Learning from preference feedback is a common practice for aligning large language models (LLMs) with human value. Conventionally, preference data is learned and encoded into a scalar reward model that connects a value head with an LLM to produce a scalar score as preference. However, scalar models lack interpretability and are known to be susceptible to biases in datasets. This paper investigates leveraging LLM itself to learn from such preference data and serve as a judge to address both limitations in one shot. Specifically, we prompt the pre-trained LLM to generate initial judgment pairs with contrastive preference in natural language form. The self-generated contrastive judgment pairs are used to train the LLM-as-a-Judge with Direct Preference Optimization (DPO) and incentivize its reasoning capability as a judge. This proposal of learning the LLMas-a-Judge using self-generated Contrastive judgments (Con-J) ensures natural interpretability through the generated rationales supporting the judgments, and demonstrates higher robustness against bias compared to scalar models. Experimental results show that Con-J outperforms the scalar reward model trained on the same collection of preference data, and outperforms a series of open-source and closed-source generative LLMs. We open-source the training process and model weights of Con-J at https://github.com/YeZiyi1998/Con-J.

ABSTRACT It is well known that subarachnoid haemorrhage caused by intracranial aneurysm rupture has a high fatality rate. Therefore, the prediction of rupture risk can help doctors make targeted diagnoses and treatments in advance. In this study, an image‐text‐based hierarchical prediction model for the rupture risk of intracranial aneurysms (IAIT) is proposed, which combines medical images with structured texts to improve the prediction accuracy. This model captures the detailed features in the images, explores the interaction of multimodal features, and achieves better performance. Specifically, ternary‐view partial attention (TPA) is introduced into the image encoder to improve the model's attention to small lesions. With two symmetric local paths and one global path, local features can be better extracted, and Kronecker product is used for full fusion of image‐text features. Experiments on a private dataset show that the proposed model substantially outperforms both unimodal and existing multimodal baselines. It achieves over 12% higher accuracy than the best unimodal text model and over 45% higher than the best unimodal image model. Moreover, the TPA module further improves classification performance, validating the model's effectiveness. Overall, this study demonstrates the potential of multimodal fusion for accurate and interpretable prediction of intracranial aneurysm rupture risk.

With the development of deep learning and natural language processing techniques, pre-trained language models have been widely used to solve information retrieval (IR) problems. Benefiting from the pre-training and fine-tuning paradigm, these models achieve state-of-the-art performance. In previous works, plain texts in Wikipedia have been widely used in the pre-training stage. However, the rich structured information in Wikipedia, such as the titles, abstracts, hierarchical heading (multi-level title) structure, relationship between articles, references, hyperlink structures, and the writing organizations, has not been fully explored. In this paper, we devise four pre-training objectives tailored for IR tasks based on the structured knowledge of Wikipedia. Compared to existing pre-training methods, our approach can better capture the semantic knowledge in the training corpus by leveraging the human-edited structured data from Wikipedia. Experimental results on multiple IR benchmark datasets show the superior performance of our model in both zero-shot and fine-tuning settings compared to existing strong retrieval baselines. Besides, experimental results in biomedical and legal domains demonstrate that our approach achieves better performance in vertical domains compared to previous models, especially in scenarios where long text similarity matching is needed. The code is available at https://github.com/oneal2000/Wikiformer.

Low-light image enhancement aims to improve the visual quality of images captured under poor illumination and has caught much attention these years. However, existing low-light enhancement methods encounter many problems, e.g., they may not be robust to diverse low-light conditions or have to sacrifice computational efficiency for enhancement performance, which hinder their practical applications. To solve these problems, this paper proposes a novel enhancement method, called Pixel-Wise Gamma Correction Mapping (PWGCM), which combines our innovative pixel-wise Gamma Correction (GC) and deep learning. Compared with conventional GC, our pixel-wise GC is characterized by a set of gamma correction maps, which have the same size as the input image and are taken to replace the single global GC parameter of conventional GC. These gamma correction maps are generated from the low-light image input by a lightweight convolutional neural network at low computational cost. New no-reference loss functions are provided to train the network, ensuring reliable unsupervised learning. Furthermore, our PWGCM is enhanced by an iterative strategy, under which the low-light input image is iteratively enhanced based on the generated gamma correction maps and can yield visually pleasant results. Extensive experiments are done to compare our PWGCM with several state-of-the-art methods in terms of visual quality, efficiency, and auxiliary effects on high-level tasks. The comparison results confirm the superiority of our PWGCM.

Passage ranking involves two stages: passage retrieval and passage re-ranking, which are important and challenging topics for both academics and industries in the area of Information Retrieval (IR). However, the commonly-used datasets for passage ranking usually focus on the English language. For non-English scenarios, such as Chinese, the existing datasets are limited in terms of data scale, fine-grained relevance annotation and false negative issues. To address this problem, we introduce T2Ranking, a large-scale Chinese benchmark for passage ranking. T2Ranking comprises more than 300K queries and over 2M unique passages from real-world search engines. Expert annotators are recruited to provide 4-level graded relevance scores (fine-grained) for query-passage pairs instead of binary relevance judgments (coarse-grained). To ease the false negative issues, more passages with higher diversities are considered when performing relevance annotations, especially in the test set, to ensure a more accurate evaluation. Apart from the textual query and passage data, other auxiliary resources are also provided, such as query types and XML files of documents which passages are generated from, to facilitate further studies. To evaluate the dataset, commonly used ranking models are implemented and tested on T2Ranking as baselines. The experimental results show that T2Ranking is challenging and there is still scope for improvement. The full data and all codes are available at https://github.com/THUIR/T2Ranking/.

This paper presents a neural information retrieval pipeline that integrates cooperative learning of query reformulation and neural retrieval models. Our pipeline first exploits an automatic query reformulator to reformulate the user-issued query and then submits the reformulated query to the neural retrieval model. We simultaneously optimize the quality of reformulated queries and ranking performance with an alternate training strategy where query reformulator and neural retrieval model learn from the feedback of each other. Besides, we incorporate knowledge information into automatic query reformulation. The reformulated queries are further improved and contribute to a better ranking performance of the following neural retrieval model. We study two representative neural retrieval models KNRM and BERT in our pipeline. Experiments on two datasets show that our pipeline consistently improves the retrieval performance of the original neural retrieval models while only increases negligible time on automatic query reformulation.

This article considers the event-triggered control for the active seat suspension system with time-varying full-state constraints. Consider that communication resources may be limited, this article proposes a dynamic relative threshold strategy to reduce the communication burden of actuator and controller. Compared with the fixed value as a trigger condition, the dynamically changing thresholds as trigger conditions are more general and universal. The time-varying full-state constraint problem is solved by using the barrier Lyapunov function. In addition, the radial basis function neural networks are employed to approximate the unknown terms. Then, all signals in the resulted system are bounded, and the Zeno behavior can be avoided successfully. Moreover, all the system states satisfy their corresponding constraint condition. Finally, the feasibility and rationality of this method are proved by the simulation analysis of a real example of a seat suspension system.

Knowledge graphs are widely used in information retrieval as they can enhance our semantic understanding of queries and documents. The main idea is to consider entities and entity relationships as side information. Although existing work has achieved improvements in retrieval effectiveness by incorporating information from knowledge graphs into retrieval models, few studies have leveraged knowledge graphs in understanding users' search behavior. We investigate user behavior during session search from the perspective of a knowledge graph. We conduct a query log-based analysis of users' query reformulation and document clicking behavior. Based on a large-scale commercial query log and a knowledge graph, we find new user behavior patterns in terms of query reformulation and document clicking. Our study deepens our understanding of user behavior in session search and provides implications to help improve retrieval models with knowledge graphs.

Relevance measures the relation between query and document which contains several different dimensions, e.g., semantic similarity, topical relatedness, cognitive relevance (the relations in the aspect of knowledge), usefulness, timeliness, utility and so on. However, existing retrieval models mainly focus on semantic similarity and cognitive relevance while ignore other possible dimensions to model relevance. Topical relatedness, as an important dimension to measure relevance, is not well studied in existing neural information retrieval. In this paper, we propose a Topic Enhanced Knowledge-aware retrieval Model (TEKM) that jointly learns semantic similarity, knowledge relevance and topical relatedness to estimate relevance between query and document. We first construct a neural topic model to learn topical information and generate topic embeddings of a query. Then we combine the topic embeddings with a knowledge-aware retrieval model to estimate different dimensions of relevance. Specifically, we exploit kernel pooling to soft match topic embeddings with word and entity in a unified embedding space to generate fine-grained topical relatedness. The whole model is trained in an end-to-end manner. Experiments on a large-scale publicly available benchmark dataset show that TEKM outperforms existing retrieval models. Further analysis also shows how topic relatedness is modeled to improve traditional retrieval model with semantic similarity and knowledge relevance.

Neural ranking models have recently gained much attention in Information Retrieval community and obtain good ranking performance. However, most of these retrieval models focus on capturing the textual matching signals between query and document but do not consider user behavior information that may be helpful for the retrieval task. Specifically, users' click and query reformulation behavior can be represented by a click-through bipartite graph and a session-flow graph, respectively. Such graph representations contain rich user behavior information and may help us better understand users' search intent beyond the textual information. In this study, we aim to incorporate this rich information encoded in these two graphs into existing neural ranking models.

Retrieval models aim to estimate the relevance of a document to a certain query. Although existing retrieval models have gained much success in both deepening our understanding of information seeking behavior and constructing practical retrieval systems (e.g. Web search engines), we have to admit that the models work in a rather different manner than how humans make relevance judgments. In this paper, we aim to reexamine the existing models as well as to propose new ones based on the findings in how human read documents during relevance judgment. First, we summarize a number of reading heuristics from practical user behavior patterns, which are categorized into implicit and explicit heuristics. By reviewing a variety of existing retrieval models, we find that most of them only satisfy a part of these reading heuristics. To evaluate the effectiveness of each heuristic, we conduct an ablation study and find that most heuristics have positive impacts on retrieval performance. We further integrate all the effective heuristics into a new retrieval model named Reading Inspired Model (RIM). Specifically, implicit reading heuristics are incorporated into the model framework and explicit reading heuristics are modeled as a Markov Decision Process and learned by reinforcement learning. Experimental results on a large-scale public available benchmark dataset and two test sets from NTCIR WWW tasks show that RIM outperforms most existing models, which illustrates the effectiveness of the reading heuristics. We believe that this work contributes to constructing retrieval models with both higher retrieval performance and better explainability.

Reading is a complex cognitive activity in many information retrieval related scenarios, such as relevance judgement and question answering. There exists plenty of works which model these processes as a matching problem, which focuses on how to estimate the relevance score between a document and a query. However, little is known about what happened during the reading process, i.e., how users allocate their attention while reading a document during a specific information retrieval task. We believe that a better understanding of this process can help us design better weighting functions inside the document and contributes to the improvement of ranking performance. In this paper, we focus on the reading process during relevance judgement task. We designed a lab-based user study to investigate human reading patterns in assessing a document, where users' eye movements and their labeled relevant text were collected, respectively. Through a systematic analysis into the collected data, we propose a two-stage reading model which consists of a preliminary relevance judgement stage (Stage 1) and a reading with preliminary relevance stage (Stage 2). In addition, we investigate how different behavior biases affect users' reading behaviors in these two stages. Taking these biases into consideration, we further build prediction models for user's reading attention. Experiment results show that query independent features outperform query dependent features, which indicates that users allocate attentions based on many signals other than query terms in this process. Our study sheds light on the understanding of users' attention allocation during relevance judgement and provides implications for improving the design of existing ranking models.

Social emotion classification aims to predict the aggregation of emotional responses embedded in online comments contributed by various users. Such a task is inherently challenging because extracting relevant semantics from free texts is a classical research problem. Moreover, online comments are typically characterized by a sparse feature space, which makes the corresponding emotion classification task very difficult. On the other hand, though deep neural networks have been shown to be effective for speech recognition and image analysis tasks because of their capabilities of transforming sparse low-level features to dense high-level features, their effectiveness on emotion classification requires further investigation. The main contribution of our work reported in this paper is the development of a novel model of semantically rich hybrid neural network (HNN) which leverages unsupervised teaching models to incorporate semantic domain knowledge into the neural network to bootstrap its inference power and interpretability. To our best knowledge, this is the first successful work of incorporating semantics into neural networks to enhance social emotion classification and network interpretability. Through empirical studies based on three real-world social media datasets, our experimental results confirm that the proposed hybrid neural networks outperform other state-of-the-art emotion classification methods.

Short text is prevalent on the Web, but it brings challenges to content analysis methods for the lack of contextual information. Biterm topic model (BTM) is a variant of latent Dirichlet allocation, which effectively infers the latent topic distribution of short text by modeling the generation of biterms in the whole corpus. However, it needs fine-tuning from labels to reduce noise when applied to supervised learning. Motivated by the transfer learning approach, we propose the hybrid neural networks based on BTM and conventional neural networks, which first make the hidden layer of neural networks approximate the inference of BTM. Following this initial pre-training phase, we then use the simple back-propagation algorithm to fine-tune the topic distribution learned from BTM, so as to improve the performance of supervised learning. Our experiment on two diverse collections of short text validates the effectiveness of the proposed hybrid neural networks for social emotion detection.