Hongyu Chen

Dejun Feng, Hongyu Chen, Suning Liu et al.

With the increasing application of deep learning in various domains, salient object detection in optical remote sensing images (ORSI-SOD) has attracted significant attention. However, most existing ORSI-SOD methods predominantly rely on local information from low-level features to infer salient boundary cues and supervise them using boundary ground truth, but fail to sufficiently optimize and protect the local information, and almost all approaches ignore the potential advantages offered by the last layer of the decoder to maintain the integrity of saliency maps. To address these issues, we propose a novel method named boundary-semantic collaborative guidance network (BSCGNet) with dual-stream feedback mechanism. First, we propose a boundary protection calibration (BPC) module, which effectively reduces the loss of edge position information during forward propagation and suppresses noise in low-level features without relying on boundary ground truth. Second, based on the BPC module, a dual feature feedback complementary (DFFC) module is proposed, which aggregates boundary-semantic dual features and provides effective feedback to coordinate features across different layers, thereby enhancing cross-scale knowledge communication. Finally, to obtain more complete saliency maps, we consider the uniqueness of the last layer of the decoder for the first time and propose the adaptive feedback refinement (AFR) module, which further refines feature representation and eliminates differences between features through a unique feedback mechanism. Extensive experiments on three benchmark datasets demonstrate that BSCGNet exhibits distinct advantages in challenging scenarios and outperforms the 17 state-of-the-art (SOTA) approaches proposed in recent years. Codes and results have been released on GitHub: https://github.com/YUHsss/BSCGNet.

Hongyu Chen, Weiming Zeng, Chengcheng Chen et al.

In the fields of affective computing (AC) and brain-machine interface (BMI), the analysis of physiological and behavioral signals to discern individual emotional states has emerged as a critical research frontier. While deep learning-based approaches have made notable strides in EEG emotion recognition, particularly in feature extraction and pattern recognition, significant challenges persist in achieving end-to-end emotion computation, including real-time processing, individual adaptation, and seamless user interaction. This paper presents the EEG Emotion Copilot, a system optimizing a lightweight large language model (LLM) with 0.5B parameters operating in a local setting, which first recognizes emotional states directly from EEG signals, subsequently generates personalized diagnostic and treatment suggestions, and finally supports the automation of assisted electronic medical records. Specifically, we demonstrate the critical techniques in the novel data structure of prompt, model pruning and fine-tuning training, and deployment strategies aiming at improving real-time performance and computational efficiency. Extensive experiments show that our optimized lightweight LLM-based copilot achieves an enhanced intuitive interface for participant interaction, superior accuracy of emotion recognition and assisted electronic medical records generation, in comparison to such models with similar scale parameters or large-scale parameters such as 1.5B, 1.8B, 3B and 7B. In summary, through these efforts, the proposed copilot is expected to advance the application of AC in the medical domain, offering innovative solution to mental health monitoring. The codes will be released at https://github.com/NZWANG/EEG_Emotion_Copilot.

Kelly Marchisio, Saurabh Dash, Hongyu Chen et al.

Quantization techniques are widely used to improve inference speed and deployment of large language models. While a wide body of work examines the impact of quantization on LLMs in English, none have evaluated across languages. We conduct a thorough analysis of quantized multilingual LLMs, focusing on performance across languages and at varying scales. We use automatic benchmarks, LLM-as-a-Judge, and human evaluation, finding that (1) harmful effects of quantization are apparent in human evaluation, which automatic metrics severely underestimate: a 1.7% average drop in Japanese across automatic tasks corresponds to a 16.0% drop reported by human evaluators on realistic prompts; (2) languages are disparately affected by quantization, with non-Latin script languages impacted worst; and (3) challenging tasks like mathematical reasoning degrade fastest. As the ability to serve low-compute models is critical for wide global adoption of NLP technologies, our results urge consideration of multilingual performance as a key evaluation criterion for efficient models.

Jia Wei, Yun Li, Meiyu Qiu et al.

Laryngo-pharyngeal cancer (LPC) is a highly fatal malignant disease affecting the head and neck region. Previous studies on endoscopic tumor detection, particularly those leveraging dual-branch network architectures, have shown significant advancements in tumor detection. These studies highlight the potential of dual-branch networks in improving diagnostic accuracy by effectively integrating global and local (lesion) feature extraction. However, they are still limited in their capabilities to accurately locate the lesion region and capture the discriminative feature information between the global and local branches. To address these issues, we propose a novel SAM-guided fusion network (SAM-FNet), a dual-branch network for laryngo-pharyngeal tumor detection. By leveraging the powerful object segmentation capabilities of the Segment Anything Model (SAM), we introduce the SAM into the SAM-FNet to accurately segment the lesion region. Furthermore, we propose a GAN-like feature optimization (GFO) module to capture the discriminative features between the global and local branches, enhancing the fusion feature complementarity. Additionally, we collect two LPC datasets from the First Affiliated Hospital (FAHSYSU) and the Sixth Affiliated Hospital (SAHSYSU) of Sun Yat-sen University. The FAHSYSU dataset is used as the internal dataset for training the model, while the SAHSYSU dataset is used as the external dataset for evaluating the model's performance. Extensive experiments on both datasets of FAHSYSU and SAHSYSU demonstrate that the SAM-FNet can achieve competitive results, outperforming the state-of-the-art counterparts. The source code of SAM-FNet is available at the URL of https://github.com/VVJia/SAM-FNet.

Yueyang Li, Weiming Zeng, Wenhao Dong et al.

Single-channel electroencephalogram (EEG) is a cost-effective, comfortable, and non-invasive method for monitoring brain activity, widely adopted by researchers, consumers, and clinicians. The increasing number and proportion of articles on single-channel EEG underscore its growing potential. This paper provides a comprehensive review of single-channel EEG, focusing on development trends, devices, datasets, signal processing methods, recent applications, and future directions. Definitions of bipolar and unipolar configurations in single-channel EEG are clarified to guide future advancements. Applications mainly span sleep staging, emotion recognition, educational research, and clinical diagnosis. Ongoing advancements of single-channel EEG in AI-based EEG generation techniques suggest potential parity or superiority over multichannel EEG performance.

Dingyi Yang, Hongyu Chen, Xinglin Hou et al.

Stylized visual captioning aims to generate image or video descriptions with specific styles, making them more attractive and emotionally appropriate. One major challenge with this task is the lack of paired stylized captions for visual content, so most existing works focus on unsupervised methods that do not rely on parallel datasets. However, these approaches still require training with sufficient examples that have style labels, and the generated captions are limited to predefined styles. To address these limitations, we explore the problem of Few-Shot Stylized Visual Captioning, which aims to generate captions in any desired style, using only a few examples as guidance during inference, without requiring further training. We propose a framework called FS-StyleCap for this task, which utilizes a conditional encoder-decoder language model and a visual projection module. Our two-step training scheme proceeds as follows: first, we train a style extractor to generate style representations on an unlabeled text-only corpus. Then, we freeze the extractor and enable our decoder to generate stylized descriptions based on the extracted style vector and projected visual content vectors. During inference, our model can generate desired stylized captions by deriving the style representation from user-supplied examples. Our automatic evaluation results for few-shot sentimental visual captioning outperform state-of-the-art approaches and are comparable to models that are fully trained on labeled style corpora. Human evaluations further confirm our model s ability to handle multiple styles.

Shaojie Zhuo, Hongyu Chen, Ramchalam Kinattinkara Ramakrishnan et al.

Tiny machine learning (tinyML) has emerged during the past few years aiming to deploy machine learning models to embedded AI processors with highly constrained memory and computation capacity. Low precision quantization is an important model compression technique that can greatly reduce both memory consumption and computation cost of model inference. In this study, we focus on post-training quantization (PTQ) algorithms that quantize a model to low-bit (less than 8-bit) precision with only a small set of calibration data and benchmark them on different tinyML use cases. To achieve a fair comparison, we build a simulated quantization framework to investigate recent PTQ algorithms. Furthermore, we break down those algorithms into essential components and re-assembled a generic PTQ pipeline. With ablation study on different alternatives of components in the pipeline, we reveal key design choices when performing low precision quantization. We hope this work could provide useful data points and shed lights on the future research of low precision quantization.

Wei Zhang, Weiming Zeng, Hongyu Chen et al.

Accurate diagnosis of depression is crucial for timely implementation of optimal treatments, preventing complications and reducing the risk of suicide. Traditional methods rely on self-report questionnaires and clinical assessment, lacking objective biomarkers. Combining fMRI with artificial intelligence can enhance depression diagnosis by integrating neuroimaging indicators. However, the specificity of fMRI acquisition for depression often results in unbalanced and small datasets, challenging the sensitivity and accuracy of classification models. In this study, we propose the Spatio-Temporal Aggregation Network (STANet) for diagnosing depression by integrating CNN and RNN to capture both temporal and spatial features of brain activity. STANet comprises the following steps:(1) Aggregate spatio-temporal information via ICA. (2) Utilize multi-scale deep convolution to capture detailed features. (3) Balance data using the SMOTE to generate new samples for minority classes. (4) Employ the AFGRU classifier, which combines Fourier transformation with GRU, to capture long-term dependencies, with an adaptive weight assignment mechanism to enhance model generalization. The experimental results demonstrate that STANet achieves superior depression diagnostic performance with 82.38% accuracy and a 90.72% AUC. The STFA module enhances classification by capturing deeper features at multiple scales. The AFGRU classifier, with adaptive weights and stacked GRU, attains higher accuracy and AUC. SMOTE outperforms other oversampling methods. Additionally, spatio-temporal aggregated features achieve better performance compared to using only temporal or spatial features. STANet outperforms traditional or deep learning classifiers, and functional connectivity-based classifiers, as demonstrated by ten-fold cross-validation.

Yu Feng, Weiming Zeng, Yifan Xie et al.

Background: Although it has been noticed that depressed patients show differences in processing emotions, the precise neural modulation mechanisms of positive and negative emotions remain elusive. FMRI is a cutting-edge medical imaging technology renowned for its high spatial resolution and dynamic temporal information, making it particularly suitable for the neural dynamics of depression research. Methods: To address this gap, our study firstly leveraged fMRI to delineate activated regions associated with positive and negative emotions in healthy individuals, resulting in the creation of positive emotion atlas (PEA) and negative emotion atlas (NEA). Subsequently, we examined neuroimaging changes in depression patients using these atlases and evaluated their diagnostic performance based on machine learning. Results: Our findings demonstrate that the classification accuracy of depressed patients based on PEA and NEA exceeded 0.70, a notable improvement compared to the whole-brain atlases. Furthermore, ALFF analysis unveiled significant differences between depressed patients and healthy controls in eight functional clusters during the NEA, focusing on the left cuneus, cingulate gyrus, and superior parietal lobule. In contrast, the PEA revealed more pronounced differences across fifteen clusters, involving the right fusiform gyrus, parahippocampal gyrus, and inferior parietal lobule. Limitations: Due to the limited sample size and subtypes of depressed patients, the efficacy may need further validation in future. Conclusions: These findings emphasize the complex interplay between emotion modulation and depression, showcasing significant alterations in both PEA and NEA among depression patients. This research enhances our understanding of emotion modulation in depression, with implications for diagnosis and treatment evaluation.

Yueyang Li, Weiming Zeng, Wenhao Dong et al.

Background: Deep learning models have shown promise in diagnosing neurodevelopmental disorders (NDD) like ASD and ADHD. However, many models either use graph neural networks (GNN) to construct single-level brain functional networks (BFNs) or employ spatial convolution filtering for local information extraction from rs-fMRI data, often neglecting high-order features crucial for NDD classification. Methods: We introduce a Multi-view High-order Network (MHNet) to capture hierarchical and high-order features from multi-view BFNs derived from rs-fMRI data for NDD prediction. MHNet has two branches: the Euclidean Space Features Extraction (ESFE) module and the Non-Euclidean Space Features Extraction (Non-ESFE) module, followed by a Feature Fusion-based Classification (FFC) module for NDD identification. ESFE includes a Functional Connectivity Generation (FCG) module and a High-order Convolutional Neural Network (HCNN) module to extract local and high-order features from BFNs in Euclidean space. Non-ESFE comprises a Generic Internet-like Brain Hierarchical Network Generation (G-IBHN-G) module and a High-order Graph Neural Network (HGNN) module to capture topological and high-order features in non-Euclidean space. Results: Experiments on three public datasets show that MHNet outperforms state-of-the-art methods using both AAL1 and Brainnetome Atlas templates. Extensive ablation studies confirm the superiority of MHNet and the effectiveness of using multi-view fMRI information and high-order features. Our study also offers atlas options for constructing more sophisticated hierarchical networks and explains the association between key brain regions and NDD. Conclusion: MHNet leverages multi-view feature learning from both Euclidean and non-Euclidean spaces, incorporating high-order information from BFNs to enhance NDD classification performance.

Ruicheng Ao, Hongyu Chen, Siyang Gao et al.

Designing service systems requires selecting among alternative configurations -- choosing the best chatbot variant, the optimal routing policy, or the most effective quality control procedure. In many service systems, the primary evidence of performance quality is textual -- customer support transcripts, complaint narratives, compliance review reports -- rather than the scalar measurements assumed by classical optimization methods. Large language models (LLMs) can read such textual evidence and produce standardized quality scores, but these automated judges exhibit systematic biases that vary across alternatives and evaluation instances. Human expert review remains accurate but costly. We study how to identify the best service configuration with high confidence while minimizing expensive human audits, given that automated evaluation is cheap but biased. We formalize this as a sequential decision problem where a biased proxy score is observed for every evaluation, and a verified outcome can be acquired selectively at additional cost. We prove that LLM-only selection fails under arm-dependent bias, and that naive selective-audit estimators can be asymptotically biased. We develop an estimator combining proxy scores with inverse-propensity-weighted residuals and construct anytime-valid confidence sequences. Our algorithm, PP-LUCB, jointly decides which alternatives to evaluate and whether to request human audits, concentrating reviews where the LLM judge is least reliable. We prove correctness and establish instance-dependent cost bounds showing near-optimal efficiency. On a customer support ticket classification task, our algorithm correctly identifies the best model in 40/40 trials while achieving 90\% audit cost reduction.

Yue Liu, Jiaying Wu, Yufei He et al. · pku, tsinghua

Large Reasoning Models (LRMs) significantly improve the reasoning ability of Large Language Models (LLMs) by learning to reason, exhibiting promising performance in solving complex tasks. However, their deliberative reasoning process leads to inefficiencies in token usage, memory consumption, and inference time. Thus, this survey provides a review of efficient inference methods designed specifically for LRMs, focusing on mitigating token inefficiency while preserving the reasoning quality. The overview structure of this paper is shown in Figure~\ref{fig:paper_structure}. First, we introduce a taxonomy to group the recent methods into two main categories: (a) explicit compact Chain-of-Thought (CoT), which reduces tokens while keeping the explicit reasoning structure, and (b) implicit latent CoT, which encodes reasoning steps within hidden representations instead of explicit tokens. Meanwhile, we discuss their strengths and weaknesses. Then, we conduct empirical analyses on existing methods from reasoning scenarios, object functions, and performance \& efficiency aspects. Besides, we present open challenges in this field, including human-centric controllable reasoning, trade-off between interpretability and efficiency of reasoning, ensuring the safety of efficient reasoning, and broader applications of efficient reasoning. In addition, we highlight key insights for enhancing LRMs' inference efficiency via techniques such as model merging, new architectures, and agent routers. We hope this work serves as a valuable guide, helping researchers overcome challenges in this vibrant field. A collection of efficient reasoning methods for LRMs (papers and codes) is provided at this link: https://github.com/yueliu1999/Awesome-Efficient-Inference-for-LRMs.

Xue Yang, Rigui Zhou, Shizheng Jia et al.

Quantum Generative Adversarial Networks (QGANs) offer a promising path for learning data distributions on near-term quantum devices. However, existing QGANs for image synthesis avoid direct full-image generation, relying on classical post-processing or patch-based methods. These approaches dilute the quantum generator's role and struggle to capture global image semantics. To address this, we propose ReQGAN, an end-to-end framework that synthesizes an entire N=2^D-pixel image using a single D-qubit quantum circuit. ReQGAN overcomes two fundamental bottlenecks hindering direct pixel generation: (1) the rigid classical-to-quantum noise interface and (2) the output mismatch between normalized quantum statistics and the desired pixel-intensity space. We introduce a learnable Neural Noise Encoder for adaptive state preparation and a differentiable Intensity Calibration module to map measurements to a stable, visually meaningful pixel domain. Experiments on MNIST and Fashion-MNIST demonstrate that ReQGAN achieves stable training and effective image synthesis under stringent qubit budgets, with ablation studies verifying the contribution of each component.

Yirong Sun, Yanjun Chen, Xin Qiu et al.

Large Audio Language Models (LALMs) excel at semantic and paralinguistic tasks, yet their ability to perceive the fundamental physical attributes of audio such as pitch, loudness, and spatial location remains under-explored. To bridge this gap, we introduce SonicBench, a psychophysically grounded benchmark that systematically evaluates 12 core physical attributes across five perceptual dimensions. Unlike previous datasets, SonicBench uses a controllable generation toolbox to construct stimuli for two complementary paradigms: recognition (absolute judgment) and comparison (relative judgment). This design allows us to probe not only sensory precision but also relational reasoning capabilities, a domain where humans typically exhibit greater proficiency. Our evaluation reveals a substantial deficiency in LALMs' foundational auditory understanding; most models perform near random guessing and, contrary to human patterns, fail to show the expected advantage on comparison tasks. Furthermore, explicit reasoning yields minimal gains. However, our linear probing analysis demonstrates crucially that frozen audio encoders do successfully capture these physical cues (accuracy at least 60%), suggesting that the primary bottleneck lies in the alignment and decoding stages, where models fail to leverage the sensory signals they have already captured.

Hongcheng Gao, Jiashu Qu, Jingyi Tang et al. · tsinghua

The hallucination of large multimodal models (LMMs), providing responses that appear correct but are actually incorrect, limits their reliability and applicability. This paper aims to study the hallucination problem of LMMs in video modality, which is dynamic and more challenging compared to static modalities like images and text. From this motivation, we first present a comprehensive benchmark termed HAVEN for evaluating hallucinations of LMMs in video understanding tasks. It is built upon three dimensions, i.e., hallucination causes, hallucination aspects, and question formats, resulting in 6K questions. Then, we quantitatively study 7 influential factors on hallucinations, e.g., duration time of videos, model sizes, and model reasoning, via experiments of 16 LMMs on the presented benchmark. In addition, inspired by recent thinking models like OpenAI o1, we propose a video-thinking model to mitigate the hallucinations of LMMs via supervised reasoning fine-tuning (SRFT) and direct preference optimization (TDPO)-- where SRFT enhances reasoning capabilities while TDPO reduces hallucinations in the thinking process. Extensive experiments and analyses demonstrate the effectiveness. Remarkably, it improves the baseline by 7.65% in accuracy on hallucination evaluation and reduces the bias score by 4.5%. The code and data are public at https://github.com/Hongcheng-Gao/HAVEN.

Jingyu Peng, Hongyu Chen, Jiancheng Dong et al.

Safety alignment in large language models (LLMs) is commonly implemented as a single static policy embedded in model parameters. However, real-world deployments often require context-dependent safety rules that vary across users, regions, and applications. Existing approaches struggle to provide such conditional control: parameter-level alignment entangles safety behaviors with general capabilities, while prompt-based methods rely on natural language instructions that provide weak enforcement. We propose MOSAIC, a modular framework that enables compositional safety alignment through learnable control tokens optimized over a frozen backbone model. Each token represents a safety constraint and can be flexibly activated and composed at inference time. To train compositional tokens efficiently, we introduce order-based task sampling and a distribution-level alignment objective that mitigates over-refusal. Experiments show that MOSAIC achieves strong defense performance with substantially lower over-refusal while preserving model utility.

Yingming Pu, Tao Lin, Hongyu Chen

Large Language Model (LLM)-based multi-agent systems (MAS) demonstrate remarkable potential for scientific discovery. Existing approaches, however, often automate scientific discovery using predefined workflows that lack rationality constraints. This often leads to aimless hypothesizing and a failure to consistently link hypotheses with evidence, thereby hindering the systematic reduction of uncertainty. Overcoming these limitations fundamentally requires a principled approach to exploration. We introduce PiFlow, an information-theoretical framework, treating automated scientific discovery as a structured uncertainty reduction problem guided by principles (e.g., scientific laws). In evaluations across three distinct scientific domains -- discovering nanomaterial structures, bio-molecules, and superconductor candidates with targeted properties -- our method significantly improves discovery efficiency, reflected by a 73.55\% increase in the Area Under the Curve (AUC) of property values versus exploration steps, and enhances solution quality by 94.06\% compared to a vanilla agent system. Overall, PiFlow serves as a Plug-and-Play method, establishing a novel paradigm shift in highly efficient automated scientific discovery, paving the way for more robust and accelerated AI-driven research. Code is publicly available at our \href{https://github.com/amair-lab/PiFlow}{GitHub}.

Shaoguang Huang, Yunzhen Wang, Haijin Zeng et al.

Snapshot compressive imaging (SCI) captures multispectral images (MSIs) using a single coded two-dimensional (2-D) measurement, but reconstructing high-fidelity MSIs from these compressed inputs remains a fundamentally ill-posed challenge. While diffusion-based reconstruction methods have recently raised the bar for quality, they face critical limitations: a lack of large-scale MSI training data, adverse domain shifts from RGB-pretrained models, and inference inefficiencies due to multi-step sampling. These drawbacks restrict their practicality in real-world applications. In contrast to existing methods, which either follow costly iterative refinement or adapt subspace-based embeddings for diffusion models (e.g. DiffSCI, PSR-SCI), we introduce a fundamentally different paradigm: a self-supervised One-Step Diffusion (OSD) framework specifically designed for SCI. The key novelty lies in using a single-step diffusion refiner to correct an initial reconstruction, eliminating iterative denoising entirely while preserving generative quality. Moreover, we adopt a self-supervised equivariant learning strategy to train both the predictor and refiner directly from raw 2-D measurements, enabling generalization to unseen domains without the need for ground-truth MSI. To further address the challenge of limited MSI data, we design a band-selection-driven distillation strategy that transfers core generative priors from large-scale RGB datasets, effectively bridging the domain gap. Extensive experiments confirm that our approach sets a new benchmark, yielding PSNR gains of 3.44 dB, 1.61 dB, and 0.28 dB on the Harvard, NTIRE, and ICVL datasets, respectively, while reducing reconstruction time by 97.5%. This remarkable improvement in efficiency and adaptability makes our method a significant advancement in SCI reconstruction, combining both accuracy and practicality for real-world deployment.

Tianyu Guo, Hongyu Chen, Hao Liang et al.

Automatic audio captioning is essential for audio understanding, enabling applications such as accessibility and content indexing. However, evaluating the quality of audio captions remains a major challenge, especially in reference-free settings where high-quality ground-truth captions are unavailable. While CLAPScore is currently the most widely used reference-free Audio Caption Evaluation Metric(ACEM), its robustness under diverse conditions has not been systematically validated. To address this gap, we introduce BRACE, a new benchmark designed to evaluate audio caption alignment quality in a reference-free setting. BRACE is primarily designed for assessing ACEMs, and can also be extended to measure the modality alignment abilities of Large Audio Language Model(LALM). BRACE consists of two sub-benchmarks: BRACE-Main for fine-grained caption comparison and BRACE-Hallucination for detecting subtle hallucinated content. We construct these datasets through high-quality filtering, LLM-based corruption, and human annotation. Given the widespread adoption of CLAPScore as a reference-free ACEM and the increasing application of LALMs in audio-language tasks, we evaluate both approaches using the BRACE benchmark, testing CLAPScore across various CLAP model variants and assessing multiple LALMs. Notably, even the best-performing CLAP-based ACEM achieves only a 70.01 F1-score on the BRACE-Main benchmark, while the best LALM reaches just 63.19. By revealing the limitations of CLAP models and LALMs, our BRACE benchmark offers valuable insights into the direction of future research.

Hongyu Chen, Chengcheng Chen, Fei Wang et al.

Recent advancements in synthetic aperture radar (SAR) ship detection using deep learning have significantly improved accuracy and speed, yet effectively detecting small objects in complex backgrounds with fewer parameters remains a challenge. This letter introduces RSNet, a lightweight framework constructed to enhance ship detection in SAR imagery. To ensure accuracy with fewer parameters, we proposed Waveletpool-ContextGuided (WCG) as its backbone, guiding global context understanding through multi-scale wavelet features for effective detection in complex scenes. Additionally, Waveletpool-StarFusion (WSF) is introduced as the neck, employing a residual wavelet element-wise multiplication structure to achieve higher dimensional nonlinear features without increasing network width. The Lightweight-Shared (LS) module is designed as detect components to achieve efficient detection through lightweight shared convolutional structure and multi-format compatibility. Experiments on the SAR Ship Detection Dataset (SSDD) and High-Resolution SAR Image Dataset (HRSID) demonstrate that RSNet achieves a strong balance between lightweight design and detection performance, surpassing many state-of-the-art detectors, reaching 72.5\% and 67.6\% in \textbf{\(\mathbf{mAP_{.50:.95}}\) }respectively with 1.49M parameters. Our code will be released soon.

Ruicheng Ao, Hongyu Chen, Siyang Gao et al.

We study fixed-confidence best-arm identification (BAI) where a cheap but potentially biased proxy (e.g., LLM judge) is available for every sample, while an expensive ground-truth label can only be acquired selectively when using a human for auditing. Unlike classical multi-fidelity BAI, the proxy is biased (arm- and context-dependent) and ground truth is selectively observed. Consequently, standard multi-fidelity methods can mis-select the best arm, and uniform auditing, though accurate, wastes scarce resources and is inefficient. We prove that without bias correction and propensity adjustment, mis-selection probability may not vanish (even with unlimited proxy data). We then develop an estimator for the mean of each arm that combines proxy scores with inverse-propensity-weighted residuals and form anytime-valid confidence sequences for that estimator. Based on the estimator and confidence sequence, we propose an algorithm that adaptively selects and audits arms. The algorithm concentrates audits on unreliable contexts and close arms and we prove that a plug-in Neyman rule achieves near-oracle audit efficiency. Numerical experiments confirm the theoretical guarantees and demonstrate the superior empirical performance of the proposed algorithm.

Yugang Chang, Hongyu Chen, Fei Wang et al.

This paper introduces the Maritime Ship Navigation Behavior Dataset (MID), designed to address challenges in ship detection within complex maritime environments using Oriented Bounding Boxes (OBB). MID contains 5,673 images with 135,884 finely annotated target instances, supporting both supervised and semi-supervised learning. It features diverse maritime scenarios such as ship encounters under varying weather, docking maneuvers, small target clustering, and partial occlusions, filling critical gaps in datasets like HRSID, SSDD, and NWPU-10. MID's images are sourced from high-definition video clips of real-world navigation across 43 water areas, with varied weather and lighting conditions (e.g., rain, fog). Manually curated annotations enhance the dataset's variety, ensuring its applicability to real-world demands in busy ports and dense maritime regions. This diversity equips models trained on MID to better handle complex, dynamic environments, supporting advancements in maritime situational awareness. To validate MID's utility, we evaluated 10 detection algorithms, providing an in-depth analysis of the dataset, detection results from various models, and a comparative study of baseline algorithms, with a focus on handling occlusions and dense target clusters. The results highlight MID's potential to drive innovation in intelligent maritime traffic monitoring and autonomous navigation systems. The dataset will be made publicly available at https://github.com/VirtualNew/MID_DataSet.

Ruicheng Ao, Hongyu Chen, Haoyang Liu et al.

We study semi-supervised stochastic optimization when labeled data is scarce but predictions from pre-trained models are available. PPI and SVRG both reduce variance through control variates -- PPI uses predictions, SVRG uses reference gradients. We show they are mathematically equivalent and develop PPI-SVRG, which combines both. Our convergence bound decomposes into the standard SVRG rate plus an error floor from prediction uncertainty. The rate depends only on loss geometry; predictions affect only the neighborhood size. When predictions are perfect, we recover SVRG exactly. When predictions degrade, convergence remains stable but reaches a larger neighborhood. Experiments confirm the theory: PPI-SVRG reduces MSE by 43--52\% under label scarcity on mean estimation benchmarks and improves test accuracy by 2.7--2.9 percentage points on MNIST with only 10\% labeled data.

Maolin Wang, Dongze Wu, Jianing Zhou et al.

Large language models (LLMs) have become an important semantic infrastructure for modern recommender systems. A prevailing paradigm integrates LLM-derived semantic embeddings with collaborative representations via representation alignment, implicitly assuming that the two views encode a shared latent entity and that stronger alignment yields better results. We formalize this assumption as the global low-complexity alignment hypothesis and argue that it is stronger than necessary and often structurally mismatched with real-world recommendation settings. We propose a complementary perspective in which semantic and collaborative representations are treated as partially shared yet fundamentally heterogeneous views, each containing both shared and view-specific factors. Under this shared-plus-private latent structure, enforcing global geometric alignment may distort local structure, suppress view-specific signals, and reduce informational diversity. To support this perspective, we develop complementarity-aware diagnostics that quantify overlap, unique-hit contribution, and theoretical fusion upper bounds. Empirical analyses on sparse recommendation benchmarks reveal low item-level agreement between semantic and collaborative views and substantial oracle fusion gains, indicating strong complementarity. Furthermore, controlled alignment probes show that low-capacity mappings capture only shared components and fail to recover full collaborative geometry, especially under distribution shift. These findings suggest that alignment should not be treated as the default integration principle. We advocate a shift from alignment-centric modeling to complementarity fusion-centric, complementarity-aware design, where shared factors are selectively integrated while private signals are preserved. This reframing provides a principled foundation for the next generation of LLM-enhanced recommender systems.

Maolin Wang, Beining Bao, Gan Yuan et al.

Electronic health records (EHRs) and other real-world clinical data are essential for clinical research, medical artificial intelligence, and life science, but their sharing is severely limited by privacy, governance, and interoperability constraints. These barriers create persistent data silos that hinder multi-center studies, large-scale model development, and broader biomedical discovery. Existing privacy-preserving approaches, including multi-party computation and related cryptographic techniques, provide strong protection but often introduce substantial computational overhead, reducing the efficiency of large-scale machine learning and foundation-model training. In addition, many such methods make data usable for restricted computation while leaving them effectively invisible to clinicians and researchers, limiting their value in workflows that still require direct inspection, exploratory analysis, and human interpretation. We propose a real-world-data transformation framework for privacy-preserving sharing of structured clinical records. Instead of converting data into opaque representations, our approach constructs transformed numeric views that preserve medical semantics and major statistical properties while, under a clearly specified threat model, provably breaking direct linkage between those views and protected patient-level attributes. Through collaboration between computer scientists and the AI agent \textbf{SciencePal}, acting as a constrained tool inventor under human guidance, we design three transformation operators that are non-reversible within this threat model, together with an additional mixing strategy for high-risk scenarios, supported by theoretical analysis and empirical evaluation under reconstruction, record linkage, membership inference, and attribute inference attacks.

Hongyu Chen, James McGowan, Michael Franz

We present Elevator, the first binary translator that statically translates entire x86-64 executables to AArch64 without debug information, source code, or assumptions about code layout. Unlike existing systems, which rely on heuristics or runtime fallbacks to handle code-versus-data decoding errors, Elevator considers all possible interpretations of every byte and produces a separate translation for each feasible one ahead of time. Any byte may be interpreted as data, an opcode, or an opcode argument; we generate separate control flow paths for all interpretations, pruning only those leading to abnormal termination. Translations are built by composing code "tiles" automatically derived from a high-level description of the source ISA, yielding a nimble translation framework. The approach is deterministic and produces complete, self-contained binaries with no runtime component in the trusted code base. The principal cost is substantial code size expansion. The key benefit is that the output is the actual code that will run, enabling testing, validation, certification, and cryptographic signing prior to deployment, reducing risk compared to emulators or JIT compilers. We evaluate Elevator on a diverse corpus of real-world binaries, including the entire SPECint 2006 suite, demonstrating that static full-program binary translation can be both reliable and practical. Elevator achieves performance on par with or better than QEMU's user-mode JIT emulation.

Hongyu Chen, David Simchi-Levi, Ruoxuan Xiong

Estimating population quantities such as mean outcomes from user feedback is fundamental to platform evaluation and social science, yet feedback is often missing not at random (MNAR): users with stronger opinions are more likely to respond, so standard estimators are biased and the estimand is not identified without additional assumptions. Existing approaches typically rely on strong parametric assumptions or bespoke auxiliary variables that may be unavailable in practice. In this paper, we develop a partial identification framework in which sharp bounds on the estimand are obtained by solving a pair of linear programs whose constraints encode the observed data structure. This formulation naturally incorporates outcome predictions from pretrained models, including large language models (LLMs), as additional linear constraints that tighten the feasible set. We call these predictions weak shadow variables: they satisfy a conditional independence assumption with respect to missingness but need not meet the completeness conditions required by classical shadow-variable methods. When predictions are sufficiently informative, the bounds collapse to a point, recovering standard identification as a special case. In finite samples, to provide valid coverage of the identified set, we propose a set-expansion estimator that achieves slower-than-$\sqrt{n}$ convergence rate in the set-identified regime and the standard $\sqrt{n}$ rate under point identification. In simulations and semi-synthetic experiments on customer-service dialogues, we find that LLM predictions are often ill-conditioned for classical shadow-variable methods yet remain highly effective in our framework. They shrink identification intervals by 75--83\% while maintaining valid coverage under realistic MNAR mechanisms.

Yingming Pu, Tao Lin, Hongyu Chen

The capacity of Large Language Models (LLMs) to generate valid scientific hypotheses for materials synthesis remains largely unquantified, hindered by the absence of benchmarks probing physicochemical logics reasoning. To address this, we introduce MatterMech, a benchmark for evaluating LLM-generated hypotheses across eight nanomaterial synthesis domains. Our analysis reveals a critical disconnect: LLMs are proficient in abstract logic yet fail to ground their reasoning in fundamental physicochemical principles. We demonstrate that our proposed principle-aware prompting methodology substantially outperforms standard Chain-of-Thought, enhancing both hypothesis accuracy and computational efficiency. This work provides a methodological framework to advance LLMs toward reliable scientific hypothesis generation in materials science. The MatterMech benchmark and associated code is publicly available at \href{https://github.com/amair-lab/MatterMech}{GitHub}.

Luhui Cai, Weiming Zeng, Hongyu Chen et al.

Graph deep learning (GDL) has demonstrated impressive performance in predicting population-based brain disorders (BDs) through the integration of both imaging and non-imaging data. However, the effectiveness of GDL based methods heavily depends on the quality of modeling the multi-modal population graphs and tends to degrade as the graph scale increases. Furthermore, these methods often constrain interactions between imaging and non-imaging data to node-edge interactions within the graph, overlooking complex inter-modal correlations, leading to suboptimal outcomes. To overcome these challenges, we propose MM-GTUNets, an end-to-end graph transformer based multi-modal graph deep learning (MMGDL) framework designed for brain disorders prediction at large scale. Specifically, to effectively leverage rich multi-modal information related to diseases, we introduce Modality Reward Representation Learning (MRRL) which adaptively constructs population graphs using a reward system. Additionally, we employ variational autoencoder to reconstruct latent representations of non-imaging features aligned with imaging features. Based on this, we propose Adaptive Cross-Modal Graph Learning (ACMGL), which captures critical modality-specific and modality-shared features through a unified GTUNet encoder taking advantages of Graph UNet and Graph Transformer, and feature fusion module. We validated our method on two public multi-modal datasets ABIDE and ADHD-200, demonstrating its superior performance in diagnosing BDs. Our code is available at https://github.com/NZWANG/MM-GTUNets.

Anjiang Wei, Jiannan Cao, Ran Li et al. · stanford

As large language models (LLMs) become integral to code-related tasks, a central question emerges: Do LLMs truly understand program semantics? We introduce EquiBench, a new benchmark for evaluating LLMs through equivalence checking, i.e., determining whether two programs produce identical outputs for all possible inputs. Unlike prior code generation benchmarks, this task directly tests a model's ability to reason about program semantics. EquiBench consists of 2400 program pairs across four languages and six categories. These pairs are generated through program analysis, compiler scheduling, and superoptimization, ensuring high-confidence labels, nontrivial difficulty, and full automation. We evaluate 19 state-of-the-art LLMs and find that in the most challenging categories, the best accuracies are 63.8% and 76.2%, only modestly above the 50% random baseline. Further analysis reveals that models often rely on syntactic similarity rather than exhibiting robust reasoning about program semantics, highlighting current limitations. Our code and dataset are publicly available at https://github.com/Anjiang-Wei/equibench

Jia Wei, Yun Li, Xiaomao Fan et al.

Laryngo-pharyngeal cancer (LPC) is a highly lethal malignancy in the head and neck region. Recent advancements in tumor detection, particularly through dual-branch network architectures, have significantly improved diagnostic accuracy by integrating global and local feature extraction. However, challenges remain in accurately localizing lesions and fully capitalizing on the complementary nature of features within these branches. To address these issues, we propose SAM-Swin, an innovative SAM-driven Dual-Swin Transformer for laryngo-pharyngeal tumor detection. This model leverages the robust segmentation capabilities of the Segment Anything Model 2 (SAM2) to achieve precise lesion segmentation. Meanwhile, we present a multi-scale lesion-aware enhancement module (MS-LAEM) designed to adaptively enhance the learning of nuanced complementary features across various scales, improving the quality of feature extraction and representation. Furthermore, we implement a multi-scale class-aware guidance (CAG) loss that delivers multi-scale targeted supervision, thereby enhancing the model's capacity to extract class-specific features. To validate our approach, we compiled three LPC datasets from the First Affiliated Hospital (FAHSYSU), the Sixth Affiliated Hospital (SAHSYSU) of Sun Yat-sen University, and Nanfang Hospital of Southern Medical University (NHSMU). The FAHSYSU dataset is utilized for internal training, while the SAHSYSU and NHSMU datasets serve for external evaluation. Extensive experiments demonstrate that SAM-Swin outperforms state-of-the-art methods, showcasing its potential for advancing LPC detection and improving patient outcomes. The source code of SAM-Swin is available at the URL of \href{https://github.com/VVJia/SAM-Swin}{https://github.com/VVJia/SAM-Swin}.

Zhiqiang Gao, Shihao Gao, Zixing Zhang et al.

Understanding sentiment in multimodal conversations is a complex yet crucial challenge toward building emotionally intelligent AI systems. The Multimodal Conversational Aspect-based Sentiment Analysis (MCABSA) Challenge invited participants to tackle two demanding subtasks: (1) extracting a comprehensive sentiment sextuple, including holder, target, aspect, opinion, sentiment, and rationale from multi-speaker dialogues, and (2) detecting sentiment flipping, which detects dynamic sentiment shifts and their underlying triggers. For Subtask-I, in the present paper, we designed a structured prompting pipeline that guided large language models (LLMs) to sequentially extract sentiment components with refined contextual understanding. For Subtask-II, we further leveraged the complementary strengths of three LLMs through ensembling to robustly identify sentiment transitions and their triggers. Our system achieved a 47.38% average score on Subtask-I and a 74.12% exact match F1 on Subtask-II, showing the effectiveness of step-wise refinement and ensemble strategies in rich, multimodal sentiment analysis tasks.

Hongyu Chen, Liang Lin, Guangrun Wang

Standard Chain-of-Thought (CoT) prompting empowers Large Language Models (LLMs) with reasoning capabilities, yet its reliance on linear natural language is inherently insufficient for effective world modeling in embodied tasks. While text offers flexibility, it fails to explicitly represent the state-space, object hierarchies, and causal dependencies required for robust robotic planning. To address these limitations, we propose Object-Oriented World Modeling (OOWM), a novel framework that structures embodied reasoning through the lens of software engineering formalisms. We redefine the world model not as a latent vector space, but as an explicit symbolic tuple $W = \langle S, T \rangle$: a State Abstraction ($G_\text{state}$) instantiating the environmental state $S$, coupled with a Control Policy ($G_\text{control}$) representing the transition logic $T: S \times A \rightarrow S'$. OOWM leverages the Unified Modeling Language (UML) to materialize this definition: it employs Class Diagrams to ground visual perception into rigorous object hierarchies, and Activity Diagrams to operationalize planning into executable control flows. Furthermore, we introduce a three-stage training pipeline combining Supervised Fine-Tuning (SFT) with Group Relative Policy Optimization (GRPO). Crucially, this method utilizes outcome-based rewards from the final plan to implicitly optimize the underlying object-oriented reasoning structure, enabling effective learning even with sparse annotations. Extensive evaluations on the MRoom-30k benchmark demonstrate that OOWM significantly outperforms unstructured textual baselines in planning coherence, execution success, and structural fidelity, establishing a new paradigm for structured embodied reasoning.

Yingming Pu, Liping Huang, Tao Lin et al.

With the rapid development of artificial intelligence (AI), large language models (LLMs) such as GPT-4 have garnered significant attention in the scientific community, demonstrating great potential in advancing scientific discovery. This progress raises a critical question: are these LLMs well-aligned with real-world physicochemical principles? Current evaluation strategies largely emphasize fact-based knowledge, such as material property prediction or name recognition, but they often lack an understanding of fundamental physicochemical mechanisms that require logical reasoning. To bridge this gap, our study developed a benchmark consisting of 775 multiple-choice questions focusing on the mechanisms of gold nanoparticle synthesis. By reflecting on existing evaluation metrics, we question whether a direct true-or-false assessment merely suggests conjecture. Hence, we propose a novel evaluation metric, the confidence-based score (c-score), which probes the output logits to derive the precise probability for the correct answer. Based on extensive experiments, our results show that in the context of gold nanoparticle synthesis, LLMs understand the underlying physicochemical mechanisms rather than relying on conjecture. This study underscores the potential of LLMs to grasp intrinsic scientific mechanisms and sets the stage for developing more reliable and effective AI tools across various scientific domains.

Yingming Pu, Tao Lin, Hongyu Chen

Large Language Model (LLM)-based scientific agents have accelerated scientific discovery, yet they often suffer from significant inefficiencies due to adherence to fixed initial priors. Existing approaches predominantly operate within a static hypothesis space, which restricts the discovery of novel phenomena, resulting in computational waste when baseline theories fail. To address this, we propose shifting the focus from searching hypotheses to evolving the underlying scientific principles. We present PiEvo, a principle-evolvable framework that treats scientific discovery as Bayesian optimization over an expanding principle space. By integrating Information-Directed Hypothesis Selection via Gaussian Process and an anomaly-driven augmentation mechanism, PiEvo enables agents to autonomously refine their theoretical worldview. Evaluation across four benchmarks demonstrates that PiEvo (1) achieves an average solution quality of up to 90.81%~93.15%, representing a 29.7%~31.1% improvement over the state-of-the-art, (2) attains an 83.3% speedup in convergence step via significantly reduced sample complexity by optimizing the compact principle space, and (3) maintains robust performance across diverse scientific domains and LLM backbones.

Team Cohere, Aakanksha, Arash Ahmadian et al. · mila

In this report we describe the development of Command A, a powerful large language model purpose-built to excel at real-world enterprise use cases. Command A is an agent-optimised and multilingual-capable model, with support for 23 languages of global business, and a novel hybrid architecture balancing efficiency with top of the range performance. It offers best-in-class Retrieval Augmented Generation (RAG) capabilities with grounding and tool use to automate sophisticated business processes. These abilities are achieved through a decentralised training approach, including self-refinement algorithms and model merging techniques. We also include results for Command R7B which shares capability and architectural similarities to Command A. Weights for both models have been released for research purposes. This technical report details our original training pipeline and presents an extensive evaluation of our models across a suite of enterprise-relevant tasks and public benchmarks, demonstrating excellent performance and efficiency.

Yakun Xie, Suning Liu, Hongyu Chen et al.

Despite significant advancements in salient object detection(SOD) in optical remote sensing images(ORSI), challenges persist due to the intricate edge structures of ORSIs and the complexity of their contextual relationships. Current deep learning approaches encounter difficulties in accurately identifying boundary features and lack efficiency in collaboratively modeling the foreground and background by leveraging contextual features. To address these challenges, we propose a stronger multifaceted collaborative salient object detector in ORSIs, termed LBA-MCNet, which incorporates aspects of localization, balance, and affinity. The network focuses on accurately locating targets, balancing detailed features, and modeling image-level global context information. Specifically, we design the Edge Feature Adaptive Balancing and Adjusting(EFABA) module for precise edge localization, using edge features to guide attention to boundaries and preserve spatial details. Moreover, we design the Global Distributed Affinity Learning(GDAL) module to model global context. It captures global context by generating an affinity map from the encoders final layer, ensuring effective modeling of global patterns. Additionally, deep supervision during deconvolution further enhances feature representation. Finally, we compared with 28 state of the art approaches on three publicly available datasets. The results clearly demonstrate the superiority of our method.

Gongning Luo, Mingwang Xu, Hongyu Chen et al.

Breast cancer is one of the most common causes of death among women worldwide. Early detection helps in reducing the number of deaths. Automated 3D Breast Ultrasound (ABUS) is a newer approach for breast screening, which has many advantages over handheld mammography such as safety, speed, and higher detection rate of breast cancer. Tumor detection, segmentation, and classification are key components in the analysis of medical images, especially challenging in the context of 3D ABUS due to the significant variability in tumor size and shape, unclear tumor boundaries, and a low signal-to-noise ratio. The lack of publicly accessible, well-labeled ABUS datasets further hinders the advancement of systems for breast tumor analysis. Addressing this gap, we have organized the inaugural Tumor Detection, Segmentation, and Classification Challenge on Automated 3D Breast Ultrasound 2023 (TDSC-ABUS2023). This initiative aims to spearhead research in this field and create a definitive benchmark for tasks associated with 3D ABUS image analysis. In this paper, we summarize the top-performing algorithms from the challenge and provide critical analysis for ABUS image examination. We offer the TDSC-ABUS challenge as an open-access platform at https://tdsc-abus2023.grand-challenge.org/ to benchmark and inspire future developments in algorithmic research.

Hongyu Chen, Yiqi Gao, Min Zhou et al.

Recently, integrating visual controls into text-to-image~(T2I) models, such as ControlNet method, has received significant attention for finer control capabilities. While various training-free methods make efforts to enhance prompt following in T2I models, the issue with visual control is still rarely studied, especially in the scenario that visual controls are misaligned with text prompts. In this paper, we address the challenge of ``Prompt Following With Visual Control" and propose a training-free approach named Mask-guided Prompt Following (MGPF). Object masks are introduced to distinct aligned and misaligned parts of visual controls and prompts. Meanwhile, a network, dubbed as Masked ControlNet, is designed to utilize these object masks for object generation in the misaligned visual control region. Further, to improve attribute matching, a simple yet efficient loss is designed to align the attention maps of attributes with object regions constrained by ControlNet and object masks. The efficacy and superiority of MGPF are validated through comprehensive quantitative and qualitative experiments.

Hongyu Chen, Seraphina Goldfarb-Tarrant

Large Language Models (LLMs) are increasingly employed as automated evaluators to assess the safety of generated content, yet their reliability in this role remains uncertain. This study evaluates a diverse set of 11 LLM judge models across critical safety domains, examining three key aspects: self-consistency in repeated judging tasks, alignment with human judgments, and susceptibility to input artifacts such as apologetic or verbose phrasing. Our findings reveal that biases in LLM judges can significantly distort the final verdict on which content source is safer, undermining the validity of comparative evaluations. Notably, apologetic language artifacts alone can skew evaluator preferences by up to 98\%. Contrary to expectations, larger models do not consistently exhibit greater robustness, while smaller models sometimes show higher resistance to specific artifacts. To mitigate LLM evaluator robustness issues, we investigate jury-based evaluations aggregating decisions from multiple models. Although this approach both improves robustness and enhances alignment to human judgements, artifact sensitivity persists even with the best jury configurations. These results highlight the urgent need for diversified, artifact-resistant methodologies to ensure reliable safety assessments.

Preethi Seshadri, Hongyu Chen, Sameer Singh et al.

Large language models (LLMs) are increasingly being deployed in high-stakes applications like hiring, yet their potential for unfair decision-making remains understudied in generative and retrieval settings. In this work, we examine the allocational fairness of LLM-based hiring systems through two tasks that reflect actual HR usage: resume summarization and applicant ranking. By constructing a synthetic resume dataset with controlled perturbations and curating job postings, we investigate whether model behavior differs across demographic groups. Our findings reveal that generated summaries exhibit meaningful differences more frequently for race than for gender perturbations. Models also display non-uniform retrieval selection patterns across demographic groups and exhibit high ranking sensitivity to both gender and race perturbations. Surprisingly, retrieval models can show comparable sensitivity to both demographic and non-demographic changes, suggesting that fairness issues may stem from broader model brittleness. Overall, our results indicate that LLM-based hiring systems, especially in the retrieval stage, can exhibit notable biases that lead to discriminatory outcomes in real-world contexts.

Hongcheng Gao, Zihao Huang, Lin Xu et al.

Achieving human-like perception and reasoning in Multimodal Large Language Models (MLLMs) remains a central challenge in artificial intelligence. While recent research has primarily focused on enhancing reasoning capabilities in MLLMs, a fundamental question persists: Can Multimodal Large Language Models truly perceive the world as humans do? This paper shifts focus from reasoning to perception. Rather than constructing benchmarks specifically for reasoning, we introduce the Turing Eye Test (TET), a challenging perception-oriented benchmark comprising four diagnostic tasks that evaluate MLLMs' performance on synthetic images that humans process intuitively. Our findings reveal that state-of-the-art MLLMs exhibit catastrophic failures on our perceptual tasks trivial for humans. Both in-context learning and training on language backbone-effective for previous benchmarks-fail to improve performance on our tasks, while fine-tuning the vision tower enables rapid adaptation, suggesting that our benchmark poses challenges for vision tower generalization rather than for the knowledge and reasoning capabilities of the language backbone-a key gap between current MLLMs and human perception. We release a representative subset of TET tasks in this version, and will introduce more diverse tasks and methods to enhance visual generalization in future work.

Fei Wang, Chengcheng Chen, Hongyu Chen et al.

Current visual question answering (VQA) tasks often require constructing multimodal datasets and fine-tuning visual language models, which demands significant time and resources. This has greatly hindered the application of VQA to downstream tasks, such as ship information analysis based on Synthetic Aperture Radar (SAR) imagery. To address this challenge, this letter proposes a novel VQA approach that integrates object detection networks with visual language models, specifically designed for analyzing ships in SAR images. This integration aims to enhance the capabilities of VQA systems, focusing on aspects such as ship location, density, and size analysis, as well as risk behavior detection. Initially, we conducted baseline experiments using YOLO networks on two representative SAR ship detection datasets, SSDD and HRSID, to assess each model's performance in terms of detection accuracy. Based on these results, we selected the optimal model, YOLOv8n, as the most suitable detection network for this task. Subsequently, leveraging the vision-language model Qwen2-VL, we designed and implemented a VQA task specifically for SAR scenes. This task employs the ship location and size information output by the detection network to generate multi-turn dialogues and scene descriptions for SAR imagery. Experimental results indicate that this method not only enables fundamental SAR scene question-answering without the need for additional datasets or fine-tuning but also dynamically adapts to complex, multi-turn dialogue requirements, demonstrating robust semantic understanding and adaptability.

Ruicheng Ao, Hongyu Chen, David Simchi-Levi

In this work, we introduce a new framework for active experimentation, the Prediction-Guided Active Experiment (PGAE), which leverages predictions from an existing machine learning model to guide sampling and experimentation. Specifically, at each time step, an experimental unit is sampled according to a designated sampling distribution, and the actual outcome is observed based on an experimental probability. Otherwise, only a prediction for the outcome is available. We begin by analyzing the non-adaptive case, where full information on the joint distribution of the predictor and the actual outcome is assumed. For this scenario, we derive an optimal experimentation strategy by minimizing the semi-parametric efficiency bound for the class of regular estimators. We then introduce an estimator that meets this efficiency bound, achieving asymptotic optimality. Next, we move to the adaptive case, where the predictor is continuously updated with newly sampled data. We show that the adaptive version of the estimator remains efficient and attains the same semi-parametric bound under certain regularity assumptions. Finally, we validate PGAE's performance through simulations and a semi-synthetic experiment using data from the US Census Bureau. The results underscore the PGAE framework's effectiveness and superiority compared to other existing methods.

Hongyu Chen, Letian Ruan, Zilin Xu et al.

LoRA enables efficient customization of LLMs and is widely used in multi-tenant and multi-task serving. However, emerging model architectures such as MoE significantly increase LoRA memory cost, making existing coupled LoRA serving designs poorly scalable and prone to tail-latency inflation. We present InfiniLoRA, a disaggregated LoRA serving system that decouples LoRA execution from base-model inference. InfiniLoRA introduces a shared LoRA Server with parallelism-aware execution, SLO-driven provisioning, and critical-path optimizations, including GPU-initiated communication and hardware-specialized LoRA kernels. Experiments show that InfiniLoRA can achieve an average $3.05\times$ increase in serviceable request rate under strict latency SLOs, and improve the percentage of LoRA adapters satisfying the SLO requirement by 54.0\%.

Siddarth Nilol Kundur Satish, Devesh Jaiswal, Hongyu Chen et al.

Current video generation models produce high-quality aesthetic videos but often struggle to learn representations of real-world physics dynamics, resulting in artifacts such as unnatural object collisions, inconsistent gravity, and temporal flickering. In this work, we propose PhysVideoGenerator, a proof-of-concept framework that explicitly embeds a learnable physics prior into the video generation process. We introduce a lightweight predictor network, PredictorP, which regresses high-level physical features extracted from a pre-trained Video Joint Embedding Predictive Architecture (V-JEPA 2) directly from noisy diffusion latents. These predicted physics tokens are injected into the temporal attention layers of a DiT-based generator (Latte) via a dedicated cross-attention mechanism. Our primary contribution is demonstrating the technical feasibility of this joint training paradigm: we show that diffusion latents contain sufficient information to recover V-JEPA 2 physical representations, and that multi-task optimization remains stable over training. This report documents the architectural design, technical challenges, and validation of training stability, establishing a foundation for future large-scale evaluation of physics-aware generative models.

Hongyu Chen, Jiping Liu, Yong Wang et al.

Unsupervised Domain Adaptation (UDA) has shown promise in effectively alleviating the performance degradation caused by domain gaps between source and target domains, and it can potentially be generalized to UAV object detection in adverse scenes. However, existing UDA studies are based on natural images or clear UAV imagery, and research focused on UAV imagery in adverse conditions is still in its infancy. Moreover, due to the unique perspective of UAVs and the interference from adverse conditions, these methods often fail to accurately align features and are influenced by limited or noisy pseudo-labels. To address this, we propose the first benchmark for UAV object detection in adverse scenes, the Statistical Feedback-Driven Threshold and Mask Adjustment Teacher-Student Framework (SF-TMAT). Specifically, SF-TMAT introduces a design called Dynamic Step Feedback Mask Adjustment Autoencoder (DSFMA), which dynamically adjusts the mask ratio and reconstructs feature maps by integrating training progress and loss feedback. This approach dynamically adjusts the learning focus at different training stages to meet the model's needs for learning features at varying levels of granularity. Additionally, we propose a unique Variance Feedback Smoothing Threshold (VFST) strategy, which statistically computes the mean confidence of each class and dynamically adjusts the selection threshold by incorporating a variance penalty term. This strategy improves the quality of pseudo-labels and uncovers potentially valid labels, thus mitigating domain bias. Extensive experiments demonstrate the superiority and generalization capability of the proposed SF-TMAT in UAV object detection under adverse scene conditions. The Code is released at https://github.com/ChenHuyoo .

Weizhi Liu, Xi Chen, Zekun Jiang et al.

Knee osteoarthritis (KOA) affects more than 600 million individuals globally and is associated with significant pain, functional impairment, and disability. While personalized multidisciplinary interventions have the potential to slow disease progression and enhance quality of life, they typically require substantial medical resources and expertise, making them difficult to implement in resource-limited settings. To address this challenge, we developed KOM, a multi-agent system designed to automate KOA evaluation, risk prediction, and treatment prescription. This system assists clinicians in performing essential tasks across the KOA care pathway and supports the generation of tailored management plans based on individual patient profiles, disease status, risk factors, and contraindications. In benchmark experiments, KOM demonstrated superior performance compared to several general-purpose large language models in imaging analysis and prescription generation. A randomized three-arm simulation study further revealed that collaboration between KOM and clinicians reduced total diagnostic and planning time by 38.5% and resulted in improved treatment quality compared to each approach used independently. These findings indicate that KOM could help facilitate automated KOA management and, when integrated into clinical workflows, has the potential to enhance care efficiency. The modular architecture of KOM may also offer valuable insights for developing AI-assisted management systems for other chronic conditions.

Haonan Chen, Jiaming Xu, Hongyu Chen et al.

Effectively integrating diverse sensory modalities is crucial for robotic manipulation. However, the typical approach of feature concatenation is often suboptimal: dominant modalities such as vision can overwhelm sparse but critical signals like touch in contact-rich tasks, and monolithic architectures cannot flexibly incorporate new or missing modalities without retraining. Our method factorizes the policy into a set of diffusion models, each specialized for a single representation (e.g., vision or touch), and employs a router network that learns consensus weights to adaptively combine their contributions, enabling incremental of new representations. We evaluate our approach on simulated manipulation tasks in {RLBench}, as well as real-world tasks such as occluded object picking, in-hand spoon reorientation, and puzzle insertion, where it significantly outperforms feature-concatenation baselines on scenarios requiring multimodal reasoning. Our policy further demonstrates robustness to physical perturbations and sensor corruption. We further conduct perturbation-based importance analysis, which reveals adaptive shifts between modalities.

Hongyu Chen, Guangrun Wang

Chain-of-Thought (CoT) prompting improves reasoning in large language models (LLMs), but its reliance on unstructured text limits interpretability and executability in embodied tasks. Prior work has explored structured CoTs using scene or logic graphs, yet these remain fundamentally limited: they model only low-order relations, lack constructs like inheritance or behavioral abstraction, and provide no standardized semantics for sequential or conditional planning. We propose UML-CoT, a structured reasoning and planning framework that leverages Unified Modeling Language (UML) to generate symbolic CoTs and executable action plans. UML class diagrams capture compositional object semantics, while activity diagrams model procedural control flow. Our three-stage training pipeline combines supervised fine-tuning with Group Relative Policy Optimization (GRPO), including reward learning from answer-only data. We evaluate UML-CoT on MRoom-30k, a new benchmark of cluttered room-cleaning scenarios. UML-CoT outperforms unstructured CoTs in interpretability, planning coherence, and execution success, highlighting UML as a more expressive and actionable structured reasoning formalism.