APIs like ChatGPT. How to effectively distill the knowledge of white-box LLMs into small models is still under-explored, which becomes more important with the prosperity of open-source LLMs. In this work, we propose a KD approach that distills LLMs into smaller language models. We first replace the forward Kullback-Leibler divergence (KLD) objective in the standard KD approaches with reverse KLD, which is more suitable for KD on generative language models, to prevent the student model from overestimating the low-probability regions of the teacher distribution. Then, we derive an effective optimization approach to learn this objective. The student models are named MiniLLM. Extensive experiments in the instruction-following setting show that MiniLLM generates more precise responses with higher overall quality, lower exposure bias, better calibration, and higher long-text generation performance than the baselines. Our method is scalable for different model families with 120M to 13B parameters. Our code, data, and model checkpoints can be found in https://github.com/microsoft/LMOps/tree/main/minillm.
Authors: Linan Yue, Qi Liu, Yichao Du, Weibo Gao, Ye Liu, Fangzhou Yao
Abstract: Large Language Models (LLMs) have gained prominence in the field of Legal Intelligence, offering potential applications in assisting legal professionals and laymen. However, the centralized training of these Legal LLMs raises data privacy concerns, as legal data is distributed among various institutions containing sensitive individual information. This paper addresses this challenge by exploring the integration of Legal LLMs with Federated Learning (FL) methodologies. By employing FL, Legal LLMs can be fine-tuned locally on devices or clients, and their parameters are aggregated and distributed on a central server, ensuring data privacy without directly sharing raw data. However, computation and communication overheads hinder the full fine-tuning of LLMs under the FL setting. Moreover, the distribution shift of legal data reduces the effectiveness of FL methods. To this end, in this paper, we propose the first Federated Legal Large Language Model (FedJudge) framework, which fine-tunes Legal LLMs efficiently and effectively. Specifically, FedJudge utilizes parameter-efficient fine-tuning methods to update only a few additional parameters during the FL training. Besides, we explore the continual learning methods to preserve the global model's important parameters when training local clients to mitigate the problem of data shifts. Extensive experimental results on three real-world datasets clearly validate the effectiveness of FedJudge. Code is released at https://github.com/yuelinan/FedJudge.
Authors: Mikhail Galkin, Xinyu Yuan, Hesham Mostafa, Jian Tang, Zhaocheng Zhu
Abstract: Foundation models in language and vision have the ability to run inference on any textual and visual inputs thanks to the transferable representations such as a vocabulary of tokens in language. Knowledge graphs (KGs) have different entity and relation vocabularies that generally do not overlap. The key challenge of designing foundation models on KGs is to learn such transferable representations that enable inference on any graph with arbitrary entity and relation vocabularies. In this work, we make a step towards such foundation models and present ULTRA, an approach for learning universal and transferable graph representations. ULTRA builds relational representations as a function conditioned on their interactions. Such a conditioning strategy allows a pre-trained ULTRA model to inductively generalize to any unseen KG with any relation vocabulary and to be fine-tuned on any graph. Conducting link prediction experiments on 57 different KGs, we find that the zero-shot inductive inference performance of a single pre-trained ULTRA model on unseen graphs of various sizes is often on par or better than strong baselines trained on specific graphs. Fine-tuning further boosts the performance.
Authors: Zhiqing Sun, Yikang Shen, Hongxin Zhang, Qinhong Zhou, Zhenfang Chen, David Cox, Yiming Yang, Chuang Gan
Abstract: Supervised Fine-Tuning (SFT) on response demonstrations combined with Reinforcement Learning from Human Feedback (RLHF) constitutes a powerful paradigm for aligning LLM-based AI agents. However, a significant limitation of such an approach is its dependency on high-quality human annotations, making its application to intricate tasks challenging due to difficulties in obtaining consistent response demonstrations and in-distribution response preferences. This paper presents a novel approach, namely SALMON, to align base language models with minimal human supervision, using only a small set of human-defined principles, yet achieving superior performance. Central to our approach is an instructable reward model. Trained on synthetic preference data, this model can generate reward scores based on arbitrary human-defined principles. By merely adjusting these principles during the RL training phase, we gain full control over the preferences with the instructable reward model, subsequently influencing the behavior of the RL-trained policy models, and reducing the reliance on the collection of online human preferences. Applying our method to the LLaMA-2-70b base language model, we developed an AI assistant named Dromedary-2. With only 6 exemplars for in-context learning and 31 human-defined principles, Dromedary-2 significantly surpasses the performance of several state-of-the-art AI systems, including LLaMA-2-Chat-70b, on various benchmark datasets. We have open-sourced the code and model weights to encourage further research into aligning LLM-based AI agents with enhanced supervision efficiency, improved controllability, and scalable oversight.
Authors: Aaron Mueller, Albert Webson, Jackson Petty, Tal Linzen
Abstract: In-context learning (ICL) is now a common method for teaching large language models (LLMs) new tasks: given labeled examples in the input context, the LLM learns to perform the task without weight updates. Do models guided via ICL infer the underlying structure of the task defined by the context, or do they rely on superficial heuristics that only generalize to identically distributed examples? We address this question using transformations tasks and an NLI task that assess sensitivity to syntax - a requirement for robust language understanding. We further investigate whether out-of-distribution generalization can be improved via chain-of-thought prompting, where the model is provided with a sequence of intermediate computation steps that illustrate how the task ought to be performed. In experiments with models from the GPT, PaLM, and Llama 2 families, we find large variance across LMs. The variance is explained more by the composition of the pre-training corpus and supervision methods than by model size; in particular, models pre-trained on code generalize better, and benefit more from chain-of-thought prompting.
Authors: Chadi Helwe, Tom Calamai, Pierre-Henri Paris, Chlo\'e Clavel, Fabian Suchanek
Abstract: We introduce MAFALDA, a benchmark for fallacy classification that merges and unites previous fallacy datasets. It comes with a taxonomy that aligns, refines, and unifies existing classifications of fallacies. We further provide a manual annotation of a part of the dataset together with manual explanations for each annotation. We propose a new annotation scheme tailored for subjective NLP tasks, and a new evaluation method designed to handle subjectivity. We then evaluate several language models under a zero-shot learning setting and human performances on MAFALDA to assess their capability to detect and classify fallacies.
Authors: Vil\'em Zouhar, Ond\v{r}ej Bojar
Abstract: Automatic machine translation metrics typically rely on human translations to determine the quality of system translations. Common wisdom in the field dictates that the human references should be of very high quality. However, there are no cost-benefit analyses that could be used to guide practitioners who plan to collect references for machine translation evaluation. We find that higher-quality references lead to better metric correlations with humans at the segment-level. Having up to 7 references per segment and taking their average (or maximum) helps all metrics. Interestingly, the references from vendors of different qualities can be mixed together and improve metric success. Higher quality references, however, cost more to create and we frame this as an optimization problem: given a specific budget, what references should be collected to maximize metric success. These findings can be used by evaluators of shared tasks when references need to be created under a certain budget.
Authors: Michael Wornow, Alejandro Lozano, Dev Dash, Jenelle Jindal, Kenneth W. Mahaffey, Nigam H. Shah
Abstract: Matching patients to clinical trials is a key unsolved challenge in bringing new drugs to market. Today, identifying patients who meet a trial's eligibility criteria is highly manual, taking up to 1 hour per patient. Automated screening is challenging, however, as it requires understanding unstructured clinical text. Large language models (LLMs) offer a promising solution. In this work, we explore their application to trial matching. First, we design an LLM-based system which, given a patient's medical history as unstructured clinical text, evaluates whether that patient meets a set of inclusion criteria (also specified as free text). Our zero-shot system achieves state-of-the-art scores on the n2c2 2018 cohort selection benchmark. Second, we improve the data and cost efficiency of our method by identifying a prompting strategy which matches patients an order of magnitude faster and more cheaply than the status quo, and develop a two-stage retrieval pipeline that reduces the number of tokens processed by up to a third while retaining high performance. Third, we evaluate the interpretability of our system by having clinicians evaluate the natural language justifications generated by the LLM for each eligibility decision, and show that it can output coherent explanations for 97% of its correct decisions and 75% of its incorrect ones. Our results establish the feasibility of using LLMs to accelerate clinical trial operations.
Authors: Taichi Aida, Danushka Bollegala
Abstract: Detecting temporal semantic changes of words is an important task for various NLP applications that must make time-sensitive predictions. Lexical Semantic Change Detection (SCD) task involves predicting whether a given target word, $w$, changes its meaning between two different text corpora, $C_1$ and $C_2$. For this purpose, we propose a supervised two-staged SCD method that uses existing Word-in-Context (WiC) datasets. In the first stage, for a target word $w$, we learn two sense-aware encoders that represent the meaning of $w$ in a given sentence selected from a corpus. Next, in the second stage, we learn a sense-aware distance metric that compares the semantic representations of a target word across all of its occurrences in $C_1$ and $C_2$. Experimental results on multiple benchmark datasets for SCD show that our proposed method consistently outperforms all previously proposed SCD methods for multiple languages, establishing a novel state-of-the-art for SCD. Interestingly, our findings imply that there are specialised dimensions that carry information related to semantic changes of words in the sense-aware embedding space. Source code is available at https://github.com/a1da4/svp-sdml .
Authors: Yiding Liu, Jingjing Wang, Jiamin Luo, Tao Zeng, Guodong Zhou
Abstract: Aspect Sentiment Understanding (ASU) in interactive scenarios (e.g., Question-Answering and Dialogue) has attracted ever-more interest in recent years and achieved important progresses. However, existing studies on interactive ASU largely ignore the coreference issue for opinion targets (i.e., aspects), while this phenomenon is ubiquitous in interactive scenarios especially dialogues, limiting the ASU performance. Recently, large language models (LLMs) shows the powerful ability to integrate various NLP tasks with the chat paradigm. In this way, this paper proposes a new Chat-based Aspect Sentiment Understanding (ChatASU) task, aiming to explore LLMs' ability in understanding aspect sentiments in dialogue scenarios. Particularly, this ChatASU task introduces a sub-task, i.e., Aspect Chain Reasoning (ACR) task, to address the aspect coreference issue. On this basis, we propose a Trusted Self-reflexion Approach (TSA) with ChatGLM as backbone to ChatASU. Specifically, this TSA treats the ACR task as an auxiliary task to boost the performance of the primary ASU task, and further integrates trusted learning into reflexion mechanisms to alleviate the LLMs-intrinsic factual hallucination problem in TSA. Furthermore, a high-quality ChatASU dataset is annotated to evaluate TSA, and extensive experiments show that our proposed TSA can significantly outperform several state-of-the-art baselines, justifying the effectiveness of TSA to ChatASU and the importance of considering the coreference and hallucination issues in ChatASU.
Authors: Gemma Team, Thomas Mesnard, Cassidy Hardin, Robert Dadashi, Surya Bhupatiraju, Shreya Pathak, Laurent Sifre, Morgane Rivi\`ere, Mihir Sanjay Kale, Juliette Love, Pouya Tafti, L\'eonard Hussenot, Pier Giuseppe Sessa, Aakanksha Chowdhery, Adam Roberts, Aditya Barua, Alex Botev, Alex Castro-Ros, Ambrose Slone, Am\'elie H\'eliou, Andrea Tacchetti, Anna Bulanova, Antonia Paterson, Beth Tsai, Bobak Shahriari, Charline Le Lan, Christopher A. Choquette-Choo, Cl\'ement Crepy, Daniel Cer, Daphne Ippolito, David Reid, Elena Buchatskaya, Eric Ni, Eric Noland, Geng Yan, George Tucker, George-Christian Muraru, Grigory Rozhdestvenskiy, Henryk Michalewski, Ian Tenney, Ivan Grishchenko, Jacob Austin, James Keeling, Jane Labanowski, Jean-Baptiste Lespiau, Jeff Stanway, Jenny Brennan, Jeremy Chen, Johan Ferret, Justin Chiu, Justin Mao-Jones, Katherine Lee, Kathy Yu, Katie Millican, Lars Lowe Sjoesund, Lisa Lee, Lucas Dixon, Machel Reid, Maciej Miku{\l}a, Mateo Wirth, Michael Sharman, Nikolai Chinaev, Nithum Thain, Olivier Bachem, Oscar Chang, Oscar Wahltinez, Paige Bailey, Paul Michel, Petko Yotov, Rahma Chaabouni, Ramona Comanescu, Reena Jana, Rohan Anil, Ross McIlroy, Ruibo Liu, Ryan Mullins, Samuel L Smith, Sebastian Borgeaud, Sertan Girgin, Sholto Douglas, Shree Pandya, Siamak Shakeri, Soham De, Ted Klimenko, Tom Hennigan, Vlad Feinberg, Wojciech Stokowiec, Yu-hui Chen, Zafarali Ahmed, Zhitao Gong, Tris Warkentin, Ludovic Peran, Minh Giang, Cl\'ement Farabet, Oriol Vinyals, Jeff Dean, Koray Kavukcuoglu, Demis Hassabis, Zoubin Ghahramani, Douglas Eck, Joelle Barral, Fernando Pereira, Eli Collins, Armand Joulin, Noah Fiedel, Evan Senter, Alek Andreev, Kathleen Kenealy
Abstract: This work introduces Gemma, a family of lightweight, state-of-the art open models built from the research and technology used to create Gemini models. Gemma models demonstrate strong performance across academic benchmarks for language understanding, reasoning, and safety. We release two sizes of models (2 billion and 7 billion parameters), and provide both pretrained and fine-tuned checkpoints. Gemma outperforms similarly sized open models on 11 out of 18 text-based tasks, and we present comprehensive evaluations of safety and responsibility aspects of the models, alongside a detailed description of model development. We believe the responsible release of LLMs is critical for improving the safety of frontier models, and for enabling the next wave of LLM innovations.
Authors: Rustem Yeshpanov, Huseyin Atakan Varol
Abstract: This paper presents KazSAnDRA, a dataset developed for Kazakh sentiment analysis that is the first and largest publicly available dataset of its kind. KazSAnDRA comprises an extensive collection of 180,064 reviews obtained from various sources and includes numerical ratings ranging from 1 to 5, providing a quantitative representation of customer attitudes. The study also pursued the automation of Kazakh sentiment classification through the development and evaluation of four machine learning models trained for both polarity classification and score classification. Experimental analysis included evaluation of the results considering both balanced and imbalanced scenarios. The most successful model attained an F1-score of 0.81 for polarity classification and 0.39 for score classification on the test sets. The dataset and fine-tuned models are open access and available for download under the Creative Commons Attribution 4.0 International License (CC BY 4.0) through our GitHub repository.
Authors: Rustem Yeshpanov, Alina Polonskaya, Huseyin Atakan Varol
Abstract: We introduce KazParC, a parallel corpus designed for machine translation across Kazakh, English, Russian, and Turkish. The first and largest publicly available corpus of its kind, KazParC contains a collection of 371,902 parallel sentences covering different domains and developed with the assistance of human translators. Our research efforts also extend to the development of a neural machine translation model nicknamed Tilmash. Remarkably, the performance of Tilmash is on par with, and in certain instances, surpasses that of industry giants, such as Google Translate and Yandex Translate, as measured by standard evaluation metrics, such as BLEU and chrF. Both KazParC and Tilmash are openly available for download under the Creative Commons Attribution 4.0 International License (CC BY 4.0) through our GitHub repository.
Authors: Ziyang Wang, Sanwoo Lee, Hsiu-Yuan Huang, Yunfang Wu
Abstract: Prompt-based methods have achieved promising results in most few-shot text classification tasks. However, for readability assessment tasks, traditional prompt methods lackcrucial linguistic knowledge, which has already been proven to be essential. Moreover, previous studies on utilizing linguistic features have shown non-robust performance in few-shot settings and may even impair model performance.To address these issues, we propose a novel prompt-based tuning framework that incorporates rich linguistic knowledge, called Feature Prompt Tuning (FPT). Specifically, we extract linguistic features from the text and embed them into trainable soft prompts. Further, we devise a new loss function to calibrate the similarity ranking order between categories. Experimental results demonstrate that our proposed method FTP not only exhibits a significant performance improvement over the prior best prompt-based tuning approaches, but also surpasses the previous leading methods that incorporate linguistic features. Also, our proposed model significantly outperforms the large language model gpt-3.5-turbo-16k in most cases. Our proposed method establishes a new architecture for prompt tuning that sheds light on how linguistic features can be easily adapted to linguistic-related tasks.
Authors: Hyewon Jang, Diego Frassinelli
Abstract: We tested the robustness of sarcasm detection models by examining their behavior when fine-tuned on four sarcasm datasets containing varying characteristics of sarcasm: label source (authors vs. third-party), domain (social media/online vs. offline conversations/dialogues), style (aggressive vs. humorous mocking). We tested their prediction performance on the same dataset (intra-dataset) and across different datasets (cross-dataset). For intra-dataset predictions, models consistently performed better when fine-tuned with third-party labels rather than with author labels. For cross-dataset predictions, most models failed to generalize well to the other datasets, implying that one type of dataset cannot represent all sorts of sarcasm with different styles and domains. Compared to the existing datasets, models fine-tuned on the new dataset we release in this work showed the highest generalizability to other datasets. With a manual inspection of the datasets and post-hoc analysis, we attributed the difficulty in generalization to the fact that sarcasm actually comes in different domains and styles. We argue that future sarcasm research should take the broad scope of sarcasm into account.
Authors: Zhenhailong Wang, Joy Hsu, Xingyao Wang, Kuan-Hao Huang, Manling Li, Jiajun Wu, Heng Ji
Abstract: While large multimodal models excel in broad vision-language benchmarks, they often struggle with tasks requiring precise perception of low-level visual details, such as comparing line lengths or solving simple mazes. In particular, this failure mode persists in question-answering tasks about vector graphics -- images composed purely of 2D objects and shapes. To address this challenge, we propose the Visually Descriptive Language Model (VDLM), which performs text-based reasoning about vector graphics. VDLM leverages Scalable Vector Graphics (SVG) for a more precise visual description and first uses an off-the-shelf raster-to-SVG algorithm for encoding. Since existing language models cannot understand raw SVGs in a zero-shot setting, VDLM then bridges SVG with pretrained language models through a newly introduced intermediate symbolic representation, Primal Visual Description (PVD), comprising primitive attributes (e.g., shape, position, measurement) with their corresponding predicted values. PVD is task-agnostic and represents visual primitives that are universal across all vector graphics. It can be learned with procedurally generated (SVG, PVD) pairs and also enables the direct use of LLMs for generalization to complex reasoning tasks. By casting an image to a text-based representation, we can leverage the power of language models to learn alignment from SVG to visual primitives and generalize to unseen question-answering tasks. Empirical results show that VDLM achieves stronger zero-shot performance compared to state-of-the-art LMMs, such as GPT-4V, in various low-level multimodal perception and reasoning tasks on vector graphics. We additionally present extensive analyses on VDLM's performance, demonstrating that our framework offers better interpretability due to its disentangled perception and reasoning processes. Project page: https://mikewangwzhl.github.io/VDLM/
Authors: Chonghua Wang, Haodong Duan, Songyang Zhang, Dahua Lin, Kai Chen
Abstract: Recently, the large language model (LLM) community has shown increasing interest in enhancing LLMs' capability to handle extremely long documents. As various long-text techniques and model architectures emerge, the precise and detailed evaluation of models' long-text capabilities has become increasingly important. Existing long-text evaluation benchmarks, such as L-Eval and LongBench, construct long-text test sets based on open-source datasets, focusing mainly on QA and summarization tasks. These datasets include test samples of varying lengths (from 2k to 32k+) entangled together, making it challenging to assess model capabilities across different length ranges. Moreover, they do not cover the ultralong settings (100k+ tokens) that the latest LLMs claim to achieve. In this paper, we introduce Ada-LEval, a length-adaptable benchmark for evaluating the long-context understanding of LLMs. Ada-LEval includes two challenging subsets, TSort and BestAnswer, which enable a more reliable evaluation of LLMs' long context capabilities. These benchmarks support intricate manipulation of the length of test cases, and can easily produce text samples up to 128k tokens. We evaluate 4 state-of-the-art closed-source API models and 6 open-source models with Ada-LEval. The evaluation results demonstrate the limitations of current LLMs, especially in ultra-long-context settings. Our code is available at https://github.com/open-compass/Ada-LEval.
Authors: Mehmet Saygin Seyfioglu, Wisdom O. Ikezogwo, Fatemeh Ghezloo, Ranjay Krishna, Linda Shapiro
Abstract: Diagnosis in histopathology requires a global whole slide images (WSIs) analysis, requiring pathologists to compound evidence from different WSI patches. The gigapixel scale of WSIs poses a challenge for histopathology multi-modal models. Training multi-model models for histopathology requires instruction tuning datasets, which currently contain information for individual image patches, without a spatial grounding of the concepts within each patch and without a wider view of the WSI. Therefore, they lack sufficient diagnostic capacity for histopathology. To bridge this gap, we introduce Quilt-Instruct, a large-scale dataset of 107,131 histopathology-specific instruction question/answer pairs, grounded within diagnostically relevant image patches that make up the WSI. Our dataset is collected by leveraging educational histopathology videos from YouTube, which provides spatial localization of narrations by automatically extracting the narrators' cursor positions. Quilt-Instruct supports contextual reasoning by extracting diagnosis and supporting facts from the entire WSI. Using Quilt-Instruct, we train Quilt-LLaVA, which can reason beyond the given single image patch, enabling diagnostic reasoning across patches. To evaluate Quilt-LLaVA, we propose a comprehensive evaluation dataset created from 985 images and 1283 human-generated question-answers. We also thoroughly evaluate Quilt-LLaVA using public histopathology datasets, where Quilt-LLaVA significantly outperforms SOTA by over 10% on relative GPT-4 score and 4% and 9% on open and closed set VQA. Our code, data, and model are publicly accessible at quilt-llava.github.io.
Authors: Xingfang Wu, Eric Laufer, Heng Li, Foutse Khomh, Santhosh Srinivasan, Jayden Luo
Abstract: With the rapid growth of the developer community, the amount of posts on online technical forums has been growing rapidly, which poses difficulties for users to filter useful posts and find important information. Tags provide a concise feature dimension for users to locate their interested posts and for search engines to index the most relevant posts according to the queries. However, most tags are only focused on the technical perspective (e.g., program language, platform, tool). In most cases, forum posts in online developer communities reveal the author's intentions to solve a problem, ask for advice, share information, etc. The modeling of the intentions of posts can provide an extra dimension to the current tag taxonomy. By referencing previous studies and learning from industrial perspectives, we create a refined taxonomy for the intentions of technical forum posts. Through manual labeling and analysis on a sampled post dataset extracted from online forums, we understand the relevance between the constitution of posts (code, error messages) and their intentions. Furthermore, inspired by our manual study, we design a pre-trained transformer-based model to automatically predict post intentions. The best variant of our intention prediction framework, which achieves a Micro F1-score of 0.589, Top 1-3 accuracy of 62.6% to 87.8%, and an average AUC of 0.787, outperforms the state-of-the-art baseline approach. Our characterization and automated classification of forum posts regarding their intentions may help forum maintainers or third-party tool developers improve the organization and retrieval of posts on technical forums. We have released our annotated dataset and codes in our supplementary material package.
Authors: Maitreya Patel, Sangmin Jung, Chitta Baral, Yezhou Yang
Abstract: Despite the recent advances in personalized text-to-image (P-T2I) generative models, it remains challenging to perform finetuning-free multi-subject-driven T2I in a resource-efficient manner. Predominantly, contemporary approaches, involving the training of Hypernetworks and Multimodal Large Language Models (MLLMs), require heavy computing resources that range from 600 to 12300 GPU hours of training. These subject-driven T2I methods hinge on Latent Diffusion Models (LDMs), which facilitate T2I mapping through cross-attention layers. While LDMs offer distinct advantages, P-T2I methods' reliance on the latent space of these diffusion models significantly escalates resource demands, leading to inconsistent results and necessitating numerous iterations for a single desired image. In this paper, we present $\lambda$-ECLIPSE, an alternative prior-training strategy that works in the latent space of a pre-trained CLIP model without relying on the diffusion UNet models. $\lambda$-ECLIPSE leverages the image-text interleaved pre-training for fast and effective multi-subject-driven P-T2I. Through extensive experiments, we establish that $\lambda$-ECLIPSE surpasses existing baselines in composition alignment while preserving concept alignment performance, even with significantly lower resource utilization. $\lambda$-ECLIPSE performs multi-subject driven P-T2I with just 34M parameters and is trained on a mere 74 GPU hours. Additionally, $\lambda$-ECLIPSE demonstrates the unique ability to perform multi-concept interpolations.
Authors: Vishaal Udandarao, Ameya Prabhu, Adhiraj Ghosh, Yash Sharma, Philip H. S. Torr, Adel Bibi, Samuel Albanie, Matthias Bethge
Abstract: Web-crawled pretraining datasets underlie the impressive "zero-shot" evaluation performance of multimodal models, such as CLIP for classification/retrieval and Stable-Diffusion for image generation. However, it is unclear how meaningful the notion of "zero-shot" generalization is for such multimodal models, as it is not known to what extent their pretraining datasets encompass the downstream concepts targeted for during "zero-shot" evaluation. In this work, we ask: How is the performance of multimodal models on downstream concepts influenced by the frequency of these concepts in their pretraining datasets? We comprehensively investigate this question across 34 models and five standard pretraining datasets (CC-3M, CC-12M, YFCC-15M, LAION-400M, LAION-Aesthetics), generating over 300GB of data artifacts. We consistently find that, far from exhibiting "zero-shot" generalization, multimodal models require exponentially more data to achieve linear improvements in downstream "zero-shot" performance, following a sample inefficient log-linear scaling trend. This trend persists even when controlling for sample-level similarity between pretraining and downstream datasets, and testing on purely synthetic data distributions. Furthermore, upon benchmarking models on long-tailed data sampled based on our analysis, we demonstrate that multimodal models across the board perform poorly. We contribute this long-tail test set as the "Let it Wag!" benchmark to further research in this direction. Taken together, our study reveals an exponential need for training data which implies that the key to "zero-shot" generalization capabilities under large-scale training paradigms remains to be found.