Multimodal large language models (MLLMs) have attracted increasing attention in the past few years, but they may still generate descriptions that include objects not present in the corresponding images, a phenomenon known as object hallucination. To eliminate hallucinations, existing methods manually annotate paired responses with and without hallucinations, and then employ various alignment algorithms to improve the alignment capability between images and text. However, they not only demand considerable computation resources during the finetuning stage but also require expensive human annotation to construct paired data needed by the alignment algorithms. To address these issues, we borrow the idea of unlearning and propose an efficient fine-grained unlearning framework (EFUF), which can eliminate hallucinations without the need for paired data. Extensive experiments show that our method consistently reduces hallucinations while preserving the generation quality with modest computational overhead. Our code and datasets will be publicly available.
Pathological image analysis is a crucial field in computer-aided diagnosis, where deep learning is widely applied. Transfer learning using pre-trained models initialized on natural images has effectively improved the downstream pathological performance. However, the lack of sophisticated domain-specific pathological initialization hinders their potential. Self-supervised learning (SSL) enables pre-training without sample-level labels, which has great potential to overcome the challenge of expensive annotations. Thus, studies focusing on pathological SSL pre-training call for a comprehensive and standardized dataset, similar to the ImageNet in computer vision. This paper presents the comprehensive pathological image analysis (CPIA) dataset, a large-scale SSL pre-training dataset combining 103 open-source datasets with extensive standardization. The CPIA dataset contains 21,427,877 standardized images, covering over 48 organs/tissues and about 100 kinds of diseases, which includes two main data types: whole slide images (WSIs) and characteristic regions of interest (ROIs). A four-scale WSI standardization process is proposed based on the uniform resolution in microns per pixel (MPP), while the ROIs are divided into three scales artificially. This multi-scale dataset is built with the diagnosis habits under the supervision of experienced senior pathologists. The CPIA dataset facilitates a comprehensive pathological understanding and enables pattern discovery explorations. Additionally, to launch the CPIA dataset, several state-of-the-art (SOTA) baselines of SSL pre-training and downstream evaluation are specially conducted. The CPIA dataset along with baselines is available at https://github.com/zhanglab2021/CPIA_Dataset.
Multimodal Aspect-based Sentiment Analysis (MABSA) is a fine-grained Sentiment Analysis task, which has attracted growing research interests recently. Existing work mainly utilizes image information to improve the performance of MABSA task. However, most of the studies overestimate the importance of images since there are many noise images unrelated to the text in the dataset, which will have a negative impact on model learning. Although some work attempts to filter low-quality noise images by setting thresholds, relying on thresholds will inevitably filter out a lot of useful image information. Therefore, in this work, we focus on whether the negative impact of noisy images can be reduced without modifying the data. To achieve this goal, we borrow the idea of Curriculum Learning and propose a Multi-grained Multi-curriculum Denoising Framework (M2DF), which can achieve denoising by adjusting the order of training data. Extensive experimental results show that our framework consistently outperforms state-of-the-art work on three sub-tasks of MABSA.
Multimodal Entity Linking (MEL) is a task that aims to link ambiguous mentions within multimodal contexts to referential entities in a multimodal knowledge base. Recent methods for MEL adopt a common framework: they first interact and fuse the text and image to obtain representations of the mention and entity respectively, and then compute the similarity between them to predict the correct entity. However, these methods still suffer from two limitations: first, as they fuse the features of text and image before matching, they cannot fully exploit the fine-grained alignment relations between the mention and entity. Second, their alignment is static, leading to low performance when dealing with complex and diverse data. To address these issues, we propose a novel framework called Dynamic Relation Interactive Network (DRIN) for MEL tasks. DRIN explicitly models four different types of alignment between a mention and entity and builds a dynamic Graph Convolutional Network (GCN) to dynamically select the corresponding alignment relations for different input samples. Experiments on two datasets show that DRIN outperforms state-of-the-art methods by a large margin, demonstrating the effectiveness of our approach.
The increasing reliance on large language models (LLMs) in academic writing has led to a rise in plagiarism. Existing AI-generated text classifiers have limited accuracy and often produce false positives. We propose a novel approach using natural language processing (NLP) techniques, offering quantifiable metrics at both sentence and document levels for easier interpretation by human evaluators. Our method employs a multi-faceted approach, generating multiple paraphrased versions of a given question and inputting them into the LLM to generate answers. By using a contrastive loss function based on cosine similarity, we match generated sentences with those from the student's response. Our approach achieves up to 94% accuracy in classifying human and AI text, providing a robust and adaptable solution for plagiarism detection in academic settings. This method improves with LLM advancements, reducing the need for new model training or reconfiguration, and offers a more transparent way of evaluating and detecting AI-generated text.
* 10 Pages, 4 Figures, 9 Tables, to be published in 18th Workshop on
Innovative Use of NLP for Building Educational Applications
Optical coherence tomography angiography (OCTA) is a non-invasive imaging modality that extends the functionality of OCT by extracting moving red blood cell signals from surrounding static biological tissues. OCTA has emerged as a valuable tool for analyzing skin microvasculature, enabling more accurate diagnosis and treatment monitoring. Most existing OCTA extraction algorithms, such as speckle variance (SV)- and eigen-decomposition (ED)-OCTA, implement a larger number of repeated (NR) OCT scans at the same position to produce high-quality angiography images. However, a higher NR requires a longer data acquisition time, leading to more unpredictable motion artifacts. In this study, we propose a vasculature extraction pipeline that uses only one-repeated OCT scan to generate OCTA images. The pipeline is based on the proposed Vasculature Extraction Transformer (VET), which leverages convolutional projection to better learn the spatial relationships between image patches. In comparison to OCTA images obtained via the SV-OCTA (PSNR: 17.809) and ED-OCTA (PSNR: 18.049) using four-repeated OCT scans, OCTA images extracted by VET exhibit moderate quality (PSNR: 17.515) and higher image contrast while reducing the required data acquisition time from ~8 s to ~2 s. Based on visual observations, the proposed VET outperforms SV and ED algorithms when using neck and face OCTA data in areas that are challenging to scan. This study represents that the VET has the capacity to extract vascularture images from a fast one-repeated OCT scan, facilitating accurate diagnosis for patients.
Pathology image analysis crucially relies on the availability and quality of annotated pathological samples, which are very difficult to collect and need lots of human effort. To address this issue, beyond traditional preprocess data augmentation methods, mixing-based approaches are effective and practical. However, previous mixing-based data augmentation methods do not thoroughly explore the essential characteristics of pathology images, including the local specificity, global distribution, and inner/outer-sample instance relationship. To further understand the pathology characteristics and make up effective pseudo samples, we propose the CellMix framework with a novel distribution-based in-place shuffle strategy. We split the images into patches with respect to the granularity of pathology instances and do the shuffle process across the same batch. In this way, we generate new samples while keeping the absolute relationship of pathology instances intact. Furthermore, to deal with the perturbations and distribution-based noise, we devise a loss-drive strategy inspired by curriculum learning during the training process, making the model fit the augmented data adaptively. It is worth mentioning that we are the first to explore data augmentation techniques in the pathology image field. Experiments show SOTA results on 7 different datasets. We conclude that this novel instance relationship-based strategy can shed light on general data augmentation for pathology image analysis. The code is available at https://github.com/sagizty/CellMix.
The curve skeleton is an important shape descriptor that has been utilized in various applications in computer graphics, machine vision, and artificial intelligence. In this study, the endpoint-based part-aware curve skeleton (EPCS) extraction method for low-quality point clouds is proposed. The novel random center shift (RCS) method is first proposed for detecting the endpoints on point clouds. The endpoints are used as the initial seed points for dividing each part into layers, and then the skeletal points are obtained by computing the center points of the oriented bounding box (OBB) of the layers. Subsequently, the skeletal points are connected, thus forming the branches. Furthermore, the multi-vector momentum-driven (MVMD) method is also proposed for locating the junction points that connect the branches. Due to the shape differences between different parts on point clouds, the global topology of the skeleton is finally optimized by removing the redundant junction points, re-connecting some branches using the proposed MVMD method, and applying an interpolation method based on the splitting operator. Consequently, a complete and smooth curve skeleton is achieved. The proposed EPCS method is compared with several state-of-the-art methods, and the experimental results verify its robustness, effectiveness, and efficiency. Furthermore, the skeleton extraction and model segmentation results on the point clouds of broken Terracotta also highlight the utility of the proposed method.
We propose a new spectrum allocation strategy, aided by unsupervised learning, for multiuser terahertz communication systems. In this strategy, adaptive sub-band bandwidth is considered such that the spectrum of interest can be divided into sub-bands with unequal bandwidths. This strategy reduces the variation in molecular absorption loss among the users, leading to the improved data rate performance. We first formulate an optimization problem to determine the optimal sub-band bandwidth and transmit power, and then propose the unsupervised learning-based approach to obtaining the near-optimal solution to this problem. In the proposed approach, we first train a deep neural network (DNN) while utilizing a loss function that is inspired by the Lagrangian of the formulated problem. Then using the trained DNN, we approximate the near-optimal solutions. Numerical results demonstrate that comparing to existing approaches, our proposed unsupervised learning-based approach achieves a higher data rate, especially when the molecular absorption coefficient within the spectrum of interest varies in a highly non-linear manner.
* This paper has been accepted for publication in IEEE Global
Communications Conferences (GLOBECOM) 2022. Copyright may be transferred
without notice, after which this version may no longer be accessible