Most interpretability research in NLP focuses on understanding the behavior and features of a fully trained model. However, certain insights into model behavior may only be accessible by observing the trajectory of the training process. We present a case study of syntax acquisition in masked language models (MLMs) that demonstrates how analyzing the evolution of interpretable artifacts throughout training deepens our understanding of emergent behavior. In particular, we study Syntactic Attention Structure (SAS), a naturally emerging property of MLMs wherein specific Transformer heads tend to focus on specific syntactic relations. We identify a brief window in pretraining when models abruptly acquire SAS, concurrent with a steep drop in loss. This breakthrough precipitates the subsequent acquisition of linguistic capabilities. We then examine the causal role of SAS by manipulating SAS during training, and demonstrate that SAS is necessary for the development of grammatical capabilities. We further find that SAS competes with other beneficial traits during training, and that briefly suppressing SAS improves model quality. These findings offer an interpretation of a real-world example of both simplicity bias and breakthrough training dynamics.
相關內容
This study investigates self-supervised learning techniques to obtain representations of Event Sequences. It is a key modality in various applications, including but not limited to banking, e-commerce, and healthcare. We perform a comprehensive study of generative and contrastive approaches in self-supervised learning, applying them both independently. We find that there is no single supreme method. Consequently, we explore the potential benefits of combining these approaches. To achieve this goal, we introduce a novel method that aligns generative and contrastive embeddings as distinct modalities, drawing inspiration from contemporary multimodal research. Generative and contrastive approaches are often treated as mutually exclusive, leaving a gap for their combined exploration. Our results demonstrate that this aligned model performs at least on par with, and mostly surpasses, existing methods and is more universal across a variety of tasks. Furthermore, we demonstrate that self-supervised methods consistently outperform the supervised approach on our datasets.
Unbiased Learning to Rank (ULTR) aims to train unbiased ranking models from biased click logs, by explicitly modeling a generation process for user behavior and fitting click data based on examination hypothesis. Previous research found empirically that the true latent relevance is mostly recoverable through perfect click fitting. However, we demonstrate that this is not always achievable, resulting in a significant reduction in ranking performance. This research investigates the conditions under which relevance can be recovered from click data at a foundational level. We initially characterize a ranking model as identifiable if it can recover the true relevance up to a scaling transformation, a criterion sufficient for the pairwise ranking objective. Subsequently, we investigate an equivalent condition for identifiability, articulated as a graph connectivity test problem: the recovery of relevance is feasible if and only if the identifiability graph (IG), derived from the underlying structure of the dataset, is connected. The presence of a disconnected IG may lead to degenerate cases and suboptimal ranking performance. To tackle this challenge, we introduce two methods, namely node intervention and node merging, designed to modify the dataset and restore the connectivity of the IG. Empirical results derived from a simulated dataset and two real-world LTR benchmark datasets not only validate our proposed theorems but also demonstrate the effectiveness of our methods in alleviating data bias when the relevance model is unidentifiable.
The present study puts forward a novel biographical knowledge graph (KG) on Prof. S. R. Ranganathan, one of the pioneering figures in the Library and Information Science (LIS) domain. It has been found that most of the relevant facts about Ranganathan exist in a variety of resources (e.g., books, essays, journal articles, websites, blogs, etc.), offering information in a fragmented and piecemeal way. With this dedicated KG (henceforth known as RKG), we hope to furnish a 360-degree view of his life and achievements. To the best of our knowledge, such a dedicated representation is unparalleled in its scope and coverage: using state-of-the-art technology for anyone to openly access, use/re-use, and contribute. Inspired by Ranganathan's theories and ideas, the KG was developed using a "facet-based methodology" at two levels: in the identification of the vital biographical aspects and the development of the ontological model. Finally, with this study, we call for a community-driven effort to enhance the KG and pay homage to the Father of Library Science on the hundredth anniversary of his revitalizing the LIS domain through his enduring participation.
The burgeoning field of Shared Autonomous Vehicles (SAVs) presents transformative potential for the transport sector, subject to public acceptance. Traditional acceptance models, primarily reliant on Structural Equation Modelling (SEM), often fall short of capturing the complex, non-linear dynamics underlying this acceptance. To address these limitations, this paper proposes a Machine Learning (ML) approach to predict public acceptance of SAVs and employs a chord diagram to visualize the influence of different predictors. This approach reveals nuanced, non-linear relationships between factors at both macro and micro levels, and identifies attitude as the primary predictor of SAV usage intention, followed by perceived risk, perceived usefulness, trust, and perceived ease of use. The framework also uncovers divergent perceptions of these factors among SAV adopters and non-adopters, providing granular insights for strategic initiatives to enhance SAV acceptance. Using SAV acceptance as a case study, our findings contribute a novel, machine learning-based perspective to the discourse on technology acceptance, underscoring the importance of nuanced, data-driven approaches in understanding and fostering public acceptance of emerging transport technologies.
This paper studies the theoretical framework of the alignment process of generative models with Reinforcement Learning from Human Feedback (RLHF). We consider a standard mathematical formulation, the reverse-KL regularized contextual bandit for RLHF. Despite its widespread practical application, a rigorous theoretical analysis of this formulation remains open. We investigate its behavior in three distinct settings -- offline, online, and hybrid -- and propose efficient algorithms with finite-sample theoretical guarantees. Moving towards practical applications, our framework, with a robust approximation of the information-theoretical policy improvement oracle, naturally gives rise to several novel RLHF algorithms. This includes an iterative version of the Direct Preference Optimization (DPO) algorithm for online settings, and a multi-step rejection sampling strategy for offline scenarios. Our empirical evaluations on real-world alignment experiment of large language model demonstrate that these proposed methods significantly surpass existing strong baselines, such as DPO and Rejection Sampling Optimization (RSO), showcasing the connections between solid theoretical foundations and their powerful practical implementations.
RoboCup represents an International testbed for advancing research in AI and robotics, focusing on a definite goal: developing a robot team that can win against the human world soccer champion team by the year 2050. To achieve this goal, autonomous humanoid robots' coordination is crucial. This paper explores novel solutions within the RoboCup Standard Platform League (SPL), where a reduction in WiFi communication is imperative, leading to the development of new coordination paradigms. The SPL has experienced a substantial decrease in network packet rate, compelling the need for advanced coordination architectures to maintain optimal team functionality in dynamic environments. Inspired by market-based task assignment, we introduce a novel distributed coordination system to orchestrate autonomous robots' actions efficiently in low communication scenarios. This approach has been tested with NAO robots during official RoboCup competitions and in the SimRobot simulator, demonstrating a notable reduction in task overlaps in limited communication settings.
We study the problem of allocating indivisible resources under the connectivity constraints of a graph $G$. This model, initially introduced by Bouveret et al. (published in IJCAI, 2017), effectively encompasses a diverse array of scenarios characterized by spatial or temporal limitations, including the division of land plots and the allocation of time plots. In this paper, we introduce a novel fairness concept that integrates local comparisons within the social network formed by a connected allocation of the item graph. Our particular focus is to achieve pairwise-maximin fair share (PMMS) among the "neighbors" within this network. For any underlying graph structure, we show that a connected allocation that maximizes Nash welfare guarantees a $(1/2)$-PMMS fairness. Moreover, for two agents, we establish that a $(3/4)$-PMMS allocation can be efficiently computed. Additionally, we demonstrate that for three agents and the items aligned on a path, a PMMS allocation is always attainable and can be computed in polynomial time. Lastly, when agents have identical additive utilities, we present a pseudo-polynomial-time algorithm for a $(3/4)$-PMMS allocation, irrespective of the underlying graph $G$. Furthermore, we provide a polynomial-time algorithm for obtaining a PMMS allocation when $G$ is a tree.
Recent studies of the emergent capabilities of transformer-based Natural Language Understanding (NLU) models have indicated that they have an understanding of lexical and compositional semantics. We provide evidence that suggests these claims should be taken with a grain of salt: we find that state-of-the-art Natural Language Inference (NLI) models are sensitive towards minor semantics preserving surface-form variations, which lead to sizable inconsistent model decisions during inference. Notably, this behaviour differs from valid and in-depth comprehension of compositional semantics, however does neither emerge when evaluating model accuracy on standard benchmarks nor when probing for syntactic, monotonic, and logically robust reasoning. We propose a novel framework to measure the extent of semantic sensitivity. To this end, we evaluate NLI models on adversarially generated examples containing minor semantics-preserving surface-form input noise. This is achieved using conditional text generation, with the explicit condition that the NLI model predicts the relationship between the original and adversarial inputs as a symmetric equivalence entailment. We systematically study the effects of the phenomenon across NLI models for \emph{in-} and \emph{out-of} domain settings. Our experiments show that semantic sensitivity causes performance degradations of $12.92\%$ and $23.71\%$ average over \emph{in-} and \emph{out-of-} domain settings, respectively. We further perform ablation studies, analysing this phenomenon across models, datasets, and variations in inference and show that semantic sensitivity can lead to major inconsistency within model predictions.
Multimodality Representation Learning, as a technique of learning to embed information from different modalities and their correlations, has achieved remarkable success on a variety of applications, such as Visual Question Answering (VQA), Natural Language for Visual Reasoning (NLVR), and Vision Language Retrieval (VLR). Among these applications, cross-modal interaction and complementary information from different modalities are crucial for advanced models to perform any multimodal task, e.g., understand, recognize, retrieve, or generate optimally. Researchers have proposed diverse methods to address these tasks. The different variants of transformer-based architectures performed extraordinarily on multiple modalities. This survey presents the comprehensive literature on the evolution and enhancement of deep learning multimodal architectures to deal with textual, visual and audio features for diverse cross-modal and modern multimodal tasks. This study summarizes the (i) recent task-specific deep learning methodologies, (ii) the pretraining types and multimodal pretraining objectives, (iii) from state-of-the-art pretrained multimodal approaches to unifying architectures, and (iv) multimodal task categories and possible future improvements that can be devised for better multimodal learning. Moreover, we prepare a dataset section for new researchers that covers most of the benchmarks for pretraining and finetuning. Finally, major challenges, gaps, and potential research topics are explored. A constantly-updated paperlist related to our survey is maintained at //github.com/marslanm/multimodality-representation-learning.
A fundamental goal of scientific research is to learn about causal relationships. However, despite its critical role in the life and social sciences, causality has not had the same importance in Natural Language Processing (NLP), which has traditionally placed more emphasis on predictive tasks. This distinction is beginning to fade, with an emerging area of interdisciplinary research at the convergence of causal inference and language processing. Still, research on causality in NLP remains scattered across domains without unified definitions, benchmark datasets and clear articulations of the remaining challenges. In this survey, we consolidate research across academic areas and situate it in the broader NLP landscape. We introduce the statistical challenge of estimating causal effects, encompassing settings where text is used as an outcome, treatment, or as a means to address confounding. In addition, we explore potential uses of causal inference to improve the performance, robustness, fairness, and interpretability of NLP models. We thus provide a unified overview of causal inference for the computational linguistics community.
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191