Existing multi-relational graph neural networks use one of two strategies for identifying informative relations: either they reduce this problem to low-level weight learning, or they rely on handcrafted chains of relational dependencies, called meta-paths. However, the former approach faces challenges in the presence of many relations (e.g., knowledge graphs), while the latter requires substantial domain expertise to identify relevant meta-paths. In this work we propose a novel approach to learn meta-paths and meta-path GNNs that are highly accurate based on a small number of informative meta-paths. Key element of our approach is a scoring function for measuring the potential informativeness of a relation in the incremental construction of the meta-path. Our experimental evaluation shows that the approach manages to correctly identify relevant meta-paths even with a large number of relations, and substantially outperforms existing multi-relational GNNs on synthetic and real-world experiments.
In the past, the dichotomy between homophily and heterophily has inspired research contributions toward a better understanding of Deep Graph Networks' inductive bias. In particular, it was believed that homophily strongly correlates with better node classification predictions of message-passing methods. More recently, however, researchers pointed out that such dichotomy is too simplistic as we can construct node classification tasks where graphs are completely heterophilic but the performances remain high. Most of these works have also proposed new quantitative metrics to understand when a graph structure is useful, which implicitly or explicitly assume the correlation between node features and target labels. Our work empirically investigates what happens when this strong assumption does not hold, by formalising two generative processes for node classification tasks that allow us to build and study ad-hoc problems. To quantitatively measure the influence of the node features on the target labels, we also use a metric we call Feature Informativeness. We construct six synthetic tasks and evaluate the performance of six models, including structure-agnostic ones. Our findings reveal that previously defined metrics are not adequate when we relax the above assumption. Our contribution to the workshop aims at presenting novel research findings that could help advance our understanding of the field.
The rise of large language models (LLMs) had a transformative impact on search, ushering in a new era of search engines that are capable of generating search results in natural language text, imbued with citations for supporting sources. Building generative information-seeking models demands openly accessible datasets, which currently remain lacking. In this paper, we introduce a new dataset, HAGRID (Human-in-the-loop Attributable Generative Retrieval for Information-seeking Dataset) for building end-to-end generative information-seeking models that are capable of retrieving candidate quotes and generating attributed explanations. Unlike recent efforts that focus on human evaluation of black-box proprietary search engines, we built our dataset atop the English subset of MIRACL, a publicly available information retrieval dataset. HAGRID is constructed based on human and LLM collaboration. We first automatically collect attributed explanations that follow an in-context citation style using an LLM, i.e. GPT-3.5. Next, we ask human annotators to evaluate the LLM explanations based on two criteria: informativeness and attributability. HAGRID serves as a catalyst for the development of information-seeking models with better attribution capabilities.
Active domain adaptation (ADA) aims to improve the model adaptation performance by incorporating active learning (AL) techniques to label a maximally-informative subset of target samples. Conventional AL methods do not consider the existence of domain shift, and hence, fail to identify the truly valuable samples in the context of domain adaptation. To accommodate active learning and domain adaption, the two naturally different tasks, in a collaborative framework, we advocate that a customized learning strategy for the target data is the key to the success of ADA solutions. We present Divide-and-Adapt (DiaNA), a new ADA framework that partitions the target instances into four categories with stratified transferable properties. With a novel data subdivision protocol based on uncertainty and domainness, DiaNA can accurately recognize the most gainful samples. While sending the informative instances for annotation, DiaNA employs tailored learning strategies for the remaining categories. Furthermore, we propose an informativeness score that unifies the data partitioning criteria. This enables the use of a Gaussian mixture model (GMM) to automatically sample unlabeled data into the proposed four categories. Thanks to the "divideand-adapt" spirit, DiaNA can handle data with large variations of domain gap. In addition, we show that DiaNA can generalize to different domain adaptation settings, such as unsupervised domain adaptation (UDA), semi-supervised domain adaptation (SSDA), source-free domain adaptation (SFDA), etc.
The process of annotating histological gigapixel-sized whole slide images (WSIs) at the pixel level for the purpose of training a supervised segmentation model is time-consuming. Region-based active learning (AL) involves training the model on a limited number of annotated image regions instead of requesting annotations of the entire images. These annotation regions are iteratively selected, with the goal of optimizing model performance while minimizing the annotated area. The standard method for region selection evaluates the informativeness of all square regions of a specified size and then selects a specific quantity of the most informative regions. We find that the efficiency of this method highly depends on the choice of AL step size (i.e., the combination of region size and the number of selected regions per WSI), and a suboptimal AL step size can result in redundant annotation requests or inflated computation costs. This paper introduces a novel technique for selecting annotation regions adaptively, mitigating the reliance on this AL hyperparameter. Specifically, we dynamically determine each region by first identifying an informative area and then detecting its optimal bounding box, as opposed to selecting regions of a uniform predefined shape and size as in the standard method. We evaluate our method using the task of breast cancer metastases segmentation on the public CAMELYON16 dataset and show that it consistently achieves higher sampling efficiency than the standard method across various AL step sizes. With only 2.6\% of tissue area annotated, we achieve full annotation performance and thereby substantially reduce the costs of annotating a WSI dataset. The source code is available at https://github.com/DeepMicroscopy/AdaptiveRegionSelection.
Given a limited labeling budget, active learning (AL) aims to sample the most informative instances from an unlabeled pool to acquire labels for subsequent model training. To achieve this, AL typically measures the informativeness of unlabeled instances based on uncertainty and diversity. However, it does not consider erroneous instances with their neighborhood error density, which have great potential to improve the model performance. To address this limitation, we propose $REAL$, a novel approach to select data instances with $\underline{R}$epresentative $\underline{E}$rrors for $\underline{A}$ctive $\underline{L}$earning. It identifies minority predictions as \emph{pseudo errors} within a cluster and allocates an adaptive sampling budget for the cluster based on estimated error density. Extensive experiments on five text classification datasets demonstrate that $REAL$ consistently outperforms all best-performing baselines regarding accuracy and F1-macro scores across a wide range of hyperparameter settings. Our analysis also shows that $REAL$ selects the most representative pseudo errors that match the distribution of ground-truth errors along the decision boundary. Our code is publicly available at https://github.com/withchencheng/ECML_PKDD_23_Real.
Despite the popularity and success of deep learning, there is limited understanding of when, how, and why neural networks generalize to unseen examples. Since learning can be seen as extracting information from data, we formally study information captured by neural networks during training. Specifically, we start with viewing learning in presence of noisy labels from an information-theoretic perspective and derive a learning algorithm that limits label noise information in weights. We then define a notion of unique information that an individual sample provides to the training of a deep network, shedding some light on the behavior of neural networks on examples that are atypical, ambiguous, or belong to underrepresented subpopulations. We relate example informativeness to generalization by deriving nonvacuous generalization gap bounds. Finally, by studying knowledge distillation, we highlight the important role of data and label complexity in generalization. Overall, our findings contribute to a deeper understanding of the mechanisms underlying neural network generalization.
We develop an Empirical Bayes grading scheme that balances the informativeness of the assigned grades against the expected frequency of ranking errors. Applying the method to a massive correspondence experiment, we grade the racial biases of 97 U.S. employers. A four-grade ranking limits the chances that a randomly selected pair of firms is mis-ranked to 5% while explaining nearly half of the variation in firms' racial contact gaps. The grades are presented alongside measures of uncertainty about each firm's contact gap in an accessible rubric that is easily adapted to other settings where ranks and levels are of simultaneous interest.
Time series classification is a task which deals with temporal sequences, a prevalent data type common in domains such as human activity recognition, sports analytics and general sensing. In this area, interest in explainability has been growing as explanation is key to understand the data and the model better. Recently, a great variety of techniques have been proposed and adapted for time series to provide explanation in the form of saliency maps, where the importance of each data point in the time series is quantified with a numerical value. However, the saliency maps can and often disagree, so it is unclear which one to use. This paper provides a novel framework to quantitatively evaluate and rank explanation methods for time series classification. We show how to robustly evaluate the informativeness of a given explanation method (i.e., relevance for the classification task), and how to compare explanations side-by-side. The goal is to recommend the best explainer for a given time series classification dataset. We propose AMEE, a Model-Agnostic Explanation Evaluation framework, for recommending saliency-based explanations for time series classification. In this approach, data perturbation is added to the input time series guided by each explanation. Our results show that perturbing discriminative parts of the time series leads to significant changes in classification accuracy, which can be used to evaluate each explanation. To be robust to different types of perturbations and different types of classifiers, we aggregate the accuracy loss across perturbations and classifiers. This novel approach allows us to recommend the best explainer among a set of different explainers, including random and oracle explainers. We provide a quantitative and qualitative analysis for synthetic datasets, a variety of timeseries datasets, as well as a real-world case study with known expert ground truth.
The success of NLP systems often relies on the availability of large, high-quality datasets. However, not all samples in these datasets are equally valuable for learning, as some may be redundant or noisy. Several methods for characterizing datasets based on model-driven meta-information (e.g., model's confidence) have been developed, but the relationship and complementary effects of these methods have received less attention. In this paper, we introduce infoVerse, a universal framework for dataset characterization, which provides a new feature space that effectively captures multidimensional characteristics of datasets by incorporating various model-driven meta-information. infoVerse reveals distinctive regions of the dataset that are not apparent in the original semantic space, hence guiding users (or models) in identifying which samples to focus on for exploration, assessment, or annotation. Additionally, we propose a novel sampling method on infoVerse to select a set of data points that maximizes informativeness. In three real-world applications (data pruning, active learning, and data annotation), the samples chosen on infoVerse space consistently outperform strong baselines in all applications. Our code and demo are publicly available.
Recent research has focused on using large language models (LLMs) to generate explanations for hate speech through fine-tuning or prompting. Despite the growing interest in this area, these generated explanations' effectiveness and potential limitations remain poorly understood. A key concern is that these explanations, generated by LLMs, may lead to erroneous judgments about the nature of flagged content by both users and content moderators. For instance, an LLM-generated explanation might inaccurately convince a content moderator that a benign piece of content is hateful. In light of this, we propose an analytical framework for examining hate speech explanations and conducted an extensive survey on evaluating such explanations. Specifically, we prompted GPT-3 to generate explanations for both hateful and non-hateful content, and a survey was conducted with 2,400 unique respondents to evaluate the generated explanations. Our findings reveal that (1) human evaluators rated the GPT-generated explanations as high quality in terms of linguistic fluency, informativeness, persuasiveness, and logical soundness, (2) the persuasive nature of these explanations, however, varied depending on the prompting strategy employed, and (3) this persuasiveness may result in incorrect judgments about the hatefulness of the content. Our study underscores the need for caution in applying LLM-generated explanations for content moderation. Code and results are available at https://github.com/Social-AI-Studio/GPT3-HateEval.
Replay-based methods have proved their effectiveness on online continual learning by rehearsing past samples from an auxiliary memory. With many efforts made on improving training schemes based on the memory, however, the information carried by each sample in the memory remains under-investigated. Under circumstances with restricted storage space, the informativeness of the memory becomes critical for effective replay. Although some works design specific strategies to select representative samples, by only employing a small number of original images, the storage space is still not well utilized. To this end, we propose to Summarize the knowledge from the Stream Data (SSD) into more informative samples by distilling the training characteristics of real images. Through maintaining the consistency of training gradients and relationship to the past tasks, the summarized samples are more representative for the stream data compared to the original images. Extensive experiments are conducted on multiple online continual learning benchmarks to support that the proposed SSD method significantly enhances the replay effects. We demonstrate that with limited extra computational overhead, SSD provides more than 3% accuracy boost for sequential CIFAR-100 under extremely restricted memory buffer. Code in https://github.com/vimar-gu/SSD.
Large language models (large LMs) are susceptible to producing text that contains hallucinated content. An important instance of this problem is self-contradiction, where the LM generates two contradictory sentences within the same context. In this work, we present a comprehensive investigation into self-contradiction for various instruction-tuned LMs, covering evaluation, detection, and mitigation. Our primary evaluation task is open-domain text generation, but we also demonstrate the applicability of our approach to shorter question answering. Our analysis reveals the prevalence of self-contradictions, e.g., in 17.7% of all sentences produced by ChatGPT. We then propose a novel prompting-based framework designed to effectively detect and mitigate self-contradictions. Our detector achieves high accuracy, e.g., around 80% F1 score when prompting ChatGPT. The mitigation algorithm iteratively refines the generated text to remove contradictory information while preserving text fluency and informativeness. Importantly, our entire framework is applicable to black-box LMs and does not require retrieval of external knowledge. Rather, our method complements retrieval-based methods, as a large portion of self-contradictions (e.g., 35.2% for ChatGPT) cannot be verified using online text. Our approach is practically effective and has been released as a push-button tool to benefit the public at https://chatprotect.ai/.
Compositional zero-shot learning (CZSL) task aims to recognize unseen compositional visual concepts, e.g., sliced tomatoes, where the model is learned only from the seen compositions, e.g., sliced potatoes and red tomatoes. Thanks to the prompt tuning on large pre-trained visual language models such as CLIP, recent literature shows impressively better CZSL performance than traditional vision-based methods. However, the key aspects that impact the generalization to unseen compositions, including the diversity and informativeness of class context, and the entanglement between visual primitives, i.e., state and object, are not properly addressed in existing CLIP-based CZSL literature. In this paper, we propose a model by prompting the language-informed distribution, aka., PLID, for the CZSL task. Specifically, the PLID leverages pre-trained large language models (LLM) to (i) formulate the language-informed class distributions which are diverse and informative, and (ii) enhance the compositionality of the class embedding. Moreover, a visual-language primitive decomposition (VLPD) module is proposed to dynamically fuse the classification decisions from the compositional and the primitive space. Orthogonal to the existing literature of soft, hard, or distributional prompts, our method advocates prompting the LLM-supported class distributions, leading to a better zero-shot generalization. Experimental results on MIT-States, UT-Zappos, and C-GQA datasets show the superior performance of the PLID to the prior arts. Our code and models are released: https://github.com/Cogito2012/PLID.
In-context learning (ICL), the ability of large language models to perform novel tasks by conditioning on a prompt with a few task examples, requires these examples to be informative about the test instance. The standard approach of independently ranking and selecting the most similar examples selects redundant examples while omitting important information. In this work, we show that BERTScore-Recall (BSR) selects better examples that demonstrate more of the salient aspects, e.g. reasoning patterns, of the test input. We further extend BSR and many standard metrics to easily optimizable set-level metrics, giving still better coverage of those salient aspects. On 15 datasets spanning 6 tasks and with 7 diverse LLMs, we show that (1) BSR is the superior metric for in-context example selection across the board, and (2) for compositional tasks, set selection using Set-BSR outperforms independent ranking by up to 17 points on average and, despite being training-free, surpasses methods that leverage task or LLM-specific training.
While there has been significant development of models for Plain Language Summarization (PLS), evaluation remains a challenge. PLS lacks a dedicated assessment metric, and the suitability of text generation evaluation metrics is unclear due to the unique transformations involved (e.g., adding background explanations, removing jargon). To address these questions, our study introduces a granular meta-evaluation testbed, APPLS, designed to evaluate metrics for PLS. We identify four PLS criteria from previous work -- informativeness, simplification, coherence, and faithfulness -- and define a set of perturbations corresponding to these criteria that sensitive metrics should be able to detect. We apply these perturbations to extractive hypotheses for two PLS datasets to form our testbed. Using APPLS, we assess performance of 14 metrics, including automated scores, lexical features, and LLM prompt-based evaluations. Our analysis reveals that while some current metrics show sensitivity to specific criteria, no single method captures all four criteria simultaneously. We therefore recommend a suite of automated metrics be used to capture PLS quality along all relevant criteria. This work contributes the first meta-evaluation testbed for PLS and a comprehensive evaluation of existing metrics. APPLS and our evaluation code is available at https://github.com/LinguisticAnomalies/APPLS.
Task-oriented communication is an emerging paradigm for next-generation communication networks, which extracts and transmits task-relevant information, instead of raw data, for downstream applications. Most existing deep learning (DL)-based task-oriented communication systems adopt a closed-world scenario, assuming either the same data distribution for training and testing, or the system could have access to a large out-of-distribution (OoD) dataset for retraining. However, in practical open-world scenarios, task-oriented communication systems need to handle unknown OoD data. Under such circumstances, the powerful approximation ability of learning methods may force the task-oriented communication systems to overfit the training data (i.e., in-distribution data) and provide overconfident judgments when encountering OoD data. Based on the information bottleneck (IB) framework, we propose a class conditional IB (CCIB) approach to address this problem in this paper, supported by information-theoretical insights. The idea is to extract distinguishable features from in-distribution data while keeping their compactness and informativeness. This is achieved by imposing the class conditional latent prior distribution and enforcing the latent of different classes to be far away from each other. Simulation results shall demonstrate that the proposed approach detects OoD data more efficiently than the baselines and state-of-the-art approaches, without compromising the rate-distortion tradeoff.
Embodied agents have achieved prominent performance in following human instructions to complete tasks. However, the potential of providing instructions informed by texts and images to assist humans in completing tasks remains underexplored. To uncover this capability, we present the multimodal procedural planning (MPP) task, in which models are given a high-level goal and generate plans of paired text-image steps, providing more complementary and informative guidance than unimodal plans. The key challenges of MPP are to ensure the informativeness, temporal coherence,and accuracy of plans across modalities. To tackle this, we propose Text-Image Prompting (TIP), a dual-modality prompting method that jointly leverages zero-shot reasoning ability in large language models (LLMs) and compelling text-to-image generation ability from diffusion-based models. TIP improves the interaction in the dual modalities using Text-to-Image Bridge and Image-to-Text Bridge, allowing LLMs to guide the textual-grounded image plan generation and leveraging the descriptions of image plans to ground the textual plan reversely. To address the lack of relevant datasets, we collect WIKIPLAN and RECIPEPLAN as a testbed for MPP. Our results show compelling human preferences and automatic scores against unimodal and multimodal baselines on WIKIPLAN and RECIPEPLAN in terms of informativeness, temporal coherence, and plan accuracy. Our code and data: https://github.com/YujieLu10/MPP.
Collaborative Filtering (CF) is a widely used and effective technique for recommender systems. In recent decades, there have been significant advancements in latent embedding-based CF methods for improved accuracy, such as matrix factorization, neural collaborative filtering, and LightGCN. However, the explainability of these models has not been fully explored. Adding explainability to recommendation models can not only increase trust in the decisionmaking process, but also have multiple benefits such as providing persuasive explanations for item recommendations, creating explicit profiles for users and items, and assisting item producers in design improvements. In this paper, we propose a neat and effective Explainable Collaborative Filtering (ECF) model that leverages interpretable cluster learning to achieve the two most demanding objectives: (1) Precise - the model should not compromise accuracy in the pursuit of explainability; and (2) Self-explainable - the model's explanations should truly reflect its decision-making process, not generated from post-hoc methods. The core of ECF is mining taste clusters from user-item interactions and item profiles.We map each user and item to a sparse set of taste clusters, and taste clusters are distinguished by a few representative tags. The user-item preference, users/items' cluster affiliations, and the generation of taste clusters are jointly optimized in an end-to-end manner. Additionally, we introduce a forest mechanism to ensure the model's accuracy, explainability, and diversity. To comprehensively evaluate the explainability quality of taste clusters, we design several quantitative metrics, including in-cluster item coverage, tag utilization, silhouette, and informativeness. Our model's effectiveness is demonstrated through extensive experiments on three real-world datasets.
Multi-step reasoning ability is fundamental to many natural language tasks, yet it is unclear what constitutes a good reasoning chain and how to evaluate them. Most existing methods focus solely on whether the reasoning chain leads to the correct conclusion, but this answer-oriented view may confound reasoning quality with other spurious shortcuts to predict the answer. To bridge this gap, we evaluate reasoning chains by viewing them as informal proofs that derive the final answer. Specifically, we propose ReCEval (Reasoning Chain Evaluation), a framework that evaluates reasoning chains via two key properties: (1) correctness, i.e., each step makes a valid inference based on information contained within the step, preceding steps, and input context, and (2) informativeness, i.e., each step provides new information that is helpful towards deriving the generated answer. We evaluate these properties by developing metrics using natural language inference models and V-Information. On multiple datasets, we show that ReCEval effectively identifies various error types and yields notable improvements compared to prior methods. We analyze the impact of step boundaries, and previous steps on evaluating correctness and demonstrate that our informativeness metric captures the expected flow of information in high-quality reasoning chains. Finally, we show that scoring reasoning chains based on ReCEval improves downstream task performance. Our code is publicly available at: https://github.com/archiki/ReCEval