Author name cluster

Siqi Li

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

10 papers

2 author rows

JBHI Journal 2025 Journal Article

Benchmarking Foundation Models with Multimodal Public Electronic Health Records

Kunyu Yu
Rui Yang
Jingchi Liao
Siqi Li
Huitao Li
Irene Li
Yifan Peng
Rishikesan Kamaleswaran

Foundation models have emerged as a powerful approach for processing electronic health records (EHRs), offering flexibility to handle diverse medical data modalities. In this study, we present a comprehensive benchmark that evaluates the performance, fairness, and interpretability of foundation models, both as unimodal encoders and as multimodal learners, using the publicly available MIMIC-IV database. To support consistent and reproducible evaluation, we developed a standardized data processing pipeline that harmonizes heterogeneous clinical records into an analysis-ready format. We systematically compared twelve foundation models, encompassing both unimodal and multimodal models, as well as domain-specific and general-purpose variants. Our findings demonstrate that incorporating multiple data modalities generally improves predictive performance without introducing additional bias. While domain-specific fine-tuning offers a cost-effective solution for unimodal foundation models, this effectiveness does not translate well to multimodal scenarios. Additionally, our experiments reveal limited task generalizability in current large vision-language models (LVLMs), emphasizing the need for more versatile and robust medical LVLMs. Through this benchmark, we aim to support the development of effective and trustworthy multimodal artificial intelligence (AI) systems for real-world clinical applications.

Details DOI

ICRA Conference 2025 Conference Paper

Chameleon: Fast-Slow Neuro-Symbolic Lane Topology Extraction

Zongzheng Zhang
Xinrun Li
Sizhe Zou
Guoxuan Chi
Siqi Li
Xuchong Qiu
Guoliang Wang 0002
Guantian Zheng

Lane topology extraction involves detecting lanes and traffic elements and determining their relationships, a key perception task for mapless autonomous driving. This task requires complex reasoning, such as determining whether it is possible to turn left into a specific lane. To address this challenge, we introduce neuro-symbolic methods powered by vision-language foundation models (VLMs). Existing approaches have notable limitations: (1) Dense visual prompting with VLMs can achieve strong performance but is costly in terms of both financial resources and carbon footprint, making it impractical for robotics applications. (2) Neuro-symbolic reasoning methods for 3D scene understanding fail to integrate visual inputs when synthesizing programs, making them ineffective in handling complex corner cases. To this end, we propose a fast-slow neuro-symbolic lane topology extraction algorithm, named Chameleon, which alternates between a fast system that directly reasons over detected instances using synthesized programs and a slow system that utilizes a VLM with a chain-of-thought design to handle corner cases. Chameleon leverages the strengths of both approaches, providing an affordable solution while maintaining high performance. We evaluate the method on the OpenLane-V2 dataset, showing consistent improvements across various baseline detectors. Our code, data, and models are publicly available at https://github.com/XR-Lee/neural-symbolic

Details

IROS Conference 2025 Conference Paper

Learning Symmetric Legged Locomotion via State Distribution Symmetrization

Chengrui Zhu
Zhen Zhang
Siqi Li
Qingpeng Li
Yong Liu

Morphological symmetry is a fundamental characteristic of legged animals and robots. Most existing Deep Reinforcement Learning approaches for legged locomotion neglect to exploit this inherent symmetry, often producing unnatural and suboptimal behaviors such as dominant legs or non-periodic gaits. To address this limitation, we propose a novel learning-based framework to systematically optimize symmetry by state distribution symmetrization. First, we introduce the degree of asymmetry (DoA), a quantitative metric that measures the discrepancy between original and mirrored state distributions. Second, we develop an efficient computation method for DoA using gradient ascent with a trained discriminator network. This metric is then incorporated into a reinforcement learning framework by introducing it to the reward function, explicitly encouraging symmetry during policy training. We validate our framework with extensive experiments on quadrupedal and humanoid robots in simulated and real-world environments. Results demonstrate the efficacy of our approach for improving policy symmetry and overall locomotion performance.

Details

NeurIPS Conference 2025 Conference Paper

SpaceServe: Spatial Multiplexing of Complementary Encoders and Decoders for Multimodal LLMs

Zhicheng Li
Shuoming Zhang
Jiacheng Zhao
Siqi Li
Xiyu Shi
Yangyu Zhang
Shuaijiang Li
Donglin Yu

Recent multimodal large language models (MLLMs) marry modality-specific vision or audio encoders with a shared text decoder. While the encoder is compute- intensive but memory-light, the decoder is the opposite, yet state-of-the-art serving stacks still time-multiplex these complementary kernels, idling SMs or HBM in turn. We introduce SpaceServe, a serving system that space-multiplexes MLLMs: it decouples all modality encoders from the decoder, and co-locates them on the same GPU using fine-grained SM partitioning available in modern runtimes. A cost-model-guided Space-Inference Scheduler (SIS) dynamically assigns SM slices, while a Time-Windowed Shortest-Remaining-First (TWSRFT) policy batches en- coder requests to minimise completion latency and smooth decoder arrivals. Evaluation shows that SpaceServe reduces time-per-output-token by 4. 81× on average and up to 28. 9× on Nvidia A100 GPUs. SpaceServe is available at https: //github. com/gofreelee/SpaceServe

PDF Details

IROS Conference 2024 Conference Paper

PreAfford: Universal Affordance-Based Pre-Grasping for Diverse Objects and Environments

Kairui Ding
Boyuan Chen 0009
Ruihai Wu
Yuyang Li
Zongzheng Zhang
Huan-ang Gao
Siqi Li
Guyue Zhou

Robotic manipulation with two-finger grippers is challenged by objects lacking distinct graspable features. Traditional pre-grasping methods, which typically involve repositioning objects or utilizing external aids like table edges, are limited in their adaptability across different object categories and environments. To overcome these limitations, we introduce PreAfford, a novel pre-grasping planning framework incorporating a point-level affordance representation and a relay training approach. Our method significantly improves adaptability, allowing effective manipulation across a wide range of environments and object types. When evaluated on the ShapeNet-v2 dataset, PreAfford not only enhances grasping success rates by 69% but also demonstrates its practicality through successful real-world experiments. These improvements highlight PreAfford’s potential to redefine standards for robotic handling of complex manipulation tasks in diverse settings.

Details

AIIM Journal 2023 Journal Article

Handling missing values in healthcare data: A systematic review of deep learning-based imputation techniques

Mingxuan Liu
Siqi Li
Han Yuan
Marcus Eng Hock Ong
Yilin Ning
Feng Xie
Seyed Ehsan Saffari
Yuqing Shang

Objective The proper handling of missing values is critical to delivering reliable estimates and decisions, especially in high-stakes fields such as clinical research. In response to the increasing diversity and complexity of data, many researchers have developed deep learning (DL)-based imputation techniques. We conducted a systematic review to evaluate the use of these techniques, with a particular focus on the types of data, intending to assist healthcare researchers from various disciplines in dealing with missing data. Materials and methods We searched five databases (MEDLINE, Web of Science, Embase, CINAHL, and Scopus) for articles published prior to February 8, 2023 that described the use of DL-based models for imputation. We examined selected articles from four perspectives: data types, model backbones (i. e. , main architectures), imputation strategies, and comparisons with non-DL-based methods. Based on data types, we created an evidence map to illustrate the adoption of DL models. Results Out of 1822 articles, a total of 111 were included, of which tabular static data (29%, 32/111) and temporal data (40%, 44/111) were the most frequently investigated. Our findings revealed a discernible pattern in the choice of model backbones and data types, for example, the dominance of autoencoder and recurrent neural networks for tabular temporal data. The discrepancy in imputation strategy usage among data types was also observed. The “integrated” imputation strategy, which solves the imputation task simultaneously with downstream tasks, was most popular for tabular temporal data (52%, 23/44) and multi-modal data (56%, 5/9). Moreover, DL-based imputation methods yielded a higher level of imputation accuracy than non-DL methods in most studies. Conclusion The DL-based imputation models are a family of techniques, with diverse network structures. Their designation in healthcare is usually tailored to data types with different characteristics. Although DL-based imputation models may not be superior to conventional approaches across all datasets, it is highly possible for them to achieve satisfactory results for a particular data type or dataset. There are, however, still issues with regard to portability, interpretability, and fairness associated with current DL-based imputation models.

Details DOI

NeurIPS Conference 2023 Conference Paper

SUBP: Soft Uniform Block Pruning for 1$\times$N Sparse CNNs Multithreading Acceleration

JINGYANG XIANG
Siqi Li
Jun Chen
Guang Dai
Shipeng Bai
Yukai Ma
Yong Liu

The study of sparsity in Convolutional Neural Networks (CNNs) has become widespread to compress and accelerate models in environments with limited resources. By constraining N consecutive weights along the output channel to be group-wise non-zero, the recent network with 1$\times$N sparsity has received tremendous popularity for its three outstanding advantages: 1) A large amount of storage space saving by a \emph{Block Sparse Row} matrix. 2) Excellent performance at a high sparsity. 3) Significant speedups on CPUs with Advanced Vector Extensions. Recent work requires selecting and fine-tuning 1$\times$N sparse weights based on dense pre-trained weights, leading to the problems such as expensive training cost and memory access, sub-optimal model quality, as well as unbalanced workload across threads (different sparsity across output channels). To overcome them, this paper proposes a novel \emph{\textbf{S}oft \textbf{U}niform \textbf{B}lock \textbf{P}runing} (SUBP) approach to train a uniform 1$\times$N sparse structured network from scratch. Specifically, our approach tends to repeatedly allow pruned blocks to regrow to the network based on block angular redundancy and importance sampling in a uniform manner throughout the training process. It not only makes the model less dependent on pre-training, reduces the model redundancy and the risk of pruning the important blocks permanently but also achieves balanced workload. Empirically, on ImageNet, comprehensive experiments across various CNN architectures show that our SUBP consistently outperforms existing 1$\times$N and structured sparsity methods based on pre-trained models or training from scratch. Source codes and models are available at \url{https: //github. com/JingyangXiang/SUBP}.

PDF Details

JBHI Journal 2021 Journal Article

AW-SDRLSE: Adaptive Weighting and Scalable Distance Regularized Level Set Evolution for Lymphoma Segmentation on PET Images

Siqi Li
Huiyan Jiang
Haoming Li
Yu-dong Yao

Accurate lymphoma segmentation on Positron Emission Tomography (PET) images is of great importance for medical diagnoses, such as for distinguishing benign and malignant. To this end, this paper proposes an adaptive weighting and scalable distance regularized level set evolution (AW-SDRLSE) method for delineating lymphoma boundaries on 2D PET slices. There are three important characteristics with respect to AW-SDRLSE: 1) A scalable distance regularization term is proposed and a parameter $q$ can control the contour's convergence rate and precision in theory. 2) A novel dynamic annular mask is proposed to calculate mean intensities of local interior and exterior regions and further define the region energy term. 3) As the level set method is sensitive to parameters, we thus propose an adaptive weighting strategy for the length and area energy terms using local region intensity and boundary direction information. AW-SDRLSE is evaluated on 90 cases of real PET data with a mean Dice coefficient of 0. 8796. Comparative results demonstrate the accuracy and robustness of AW-SDRLSE as well as its performance advantages as compared with related level set methods. In addition, experimental results indicate that AW-SDRLSE can be a fine segmentation method for improving the lymphoma segmentation results obtained by deep learning (DL) methods significantly.

Details DOI

AAAI Conference 2020 Conference Paper

Attention-Based Multi-Modal Fusion Network for Semantic Scene Completion

Siqi Li
Changqing Zou
Yipeng Li
Xibin Zhao
Yue Gao

This paper presents an end-to-end 3D convolutional network named attention-based multi-modal fusion network (AMFNet) for the semantic scene completion (SSC) task of inferring the occupancy and semantic labels of a volumetric 3D scene from single-view RGB-D images. Compared with previous methods which use only the semantic features extracted from RGB-D images, the proposed AMFNet learns to perform effective 3D scene completion and semantic segmentation simultaneously via leveraging the experience of inferring 2D semantic segmentation from RGB-D images as well as the reliable depth cues in spatial dimension. It is achieved by employing a multi-modal fusion architecture boosted from 2D semantic segmentation and a 3D semantic completion network empowered by residual attention blocks. We validate our method on both the synthetic SUNCG-RGBD dataset and the real NYUv2 dataset and the results show that our method respectively achieves the gains of 2. 5% and 2. 6% on the synthetic SUNCG-RGBD dataset and the real NYUv2 dataset against the state-of-the-art method.

PDF Details

JBHI Journal 2018 Journal Article

Organ Location Determination and Contour Sparse Representation for Multiorgan Segmentation

Siqi Li
Huiyan Jiang
Yu-dong Yao
Benqiang Yang

Organ segmentation on computed tomography (CT) images is of great importance in medical diagnoses and treatment. This paper proposes organ location determination and contour sparse representation methods (OLD-CSR) for multiorgan segmentation (liver, kidney, and spleen) on abdomen CT images using an extreme learning machine classifier. First, a location determination method is designed to obtain location information of each organ, which is used for coarse segmentation. Second, for coarse-to-fine segmentation, a contour gradient and rate change based feature point extraction method is proposed. A sparse optimization model is developed for refining the contour feature points. Experimentations with 153 CT images demonstrate the performance advantages of OLD-CSR as compared with related work.

Details DOI