Lili Zhao

Wei Chen, Lili Zhao, Zhi Zheng et al.

Multi-hop question answering (MHQA) enables accurate answers to complex queries by retrieving and reasoning over evidence dispersed across multiple documents. Existing MHQA approaches mainly rely on iterative retrieval-augmented generation, which suffer from the following two major issues. 1) Existing methods prematurely commit to surface-level entities rather than underlying reasoning structures, making question decomposition highly vulnerable to lexical ambiguity. 2) Existing methods overlook the logical dependencies among reasoning steps, resulting in uncoordinated execution. To address these issues, we propose STRIDE, a framework that separates strategic planning, dynamic control, and grounded execution. At its core, a Meta-Planner first constructs an entity-agnostic reasoning skeleton to capture the abstract logic of the query, thereby deferring entity grounding until after the reasoning structure is established, which mitigates disambiguation errors caused by premature lexical commitment. A Supervisor then orchestrates sub-question execution in a dependency-aware manner, enabling efficient parallelization where possible and sequential coordination when necessary. By dynamically deciding whether to retrieve new evidence or infer from existing facts, it avoids redundant queries and error propagation, while fusing cross-branch information and reformulating failed queries to enhance robustness. Grounded fact extraction and logical inference are delegated to specialized execution modules, ensuring faithfulness through explicit separation of retrieval and reasoning. We further propose STRIDE-FT, a modular fine-tuning framework that uses self-generated execution trajectories from STRIDE, requiring neither human annotations nor stronger teacher models. Experiments show that STRIDE achieves robust and accurate reasoning, while STRIDE-FT effectively enhances open-source LLMs.

Yu Yuan, Lili Zhao, Kai Zhang et al.

Large Language Models (LLMs) have shown remarkable capabilities in various natural language processing tasks. However, LLMs may rely on dataset biases as shortcuts for prediction, which can significantly impair their robustness and generalization capabilities. This paper presents Shortcut Suite, a comprehensive test suite designed to evaluate the impact of shortcuts on LLMs' performance, incorporating six shortcut types, five evaluation metrics, and four prompting strategies. Our extensive experiments yield several key findings: 1) LLMs demonstrate varying reliance on shortcuts for downstream tasks, significantly impairing their performance. 2) Larger LLMs are more likely to utilize shortcuts under zero-shot and few-shot in-context learning prompts. 3) Chain-of-thought prompting notably reduces shortcut reliance and outperforms other prompting strategies, while few-shot prompts generally underperform compared to zero-shot prompts. 4) LLMs often exhibit overconfidence in their predictions, especially when dealing with datasets that contain shortcuts. 5) LLMs generally have a lower explanation quality in shortcut-laden datasets, with errors falling into three types: distraction, disguised comprehension, and logical fallacy. Our findings offer new insights for evaluating robustness and generalization in LLMs and suggest potential directions for mitigating the reliance on shortcuts. The code is available at \url {https://github.com/yyhappier/ShortcutSuite.git}.

Lili Zhao, Dai Feng

There has been increasing interest in modelling survival data using deep learning methods in medical research. Current approaches have focused on designing special cost functions to handle censored survival data. We propose a very different method with two steps. In the first step, we transform each subject's survival time into a series of jackknife pseudo conditional survival probabilities and then use these pseudo probabilities as a quantitative response variable in the deep neural network model. By using the pseudo values, we reduce a complex survival analysis to a standard regression problem, which greatly simplifies the neural network construction. Our two-step approach is simple, yet very flexible in making risk predictions for survival data, which is very appealing from the practice point of view. The source code is freely available at http://github.com/lilizhaoUM/DNNSurv.

Linan Yue, Qi Liu, Lili Zhao et al.

With the development of legal intelligence, Criminal Court View Generation has attracted much attention as a crucial task of legal intelligence, which aims to generate concise and coherent texts that summarize case facts and provide explanations for verdicts. Existing researches explore the key information in case facts to yield the court views. Most of them employ a coarse-grained approach that partitions the facts into broad segments (e.g., verdict-related sentences) to make predictions. However, this approach fails to capture the complex details present in the case facts, such as various criminal elements and legal events. To this end, in this paper, we propose an Event Grounded Generation (EGG) method for criminal court view generation with cooperative (Large) Language Models, which introduces the fine-grained event information into the generation. Specifically, we first design a LLMs-based extraction method that can extract events in case facts without massive annotated events. Then, we incorporate the extracted events into court view generation by merging case facts and events. Besides, considering the computational burden posed by the use of LLMs in the extraction phase of EGG, we propose a LLMs-free EGG method that can eliminate the requirement for event extraction using LLMs in the inference phase. Extensive experimental results on a real-world dataset clearly validate the effectiveness of our proposed method.

Lili Zhao, Qi Liu, Wei Chen et al.

Empirical Risk Minimization (ERM) models often rely on spurious correlations between features and labels during the learning process, leading to shortcut learning behavior that undermines robustness generalization performance. Current research mainly targets identifying or mitigating a single shortcut; however, in real-world scenarios, cues within the data are diverse and unknown. In empirical studies, we reveal that the models rely to varying extents on different shortcuts. Compared to weak shortcuts, models depend more heavily on strong shortcuts, resulting in their poor generalization ability. To address these challenges, we propose MiMu, a novel method integrated with Transformer-based ERMs designed to Mitigate Multiple shortcut learning behavior, which incorporates self-calibration strategy and self-improvement strategy. In the source model, we preliminarily propose the self-calibration strategy to prevent the model from relying on shortcuts and make overconfident predictions. Then, we further design self-improvement strategy in target model to reduce the reliance on multiple shortcuts. The random mask strategy involves randomly masking partial attention positions to diversify the focus of target model other than concentrating on a fixed region. Meanwhile, the adaptive attention alignment module facilitates the alignment of attention weights to the calibrated source model, without the need for post-hoc attention maps or supervision. Finally, extensive experiments conducted on Natural Language Processing (NLP) and Computer Vision (CV) demonstrate the effectiveness of MiMu in improving robustness generalization abilities.

Yu Yuan, Lili Zhao, Wei Chen et al.

Capturing human learning behavior based on deep learning methods has become a major research focus in both psychology and intelligent systems. Recent approaches rely on controlled experiments or rule-based models to explore cognitive processes. However, they struggle to capture learning dynamics, track progress over time, or provide explainability. To address these challenges, we introduce LearnerAgent, a novel multi-agent framework based on Large Language Models (LLMs) to simulate a realistic teaching environment. To explore human-like learning dynamics, we construct learners with psychologically grounded profiles-such as Deep, Surface, and Lazy-as well as a persona-free General Learner to inspect the base LLM's default behavior. Through weekly knowledge acquisition, monthly strategic choices, periodic tests, and peer interaction, we can track the dynamic learning progress of individual learners over a full-year journey. Our findings are fourfold: 1) Longitudinal analysis reveals that only Deep Learner achieves sustained cognitive growth. Our specially designed "trap questions" effectively diagnose Surface Learner's shallow knowledge. 2) The behavioral and cognitive patterns of distinct learners align closely with their psychological profiles. 3) Learners' self-concept scores evolve realistically, with the General Learner developing surprisingly high self-efficacy despite its cognitive limitations. 4) Critically, the default profile of base LLM is a "diligent but brittle Surface Learner"-an agent that mimics the behaviors of a good student but lacks true, generalizable understanding. Extensive simulation experiments demonstrate that LearnerAgent aligns well with real scenarios, yielding more insightful findings about LLMs' behavior.

Jincheng Li, Danyang Dong, Menglin Zheng et al.

Automatic detection of abnormal cervical cells from Thinprep Cytologic Test (TCT) images is a critical component in the development of intelligent computer-aided diagnostic systems. However, existing algorithms typically fail to effectively model the correlations of visual features, while these spatial correlation features actually contain critical diagnostic information. Furthermore, no detection algorithm has the ability to integrate inter-correlation features of cells with intra-discriminative features of cells, lacking a fusion strategy for the end-to-end detection model. In this work, we propose a hypergraph-based cell detection network that effectively fuses different types of features, combining spatial correlation features and deep discriminative features. Specifically, we use a Multi-level Fusion Sub-network (MLF-SNet) to enhance feature extractioncapabilities. Then we introduce a Cross-level Feature Fusion Strategy with Hypergraph Computation module (CLFFS-HC), to integrate mixed features. Finally, we conducted experiments on three publicly available datasets, and the results demonstrate that our method significantly improves the performance of cervical abnormal cell detection.

Hongyi Pan, Gorkem Durak, Elif Keles et al.

Pancreatic cancer is projected to be the second-deadliest cancer by 2030, making early detection critical. Intraductal papillary mucinous neoplasms (IPMNs), key cancer precursors, present a clinical dilemma, as current guidelines struggle to stratify malignancy risk, leading to unnecessary surgeries or missed diagnoses. Here, we developed Cyst-X, an AI framework for IPMN risk prediction trained on a unique, multi-center dataset of 1,461 MRI scans from 764 patients. Cyst-X achieves significantly higher accuracy (AUC = 0.82) than both the established Kyoto guidelines (AUC = 0.75) and expert radiologists, particularly in correct identification of high-risk lesions. Clinically, this translates to a 20% increase in cancer detection sensitivity (87.8% vs. 64.1%) for high-risk lesions. We demonstrate that this performance is maintained in a federated learning setting, allowing for collaborative model training without compromising patient privacy. To accelerate research in early pancreatic cancer detection, we publicly release the Cyst-X dataset and models, providing the first large-scale, multi-center MRI resource for pancreatic cyst analysis.

Zijian Gao, Jingyu Liu, Weiqi Sun et al.

Modern video-text retrieval frameworks basically consist of three parts: video encoder, text encoder and the similarity head. With the success on both visual and textual representation learning, transformer based encoders and fusion methods have also been adopted in the field of video-text retrieval. In this report, we present CLIP2TV, aiming at exploring where the critical elements lie in transformer based methods. To achieve this, We first revisit some recent works on multi-modal learning, then introduce some techniques into video-text retrieval, finally evaluate them through extensive experiments in different configurations. Notably, CLIP2TV achieves 52.9@R1 on MSR-VTT dataset, outperforming the previous SOTA result by 4.1%.

Lili Zhao, Zezhi Zhu, Xuhu Lin et al.

LiDAR point cloud frame interpolation, which synthesizes the intermediate frame between the captured frames, has emerged as an important issue for many applications. Especially for reducing the amounts of point cloud transmission, it is by predicting the intermediate frame based on the reference frames to upsample data to high frame rate ones. However, due to high-dimensional and sparse characteristics of point clouds, it is more difficult to predict the intermediate frame for LiDAR point clouds than videos. In this paper, we propose a novel LiDAR point cloud frame interpolation method, which exploits range images (RIs) as an intermediate representation with CNNs to conduct the frame interpolation process. Considering the inherited characteristics of RIs differ from that of color images, we introduce spatially adaptive convolutions to extract range features adaptively, while a high-efficient flow estimation method is presented to generate optical flows. The proposed model then warps the input frames and range features, based on the optical flows to synthesize the interpolated frame. Extensive experiments on the KITTI dataset have clearly demonstrated that our method consistently achieves superior frame interpolation results with better perceptual quality to that of using state-of-the-art video frame interpolation methods. The proposed method could be integrated into any LiDAR point cloud compression systems for inter prediction.

Dai Feng, Lili Zhao

There has been increasing interest in modeling survival data using deep learning methods in medical research. In this paper, we proposed a Bayesian hierarchical deep neural networks model for modeling and prediction of survival data. Compared with previously studied methods, the new proposal can provide not only point estimate of survival probability but also quantification of the corresponding uncertainty, which can be of crucial importance in predictive modeling and subsequent decision making. The favorable statistical properties of point and uncertainty estimates were demonstrated by simulation studies and real data analysis. The Python code implementing the proposed approach was provided.

Yongle Luo, Kun Dong, Lili Zhao et al.

Efficient and effective learning is one of the ultimate goals of the deep reinforcement learning (DRL), although the compromise has been made in most of the time, especially for the application of robot manipulations. Learning is always expensive for robot manipulation tasks and the learning effectiveness could be affected by the system uncertainty. In order to solve above challenges, in this study, we proposed a simple but powerful reward shaping method, namely Dense2Sparse. It combines the advantage of fast convergence of dense reward and the noise isolation of the sparse reward, to achieve a balance between learning efficiency and effectiveness, which makes it suitable for robot manipulation tasks. We evaluated our Dense2Sparse method with a series of ablation experiments using the state representation model with system uncertainty. The experiment results show that the Dense2Sparse method obtained higher expected reward compared with the ones using standalone dense reward or sparse reward, and it also has a superior tolerance of system uncertainty.

Yansong Tang, Dajun Ding, Yongming Rao et al.

There are substantial instructional videos on the Internet, which enables us to acquire knowledge for completing various tasks. However, most existing datasets for instructional video analysis have the limitations in diversity and scale,which makes them far from many real-world applications where more diverse activities occur. Moreover, it still remains a great challenge to organize and harness such data. To address these problems, we introduce a large-scale dataset called "COIN" for COmprehensive INstructional video analysis. Organized with a hierarchical structure, the COIN dataset contains 11,827 videos of 180 tasks in 12 domains (e.g., vehicles, gadgets, etc.) related to our daily life. With a new developed toolbox, all the videos are annotated effectively with a series of step descriptions and the corresponding temporal boundaries. Furthermore, we propose a simple yet effective method to capture the dependencies among different steps, which can be easily plugged into conventional proposal-based action detection methods for localizing important steps in instructional videos. In order to provide a benchmark for instructional video analysis, we evaluate plenty of approaches on the COIN dataset under different evaluation criteria. We expect the introduction of the COIN dataset will promote the future in-depth research on instructional video analysis for the community.

Yang Du, Chunfeng Yuan, Bing Li et al.

Local features at neighboring spatial positions in feature maps have high correlation since their receptive fields are often overlapped. Self-attention usually uses the weighted sum (or other functions) with internal elements of each local feature to obtain its weight score, which ignores interactions among local features. To address this, we propose an effective interaction-aware self-attention model inspired by PCA to learn attention maps. Furthermore, since different layers in a deep network capture feature maps of different scales, we use these feature maps to construct a spatial pyramid and then utilize multi-scale information to obtain more accurate attention scores, which are used to weight the local features in all spatial positions of feature maps to calculate attention maps. Moreover, our spatial pyramid attention is unrestricted to the number of its input feature maps so it is easily extended to a spatio-temporal version. Finally, our model is embedded in general CNNs to form end-to-end attention networks for action classification. Experimental results show that our method achieves the state-of-the-art results on the UCF101, HMDB51 and untrimmed Charades.

Xu Liu, Lili Zhao, Dajun Ding et al.

This paper proposes an end-to-end deep hashing framework with category mask for fast video retrieval. We train our network in a supervised way by fully exploiting inter-class diversity and intra-class identity. Classification loss is optimized to maximize inter-class diversity, while intra-pair is introduced to learn representative intra-class identity. We investigate the binary bits distribution related to categories and find out that the effectiveness of binary bits is highly correlated with data categories, and some bits may degrade classification performance of some categories. We then design hash code generation scheme with category mask to filter out bits with negative contribution. Experimental results demonstrate the proposed method outperforms several state-of-the-arts under various evaluation metrics on public datasets.

Zhongqi Lu, Erheng Zhong, Lili Zhao et al.

Collaborative filtering (CF) aims to predict users' ratings on items according to historical user-item preference data. In many real-world applications, preference data are usually sparse, which would make models overfit and fail to give accurate predictions. Recently, several research works show that by transferring knowledge from some manually selected source domains, the data sparseness problem could be mitigated. However for most cases, parts of source domain data are not consistent with the observations in the target domain, which may misguide the target domain model building. In this paper, we propose a novel criterion based on empirical prediction error and its variance to better capture the consistency across domains in CF settings. Consequently, we embed this criterion into a boosting framework to perform selective knowledge transfer. Comparing to several state-of-the-art methods, we show that our proposed selective transfer learning framework can significantly improve the accuracy of rating prediction tasks on several real-world recommendation tasks.