Knowledge-Aware Neuron Interpretation for Scene Classification

Although neural models have achieved remarkable performance, they still encounter doubts due to the intransparency. To this end, model prediction explanation is attracting more and more attentions. However, current methods rarely incorporate external knowledge and still suffer from three limitations: (1) Neglecting concept completeness. Merely selecting concepts may not sufficient for prediction. (2) Lacking concept fusion. Failure to merge semantically-equivalent concepts. (3) Difficult in manipulating model behavior. Lack of verification for explanation on original model. To address these issues, we propose a novel knowledge-aware neuron interpretation framework to explain model predictions for image scene classification. Specifically, for concept completeness, we present core concepts of a scene based on knowledge graph, ConceptNet, to gauge the completeness of concepts. Our method, incorporating complete concepts, effectively provides better prediction explanations compared to baselines. Furthermore, for concept fusion, we introduce a knowledge graph-based method known as Concept Filtering, which produces over 23% point gain on neuron behaviors for neuron interpretation. At last, we propose Model Manipulation, which aims to study whether the core concepts based on ConceptNet could be employed to manipulate model behavior. The results show that core concepts can effectively improve the performance of original model by over 26%.

Formalization of Robot Collision Detection Method based on Conformal Geometric Algebra

Cooperative robots can significantly assist people in their productive activities, improving the quality of their works. Collision detection is vital to ensure the safe and stable operation of cooperative robots in productive activities. As an advanced geometric language, conformal geometric algebra can simplify the construction of the robot collision model and the calculation of collision distance. Compared with the formal method based on conformal geometric algebra, the traditional method may have some defects which are difficult to find in the modelling and calculation. We use the formal method based on conformal geometric algebra to study the collision detection problem of cooperative robots. This paper builds formal models of geometric primitives and the robot body based on the conformal geometric algebra library in HOL Light. We analyse the shortest distance between geometric primitives and prove their collision determination conditions. Based on the above contents, we construct a formal verification framework for the robot collision detection method. By the end of this paper, we apply the proposed framework to collision detection between two single-arm industrial cooperative robots. The flexibility and reliability of the proposed framework are verified by constructing a general collision model and a special collision model for two single-arm industrial cooperative robots.

MAVEN-Arg: Completing the Puzzle of All-in-One Event Understanding Dataset with Event Argument Annotation

Understanding events in texts is a core objective of natural language understanding, which requires detecting event occurrences, extracting event arguments, and analyzing inter-event relationships. However, due to the annotation challenges brought by task complexity, a large-scale dataset covering the full process of event understanding has long been absent. In this paper, we introduce MAVEN-Arg, which augments MAVEN datasets with event argument annotations, making the first all-in-one dataset supporting event detection, event argument extraction (EAE), and event relation extraction. As an EAE benchmark, MAVEN-Arg offers three main advantages: (1) a comprehensive schema covering 162 event types and 612 argument roles, all with expert-written definitions and examples; (2) a large data scale, containing 98,591 events and 290,613 arguments obtained with laborious human annotation; (3) the exhaustive annotation supporting all task variants of EAE, which annotates both entity and non-entity event arguments in document level. Experiments indicate that MAVEN-Arg is quite challenging for both fine-tuned EAE models and proprietary large language models (LLMs). Furthermore, to demonstrate the benefits of an all-in-one dataset, we preliminarily explore a potential application, future event prediction, with LLMs. MAVEN-Arg and our code can be obtained from https://github.com/THU-KEG/MAVEN-Argument.

A Comprehensive and Reliable Feature Attribution Method: Double-sided Remove and Reconstruct (DoRaR)

The limited transparency of the inner decision-making mechanism in deep neural networks (DNN) and other machine learning (ML) models has hindered their application in several domains. In order to tackle this issue, feature attribution methods have been developed to identify the crucial features that heavily influence decisions made by these black box models. However, many feature attribution methods have inherent downsides. For example, one category of feature attribution methods suffers from the artifacts problem, which feeds out-of-distribution masked inputs directly through the classifier that was originally trained on natural data points. Another category of feature attribution method finds explanations by using jointly trained feature selectors and predictors. While avoiding the artifacts problem, this new category suffers from the Encoding Prediction in the Explanation (EPITE) problem, in which the predictor's decisions rely not on the features, but on the masks that selects those features. As a result, the credibility of attribution results is undermined by these downsides. In this research, we introduce the Double-sided Remove and Reconstruct (DoRaR) feature attribution method based on several improvement methods that addresses these issues. By conducting thorough testing on MNIST, CIFAR10 and our own synthetic dataset, we demonstrate that the DoRaR feature attribution method can effectively bypass the above issues and can aid in training a feature selector that outperforms other state-of-the-art feature attribution methods. Our code is available at https://github.com/dxq21/DoRaR.

KoLA: Carefully Benchmarking World Knowledge of Large Language Models

The unprecedented performance of large language models (LLMs) necessitates improvements in evaluations. Rather than merely exploring the breadth of LLM abilities, we believe meticulous and thoughtful designs are essential to thorough, unbiased, and applicable evaluations. Given the importance of world knowledge to LLMs, we construct a Knowledge-oriented LLM Assessment benchmark (KoLA), in which we carefully design three crucial factors: (1) For ability modeling, we mimic human cognition to form a four-level taxonomy of knowledge-related abilities, covering $19$ tasks. (2) For data, to ensure fair comparisons, we use both Wikipedia, a corpus prevalently pre-trained by LLMs, along with continuously collected emerging corpora, aiming to evaluate the capacity to handle unseen data and evolving knowledge. (3) For evaluation criteria, we adopt a contrastive system, including overall standard scores for better numerical comparability across tasks and models and a unique self-contrast metric for automatically evaluating knowledge hallucination. We evaluate $21$ open-source and commercial LLMs and obtain some intriguing findings. The KoLA dataset and open-participation leaderboard are publicly released at https://kola.xlore.cn and will be continuously updated to provide references for developing LLMs and knowledge-related systems.

Limited Angle Tomography for Transmission X-Ray Microscopy Using Deep Learning

In transmission X-ray microscopy (TXM) systems, the rotation of a scanned sample might be restricted to a limited angular range to avoid collision to other system parts or high attenuation at certain tilting angles. Image reconstruction from such limited angle data suffers from artifacts due to missing data. In this work, deep learning is applied to limited angle reconstruction in TXMs for the first time. With the challenge to obtain sufficient real data for training, training a deep neural network from synthetic data is investigated. Particularly, the U-Net, the state-of-the-art neural network in biomedical imaging, is trained from synthetic ellipsoid data and multi-category data to reduce artifacts in filtered back-projection (FBP) reconstruction images. The proposed method is evaluated on synthetic data and real scanned chlorella data in $100^\circ$ limited angle tomography. For synthetic test data, the U-Net significantly reduces root-mean-square error (RMSE) from $2.55 \times 10^{-3}$ {\mu}m$^{-1}$ in the FBP reconstruction to $1.21 \times 10^{-3}$ {\mu}m$^{-1}$ in the U-Net reconstruction, and also improves structural similarity (SSIM) index from 0.625 to 0.920. With penalized weighted least square denoising of measured projections, the RMSE and SSIM are further improved to $1.16 \times 10^{-3}$ {\mu}m$^{-1}$ and 0.932, respectively. For real test data, the proposed method remarkably improves the 3-D visualization of the subcellular structures in the chlorella cell, which indicates its important value for nano-scale imaging in biology, nanoscience and materials science.

Unsupervised Trajectory Segmentation and Promoting of Multi-Modal Surgical Demonstrations

To improve the efficiency of surgical trajectory segmentation for robot learning in robot-assisted minimally invasive surgery, this paper presents a fast unsupervised method using video and kinematic data, followed by a promoting procedure to address the over-segmentation issue. Unsupervised deep learning network, stacking convolutional auto-encoder, is employed to extract more discriminative features from videos in an effective way. To further improve the accuracy of segmentation, on one hand, wavelet transform is used to filter out the noises existed in the features from video and kinematic data. On the other hand, the segmentation result is promoted by identifying the adjacent segments with no state transition based on the predefined similarity measurements. Extensive experiments on a public dataset JIGSAWS show that our method achieves much higher accuracy of segmentation than state-of-the-art methods in the shorter time.