Large Language Models (LLMs) are susceptible to security and safety threats, such as prompt injection, prompt extraction, and harmful requests. One major cause of these vulnerabilities is the lack of an instruction hierarchy. Modern LLM architectures treat all inputs equally, failing to distinguish between and prioritize various types of instructions, such as system messages, user prompts, and data. As a result, lower-priority user prompts may override more critical system instructions, including safety protocols. Existing approaches to achieving instruction hierarchy, such as delimiters and instruction-based training, do not address this issue at the architectural level. We introduce the $\textbf{I}$nstructional $\textbf{S}$egment $\textbf{E}$mbedding (ISE) technique, inspired by BERT, to modern large language models, which embeds instruction priority information directly into the model. This approach enables models to explicitly differentiate and prioritize various instruction types, significantly improving safety against malicious prompts that attempt to override priority rules. Our experiments on the Structured Query and Instruction Hierarchy benchmarks demonstrate an average robust accuracy increase of up to 15.75\% and 18.68\%, respectively. Furthermore, we observe an improvement in the instruction-following capability of up to 4.1\% on AlpacaEval. Overall, our approach offers a promising direction for enhancing the safety and effectiveness of LLM architectures.

Wenxuan Zhou, Ravi Agrawal, Shujian Zhang, Sathish Reddy Indurthi, Sanqiang Zhao, Kaiqiang Song, Silei Xu, Chenguang Zhu. Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 2024.

Pretrained language models like BERT have achieved good results on NLP tasks, but are impractical on resource-limited devices due to memory footprint.A large fraction of this footprint comes from the input embeddings with large input vocabulary and embedding dimensions.Existing knowledge distillation methods used for model compression cannot be directly applied to train student models with reduced vocabulary sizes.To this end, we propose a distillation method to align the teacher and student embeddings via mixed-vocabulary training.Our method compresses BERT LARGE to a task-agnostic model with smaller vocabulary and hidden dimensions, which is an order of magnitude smaller than other distilled BERT models and offers a better size-accuracy trade-off on language understanding benchmarks as well as a practical dialogue task.

Rui Meng, Xingdi Yuan, Tong Wang, Sanqiang Zhao, Adam Trischler, Daqing He. Proceedings of the 2021 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies. 2021.

Neural network-based sequence-to-sequence (seq2seq) models strongly suffer from the lowdiversity problem when it comes to opendomain dialogue generation.As bland and generic utterances usually dominate the frequency distribution in our daily chitchat, avoiding them to generate more interesting responses requires complex data filtering, sampling techniques or modifying the training objective.In this paper, we propose a new perspective to diversify dialogue generation by leveraging non-conversational text.Compared with bilateral conversations, nonconversational text are easier to obtain, more diverse and cover a much broader range of topics.We collect a large-scale nonconversational corpus from multi sources including forum comments, idioms and book snippets.We further present a training paradigm to effectively incorporate these text via iterative back translation.The resulting model is tested on two conversational datasets and is shown to produce significantly more diverse responses without sacrificing the relevance with context.

An image caption should fluently present the essential information in a given image, including informative, fine-grained entity mentions and the manner in which these entities interact. However, current captioning models are usually trained to generate captions that only contain common object names, thus falling short on an important “informativeness” dimension. We present a mechanism for integrating image information together with fine-grained labels (assumed to be generated by some upstream models) into a caption that describes the image in a fluent and informative manner. We introduce a multimodal, multi-encoder model based on Transformer that ingests both image features and multiple sources of entity labels. We demonstrate that we can learn to control the appearance of these entity labels in the output, resulting in captions that are both fluent and informative.

Sentence simplification aims to reduce the complexity of a sentence while retaining its original meaning.Current models for sentence simplification adopted ideas from machine translation studies and implicitly learned simplification mapping rules from normalsimple sentence pairs.In this paper, we explore a novel model based on a multi-layer and multi-head attention architecture and we propose two innovative approaches to integrate the Simple PPDB (A Paraphrase Database for Simplification), an external paraphrase knowledge base for simplification that covers a wide range of real-world simplification rules.The experiments show that the integration provides two major benefits: (1) the integrated model outperforms multiple stateof-the-art baseline models for sentence simplification in the literature (2) through analysis of the rule utilization, the model seeks to select more accurate simplification rules.The code and models used in the paper are available at https://github.com/ Sanqiang/text_simplification.

Keyphrase provides highly-summative information that can be effectively used for understanding, organizing and retrieving text content.Though previous studies have provided many workable solutions for automated keyphrase extraction, they commonly divided the to-be-summarized content into multiple text chunks, then ranked and selected the most meaningful ones.These approaches could neither identify keyphrases that do not appear in the text, nor capture the real semantic meaning behind the text.We propose a generative model for keyphrase prediction with an encoder-decoder framework, which can effectively overcome the above drawbacks.We name it as deep keyphrase generation since it attempts to capture the deep semantic meaning of the content with a deep learning method.Empirical analysis on six datasets demonstrates that our proposed model not only achieves a significant performance boost on extracting keyphrases that appear in the source text, but also can generate absent keyphrases based on the semantic meaning of the text.Code and dataset are available at https://github.com/memray/seq2seq-keyphrase.

This paper introduces a new BeiHang (BH) Keystroke Dynamics Database for testing and evaluation of biometric approaches. Different from the existing keystroke dynamics researches which solely rely on laboratory experiments, the developed database is collected from a real commercialized system and thus is more comprehensive and more faithful to human behavior. Moreover, our database comes with ready-to-use benchmark results of three keystroke dynamics methods, Nearest Neighbor classifier, Gaussian Model and One-Class Support Vector Machine. Both the database and benchmark results are open to the public and provide a significant experimental platform for international researchers in the keystroke dynamics area.

This paper proposes a novel kernel similarity modeling of texture pattern flow (KSM-TPF) for background modeling and motion detection in complex and dynamic environments. The texture pattern flow encodes the binary pattern changes in both spatial and temporal neighborhoods. The integral histogram of texture pattern flow is employed to extract the discriminative features from the input videos. Different from existing uniform threshold based motion detection approaches which are only effective for simple background, the kernel similarity modeling is proposed to produce an adaptive threshold for complex background. The adaptive threshold is computed from the mean and variance of an extended Gaussian mixture model. The proposed KSM-TPF approach incorporates machine learning method with feature extraction method in a homogenous way. Experimental results on the publicly available video sequences demonstrate that the proposed approach provides an effective and efficient way for background modeling and motion detection.

Micropattern based image representation and recognition, e.g. Local Binary Pattern (LBP), has been proved successful over the past few years due to its advantages of illumination tolerance and computational efficiency. However, LBP only encodes the first-order radial-directional derivatives of spatial images and is inadequate to completely describe the discriminative features for classification. This paper proposes a new Circular Derivative Pattern (CDP) which extracts high-order derivative information of images along circular directions. We argue that the high-order circular derivatives contain more detailed and more discriminative information than the first-order LBP in terms of recognition accuracy. Experimental evaluation through face recognition on the FERET database and insect classification on the NICTA Biosecurity Dataset demonstrated the effectiveness of the proposed method.

Face recognition using micropattern representation has recently received much attention in the computer vision and pattern recognition community. Previous researches demonstrated that micropattern representation based on Gabor features achieves better performance than its direct usage on gray-level images. This paper conducts a comparative performance evaluation of micropattern representations on four forms of Gabor features for face recognition. Three evaluation rules are proposed and observed for a fair comparison. To reduce the high feature dimensionality problem, uniform quantization is used to partition the spatial histograms. The experimental results reveal that: 1) micropattern representation based on Gabor magnitude features outperforms the other three representations, and the performances of the other three are comparable; and 2) micropattern representation based on the combination of Gabor magnitude and phase features performs the best.

This paper proposes a novel high-order local pattern descriptor, local derivative pattern (LDP), for face recognition. LDP is a general framework to encode directional pattern features based on local derivative variations. The n(th)-order LDP is proposed to encode the (n-1)(th) -order local derivative direction variations, which can capture more detailed information than the first-order local pattern used in local binary pattern (LBP). Different from LBP encoding the relationship between the central point and its neighbors, the LDP templates extract high-order local information by encoding various distinctive spatial relationships contained in a given local region. Both gray-level images and Gabor feature images are used to evaluate the comparative performances of LDP and LBP. Extensive experimental results on FERET, CAS-PEAL, CMU-PIE, Extended Yale B, and FRGC databases show that the high-order LDP consistently performs much better than LBP for both face identification and face verification under various conditions.

This paper addresses two critical but rarely concerned issues in 2D face recognition: wider-range tolerance to pose variation and misalignment. We propose a new Textural Hausdorff Distance (THD), which is a compound measurement integrating both spatial and textural features. The THD is applied to a Significant Jet Point (SJP) representation of face images, where a varied number of shape-driven SJPs are detected automatically from low-level edge map with rich information content. The comparative experiments conducted on publicly available FERET and AR face databases demonstrated that the proposed approach has a considerably wider range of tolerance against both in-depth head rotation and face misalignment.

Gabor wavelet related feature extraction and classification is an important topic in image analysis and pattern recognition. Gabor features can be used either holistically or analytically. While holistic approaches involve significant computational complexity, existing analytic approaches require explicit correspondence of predefined feature points for classification. Different from these approaches, this paper presents a new analytic Gabor method for face recognition. The proposed method attaches Gabor features on a set of shape-driven sparse points to describe both geometric and textural information. Neither the number nor the correspondence of these points is needed. A variant of Hausdorff distance is employed to recognize faces. The experiments performed on AR database demonstrated that the proposed algorithm is effective to identify individuals in various circumstances, such as under expression and illumination changes.

This paper presents a new Sobel-LBP, an extension of existing Local Binary Pattern (LBP), for facial image representation. The face image is filtered by Sobel operator to enhance the edge information. Sobel-LBP feature distributions are then extracted and concatenated into a spatial histogram to be used as a face descriptor. The proposed method is compared with the original LBP on both gray-level images and Gabor real and imaginary features for face recognition. The experimental results indicate that Sobel-LBP provides a significantly better performance than LBP under various conditions.

Recently, psychological studies showed that averaging human face images greatly improves the performance of face recognition under various pose, illumination, expression, and/or aging conditions. This paper investigates quantitatively the mechanism of the face averaging process in face recognition specifically against pose variations. Facilitated with 3D face dataset, the process of face averaging is tested on face images free from human errors and misalignments. Single images are chosen as gallery and the averaged views as probe in all the experiments. Three different scenarios are experimented, i.e., identification using single gallery images, identification using different gallery images, and averaging using unbalanced range of input image. The experimental results show that the averaging process under pose variations is equivalent to generating a face view in an average pose and the improvement in face recognition is subject to the conditions that the gallery pose is close to the average probe pose.

This paper presents a new Multidirectional Binary Pattern (MBP) for face recognition. Different from most Local Binary Pattern (LBP) related approaches which cluster LBP occurrences from whole image or partitioned subimage patches and use single or concatenated histogram measurement for recognition, MBP is applied on a sparse set of shape-driven points. The new representation is designed for describing both global structure and local texture, and also significantly reduces the high dimensionality of LBP histogram description. Composed of binary patterns from multiple directions, MBP is capable of extracting more discriminative features than LBP. The experiments on face recognition demonstrated the effectiveness of the proposed algorithm against expression and lighting variations.

This study focuses on a recent paper “100% Accuracy in Automatic Face Recognition” published on Science, in which an “Average Face” is proposed and claimed to be capable of dramatically improving performance of a face recognition system. To reveal its working mechanism, we perform the averaging process using pose-varied synthetic images generated from 3D face database and conduct a comparative study to observe its effectiveness on holistic and local face recognition approaches. Two representative methods, i.e. Eigenface and Local Binary Pattern (LBP) are employed to perform the experiments. It is interesting to find from our experiments that the performance of the “Average Face” is not independent of the face recognition approaches. Although face averaging increases the recognition accuracy of Eigenface method, it impairs the performance of LBP method.