Progress on many Natural Language Processing (NLP) tasks, such as text classification, is driven by objective, reproducible and scalable evaluation via publicly available benchmarks. However, these are not always representative of real-world scenarios where text classifiers are employed, such as sentiment analysis or misinformation detection. In this position paper, we put forward two points that aim to alleviate this problem. First, we propose to extend text classification benchmarks to evaluate the explainability of text classifiers. We review challenges associated with objectively evaluating the capabilities to produce valid explanations which leads us to the second main point: We propose to ground these benchmarks in human-centred applications, for example by using social media, gamification or to learn explainability metrics from human judgements.
Pre-trained Language Model (PLM) has become a representative foundation model in the natural language processing field. Most PLMs are trained with linguistic-agnostic pre-training tasks on the surface form of the text, such as the masked language model (MLM). To further empower the PLMs with richer linguistic features, in this paper, we aim to propose a simple but effective way to learn linguistic features for pre-trained language models. We propose LERT, a pre-trained language model that is trained on three types of linguistic features along with the original MLM pre-training task, using a linguistically-informed pre-training (LIP) strategy. We carried out extensive experiments on ten Chinese NLU tasks, and the experimental results show that LERT could bring significant improvements over various comparable baselines. Furthermore, we also conduct analytical experiments in various linguistic aspects, and the results prove that the design of LERT is valid and effective. Resources are available at https://github.com/ymcui/LERT
The domain of joint vision-language understanding, especially in the context of reasoning in Visual Question Answering (VQA) models, has garnered significant attention in the recent past. While most of the existing VQA models focus on improving the accuracy of VQA, the way models arrive at an answer is oftentimes a black box. As a step towards making the VQA task more explainable and interpretable, our method is built upon the SOTA VQA framework by augmenting it with an end-to-end explanation generation module. In this paper, we investigate two network architectures, including Long Short-Term Memory (LSTM) and Transformer decoder, as the explanation generator. Our method generates human-readable textual explanations while maintaining SOTA VQA accuracy on the GQA-REX (77.49%) and VQA-E (71.48%) datasets. Approximately 65.16% of the generated explanations are approved by humans as valid. Roughly 60.5% of the generated explanations are valid and lead to the correct answers.
State-of-the-art deep-learning-based approaches to Natural Language Processing (NLP) are credited with various capabilities that involve reasoning with natural language texts. In this paper we carry out a large-scale empirical study investigating the detection of formally valid inferences in controlled fragments of natural language for which the satisfiability problem becomes increasingly complex. We find that, while transformer-based language models perform surprisingly well in these scenarios, a deeper analysis re-veals that they appear to overfit to superficial patterns in the data rather than acquiring the logical principles governing the reasoning in these fragments.
This paper develops conformal inference methods to construct a confidence interval for the frequency of a queried object in a very large discrete data set, based on a sketch with a lower memory footprint. This approach requires no knowledge of the data distribution and can be combined with any sketching algorithm, including but not limited to the renowned count-min sketch, the count-sketch, and variations thereof. After explaining how to achieve marginal coverage for exchangeable random queries, we extend our solution to provide stronger inferences that can account for the discreteness of the data and for heterogeneous query frequencies, increasing also robustness to possible distribution shifts. These results are facilitated by a novel conformal calibration technique that guarantees valid coverage for a large fraction of distinct random queries. Finally, we show our methods have improved empirical performance compared to existing frequentist and Bayesian alternatives in simulations as well as in examples of text and SARS-CoV-2 DNA data.
Virtual, make-on-demand chemical libraries have transformed early-stage drug discovery by unlocking vast, synthetically accessible regions of chemical space. Recent years have witnessed rapid growth in these libraries from millions to trillions of compounds, hiding undiscovered, potent hits for a variety of therapeutic targets. However, they are quickly approaching a size beyond that which permits explicit enumeration, presenting new challenges for virtual screening. To overcome these challenges, we propose the Combinatorial Synthesis Library Variational Auto-Encoder (CSLVAE). The proposed generative model represents such libraries as a differentiable, hierarchically-organized database. Given a compound from the library, the molecular encoder constructs a query for retrieval, which is utilized by the molecular decoder to reconstruct the compound by first decoding its chemical reaction and subsequently decoding its reactants. Our design minimizes autoregression in the decoder, facilitating the generation of large, valid molecular graphs. Our method performs fast and parallel batch inference for ultra-large synthesis libraries, enabling a number of important applications in early-stage drug discovery. Compounds proposed by our method are guaranteed to be in the library, and thus synthetically and cost-effectively accessible. Importantly, CSLVAE can encode out-of-library compounds and search for in-library analogues. In experiments, we demonstrate the capabilities of the proposed method in the navigation of massive combinatorial synthesis libraries.
Robots operating in human environments must be able to rearrange objects into semantically-meaningful configurations, even if these objects are previously unseen. In this work, we focus on the problem of building physically-valid structures without step-by-step instructions. We propose StructDiffusion, which combines a diffusion model and an object-centric transformer to construct structures given partial-view point clouds and high-level language goals, such as "set the table". Our method can perform multiple challenging language-conditioned multi-step 3D planning tasks using one model. StructDiffusion even improves the success rate of assembling physically-valid structures out of unseen objects by on average 16% over an existing multi-modal transformer model trained on specific structures. We show experiments on held-out objects in both simulation and on real-world rearrangement tasks. Importantly, we show how integrating both a diffusion model and a collision-discriminator model allows for improved generalization over other methods when rearranging previously-unseen objects. For videos and additional results, see our website: https://structdiffusion.github.io/.
This paper develops valid bootstrap inference methods for the dynamic short panel threshold regression. We demonstrate that the standard nonparametric bootstrap is inconsistent for the first-differenced generalized method of moments (GMM) estimator. The inconsistency is due to an $n^{1/4}$-consistent non-normal asymptotic distribution for the threshold estimate when the parameter resides within the continuity region of the parameter space. It stems from the rank deficiency of the approximate Jacobian of the sample moment conditions on the continuity region. To address this, we propose a grid bootstrap to construct confidence intervals of the threshold, a residual bootstrap to construct confidence intervals of the coefficients, and a bootstrap for testing continuity. They are shown to be valid under uncertain continuity, while the grid bootstrap is additionally shown to be uniformly valid. A set of Monte Carlo experiments demonstrate that the proposed bootstraps perform well in the finite samples and improve upon the standard nonparametric bootstrap.
Consider two brands that want to jointly test alternate web experiences for their customers with an A/B test. Such collaborative tests are today enabled using \textit{third-party cookies}, where each brand has information on the identity of visitors to another website. With the imminent elimination of third-party cookies, such A/B tests will become untenable. We propose a two-stage experimental design, where the two brands only need to agree on high-level aggregate parameters of the experiment to test the alternate experiences. Our design respects the privacy of customers. We propose an estimater of the Average Treatment Effect (ATE), show that it is unbiased and theoretically compute its variance. Our demonstration describes how a marketer for a brand can design such an experiment and analyze the results. On real and simulated data, we show that the approach provides valid estimate of the ATE with low variance and is robust to the proportion of visitors overlapping across the brands.
Assessing the statistical significance of parameter estimates is an important step in high-dimensional vector autoregression modeling. Using the least-squares boosting method, we compute the p-value for each selected parameter at every boosting step in a linear model. The p-values are asymptotically valid and also adapt to the iterative nature of the boosting procedure. Our simulation experiment shows that the p-values can keep false positive rate under control in high-dimensional vector autoregressions. In an application with more than 100 macroeconomic time series, we further show that the p-values can not only select a sparser model with good prediction performance but also help control model stability. A companion R package boostvar is developed.
The Transformer architecture is shown to provide a powerful framework as an end-to-end model for building expression trees from online handwritten gestures corresponding to glyph strokes. In particular, the attention mechanism was successfully used to encode, learn and enforce the underlying syntax of expressions creating latent representations that are correctly decoded to the exact mathematical expression tree, providing robustness to ablated inputs and unseen glyphs. For the first time, the encoder is fed with spatio-temporal data tokens potentially forming an infinitely large vocabulary, which finds applications beyond that of online gesture recognition. A new supervised dataset of online handwriting gestures is provided for training models on generic handwriting recognition tasks and a new metric is proposed for the evaluation of the syntactic correctness of the output expression trees. A small Transformer model suitable for edge inference was successfully trained to an average normalised Levenshtein accuracy of 94%, resulting in valid postfix RPN tree representation for 94% of predictions.
We study the necessary and sufficient conditions under which the Mean-Variance Criterion (MVC) is equivalent to the Maximum Expected Utility Criterion (MEUC), for two lotteries. Based on Chamberlain (1983), we conclude that the MVC is equivalent to the Second-order Stochastic Dominance Rule (SSDR) under any symmetric Elliptical distribution. We then discuss the work of Schuhmacher et al. (2021). Although their theoretical findings deduce that the Mean-Variance Analysis remains valid under Skew-Elliptical distributions, we argue that this does not entail that the MVC coincides with the SSDR. In fact, generating multiple MV-pairs that follow a Skew-Normal distribution it becomes evident that the MVC fails to coincide with the SSDR for some types of risk-averse investors. In the second part of this work, we examine the premise of Levy and Markowitz (1979) that "the MVC deduces the maximization of the expected utility of an investor, under any approximately quadratic utility function, without making any further assumption on the distribution of the lotteries". Using Monte Carlo Simulations, we find out that the set of approximately quadratic utility functions is too narrow. Specifically, our simulations indicate that $\log{(a+Z)}$ and $(1+Z)^a$ are almost quadratic, while $-e^{-a(1+Z)}$ and $-(1+Z)^{-a}$ fail to approximate a quadratic utility function under either an Extreme Value or a Stable Pareto distribution.
The proliferation of unmanned aerial vehicles (UAVs) opens up new opportunities for on-demand service provisioning anywhere and anytime, but also exposes UAVs to a variety of cyber threats. Low/medium interaction honeypots offer a promising lightweight defense for actively protecting mobile Internet of things, particularly UAV networks. While previous research has primarily focused on honeypot system design and attack pattern recognition, the incentive issue for motivating UAV's participation (e.g., sharing trapped attack data in honeypots) to collaboratively resist distributed and sophisticated attacks remains unexplored. This paper proposes a novel game-theoretical collaborative defense approach to address optimal, fair, and feasible incentive design, in the presence of network dynamics and UAVs' multi-dimensional private information (e.g., valid defense data (VDD) volume, communication delay, and UAV cost). Specifically, we first develop a honeypot game between UAVs and the network operator under both partial and complete information asymmetry scenarios. The optimal VDD-reward contract design problem with partial information asymmetry is then solved using a contract-theoretic approach that ensures budget feasibility, truthfulness, fairness, and computational efficiency. In addition, under complete information asymmetry, we devise a distributed reinforcement learning algorithm to dynamically design optimal contracts for distinct types of UAVs in the time-varying UAV network. Extensive simulations demonstrate that the proposed scheme can motivate UAV's cooperation in VDD sharing and improve defensive effectiveness, compared with conventional schemes.
Existing metrics for evaluating the quality of automatically generated questions such as BLEU, ROUGE, BERTScore, and BLEURT compare the reference and predicted questions, providing a high score when there is a considerable lexical overlap or semantic similarity between the candidate and the reference questions. This approach has two major shortcomings. First, we need expensive human-provided reference questions. Second, it penalises valid questions that may not have high lexical or semantic similarity to the reference questions. In this paper, we propose a new metric, RQUGE, based on the answerability of the candidate question given the context. The metric consists of a question-answering and a span scorer modules, using pre-trained models from existing literature, thus it can be used without any further training. We demonstrate that RQUGE has a higher correlation with human judgment without relying on the reference question. Additionally, RQUGE is shown to be more robust to several adversarial corruptions. Furthermore, we illustrate that we can significantly improve the performance of QA models on out-of-domain datasets by fine-tuning on synthetic data generated by a question generation model and re-ranked by RQUGE.
In this study, we propose a novel method called region-guided masked image modeling (RGMIM) for learning meaningful representations from X-ray images. Our method adopts a new masking strategy that utilizes organ mask information to identify valid regions for learning more meaningful representations. We conduct quantitative evaluations on an open lung X-ray image dataset as well as masking ratio hyperparameter studies. When using the entire training set, RGMIM outperformed other comparable methods, achieving a 0.962 lung disease detection accuracy. Specifically, RGMIM significantly improved performance in small data volumes, such as 5% and 10% of the training set compared to other methods. RGMIM can mask more valid regions, facilitating the learning of discriminative representations and the subsequent high-accuracy lung disease detection. RGMIM outperforms other state-of-the-art self-supervised learning methods in experiments, particularly when limited training data is used.
Large datasets make it possible to build predictive models that can capture heterogenous relationships between the response variable and features. The mixture of high-dimensional linear experts model posits that observations come from a mixture of high-dimensional linear regression models, where the mixture weights are themselves feature-dependent. In this paper, we show how to construct valid prediction sets for an $\ell_1$-penalized mixture of experts model in the high-dimensional setting. We make use of a debiasing procedure to account for the bias induced by the penalization and propose a novel strategy for combining intervals to form a prediction set with coverage guarantees in the mixture setting. Synthetic examples and an application to the prediction of critical temperatures of superconducting materials show our method to have reliable practical performance.
The topic of summarization evaluation has recently attracted a surge of attention due to the rapid development of abstractive summarization systems. However, the formulation of the task is rather ambiguous, neither the linguistic nor the natural language processing community has succeeded in giving a mutually agreed-upon definition. Due to this lack of well-defined formulation, a large number of popular abstractive summarization datasets are constructed in a manner that neither guarantees validity nor meets one of the most essential criteria of summarization: factual consistency. In this paper, we address this issue by combining state-of-the-art factual consistency models to identify the problematic instances present in popular summarization datasets. We release SummFC, a filtered summarization dataset with improved factual consistency, and demonstrate that models trained on this dataset achieve improved performance in nearly all quality aspects. We argue that our dataset should become a valid benchmark for developing and evaluating summarization systems.
In this paper, we investigated the semantic segmentation of Polarimetric Synthetic Aperture Radar (PolSAR) using Complex-Valued Neural Network (CVNN). Although the coherency matrix is more widely used as the input of CVNN, the Pauli vector has recently been shown to be a valid alternative. We exhaustively compare both methods for six model architectures, three complex-valued, and their respective real-equivalent models. We are comparing, therefore, not only the input representation impact but also the complex- against the real-valued models. We then argue that the dataset splitting produces a high correlation between training and validation sets, saturating the task and thus achieving very high performance. We, therefore, use a different data pre-processing technique designed to reduce this effect and reproduce the results with the same configurations as before (input representation and model architectures). After seeing that the performance per class is highly different according to class occurrences, we propose two methods for reducing this gap and performing the results for all input representations, models, and dataset pre-processing.
Weakly Supervised Semantic Segmentation (WSSS) research has explored many directions to improve the typical pipeline CNN plus class activation maps (CAM) plus refinements, given the image-class label as the only supervision. Though the gap with the fully supervised methods is reduced, further abating the spread seems unlikely within this framework. On the other hand, WSSS methods based on Vision Transformers (ViT) have not yet explored valid alternatives to CAM. ViT features have been shown to retain a scene layout, and object boundaries in self-supervised learning. To confirm these findings, we prove that the advantages of transformers in self-supervised methods are further strengthened by Global Max Pooling (GMP), which can leverage patch features to negotiate pixel-label probability with class probability. This work proposes a new WSSS method dubbed ViT-PCM (ViT Patch-Class Mapping), not based on CAM. The end-to-end presented network learns with a single optimization process, refined shape and proper localization for segmentation masks. Our model outperforms the state-of-the-art on baseline pseudo-masks (BPM), where we achieve $69.3\%$ mIoU on PascalVOC 2012 $val$ set. We show that our approach has the least set of parameters, though obtaining higher accuracy than all other approaches. In a sentence, quantitative and qualitative results of our method reveal that ViT-PCM is an excellent alternative to CNN-CAM based architectures.
We propose a new estimator for heterogeneous treatment effects in a partially linear model (PLM) with multiple exogenous covariates and a potentially endogenous treatment variable. Our approach integrates a Robinson transformation to handle the nonparametric component, the Smooth Minimum Distance (SMD) method to leverage conditional mean independence restrictions, and a Neyman-Orthogonalized first-order condition (FOC). By employing regularized model selection techniques like the Lasso method, our estimator accommodates numerous covariates while exhibiting reduced bias, consistency, and asymptotic normality. Simulations demonstrate its robust performance with diverse instrument sets compared to traditional GMM-type estimators. Applying this method to estimate Medicaid's heterogeneous treatment effects from the Oregon Health Insurance Experiment reveals more robust and reliable results than conventional GMM approaches.