Haoyang Li

Juexiao Zhou, Longxi Zhou, Di Wang et al. · tsinghua

Heterogeneous data is endemic due to the use of diverse models and settings of devices by hospitals in the field of medical imaging. However, there are few open-source frameworks for federated heterogeneous medical image analysis with personalization and privacy protection simultaneously without the demand to modify the existing model structures or to share any private data. In this paper, we proposed PPPML-HMI, an open-source learning paradigm for personalized and privacy-preserving federated heterogeneous medical image analysis. To our best knowledge, personalization and privacy protection were achieved simultaneously for the first time under the federated scenario by integrating the PerFedAvg algorithm and designing our novel cyclic secure aggregation with the homomorphic encryption algorithm. To show the utility of PPPML-HMI, we applied it to a simulated classification task namely the classification of healthy people and patients from the RAD-ChestCT Dataset, and one real-world segmentation task namely the segmentation of lung infections from COVID-19 CT scans. For the real-world task, PPPML-HMI achieved $\sim$5\% higher Dice score on average compared to conventional FL under the heterogeneous scenario. Meanwhile, we applied the improved deep leakage from gradients to simulate adversarial attacks and showed the solid privacy-preserving capability of PPPML-HMI. By applying PPPML-HMI to both tasks with different neural networks, a varied number of users, and sample sizes, we further demonstrated the strong robustness of PPPML-HMI.

Haoyang Li, Jing Zhang, Cuiping Li et al.

One of the recent best attempts at Text-to-SQL is the pre-trained language model. Due to the structural property of the SQL queries, the seq2seq model takes the responsibility of parsing both the schema items (i.e., tables and columns) and the skeleton (i.e., SQL keywords). Such coupled targets increase the difficulty of parsing the correct SQL queries especially when they involve many schema items and logic operators. This paper proposes a ranking-enhanced encoding and skeleton-aware decoding framework to decouple the schema linking and the skeleton parsing. Specifically, for a seq2seq encoder-decode model, its encoder is injected by the most relevant schema items instead of the whole unordered ones, which could alleviate the schema linking effort during SQL parsing, and its decoder first generates the skeleton and then the actual SQL query, which could implicitly constrain the SQL parsing. We evaluate our proposed framework on Spider and its three robustness variants: Spider-DK, Spider-Syn, and Spider-Realistic. The experimental results show that our framework delivers promising performance and robustness. Our code is available at https://github.com/RUCKBReasoning/RESDSQL.

Juexiao Zhou, Haoyang Li, Xingyu Liao et al. · tsinghua

Revoking personal private data is one of the basic human rights, which has already been sheltered by several privacy-preserving laws in many countries. However, with the development of data science, machine learning and deep learning techniques, this right is usually neglected or violated as more and more patients' data are being collected and used for model training, especially in intelligent healthcare, thus making intelligent healthcare a sector where technology must meet the law, regulations, and privacy principles to ensure that the innovation is for the common good. In order to secure patients' right to be forgotten, we proposed a novel solution by using auditing to guide the forgetting process, where auditing means determining whether a dataset has been used to train the model and forgetting requires the information of a query dataset to be forgotten from the target model. We unified these two tasks by introducing a new approach called knowledge purification. To implement our solution, we developed AFS, a unified open-source software, which is able to evaluate and revoke patients' private data from pre-trained deep learning models. We demonstrated the generality of AFS by applying it to four tasks on different datasets with various data sizes and architectures of deep learning networks. The software is publicly available at \url{https://github.com/JoshuaChou2018/AFS}.

Sha Yuan, Hanyu Zhao, Shuai Zhao et al. · bytedance, pku

With the rapid development of deep learning, training Big Models (BMs) for multiple downstream tasks becomes a popular paradigm. Researchers have achieved various outcomes in the construction of BMs and the BM application in many fields. At present, there is a lack of research work that sorts out the overall progress of BMs and guides the follow-up research. In this paper, we cover not only the BM technologies themselves but also the prerequisites for BM training and applications with BMs, dividing the BM review into four parts: Resource, Models, Key Technologies and Application. We introduce 16 specific BM-related topics in those four parts, they are Data, Knowledge, Computing System, Parallel Training System, Language Model, Vision Model, Multi-modal Model, Theory&Interpretability, Commonsense Reasoning, Reliability&Security, Governance, Evaluation, Machine Translation, Text Generation, Dialogue and Protein Research. In each topic, we summarize clearly the current studies and propose some future research directions. At the end of this paper, we conclude the further development of BMs in a more general view.

Kimi Team, Tongtong Bai, Yifan Bai et al.

We introduce Kimi K2.5, an open-source multimodal agentic model designed to advance general agentic intelligence. K2.5 emphasizes the joint optimization of text and vision so that two modalities enhance each other. This includes a series of techniques such as joint text-vision pre-training, zero-vision SFT, and joint text-vision reinforcement learning. Building on this multimodal foundation, K2.5 introduces Agent Swarm, a self-directed parallel agent orchestration framework that dynamically decomposes complex tasks into heterogeneous sub-problems and executes them concurrently. Extensive evaluations show that Kimi K2.5 achieves state-of-the-art results across various domains including coding, vision, reasoning, and agentic tasks. Agent Swarm also reduces latency by up to $4.5\times$ over single-agent baselines. We release the post-trained Kimi K2.5 model checkpoint to facilitate future research and real-world applications of agentic intelligence.

Jianxin Yan, Wangze Ni, Zhenxin Li et al.

Retrieval-augmented generation (RAG) improves large language model (LLM) answer quality by grounding generation in external evidence, but processing retrieved contexts makes the prefill stage a dominant serving cost. RAG cache fusion reduces this cost by reusing precomputed key-value (KV) caches for retrieved chunks and selectively recomputing tokens under the current prompt. Existing selectors, however, face a dilemma between quality and efficiency: fast query-agnostic or final-layer query-to-context selectors can miss request-relevant evidence, whereas full-view query-aware selectors require broad context and layer visibility before recomputation and therefore stall the layer-wise cache-fusion pipeline. We present QCFuse, a compressed-view query-aware selector for RAG cache fusion. QCFuse uses chunk-anchor query probing to condition user-query states on compact per-chunk anchors and critical-layer profiling to identify recomputation tokens without all-layer inspection. We implement QCFuse in SGLang and evaluate it on four open-weight LLMs across six datasets. QCFuse reaches full-prefill-level quality. At matched quality, QCFuse achieves an average prefill-time speedup of 1.7x over full prefill and 1.5x over ProphetKV, the strongest quality-preserving baseline.

Yufei Jia, Zhanxiang Cao, Mingrui Yu et al.

Simulation-based RL for contemporary robot control is increasingly organized around GPU-resident simulation: physics, rollout collection, and learning are placed on a single GPU-centric execution path. This paradigm has greatly improved training speed, but it has also encouraged a default assumption that efficient training requires physics to reside on the GPU. We revisit this assumption. Our view is that, in simulation-dominated robot control, the essential question is not which processor runs physics, but whether simulation throughput, policy learning, and runtime synchronization form an efficient end-to-end loop. We present UniLab, a heterogeneous CPU-simulation / GPU-learning architecture that decouples CPU-parallel simulation from GPU policy updates through a unified runtime for data movement, buffering, and synchronization. UniLab is implemented as a complete and extensible training system using MuJoCoUni and MotrixSim CPU-batched physics backends, supporting PPO, SAC, FlashSAC, TD3, and APPO. On representative simulation-based robot control tasks, UniLab improves end-to-end training efficiency by 3--10$\times$ under the same hardware configuration, while reducing dependence on the NVIDIA CUDA-based software stack and supporting cross-platform execution on the Apple macOS platform and the AMD ROCm and Intel XPU accelerator backends. These results show that GPU simulation is an effective path to efficient training, but not a necessary one, broadening the practical system choices available for robot RL training. Project page: https://github.com/unilabsim/UniLab.

Xu Wenxin, Chen Jason Zhang, Xiaoyong Wei et al.

Text-to-Visualization (Text-to-Vis) translates natural language queries into visualization query languages, enabling non-expert users to perform data analysis. However, most existing methods follow a one-shot paradigm that requires users to specify all visualization details in a single round, often leading to cognitive overload and incorrect visualizations. In this paper, we propose PMVis, a progressive multi-turn paradigm for text-to-vis, where users' intents are refined through multi-turn interactions. To support research in this paradigm, we construct PMVisBench, the first dataset designed to capture the progressive and iterative nature of real-world user queries. It is built through VQL simplification and NLQ reconstruction, with explicit rule constraints to ensure each intermediate VQL remains valid and meaningful. Building upon PMVis, we further introduce PMVisAgent, an agent-based framework that simulates realistic user-system dialogues. PMVisAgent consists of a User, a System, and a Validation Agent that performs verification and repair via a ReAct-style tool-use loop to mitigate error accumulation across rounds, with explicit interaction and verification rules to ensure reliability of the multi-agent system. Extensive experiments on PMVisBench demonstrate that PMVisAgent significantly outperforms state-of-the-art text-to-vis baselines. It achieves up to 17.57\% and 23.21\% improvements in execution accuracy in single-table and multi-table settings, respectively, while ablation studies confirm the importance of combining progressive interaction with clarification. The code is available at https://github.com/wxxv/PMVis.

Jincheng Yu, Haoyang Li, Yiwen Liu et al.

Purpose: To develop an interpretable and trustworthy AI framework that combines deep learning based MRI Osteoarthritis Knee Score (MOAKS) prediction with interpretable statistical modeling to study structure-pain relationships at scale using data from the Osteoarthritis Initiative (OAI). Materials and Methods: We first developed a deep learning framework to predict MOAKS features directly from knee MRIs and incorporated conformal prediction to provide prediction uncertainty quantification. This uncertainty-aware strategy enables explicit filtering of model outputs, retaining only high-confidence MOAKS predictions at the knee level. Second, we applied a longitudinal latent class mixed model (LCMM) to examine associations between key structural abnormalities and four complementary knee pain measurements. Results: Among the three MRI-defined abnormalities (i.e., bone marrow lesions (BML), cartilage loss (CART), and meniscal extrusion (ME)), our framework substantially improved the Matthews correlation coefficient (MCC) and some other metrics. For example, MCC increased from 0.69 to 0.91 for BML, from 0.45 to 0.80 for CART, and from 0.59 to 0.89 for ME. Using these high-confidence predictions, we expanded the sample size to 2,175 knees for the LCMM analysis. Two distinct pain trajectories were identified (rapid and stable pain progression). The estimated odds ratios (95% CI) for the rapid progression group were 1.62 (1.12-2.35) for BML, 1.83 (1.24-2.70) for CART loss, and 2.50 (1.75-3.57) for ME. Conclusion: These results highlight the importance of these structural abnormalities as risk factors for pain and functional progression in osteoarthritis.

Yiyou Sun, Xinyang Han, Weichen Zhang et al.

Recent AI systems have achieved strong results on a wide range of benchmarks, yet these gains have not translated into economically meaningful deployment across many professional domains. We argue that this gap is largely an evaluation problem: widely used benchmarks lack sustained performance measurement on real and economically valuable workflows. This paper introduces Agents' Last Exam (ALE), a benchmark designed to evaluate AI agents on long-horizon, economically valuable, real-world tasks with verifiable outcomes. Developed in collaboration with 250+ industry experts, ALE covers non-physical industries defined with reference to O*NET / SOC 2018 (the U.S. federal occupational taxonomy). It is organized around a task taxonomy with 55 subfields grouped into 13 industry clusters covering 1K+ tasks. Current results show that the hardest tier remains far from saturated: across mainstream harness and backbone configurations, the average full pass rate is 2.6%. ALE is designed as a living benchmark: its task pool grows continuously as new workflows and industries are onboarded. More broadly, ALE is intended not merely as another leaderboard, but as an instrument for closing the gap between benchmark success and GDP-relevant impact.

Zeyang Zhang, Xin Wang, Ziwei Zhang et al. · tsinghua

In an era marked by the increasing adoption of Large Language Models (LLMs) for various tasks, there is a growing focus on exploring LLMs' capabilities in handling web data, particularly graph data. Dynamic graphs, which capture temporal network evolution patterns, are ubiquitous in real-world web data. Evaluating LLMs' competence in understanding spatial-temporal information on dynamic graphs is essential for their adoption in web applications, which remains unexplored in the literature. In this paper, we bridge the gap via proposing to evaluate LLMs' spatial-temporal understanding abilities on dynamic graphs, to the best of our knowledge, for the first time. Specifically, we propose the LLM4DyG benchmark, which includes nine specially designed tasks considering the capability evaluation of LLMs from both temporal and spatial dimensions. Then, we conduct extensive experiments to analyze the impacts of different data generators, data statistics, prompting techniques, and LLMs on the model performance. Finally, we propose Disentangled Spatial-Temporal Thoughts (DST2) for LLMs on dynamic graphs to enhance LLMs' spatial-temporal understanding abilities. Our main observations are: 1) LLMs have preliminary spatial-temporal understanding abilities on dynamic graphs, 2) Dynamic graph tasks show increasing difficulties for LLMs as the graph size and density increase, while not sensitive to the time span and data generation mechanism, 3) the proposed DST2 prompting method can help to improve LLMs' spatial-temporal understanding abilities on dynamic graphs for most tasks. The data and codes are publicly available at Github.

Zi Liang, Haibo Hu, Qingqing Ye et al.

The drastic increase of large language models' (LLMs) parameters has led to a new research direction of fine-tuning-free downstream customization by prompts, i.e., task descriptions. While these prompt-based services (e.g. OpenAI's GPTs) play an important role in many businesses, there has emerged growing concerns about the prompt leakage, which undermines the intellectual properties of these services and causes downstream attacks. In this paper, we analyze the underlying mechanism of prompt leakage, which we refer to as prompt memorization, and develop corresponding defending strategies. By exploring the scaling laws in prompt extraction, we analyze key attributes that influence prompt extraction, including model sizes, prompt lengths, as well as the types of prompts. Then we propose two hypotheses that explain how LLMs expose their prompts. The first is attributed to the perplexity, i.e. the familiarity of LLMs to texts, whereas the second is based on the straightforward token translation path in attention matrices. To defend against such threats, we investigate whether alignments can undermine the extraction of prompts. We find that current LLMs, even those with safety alignments like GPT-4, are highly vulnerable to prompt extraction attacks, even under the most straightforward user attacks. Therefore, we put forward several defense strategies with the inspiration of our findings, which achieve 83.8\% and 71.0\% drop in the prompt extraction rate for Llama2-7B and GPT-3.5, respectively. Source code is avaliable at https://github.com/liangzid/PromptExtractionEval.

Ziwei Zhang, Haoyang Li, Zeyang Zhang et al. · tsinghua

Large models have emerged as the most recent groundbreaking achievements in artificial intelligence, and particularly machine learning. However, when it comes to graphs, large models have not achieved the same level of success as in other fields, such as natural language processing and computer vision. In order to promote applying large models for graphs forward, we present a perspective paper to discuss the challenges and opportunities associated with developing large graph models. First, we discuss the desired characteristics of large graph models. Then, we present detailed discussions from three key perspectives: representation basis, graph data, and graph models. In each category, we provide a brief overview of recent advances and highlight the remaining challenges together with our visions. Finally, we discuss valuable applications of large graph models. We believe this perspective can encourage further investigations into large graph models, ultimately pushing us one step closer towards artificial general intelligence (AGI). We are the first to comprehensively study large graph models, to the best of our knowledge.

Haoyang Li, Xin Wang, Wenwu Zhu · tsinghua

Graph machine learning has been extensively studied in both academia and industry. However, in the literature, most existing graph machine learning models are designed to conduct training with data samples in a random order, which may suffer from suboptimal performance due to ignoring the importance of different graph data samples and their training orders for the model optimization status. To tackle this critical problem, curriculum graph machine learning (Graph CL), which integrates the strength of graph machine learning and curriculum learning, arises and attracts an increasing amount of attention from the research community. Therefore, in this paper, we comprehensively overview approaches on Graph CL and present a detailed survey of recent advances in this direction. Specifically, we first discuss the key challenges of Graph CL and provide its formal problem definition. Then, we categorize and summarize existing methods into three classes based on three kinds of graph machine learning tasks, i.e., node-level, link-level, and graph-level tasks. Finally, we share our thoughts on future research directions. To the best of our knowledge, this paper is the first survey for curriculum graph machine learning.

Javier Duarte, Haoyang Li, Avik Roy et al.

The findable, accessible, interoperable, and reusable (FAIR) data principles provide a framework for examining, evaluating, and improving how data is shared to facilitate scientific discovery. Generalizing these principles to research software and other digital products is an active area of research. Machine learning (ML) models -- algorithms that have been trained on data without being explicitly programmed -- and more generally, artificial intelligence (AI) models, are an important target for this because of the ever-increasing pace with which AI is transforming scientific domains, such as experimental high energy physics (HEP). In this paper, we propose a practical definition of FAIR principles for AI models in HEP and describe a template for the application of these principles. We demonstrate the template's use with an example AI model applied to HEP, in which a graph neural network is used to identify Higgs bosons decaying to two bottom quarks. We report on the robustness of this FAIR AI model, its portability across hardware architectures and software frameworks, and its interpretability.

Yapeng Teng, Haoyang Li, Fuzhen Cai et al.

Anomaly Detection (AD), as a critical problem, has been widely discussed. In this paper, we specialize in one specific problem, Visual Defect Detection (VDD), in many industrial applications. And in practice, defect image samples are very rare and difficult to collect. Thus, we focus on the unsupervised visual defect detection and localization tasks and propose a novel framework based on the recent score-based generative models, which synthesize the real image by iterative denoising through stochastic differential equations (SDEs). Our work is inspired by the fact that with noise injected into the original image, the defects may be changed into normal cases in the denoising process (i.e., reconstruction). First, based on the assumption that the anomalous data lie in the low probability density region of the normal data distribution, we explain a common phenomenon that occurs when reconstruction-based approaches are applied to VDD: normal pixels also change during the reconstruction process. Second, due to the differences in normal pixels between the reconstructed and original images, a time-dependent gradient value (i.e., score) of normal data distribution is utilized as a metric, rather than reconstruction loss, to gauge the defects. Third, a novel $T$ scales approach is developed to dramatically reduce the required number of iterations, accelerating the inference process. These practices allow our model to generalize VDD in an unsupervised manner while maintaining reasonably good performance. We evaluate our method on several datasets to demonstrate its effectiveness.

Yihan Lin, Haoyang Li, Yang Li et al.

Latent actions serve as an intermediate representation that enables consistent modeling of vision-language-action (VLA) models across heterogeneous datasets. However, approaches to supervising VLAs with latent actions are fragmented and lack a systematic comparison. This work structures the study of latent action supervision from two perspectives: (i) regularizing the trajectory via image-based latent actions, and (ii) unifying the target space with action-based latent actions. Under a unified VLA baseline, we instantiate and compare four representative integration strategies. Our results reveal a formulation-task correspondence: image-based latent actions benefit long-horizon reasoning and scene-level generalization, whereas action-based latent actions excel at complex motor coordination. Furthermore, we find that directly supervising the VLM with discrete latent action tokens yields the most effective performance. Finally, our experiments offer initial insights into the benefits of latent action supervision in mixed-data, suggesting a promising direction for VLA training. Code is available at https://github.com/RUCKBReasoning/From_Pixels_to_Tokens.

Qiuyue Wang, Mingsheng Li, Jian Guan et al.

Embodied intelligence is often studied through specialized models for individual tasks such as manipulation or navigation, resulting in fragmented capabilities and limited generalization across tasks, environments, and robot embodiments. In this work, we study whether heterogeneous embodied decision-making problems can be unified within a single vision-language-action model. We present Qwen-VLA, a unified embodied foundation model that extends Qwen's vision-language modeling stack from perception, understanding, and reasoning to continuous action and trajectory generation through a DiT-based action decoder. Qwen-VLA is trained with a large-scale joint pretraining recipe over diverse data sources, including robotics manipulation trajectories, human egocentric demonstrations, synthetic simulation data, vision-and-language navigation data, trajectory-centric supervision, and auxiliary vision-language data. To support multiple robot platforms, we introduce embodiment-aware prompt conditioning, where robot-specific textual descriptions specify the current embodiment and control convention. We further cast manipulation, navigation, and trajectory prediction into a unified action-and-trajectory prediction framework, enabling transferable visual grounding, spatial reasoning, and continuous action generation across robot morphologies, task families, and environments. Experiments on manipulation, navigation, and trajectory-centric benchmarks show consistent multi-task performance and out-of-distribution generalization under variations in scene layout, background, lighting, object configuration, and robot embodiment. Qwen-VLA-Instruct achieves 97.9% on LIBERO, 73.7% on Simpler-WidowX, 86.1%/87.2% on RoboTwin-Easy/Hard, 69.0% OSR on R2R, 59.6% SR on RxR, 76.9% average OOD success in real-world ALOHA experiments, and 26.6% zero-shot success on DOMINO dynamic manipulation.

Dingjie Song, Tianlong Xu, Yi-Fan Zhang et al.

Assessing student handwritten scratchwork is crucial for personalized educational feedback but presents unique challenges due to diverse handwriting, complex layouts, and varied problem-solving approaches. Existing educational NLP primarily focuses on textual responses and neglects the complexity and multimodality inherent in authentic handwritten scratchwork. Current multimodal large language models (MLLMs) excel at visual reasoning but typically adopt an "examinee perspective", prioritizing generating correct answers rather than diagnosing student errors. To bridge these gaps, we introduce ScratchMath, a novel benchmark specifically designed for explaining and classifying errors in authentic handwritten mathematics scratchwork. Our dataset comprises 1,720 mathematics samples from Chinese primary and middle school students, supporting two key tasks: Error Cause Explanation (ECE) and Error Cause Classification (ECC), with seven defined error types. The dataset is meticulously annotated through rigorous human-machine collaborative approaches involving multiple stages of expert labeling, review, and verification. We systematically evaluate 16 leading MLLMs on ScratchMath, revealing significant performance gaps relative to human experts, especially in visual recognition and logical reasoning. Proprietary models notably outperform open-source models, with large reasoning models showing strong potential for error explanation. All evaluation data and frameworks are publicly available to facilitate further research.

Zeyang Zhang, Xin Wang, Ziwei Zhang et al.

Dynamic graph neural networks (DyGNNs) have demonstrated powerful predictive abilities by exploiting graph structural and temporal dynamics. However, the existing DyGNNs fail to handle distribution shifts, which naturally exist in dynamic graphs, mainly because the patterns exploited by DyGNNs may be variant with respect to labels under distribution shifts. In this paper, we propose Disentangled Intervention-based Dynamic graph Attention networks with Invariance Promotion (I-DIDA) to handle spatio-temporal distribution shifts in dynamic graphs by discovering and utilizing invariant patterns, i.e., structures and features whose predictive abilities are stable across distribution shifts. Specifically, we first propose a disentangled spatio-temporal attention network to capture the variant and invariant patterns. By utilizing the disentangled patterns, we design a spatio-temporal intervention mechanism to create multiple interventional distributions and an environment inference module to infer the latent spatio-temporal environments, and minimize the variance of predictions among these intervened distributions and environments, so that our model can make predictions based on invariant patterns with stable predictive abilities under distribution shifts. Extensive experiments demonstrate the superiority of our method over state-of-the-art baselines under distribution shifts. Our work is the first study of spatio-temporal distribution shifts in dynamic graphs, to the best of our knowledge.

Haoyang Li

Image recognition/classification is a widely studied problem, but its reverse problem, image generation, has drawn much less attention until recently. But the vast majority of current methods for image generation require training/retraining a classifier and/or a generator with certain constraints, which can be hard to achieve. In this paper, we propose a simple approach to directly use a normally trained classifier to generate images. We evaluate our method on MNIST and show that it produces recognizable results for human eyes with limited quality with experiments.

Yiyan Li, Haoyang Li, Zhao Pu et al.

Knob tuning plays a crucial role in optimizing databases by adjusting knobs to enhance database performance. However, traditional tuning methods often follow a Try-Collect-Adjust approach, proving inefficient and database-specific. Moreover, these methods are often opaque, making it challenging for DBAs to grasp the underlying decision-making process. The emergence of large language models (LLMs) like GPT-4 and Claude-3 has excelled in complex natural language tasks, yet their potential in database knob tuning remains largely unexplored. This study harnesses LLMs as experienced DBAs for knob-tuning tasks with carefully designed prompts. We identify three key subtasks in the tuning system: knob pruning, model initialization, and knob recommendation, proposing LLM-driven solutions to replace conventional methods for each subtask. We conduct extensive experiments to compare LLM-driven approaches against traditional methods across the subtasks to evaluate LLMs' efficacy in the knob tuning domain. Furthermore, we explore the adaptability of LLM-based solutions in diverse evaluation settings, encompassing new benchmarks, database engines, and hardware environments. Our findings reveal that LLMs not only match or surpass traditional methods but also exhibit notable interpretability by generating responses in a coherent ``chain-of-thought'' manner. We further observe that LLMs exhibit remarkable generalizability through simple adjustments in prompts, eliminating the necessity for additional training or extensive code modifications. Drawing insights from our experimental findings, we identify several opportunities for future research aimed at advancing the utilization of LLMs in the realm of database management.

Haoyang Li, Jing Zhang, Hanbing Liu et al.

Language models have shown promising performance on the task of translating natural language questions into SQL queries (Text-to-SQL). However, most of the state-of-the-art (SOTA) approaches rely on powerful yet closed-source large language models (LLMs), such as ChatGPT and GPT-4, which may have the limitations of unclear model architectures, data privacy risks, and expensive inference overheads. To address the limitations, we introduce CodeS, a series of pre-trained language models with parameters ranging from 1B to 15B, specifically designed for the text-to-SQL task. CodeS is a fully open-source language model, which achieves superior accuracy with much smaller parameter sizes. This paper studies the research challenges in building CodeS. To enhance the SQL generation abilities of CodeS, we adopt an incremental pre-training approach using a specifically curated SQL-centric corpus. Based on this, we address the challenges of schema linking and rapid domain adaptation through strategic prompt construction and a bi-directional data augmentation technique. We conduct comprehensive evaluations on multiple datasets, including the widely used Spider benchmark, the newly released BIRD benchmark, robustness-diagnostic benchmarks such as Spider-DK, Spider-Syn, Spider-Realistic, and Dr.Spider, as well as two real-world datasets created for financial and academic applications. The experimental results show that our CodeS achieves new SOTA accuracy and robustness on nearly all challenging text-to-SQL benchmarks.

Juexiao Zhou, Bin Zhang, Xiuying Chen et al.

With the fast-growing and evolving omics data, the demand for streamlined and adaptable tools to handle the analysis continues to grow. In response to this need, we introduce Auto Bioinformatics Analysis (AutoBA), an autonomous AI agent based on a large language model designed explicitly for conventional omics data analysis. AutoBA simplifies the analytical process by requiring minimal user input while delivering detailed step-by-step plans for various bioinformatics tasks. Through rigorous validation by expert bioinformaticians, AutoBA's robustness and adaptability are affirmed across a diverse range of omics analysis cases, including whole genome sequencing (WGS), RNA sequencing (RNA-seq), single-cell RNA-seq, ChIP-seq, and spatial transcriptomics. AutoBA's unique capacity to self-design analysis processes based on input data variations further underscores its versatility. Compared with online bioinformatic services, AutoBA deploys the analysis locally, preserving data privacy. Moreover, different from the predefined pipeline, AutoBA has adaptability in sync with emerging bioinformatics tools. Overall, AutoBA represents a convenient tool, offering robustness and adaptability for complex omics data analysis.

Xiuying Chen, Tairan Wang, Taicheng Guo et al.

Question Answering (QA) effectively evaluates language models' reasoning and knowledge depth. While QA datasets are plentiful in areas like general domain and biomedicine, academic chemistry is less explored. Chemical QA plays a crucial role in both education and research by effectively translating complex chemical information into readily understandable format. Addressing this gap, we introduce ScholarChemQA, a large-scale QA dataset constructed from chemical papers. This dataset reflects typical real-world challenges, including an imbalanced data distribution and a substantial amount of unlabeled data that can be potentially useful. Correspondingly, we introduce a QAMatch model, specifically designed to effectively answer chemical questions by fully leveraging our collected data. We first address the issue of imbalanced label distribution by re-weighting the instance-wise loss based on the inverse frequency of each class, ensuring minority classes are not dominated by majority ones during optimization. Next, we utilize the unlabeled data to enrich the learning process, generating a variety of augmentations based on a SoftMix operation and ensuring their predictions align with the same target, i.e., pseudo-labels. To ensure the quality of the pseudo-labels, we propose a calibration procedure aimed at closely aligning the pseudo-label estimates of individual samples with a desired ground truth distribution. Experiments show that our QAMatch significantly outperforms the recent similar-scale baselines and Large Language Models (LLMs) not only on our ScholarChemQA dataset but also on four benchmark datasets. We hope our benchmark and model can facilitate and promote more research on chemical QA.

Haoyang Li, Yiming Li, Anxin Tian et al.

Large Language Models (LLMs) have revolutionized a wide range of domains such as natural language processing, computer vision, and multi-modal tasks due to their ability to comprehend context and perform logical reasoning. However, the computational and memory demands of LLMs, particularly during inference, pose significant challenges when scaling them to real-world, long-context, and real-time applications. Key-Value (KV) cache management has emerged as a critical optimization technique for accelerating LLM inference by reducing redundant computations and improving memory utilization. This survey provides a comprehensive overview of KV cache management strategies for LLM acceleration, categorizing them into token-level, model-level, and system-level optimizations. Token-level strategies include KV cache selection, budget allocation, merging, quantization, and low-rank decomposition, while model-level optimizations focus on architectural innovations and attention mechanisms to enhance KV reuse. System-level approaches address memory management, scheduling, and hardware-aware designs to improve efficiency across diverse computing environments. Additionally, the survey provides an overview of both text and multimodal datasets and benchmarks used to evaluate these strategies. By presenting detailed taxonomies and comparative analyses, this work aims to offer useful insights for researchers and practitioners to support the development of efficient and scalable KV cache management techniques, contributing to the practical deployment of LLMs in real-world applications. The curated paper list for KV cache management is in: \href{https://github.com/TreeAI-Lab/Awesome-KV-Cache-Management}{https://github.com/TreeAI-Lab/Awesome-KV-Cache-Management}.

Bo Han, Heqing Zou, Haoyang Li et al.

Text-based talking-head video editing aims to efficiently insert, delete, and substitute segments of talking videos through a user-friendly text editing approach. It is challenging because of \textbf{1)} generalizable talking-face representation, \textbf{2)} seamless audio-visual transitions, and \textbf{3)} identity-preserved talking faces. Previous works either require minutes of talking-face video training data and expensive test-time optimization for customized talking video editing or directly generate a video sequence without considering in-context information, leading to a poor generalizable representation, or incoherent transitions, or even inconsistent identity. In this paper, we propose an efficient cascaded conditional diffusion-based framework, which consists of two stages: audio to dense-landmark motion and motion to video. \textit{\textbf{In the first stage}}, we first propose a dynamic weighted in-context diffusion module to synthesize dense-landmark motions given an edited audio. \textit{\textbf{In the second stage}}, we introduce a warping-guided conditional diffusion module. The module first interpolates between the start and end frames of the editing interval to generate smooth intermediate frames. Then, with the help of the audio-to-dense motion images, these intermediate frames are warped to obtain coarse intermediate frames. Conditioned on the warped intermedia frames, a diffusion model is adopted to generate detailed and high-resolution target frames, which guarantees coherent and identity-preserved transitions. The cascaded conditional diffusion model decomposes the complex talking editing task into two flexible generation tasks, which provides a generalizable talking-face representation, seamless audio-visual transitions, and identity-preserved faces on a small dataset. Experiments show the effectiveness and superiority of the proposed method.

Haoyang Li, Shang Wu, Xiaokang Zhang et al.

Text-to-SQL, the task of translating natural language questions into SQL queries, plays a crucial role in enabling non-experts to interact with databases. While recent advancements in large language models (LLMs) have significantly enhanced text-to-SQL performance, existing approaches face notable limitations in real-world text-to-SQL applications. Prompting-based methods often depend on closed-source LLMs, which are expensive, raise privacy concerns, and lack customization. Fine-tuning-based methods, on the other hand, suffer from poor generalizability due to the limited coverage of publicly available training data. To overcome these challenges, we propose a novel and scalable text-to-SQL data synthesis framework for automatically synthesizing large-scale, high-quality, and diverse datasets without extensive human intervention. Using this framework, we introduce SynSQL-2.5M, the first million-scale text-to-SQL dataset, containing 2.5 million samples spanning over 16,000 synthetic databases. Each sample includes a database, SQL query, natural language question, and chain-of-thought (CoT) solution. Leveraging SynSQL-2.5M, we develop OmniSQL, a powerful open-source text-to-SQL model available in three sizes: 7B, 14B, and 32B. Extensive evaluations across nine datasets demonstrate that OmniSQL achieves state-of-the-art performance, matching or surpassing leading closed-source and open-source LLMs, including GPT-4o and DeepSeek-V3, despite its smaller size. We release all code, datasets, and models to support further research.

Kimi Team, Yifan Bai, Yiping Bao et al. · tsinghua

We introduce Kimi K2, a Mixture-of-Experts (MoE) large language model with 32 billion activated parameters and 1 trillion total parameters. We propose the MuonClip optimizer, which improves upon Muon with a novel QK-clip technique to address training instability while enjoying the advanced token efficiency of Muon. Based on MuonClip, K2 was pre-trained on 15.5 trillion tokens with zero loss spike. During post-training, K2 undergoes a multi-stage post-training process, highlighted by a large-scale agentic data synthesis pipeline and a joint reinforcement learning (RL) stage, where the model improves its capabilities through interactions with real and synthetic environments. Kimi K2 achieves state-of-the-art performance among open-source non-thinking models, with strengths in agentic capabilities. Notably, K2 obtains 66.1 on Tau2-Bench, 76.5 on ACEBench (En), 65.8 on SWE-Bench Verified, and 47.3 on SWE-Bench Multilingual -- surpassing most open and closed-sourced baselines in non-thinking settings. It also exhibits strong capabilities in coding, mathematics, and reasoning tasks, with a score of 53.7 on LiveCodeBench v6, 49.5 on AIME 2025, 75.1 on GPQA-Diamond, and 27.1 on OJBench, all without extended thinking. These results position Kimi K2 as one of the most capable open-source large language models to date, particularly in software engineering and agentic tasks. We release our base and post-trained model checkpoints to facilitate future research and applications of agentic intelligence.

Zongyuan Ge, Chenwaner Zhang, Haoyang Li et al.

Forecasting aurora borealis visibility matters for space weather research and aurora tourism. Visibility at a site and night depends on two distinct factors: (1) whether aurora is physically occurring, driven by solar wind-magnetosphere coupling, and (2) whether observing conditions allow naked-eye detection, mainly cloud cover and lunar illumination. We present Aurora Hunter, a two-stage cascade that decouples these factors. Stage 1 predicts P(occurring) with XGBoost using 51 physics-driven features trained on joint Tromso+Kiruna data (about 16,600 hourly samples, 2015-2023) with labels from the Tromso AI all-sky image classifier. Stage 2 predicts P(clear observation given occurring) with logistic regression using 21 cloud-cover and lunar-illumination features trained only on aurora-occurring hours. The cascade P(visible)=P(occurring)*P(clear|occurring) reaches ROC-AUC 0.937 (Tromso test, 2019-2020) and 0.905 (independent Kiruna, 2024), improving a single-stage baseline by +0.087. Held-out Skibotn data (2022-2025) confirm cross-site generalization. SHAP identifies the Kp x nightside interaction, MLT position, and auroral oval distance as dominant predictors (39% combined). Prototype: https://aurora-hunter.onrender.com.

Haoyang Li, Changsong Liu, Wei Rao et al.

Automatic speech recognition (ASR) degrades severely in noisy environments. Although speech enhancement (SE) front-ends effectively suppress background noise, they often introduce artifacts that harm recognition. Observation addition (OA) addressed this issue by fusing noisy and SE enhanced speech, improving recognition without modifying the parameters of the SE or ASR models. This paper proposes an intelligibility-guided OA method, where fusion weights are derived from intelligibility estimates obtained directly from the backend ASR. Unlike prior OA methods based on trained neural predictors, the proposed method is training-free, reducing complexity and enhances generalization. Extensive experiments across diverse SE-ASR combinations and datasets demonstrate strong robustness and improvements over existing OA baselines. Additional analyses of intelligibility-guided switching-based alternatives and frame versus utterance-level OA further validate the proposed design.

Guodong Du, Runhua Jiang, Senqiao Yang et al.

Darwinian evolution of the biological brain is documented through multiple lines of evidence, although the modes of evolutionary changes remain unclear. Drawing inspiration from the evolved neural systems (e.g., visual cortex), deep learning models have demonstrated superior performance in visual tasks, among others. While the success of training deep neural networks has been relying on back-propagation (BP) and its variants to learn representations from data, BP does not incorporate the evolutionary processes that govern biological neural systems. This work proposes a neural network optimization framework based on evolutionary theory. Specifically, BP-trained deep neural networks for visual recognition tasks obtained from the ending epochs are considered the primordial ancestors (initial population). Subsequently, the population evolved with differential evolution. Extensive experiments are carried out to examine the relationships between Darwinian evolution and neural network optimization, including the correspondence between datasets, environment, models, and living species. The empirical results show that the proposed framework has positive impacts on the network, with reduced over-fitting and an order of magnitude lower time complexity compared to BP. Moreover, the experiments show that the proposed framework performs well on deep neural networks and big datasets.

Yaxin Xiao, Qingqing Ye, Zi Liang et al.

Machine learning models constitute valuable intellectual property, yet remain vulnerable to model extraction attacks (MEA), where adversaries replicate their functionality through black-box queries. Model watermarking counters MEAs by embedding forensic markers for ownership verification. Current black-box watermarks prioritize MEA survival through representation entanglement, yet inadequately explore resilience against sequential MEAs and removal attacks. Our study reveals that this risk is underestimated because existing removal methods are weakened by entanglement. To address this gap, we propose Watermark Removal attacK (WRK), which circumvents entanglement constraints by exploiting decision boundaries shaped by prevailing sample-level watermark artifacts. WRK effectively reduces watermark success rates by at least 88.79% across existing watermarking benchmarks. For robust protection, we propose Class-Feature Watermarks (CFW), which improve resilience by leveraging class-level artifacts. CFW constructs a synthetic class using out-of-domain samples, eliminating vulnerable decision boundaries between original domain samples and their artifact-modified counterparts (watermark samples). CFW concurrently optimizes both MEA transferability and post-MEA stability. Experiments across multiple domains show that CFW consistently outperforms prior methods in resilience, maintaining a watermark success rate of at least 70.15% in extracted models even under the combined MEA and WRK distortion, while preserving the utility of protected models.

Jianxin Yan, Zeheng Qian, Wangze Ni et al.

Cache fusion accelerates generation process of LLMs equipped with RAG through KV caching and selective token recomputation, thereby reducing computational costs and improving efficiency. However, existing methods primarily rely on local perspectives for token selection and lack global awareness from the user query. Utilizing this global awareness is challenging due to the high cost of obtaining context-aware query representations and the strict pipeline constraints required for efficient attention analysis. Thus, this demonstration introduces QCFuse, an innovative KV cache fusion system centered on the user query. QCFuse leverages semantic summary anchors to enhance query representations and selectively recomputes query-related tokens to improve accuracy, updating tokens based on the attention distribution of the most critical Transformer layer to preserve the high efficiency of the pipeline structure. Evaluations on real-world datasets demonstrate that QCFuse significantly improves the response efficiency of LLMs by 40\% while maintaining equivalent accuracy compared to current methods. Additionally, in certain scenarios, QCFuse achieves an attention denoising effect that yields higher response accuracy, demonstrating substantial potential in the optimization of LLM inference.

Tianhao Tang, Haoyang Li, Lei Chen

Large Language Models (LLMs) have demonstrated remarkable capabilities in natural language understanding and generation. However, their immense number of parameters and complex transformer-based architectures result in significant resource demands and computational complexity during training, making it challenging to optimize them efficiently on large datasets. To reduce training costs while preserving performance, researchers have investigated coreset selection techniques, which aim to identify small, representative subsets of the entire training dataset to accelerate LLM training. However, existing coreset selection methods fail to adapt to the dynamic nature of LLM training and often struggle with scalability for models of this size. To address these limitations, we propose a graph-guided adaptive and dynamic coreset selection framework for LLMs, namely GRACE. GRACE dynamically constructs and updates coresets by combining representation diversity with gradient-based importance metrics, ensuring both informativeness and efficiency. To mitigate the computational cost of frequent updates, GRACE leverages a $k$-NN graph-based propagation mechanism and selectively updates scores and embeddings, adapting to evolving training dynamics. Extensive experiments on three benchmarks demonstrate that GRACE significantly improves training efficiency and downstream performance across diverse LLMs and tasks.

Haoyang Li, Xuyi Zhuang, Azmat Adnan et al.

Language Model (LM)-based generative modeling has emerged as a promising direction for TSE, offering potential for improved generalization and high-fidelity speech. We present GenTSE, a two-stage decoder-only generative LM approach for TSE: Stage-1 predicts coarse semantic tokens, and Stage-2 generates fine acoustic tokens. Separating semantics and acoustics stabilizes decoding and yields more faithful, content-aligned target speech. Both stages use continuous SSL or codec embeddings, offering richer context than discretized-prompt methods. To reduce exposure bias, we employ a Frozen-LM Conditioning training strategy that conditions the LMs on predicted tokens from earlier checkpoints to reduce the gap between teacher-forcing training and autoregressive inference. We further employ DPO to better align outputs with human perceptual preferences. Experiments on Libri2Mix show that GenTSE surpasses previous LM-based systems in speech quality, intelligibility, and speaker consistency.

Xiyin Zeng, Yuyu Sun, Haoyang Li et al.

Vision-Language-Action systems follow instructions to execute multi-step tasks in multimodal environments. Recent VLA approaches typically rely on post-hoc correction mechanisms or operate under fixed task decompositions and alignment schemes. However, once an intermediate step is mis-specified, local errors propagate through subsequent steps and eventually accumulate into cascading failures. To mitigate this compounding effect, we propose Predictive Alignment and Planning Architecture, a framework that uses prediction and contrast to adjust deviations across three levels: actions, subgoals, and trajectories. Semantic alignment is enforced at all levels using a Sinkhorn-based module and a Score-field module. The predictive correction and alignment jointly update the action generator during training, enabling it to adjust fine-grained steps to remain aligned with the overall intent. We further introduce two new metrics to quantify error propagation and recovery processes in tasks, capturing how mistakes spread and fade over long-horizon execution. Experiments show that ReCAPA achieves competitive results on embodied agent benchmarks such as VisualAgentBench, MineDojo, and AI2-THOR, outperforming strong proprietary and open-source Large Language Model baselines.

Haoyang Li, Liang Wang, Chao Wang et al.

The Base-New Trade-off (BNT) problem universally exists during the optimization of CLIP-based prompt tuning, where continuous fine-tuning on base (target) classes leads to a simultaneous decrease of generalization ability on new (unseen) classes. Existing approaches attempt to regulate the prompt tuning process to balance BNT by appending constraints. However, imposed on the same target prompt, these constraints fail to fully avert the mutual exclusivity between the optimization directions for base and new. As a novel solution to this challenge, we propose the plug-and-play Dual-Prompt Collaboration (DPC) framework, the first that decoupling the optimization processes of base and new tasks at the prompt level. Specifically, we clone a learnable parallel prompt based on the backbone prompt, and introduce a variable Weighting-Decoupling framework to independently control the optimization directions of dual prompts specific to base or new tasks, thus avoiding the conflict in generalization. Meanwhile, we propose a Dynamic Hard Negative Optimizer, utilizing dual prompts to construct a more challenging optimization task on base classes for enhancement. For interpretability, we prove the feature channel invariance of the prompt vector during the optimization process, providing theoretical support for the Weighting-Decoupling of DPC. Extensive experiments on multiple backbones demonstrate that DPC can significantly improve base performance without introducing any external knowledge beyond the base classes, while maintaining generalization to new classes. Code is available at: https://github.com/JREion/DPC.

Haoyang Li, Xuejia Chen, Zhanchao XU et al.

Large Language Models (LLMs) have demonstrated impressive capabilities in natural language processing tasks, such as text generation and semantic understanding. However, their performance on numerical reasoning tasks, such as basic arithmetic, numerical retrieval, and magnitude comparison, remains surprisingly poor. This gap arises from their reliance on surface-level statistical patterns rather than understanding numbers as continuous magnitudes. Existing benchmarks primarily focus on either linguistic competence or structured mathematical problem-solving, neglecting fundamental numerical reasoning required in real-world scenarios. To bridge this gap, we propose NumericBench, a comprehensive benchmark to evaluate six fundamental numerical capabilities: number recognition, arithmetic operations, contextual retrieval, comparison, summary, and logical reasoning. NumericBench includes datasets ranging from synthetic number lists to the crawled real-world data, addressing challenges like long contexts, noise, and multi-step reasoning. Extensive experiments on state-of-the-art LLMs, including GPT-4 and DeepSeek, reveal persistent weaknesses in numerical reasoning, highlighting the urgent need to improve numerically-aware language modeling. The benchmark is released in: https://github.com/TreeAI-Lab/NumericBench.

Yubo Wang, Haoyang Li, Fei Teng et al.

Graph-based retrieval-augmented generation (Graph-based RAG) has demonstrated significant potential in enhancing Large Language Models (LLMs) with structured knowledge. However, existing methods face three critical challenges: Inaccurate Graph Construction, caused by LLM hallucination; Poor Reasoning Ability, caused by failing to generate explicit reasons telling LLM why certain chunks were selected; and Inadequate Answering, which only partially answers the query due to the inadequate LLM reasoning, making their performance lag behind NaiveRAG on certain tasks. To address these issues, we propose AGRAG, an advanced graph-based retrieval-augmented generation framework. When constructing the graph, AGRAG substitutes the widely used LLM entity extraction method with a statistics-based method, avoiding hallucination and error propagation. When retrieval, AGRAG formulates the graph reasoning procedure as the Minimum Cost Maximum Influence (MCMI) subgraph generation problem, where we try to include more nodes with high influence score, but with less involving edge cost, to make the generated reasoning paths more comprehensive. We prove this problem to be NP-hard, and propose a greedy algorithm to solve it. The MCMI subgraph generated can serve as explicit reasoning paths to tell LLM why certain chunks were retrieved, thereby making the LLM better focus on the query-related part contents of the chunks, reducing the impact of noise, and improving AGRAG's reasoning ability. Furthermore, compared with the simple tree-structured reasoning paths, our MCMI subgraph can allow more complex graph structures, such as cycles, and improve the comprehensiveness of the generated reasoning paths.

Kai Li, Can Shen, Yile Liu et al.

Audio Large Language Models (ALLMs) have gained widespread adoption, yet their trustworthiness remains underexplored. Existing evaluation frameworks, designed primarily for text, fail to address unique vulnerabilities introduced by audio's acoustic properties. We identify significant trustworthiness risks in ALLMs arising from non-semantic acoustic cues, including timbre, accent, and background noise, which can manipulate model behavior. We propose AudioTrust, a comprehensive framework for systematic evaluation of ALLM trustworthiness across audio-specific risks. AudioTrust encompasses six key dimensions: fairness, hallucination, safety, privacy, robustness, and authentication. The framework implements 26 distinct sub-tasks using a curated dataset of over 4,420 audio samples from real-world scenarios, including daily conversations, emergency calls, and voice assistant interactions. We conduct comprehensive evaluations across 18 experimental configurations using human-validated automated pipelines. Our evaluation of 14 state-of-the-art open-source and closed-source ALLMs reveals significant limitations when confronted with diverse high-risk audio scenarios, providing insights for secure deployment of audio models. Code and data are available at https://github.com/JusperLee/AudioTrust.

Jianhong Li, Zeheng Qian, Wangze Ni et al.

LLM development has aroused great interest in Sequential Recommendation (SR) applications. However, comprehensive evaluation of SR models remains lacking due to the limitations of the existing benchmarks: 1) an overemphasis on accuracy, ignoring other real-world demands (e.g., fairness); 2) existing datasets fail to unleash LLMs' potential, leading to unfair comparison between Neural-Network-based SR (NN-SR) models and LLM-based SR (LLM-SR) models; and 3) no reliable mechanism for extracting task-specific answers from unstructured LLM outputs. To address these limitations, we propose SRBench, a comprehensive SR benchmark with three core designs: 1) a multi-dimensional framework covering accuracy, fairness, stability and efficiency, aligned with practical demands; 2) a unified input paradigm via prompt engineering to boost LLM-SR performance and enable fair comparisons between models; 3) a novel prompt-extractor-coupled extraction mechanism, which captures answers from LLM outputs through prompt-enforced output formatting and a numeric-oriented extractor. We have used SRBench to evaluate 13 mainstream models and discovered some meaningful insights (e.g., LLM-SR models overfocus on item popularity but lack deep understanding of item quality). Concisely, SRBench enables fair and comprehensive assessments for SR models, underpinning future research and practical application.

Banglei Guan, Yifei Bian, Zibin Liu et al.

Background: Large engineering structures, such as space launch towers and suspension bridges, are subjected to extreme forces that cause high-speed 3D deformation and compromise safety. These structures typically operate under extreme illumination conditions. Traditional cameras often struggle to handle strong light intensity, leading to overexposure due to their limited dynamic range. Objective: Event cameras have emerged as a compelling alternative to traditional cameras in high dynamic range and low-latency applications. This paper presents an integrated method, from calibration to measurement, using a multi-event camera array for high-speed 3D deformation monitoring of structures in extreme illumination conditions. Methods: Firstly, the proposed method combines the characteristics of the asynchronous event stream and temporal correlation analysis to extract the corresponding marker center point. Subsequently, the method achieves rapid calibration by solving the Kruppa equations in conjunction with a parameter optimization framework. Finally, by employing a unified coordinate transformation and linear intersection, the method enables the measurement of 3D deformation of the target structure. Results: Experiments confirmed that the relative measurement error is below 0.08%. Field experiments under extreme illumination conditions, including self-calibration of a multi-event camera array and 3D deformation measurement, verified the performance of the proposed method. Conclusions: This paper addressed the critical limitation of traditional cameras in measuring high-speed 3D deformations under extreme illumination conditions. The experimental results demonstrate that, compared to other methods, the proposed method can accurately measure 3D deformations of structures under harsh lighting conditions, and the relative error of the measured deformation is less than 0.1%.

Jiacheng Sun, Jiaqi Lin, Wenlong Hu et al.

One of the bottlenecks for instance segmentation today lies in the conflicting requirements of high-resolution inputs and lightweight, real-time inference. To address this bottleneck, we present a Polygon Detection Transformer (Poly-DETR) to reformulate instance segmentation as sparse vertex regression via Polar Representation, thereby eliminating the reliance on dense pixel-wise mask prediction. Considering the box-to-polygon reference shift in Detection Transformers, we propose Polar Deformable Attention and Position-Aware Training Scheme to dynamically update supervision and focus attention on boundary cues. Compared with state-of-the-art polar-based methods, Poly-DETR achieves a 4.7 mAP improvement on MS COCO test-dev. Moreover, we construct a parallel mask-based counterpart to support a systematic comparison between polar and mask representations. Experimental results show that Poly-DETR is more lightweight in high-resolution scenarios, reducing memory consumption by almost half on Cityscapes dataset. Notably, on PanNuke (cell segmentation) and SpaceNet (building footprints) datasets, Poly-DETR surpasses its mask-based counterpart on all metrics, which validates its advantage on regular-shaped instances in domain-specific settings.

Cheng Deng, Luoyang Sun, Jiwen Jiang et al.

While scaling laws have been continuously validated in large language models (LLMs) with increasing model parameters, the inherent tension between the inference demands of LLMs and the limited resources of edge devices poses a critical challenge to the development of edge intelligence. Recently, numerous small language models have emerged, aiming to distill the capabilities of LLMs into smaller footprints. However, these models often retain the fundamental architectural principles of their larger counterparts, still imposing considerable strain on the storage and bandwidth capacities of edge devices. In this paper, we introduce the PLM, a Peripheral Language Model, developed through a co-design process that jointly optimizes model architecture and edge system constraints. The PLM utilizes a Multi-head Latent Attention mechanism and employs the squared ReLU activation function to encourage sparsity, thereby reducing peak memory footprint during inference. During training, we collect and reorganize open-source datasets, implement a multi-phase training strategy, and empirically investigate the Warmup-Stable-Decay-Constant (WSDC) learning rate scheduler. Additionally, we incorporate Reinforcement Learning from Human Feedback (RLHF) by adopting the ARIES preference learning approach. Following a two-phase SFT process, this method yields performance gains of 2% in general tasks, 9% in the GSM8K task, and 11% in coding tasks. In addition to its novel architecture, evaluation results demonstrate that PLM outperforms existing small language models trained on publicly available data while maintaining the lowest number of activated parameters. Furthermore, deployment across various edge devices, including consumer-grade GPUs, mobile phones, and Raspberry Pis, validates PLM's suitability for peripheral applications. The PLM series models are publicly available at https://github.com/plm-team/PLM.

Haoyang Li, Yang You, Hao Su et al.

Reliable object manipulation requires understanding physical properties that vary across objects and environments. Vision-language model (VLM) planners can reason about friction and stability in general terms; however, they often cannot predict how a specific ball will roll on a particular surface or which stone will provide a stable foundation without direct experience. We present PhysMem, a memory framework that enables VLM robot planners to learn physical principles from interaction at test time, without updating model parameters. The system records experiences, generates candidate hypotheses, and verifies them through targeted interaction before promoting validated knowledge to guide future decisions. A central design choice is verification before application: the system tests hypotheses against new observations rather than applying retrieved experience directly, reducing rigid reliance on prior experience when physical conditions change. We evaluate PhysMem on three real-world manipulation tasks and simulation benchmarks across four VLM backbones. On a controlled brick insertion task, principled abstraction achieves 76% success compared to 23% for direct experience retrieval, and real-world experiments show consistent improvement over 30-minute deployment sessions.

Chen Zhao, Zhuoran Wang, Haoyang Li et al.

Vision-Language-Action (VLA) models have recently demonstrated strong performance across embodied tasks. Modern VLAs commonly employ diffusion action experts to efficiently generate high-precision continuous action chunks, while auto-regressive generation can be slower and less accurate at low-level control. Yet auto-regressive paradigms still provide complementary priors that can improve robustness and generalization in out-of-distribution environments. To leverage both paradigms, we propose Action-Draft-and-Verify (ADV): diffusion action expert drafts multiple candidate action chunks, and the VLM selects one by scoring all candidates in a single forward pass with a perplexity-style metric. Under matched backbones, training data, and action-chunk length, ADV improves success rate by +4.3 points in simulation and +19.7 points in real-world over diffusion-based baseline, with a single-pass VLM reranking overhead.

Yiwen Hu, Jun Wei, Yuncheng Jiang et al.

Limited by the expensive labeling, polyp segmentation models are plagued by data shortages. To tackle this, we propose the mixed supervised polyp segmentation paradigm (MixPolyp). Unlike traditional models relying on a single type of annotation, MixPolyp combines diverse annotation types (mask, box, and scribble) within a single model, thereby expanding the range of available data and reducing labeling costs. To achieve this, MixPolyp introduces three novel supervision losses to handle various annotations: Subspace Projection loss (L_SP), Binary Minimum Entropy loss (L_BME), and Linear Regularization loss (L_LR). For box annotations, L_SP eliminates shape inconsistencies between the prediction and the supervision. For scribble annotations, L_BME provides supervision for unlabeled pixels through minimum entropy constraint, thereby alleviating supervision sparsity. Furthermore, L_LR provides dense supervision by enforcing consistency among the predictions, thus reducing the non-uniqueness. These losses are independent of the model structure, making them generally applicable. They are used only during training, adding no computational cost during inference. Extensive experiments on five datasets demonstrate MixPolyp's effectiveness.

Zhuohan Ge, Haoyang Li, Yubo Wang et al.

While Large Language Models (LLMs) have demonstrated potential in healthcare, they often struggle with the complex, non-linear reasoning required for accurate clinical diagnosis. Existing methods typically rely on static, linear mappings from symptoms to diagnoses, failing to capture the iterative, hypothesis-driven reasoning inherent to human clinicians. To bridge this gap, we introduce ClinicalAgents, a novel multi-agent framework designed to simulate the cognitive workflow of expert clinicians. Unlike rigid sequential chains, ClinicalAgents employs a dynamic orchestration mechanism modeled as a Monte Carlo Tree Search (MCTS) process. This allows an Orchestrator to iteratively generate hypotheses, actively verify evidence, and trigger backtracking when critical information is missing. Central to this framework is a Dual-Memory architecture: a mutable Working Memory that maintains the evolving patient state for context-aware reasoning, and a static Experience Memory that retrieves clinical guidelines and historical cases via an active feedback loop. Extensive experiments demonstrate that ClinicalAgents achieves state-of-the-art performance, significantly enhancing both diagnostic accuracy and explainability compared to strong single-agent and multi-agent baselines.

Haoyang Li, Liang Wang, Chao Wang et al.

For CLIP-based prompt tuning, introducing more data as additional knowledge for enhancing fine-tuning process is proved to be an effective approach. Existing data amplification strategies for prompt tuning typically rely on external knowledge (e.g., large language models or pre-structured knowledge bases), resulting in higher costs for data collection and processing, while generally ignoring further utilization of features in image modality. To address this, we propose Augmentation-driven Prompt Tuning (AugPT), a self-contained distillation-based prompt tuning approach using only internal augmentation on raw dataset to better exploit known features. Specifically, AugPT employs self-supervised augmentation on unlabeled images in the training set, and introduces a novel gating mechanism based on consensus test, reusing the pre-trained prompt tuning backbone model to spontaneously filter noisy samples, further enhancing the quality of augmented views. Extensive experiments validate that AugPT simultaneously enhances model performance and generalization capability without using appended external knowledge. The code of AugPT is available at: https://github.com/JREion/AugPT .