Dazhong Shen

Yang Ji, Ying Sun, Yuting Zhang et al.

Neural networks have achieved remarkable success across various fields. However, the lack of interpretability limits their practical use, particularly in critical decision-making scenarios. Post-hoc interpretability, which provides explanations for pre-trained models, is often at risk of robustness and fidelity. This has inspired a rising interest in self-interpretable neural networks, which inherently reveal the prediction rationale through the model structures. Although there exist surveys on post-hoc interpretability, a comprehensive and systematic survey of self-interpretable neural networks is still missing. To address this gap, we first collect and review existing works on self-interpretable neural networks and provide a structured summary of their methodologies from five key perspectives: attribution-based, function-based, concept-based, prototype-based, and rule-based self-interpretation. We also present concrete, visualized examples of model explanations and discuss their applicability across diverse scenarios, including image, text, graph data, and deep reinforcement learning. Additionally, we summarize existing evaluation metrics for self-interpretability and identify open challenges in this field, offering insights for future research. To support ongoing developments, we present a publicly accessible resource to track advancements in this domain: https://github.com/yangji721/Awesome-Self-Interpretable-Neural-Network.

Zhuofan Zong, Bingqi Ma, Dazhong Shen et al. · tsinghua

As the key component in multimodal large language models (MLLMs), the ability of the visual encoder greatly affects MLLM's understanding on diverse image content. Although some large-scale pretrained vision encoders such as vision encoders in CLIP and DINOv2 have brought promising performance, we found that there is still no single vision encoder that can dominate various image content understanding, e.g., the CLIP vision encoder leads to outstanding results on general image understanding but poor performance on document or chart content. To alleviate the bias of CLIP vision encoder, we first delve into the inherent behavior of different pre-trained vision encoders and then propose the MoVA, a powerful and novel MLLM, adaptively routing and fusing task-specific vision experts with a coarse-to-fine mechanism. In the coarse-grained stage, we design a context-aware expert routing strategy to dynamically select the most suitable vision experts according to the user instruction, input image, and expertise of vision experts. This benefits from the powerful model function understanding ability of the large language model (LLM). In the fine-grained stage, we elaborately conduct the mixture-of-vision-expert adapter (MoV-Adapter) to extract and fuse task-specific knowledge from various experts. This coarse-to-fine paradigm effectively leverages representations from experts based on multimodal context and model expertise, further enhancing the generalization ability. We conduct extensive experiments to evaluate the effectiveness of the proposed approach. Without any bells and whistles, MoVA can achieve significant performance gains over current state-of-the-art methods in a wide range of challenging multimodal benchmarks.

Dongzhi Jiang, Guanglu Song, Xiaoshi Wu et al.

Diffusion models have demonstrated great success in the field of text-to-image generation. However, alleviating the misalignment between the text prompts and images is still challenging. The root reason behind the misalignment has not been extensively investigated. We observe that the misalignment is caused by inadequate token attention activation. We further attribute this phenomenon to the diffusion model's insufficient condition utilization, which is caused by its training paradigm. To address the issue, we propose CoMat, an end-to-end diffusion model fine-tuning strategy with an image-to-text concept matching mechanism. We leverage an image captioning model to measure image-to-text alignment and guide the diffusion model to revisit ignored tokens. A novel attribute concentration module is also proposed to address the attribute binding problem. Without any image or human preference data, we use only 20K text prompts to fine-tune SDXL to obtain CoMat-SDXL. Extensive experiments show that CoMat-SDXL significantly outperforms the baseline model SDXL in two text-to-image alignment benchmarks and achieves start-of-the-art performance.

Zhuofan Zong, Dongzhi Jiang, Bingqi Ma et al. · tsinghua

Significant achievements in personalization of diffusion models have been witnessed. Conventional tuning-free methods mostly encode multiple reference images by averaging their image embeddings as the injection condition, but such an image-independent operation cannot perform interaction among images to capture consistent visual elements within multiple references. Although the tuning-based Low-Rank Adaptation (LoRA) can effectively extract consistent elements within multiple images through the training process, it necessitates specific finetuning for each distinct image group. This paper introduces EasyRef, a novel plug-and-play adaptation method that enables diffusion models to be conditioned on multiple reference images and the text prompt. To effectively exploit consistent visual elements within multiple images, we leverage the multi-image comprehension and instruction-following capabilities of the multimodal large language model (MLLM), prompting it to capture consistent visual elements based on the instruction. Besides, injecting the MLLM's representations into the diffusion process through adapters can easily generalize to unseen domains, mining the consistent visual elements within unseen data. To mitigate computational costs and enhance fine-grained detail preservation, we introduce an efficient reference aggregation strategy and a progressive training scheme. Finally, we introduce MRBench, a new multi-reference image generation benchmark. Experimental results demonstrate EasyRef surpasses both tuning-free methods like IP-Adapter and tuning-based methods like LoRA, achieving superior aesthetic quality and robust zero-shot generalization across diverse domains.

Yunfei Zhang, Chuan Qin, Dazhong Shen et al.

During the past few decades, cognitive diagnostics modeling has attracted increasing attention in computational education communities, which is capable of quantifying the learning status and knowledge mastery levels of students. Indeed, the recent advances in neural networks have greatly enhanced the performance of traditional cognitive diagnosis models through learning the deep representations of students and exercises. Nevertheless, existing approaches often suffer from the issue of overconfidence in predicting students' mastery levels, which is primarily caused by the unavoidable noise and sparsity in realistic student-exercise interaction data, severely hindering the educational application of diagnostic feedback. To address this, in this paper, we propose a novel Reliable Cognitive Diagnosis(ReliCD) framework, which can quantify the confidence of the diagnosis feedback and is flexible for different cognitive diagnostic functions. Specifically, we first propose a Bayesian method to explicitly estimate the state uncertainty of different knowledge concepts for students, which enables the confidence quantification of diagnostic feedback. In particular, to account for potential differences, we suggest modeling individual prior distributions for the latent variables of different ability concepts using a pre-trained model. Additionally, we introduce a logical hypothesis for ranking confidence levels. Along this line, we design a novel calibration loss to optimize the confidence parameters by modeling the process of student performance prediction. Finally, extensive experiments on four real-world datasets clearly demonstrate the effectiveness of our ReliCD framework.

Feihu Jiang, Chuan Qin, Jingshuai Zhang et al. · baidu

In the contemporary era of widespread online recruitment, resume understanding has been widely acknowledged as a fundamental and crucial task, which aims to extract structured information from resume documents automatically. Compared to the traditional rule-based approaches, the utilization of recently proposed pre-trained document understanding models can greatly enhance the effectiveness of resume understanding. The present approaches have, however, disregarded the hierarchical relations within the structured information presented in resumes, and have difficulty parsing resumes in an efficient manner. To this end, in this paper, we propose a novel model, namely ERU, to achieve efficient resume understanding. Specifically, we first introduce a layout-aware multi-modal fusion transformer for encoding the segments in the resume with integrated textual, visual, and layout information. Then, we design three self-supervised tasks to pre-train this module via a large number of unlabeled resumes. Next, we fine-tune the model with a multi-granularity sequence labeling task to extract structured information from resumes. Finally, extensive experiments on a real-world dataset clearly demonstrate the effectiveness of ERU.

Shengzhe Zhang, Liyi Chen, Dazhong Shen et al.

Multi-modal sequential recommendation (SR) leverages multi-modal data to learn more comprehensive item features and user preferences than traditional SR methods, which has become a critical topic in both academia and industry. Existing methods typically focus on enhancing multi-modal information utility through adaptive modality fusion to capture the evolving of user preference from user-item interaction sequences. However, most of them overlook the interference caused by redundant interest-irrelevant information contained in rich multi-modal data. Additionally, they primarily rely on implicit temporal information based solely on chronological ordering, neglecting explicit temporal signals that could more effectively represent dynamic user interest over time. To address these limitations, we propose a Hierarchical time-aware Mixture of experts for multi-modal Sequential Recommendation (HM4SR) with a two-level Mixture of Experts (MoE) and a multi-task learning strategy. Specifically, the first MoE, named Interactive MoE, extracts essential user interest-related information from the multi-modal data of each item. Then, the second MoE, termed Temporal MoE, captures user dynamic interests by introducing explicit temporal embeddings from timestamps in modality encoding. To further address data sparsity, we propose three auxiliary supervision tasks: sequence-level category prediction (CP) for item feature understanding, contrastive learning on ID (IDCL) to align sequence context with user interests, and placeholder contrastive learning (PCL) to integrate temporal information with modalities for dynamic interest modeling. Extensive experiments on four public datasets verify the effectiveness of HM4SR compared to several state-of-the-art approaches.

Dazhong Shen, Guanglu Song, Yi Zhang et al.

Diffusion models have achieved outstanding image generation by reversing a forward noising process to approximate true data distributions. During training, these models predict diffusion scores from noised versions of true samples in a single forward pass, while inference requires iterative denoising starting from white noise. This training-inference divergences hinder the alignment between inference and training data distributions, due to potential prediction biases and cumulative error accumulation. To address this problem, we propose an intuitive but effective fine-tuning framework, called Adversarial Diffusion Tuning (ADT), by stimulating the inference process during optimization and aligning the final outputs with training data by adversarial supervision. Specifically, to achieve robust adversarial training, ADT features a siamese-network discriminator with a fixed pre-trained backbone and lightweight trainable parameters, incorporates an image-to-image sampling strategy to smooth discriminative difficulties, and preserves the original diffusion loss to prevent discriminator hacking. In addition, we carefully constrain the backward-flowing path for back-propagating gradients along the inference path without incurring memory overload or gradient explosion. Finally, extensive experiments on Stable Diffusion models (v1.5, XL, and v3), demonstrate that ADT significantly improves both distribution alignment and image quality.

Yongwen Ren, Chao Wang, Peng Du et al.

Recent advances in pretrained language models (PLMs) have significantly improved conversational recommender systems (CRS), enabling more fluent and context-aware interactions. To further enhance accuracy and mitigate hallucination, many methods integrate PLMs with knowledge graphs (KGs), but face key challenges: failing to fully exploit PLM reasoning over graph relationships, indiscriminately incorporating retrieved knowledge without context filtering, and neglecting collaborative preferences in multi-turn dialogues. To this end, we propose PCRS-TKA, a prompt-based framework employing retrieval-augmented generation to integrate PLMs with KGs. PCRS-TKA constructs dialogue-specific knowledge trees from KGs and serializes them into texts, enabling structure-aware reasoning while capturing rich entity semantics. Our approach selectively filters context-relevant knowledge and explicitly models collaborative preferences using specialized supervision signals. A semantic alignment module harmonizes heterogeneous inputs, reducing noise and enhancing accuracy. Extensive experiments demonstrate that PCRS-TKA consistently outperforms all baselines in both recommendation and conversational quality.

Yuanxin Zhuang, Dazhong Shen, Ying Sun

Graph generation plays a pivotal role across numerous domains, including molecular design and knowledge graph construction. Although existing methods achieve considerable success in generating realistic graphs, their interpretability remains limited, often obscuring the rationale behind structural decisions. To address this challenge, we propose the Neural Graph Topic Model (NGTM), a novel generative framework inspired by topic modeling in natural language processing. NGTM represents graphs as mixtures of latent topics, each defining a distribution over semantically meaningful substructures, which facilitates explicit interpretability at both local and global scales. The generation process transparently integrates these topic distributions with a global structural variable, enabling clear semantic tracing of each generated graph. Experiments demonstrate that NGTM achieves competitive generation quality while uniquely enabling fine-grained control and interpretability, allowing users to tune structural features or induce biological properties through topic-level adjustments.

Rui Cai, Chao Wang, Qianyi Cai et al.

Knowledge Graph-based recommendations have gained significant attention due to their ability to leverage rich semantic relationships. However, constructing and maintaining Knowledge Graphs (KGs) is resource-intensive, and the accuracy of KGs can suffer from noisy, outdated, or irrelevant triplets. Recent advancements in Large Language Models (LLMs) offer a promising way to improve the quality and relevance of KGs for recommendation tasks. Despite this, integrating LLMs into KG-based systems presents challenges, such as efficiently augmenting KGs, addressing hallucinations, and developing effective joint learning methods. In this paper, we propose the Confidence-aware KG-based Recommendation Framework with LLM Augmentation (CKG-LLMA), a novel framework that combines KGs and LLMs for recommendation task. The framework includes: (1) an LLM-based subgraph augmenter for enriching KGs with high-quality information, (2) a confidence-aware message propagation mechanism to filter noisy triplets, and (3) a dual-view contrastive learning method to integrate user-item interactions and KG data. Additionally, we employ a confidence-aware explanation generation process to guide LLMs in producing realistic explanations for recommendations. Finally, extensive experiments demonstrate the effectiveness of CKG-LLMA across multiple public datasets.

Xiaoshan Yu, Chuan Qin, Dazhong Shen et al.

In the realm of education, both independent learning and group learning are esteemed as the most classic paradigms. The former allows learners to self-direct their studies, while the latter is typically characterized by teacher-directed scenarios. Recent studies in the field of intelligent education have leveraged deep temporal models to trace the learning process, capturing the dynamics of students' knowledge states, and have achieved remarkable performance. However, existing approaches have primarily focused on modeling the independent learning process, with the group learning paradigm receiving less attention. Moreover, the reciprocal effect between the two learning processes, especially their combined potential to foster holistic student development, remains inadequately explored. To this end, in this paper, we propose RIGL, a unified Reciprocal model to trace knowledge states at both the individual and group levels, drawing from the Independent and Group Learning processes. Specifically, we first introduce a time frame-aware reciprocal embedding module to concurrently model both student and group response interactions across various time frames. Subsequently, we employ reciprocal enhanced learning modeling to fully exploit the comprehensive and complementary information between the two behaviors. Furthermore, we design a relation-guided temporal attentive network, comprised of dynamic graph modeling coupled with a temporal self-attention mechanism. It is used to delve into the dynamic influence of individual and group interactions throughout the learning processes. Conclusively, we introduce a bias-aware contrastive learning module to bolster the stability of the model's training. Extensive experiments on four real-world educational datasets clearly demonstrate the effectiveness of the proposed RIGL model.

Chao Wang, Hengshu Zhu, Dazhong Shen et al.

People usually have different intents for choosing items, while their preferences under the same intent may also different. In traditional collaborative filtering approaches, both intent and preference factors are usually entangled in the modeling process, which significantly limits the robustness and interpretability of recommendation performances. For example, the low-rating items are always treated as negative feedback while they actually could provide positive information about user intent. To this end, in this paper, we propose a two-fold representation learning approach, namely Double Disentangled Collaborative Filtering (DDCF), for personalized recommendations. The first-level disentanglement is for separating the influence factors of intent and preference, while the second-level disentanglement is performed to build independent sparse preference representations under individual intent with limited computational complexity. Specifically, we employ two variational autoencoder networks, intent recognition network and preference decomposition network, to learn the intent and preference factors, respectively. In this way, the low-rating items will be treated as positive samples for modeling intents while the negative samples for modeling preferences. Finally, extensive experiments on three real-world datasets and four evaluation metrics clearly validate the effectiveness and the interpretability of DDCF.

Zhi Zheng, Chao Wang, Tong Xu et al.

Recent years have witnessed the rapid accumulation of massive electronic medical records (EMRs), which highly support the intelligent medical services such as drug recommendation. However, prior arts mainly follow the traditional recommendation strategies like collaborative filtering, which usually treat individual drugs as mutually independent, while the latent interactions among drugs, e.g., synergistic or antagonistic effect, have been largely ignored. To that end, in this paper, we target at developing a new paradigm for drug package recommendation with considering the interaction effect within drugs, in which the interaction effects could be affected by patient conditions. Specifically, we first design a pre-training method based on neural collaborative filtering to get the initial embedding of patients and drugs. Then, the drug interaction graph will be initialized based on medical records and domain knowledge. Along this line, we propose a new Drug Package Recommendation (DPR) framework with two variants, respectively DPR on Weighted Graph (DPR-WG) and DPR on Attributed Graph (DPR-AG) to solve the problem, in which each the interactions will be described as signed weights or attribute vectors. In detail, a mask layer is utilized to capture the impact of patient condition, and graph neural networks (GNNs) are leveraged for the final graph induction task to embed the package. Extensive experiments on a real-world data set from a first-rate hospital demonstrate the effectiveness of our DPR framework compared with several competitive baseline methods, and further support the heuristic study for the drug package generation task with adequate performance.

Dazhong Shen, Qi Zhang, Tong Xu et al.

Identifying the arrival times of seismic P-phases plays a significant role in real-time seismic monitoring, which provides critical guidance for emergency response activities. While considerable research has been conducted on this topic, efficiently capturing the arrival times of seismic P-phases hidden within intensively distributed and noisy seismic waves, such as those generated by the aftershocks of destructive earthquakes, remains a real challenge since most common existing methods in seismology rely on laborious expert supervision. To this end, in this paper, we present a machine learning-enhanced framework based on ensemble learning strategy, EL-Picker, for the automatic identification of seismic P-phase arrivals on continuous and massive waveforms. More specifically, EL-Picker consists of three modules, namely, Trigger, Classifier, and Refiner, and an ensemble learning strategy is exploited to integrate several machine learning classifiers. An evaluation of the aftershocks following the MS 8.0 Wenchuan earthquake demonstrates that EL-Picker can not only achieve the best identification performance but also identify 120% more seismic P-phase arrivals as complementary data. Meanwhile, experimental results also reveal both the applicability of different machine learning models for waveforms collected from different seismic stations and the regularities of seismic P-phase arrivals that might be neglected during manual inspection. These findings clearly validate the effectiveness, efficiency, flexibility and stability of EL-Picker.