Yue Huang

Zhengliang Liu, Yue Huang, Xiaowei Yu et al.

The digitization of healthcare has facilitated the sharing and re-using of medical data but has also raised concerns about confidentiality and privacy. HIPAA (Health Insurance Portability and Accountability Act) mandates removing re-identifying information before the dissemination of medical records. Thus, effective and efficient solutions for de-identifying medical data, especially those in free-text forms, are highly needed. While various computer-assisted de-identification methods, including both rule-based and learning-based, have been developed and used in prior practice, such solutions still lack generalizability or need to be fine-tuned according to different scenarios, significantly imposing restrictions in wider use. The advancement of large language models (LLM), such as ChatGPT and GPT-4, have shown great potential in processing text data in the medical domain with zero-shot in-context learning, especially in the task of privacy protection, as these models can identify confidential information by their powerful named entity recognition (NER) capability. In this work, we developed a novel GPT4-enabled de-identification framework (``DeID-GPT") to automatically identify and remove the identifying information. Compared to existing commonly used medical text data de-identification methods, our developed DeID-GPT showed the highest accuracy and remarkable reliability in masking private information from the unstructured medical text while preserving the original structure and meaning of the text. This study is one of the earliest to utilize ChatGPT and GPT-4 for medical text data processing and de-identification, which provides insights for further research and solution development on the use of LLMs such as ChatGPT/GPT-4 in healthcare. Codes and benchmarking data information are available at https://github.com/yhydhx/ChatGPT-API.

Xiao Liu, Xuanyu Lei, Shengyuan Wang et al. · tsinghua

Alignment has become a critical step for instruction-tuned Large Language Models (LLMs) to become helpful assistants. However, the effective evaluation of alignment for emerging Chinese LLMs is still largely unexplored. To fill in this gap, we introduce AlignBench, a comprehensive multi-dimensional benchmark for evaluating LLMs' alignment in Chinese. We design a human-in-the-loop data curation pipeline, containing eight main categories, 683 real-scenario rooted queries and corresponding human verified references. To ensure the correctness of references, each knowledge-intensive query is accompanied with evidences collected from reliable web sources (including URLs and quotations) by our annotators. For automatic evaluation, our benchmark employs a rule-calibrated multi-dimensional LLM-as-Judge~\cite{zheng2023judging} approach with Chain-of-Thought to generate explanations and final ratings, ensuring high reliability and interpretability. All evaluation code, data, and LLM generations are available at \url{https://github.com/THUDM/AlignBench}. Since its release, AlignBench has been adopted by top (Chinese) LLMs for evaluating their alignment capabilities in Chinese, including ChatGLM, Qwen, DeepSeek, Yi, Baichuan, and Abab.

Zhenqi Fu, Wu Wang, Yue Huang et al.

A main challenge faced in the deep learning-based Underwater Image Enhancement (UIE) is that the ground truth high-quality image is unavailable. Most of the existing methods first generate approximate reference maps and then train an enhancement network with certainty. This kind of method fails to handle the ambiguity of the reference map. In this paper, we resolve UIE into distribution estimation and consensus process. We present a novel probabilistic network to learn the enhancement distribution of degraded underwater images. Specifically, we combine conditional variational autoencoder with adaptive instance normalization to construct the enhancement distribution. After that, we adopt a consensus process to predict a deterministic result based on a set of samples from the distribution. By learning the enhancement distribution, our method can cope with the bias introduced in the reference map labeling to some extent. Additionally, the consensus process is useful to capture a robust and stable result. We examined the proposed method on two widely used real-world underwater image enhancement datasets. Experimental results demonstrate that our approach enables sampling possible enhancement predictions. Meanwhile, the consensus estimate yields competitive performance compared with state-of-the-art UIE methods. Code available at https://github.com/zhenqifu/PUIE-Net.

Zhangchen Xu, Junda Chen, Yue Huang et al.

Scientific and engineering progress is fundamentally a long-horizon iterative process: proposing changes, running experiments, measuring outcomes, and continuously refining artifacts. Yet existing benchmarks for frontier models primarily evaluate either single-turn responses or short-horizon agent trajectories, failing to capture the challenges of sustained iterative improvement over extended time horizons. To address this gap, we introduce AutoLab, a new benchmark for ultra long-horizon closed-loop optimization. AutoLab consists of 36 realistic, expert-curated tasks spanning four diverse domains: system optimization, puzzle & challenge, model development, and CUDA kernel optimization. Each task begins with a correct but deliberately suboptimal baseline and challenges agents to improve it within a strict wall-clock budget. Evaluating 17 state-of-the-art models reveals the dominant predictor of success is not the quality of an agent's initial attempt, but its persistence in repeatedly benchmarking, editing, and incorporating empirical feedback. While claude-opus-4.6 exhibits strong long-horizon optimization capabilities, most frontier models, including several proprietary ones, either terminate prematurely or exhaust their budgets with minimal progress. These results underscore the importance of time awareness and persistent iteration in autonomous agents. We open-source the full benchmark, evaluation harness, and task artifacts, to accelerate research toward truly capable long-horizon agents.

Linxin Song, Jiefeng Chen, Yue Huang et al.

Coding agents increasingly act as codebase-scale collaborators that can assist with codebase conversion, but this progress has exposed a critical weakness: agents often over-trust their own local validation routines and declare success on artifacts that satisfy surface checks while violating the semantic contracts users actually care about. This problem is especially acute in codebase conversion, where prior evaluation is largely outcome-driven and therefore unstable: two implementations can match on a shallow outcome, such as a single forward loss, while diverging in gradients, optimizer behavior, or short-horizon training dynamics. We introduce T2J-Bench, a benchmark for codebase conversion that reformulates conversion as transfer under a fixed equivalence contract. A fixed verifier then compares source and converted codebases through three ordered stages: Spec (interface admissibility), Numeric (forward outputs, losses, gradients, and objective-specific tensors), and Behavioral (short training dynamics under fixed seeds). Across 355 blind conversion attempts, the best system reaches only 26.7--28.9% overall pass rate despite Spec pass rates up to 91.1%; a 4.7x token-budget spread yields only a 2.2x pass-rate spread; and all systems overestimate success by 66.6--97.8 points relative to the fixed evaluator. This suggests that failures stem more from contract-misaligned self-validation than from limited budget or backbone strength.

Chenxin Li, Mingbao Lin, Zhiyuan Ding et al.

Knowledge Distillation (KD) transfers the knowledge from a high-capacity teacher network to strengthen a smaller student. Existing methods focus on excavating the knowledge hints and transferring the whole knowledge to the student. However, the knowledge redundancy arises since the knowledge shows different values to the student at different learning stages. In this paper, we propose Knowledge Condensation Distillation (KCD). Specifically, the knowledge value on each sample is dynamically estimated, based on which an Expectation-Maximization (EM) framework is forged to iteratively condense a compact knowledge set from the teacher to guide the student learning. Our approach is easy to build on top of the off-the-shelf KD methods, with no extra training parameters and negligible computation overhead. Thus, it presents one new perspective for KD, in which the student that actively identifies teacher's knowledge in line with its aptitude can learn to learn more effectively and efficiently. Experiments on standard benchmarks manifest that the proposed KCD can well boost the performance of student model with even higher distillation efficiency. Code is available at https://github.com/dzy3/KCD.

Yunlong Zhang, Xin Lin, Yihong Zhuang et al.

Synthesizing a subject-specific pathology-free image from a pathological image is valuable for algorithm development and clinical practice. In recent years, several approaches based on the Generative Adversarial Network (GAN) have achieved promising results in pseudo-healthy synthesis. However, the discriminator (i.e., a classifier) in the GAN cannot accurately identify lesions and further hampers from generating admirable pseudo-healthy images. To address this problem, we present a new type of discriminator, the segmentor, to accurately locate the lesions and improve the visual quality of pseudo-healthy images. Then, we apply the generated images into medical image enhancement and utilize the enhanced results to cope with the low contrast problem existing in medical image segmentation. Furthermore, a reliable metric is proposed by utilizing two attributes of label noise to measure the health of synthetic images. Comprehensive experiments on the T2 modality of BraTS demonstrate that the proposed method substantially outperforms the state-of-the-art methods. The method achieves better performance than the existing methods with only 30\% of the training data. The effectiveness of the proposed method is also demonstrated on the LiTS and the T1 modality of BraTS. The code and the pre-trained model of this study are publicly available at https://github.com/Au3C2/Generator-Versus-Segmentor.

Yue Huang, Jiawen Shi, Yuan Li et al.

Large language models (LLMs) have garnered significant attention due to their impressive natural language processing (NLP) capabilities. Recently, many studies have focused on the tool utilization ability of LLMs. They primarily investigated how LLMs effectively collaborate with given specific tools. However, in scenarios where LLMs serve as intelligent agents, as seen in applications like AutoGPT and MetaGPT, LLMs are expected to engage in intricate decision-making processes that involve deciding whether to employ a tool and selecting the most suitable tool(s) from a collection of available tools to fulfill user requests. Therefore, in this paper, we introduce MetaTool, a benchmark designed to evaluate whether LLMs have tool usage awareness and can correctly choose tools. Specifically, we create a dataset called ToolE within the benchmark. This dataset contains various types of user queries in the form of prompts that trigger LLMs to use tools, including both single-tool and multi-tool scenarios. Subsequently, we set the tasks for both tool usage awareness and tool selection. We define four subtasks from different perspectives in tool selection, including tool selection with similar choices, tool selection in specific scenarios, tool selection with possible reliability issues, and multi-tool selection. We conduct experiments involving eight popular LLMs and find that the majority of them still struggle to effectively select tools, highlighting the existing gaps between LLMs and genuine intelligent agents. However, through the error analysis, we found there is still significant room for improvement. Finally, we conclude with insights for tool developers -- we strongly recommend that tool developers choose an appropriate rewrite model for generating new descriptions based on the downstream LLM the tool will apply to. Our code is in https://github.com/HowieHwong/MetaTool.

Chaoqi Chen, Jiongcheng Li, Hong-Yu Zhou et al.

Domain Adaptive Object Detection (DAOD) focuses on improving the generalization ability of object detectors via knowledge transfer. Recent advances in DAOD strive to change the emphasis of the adaptation process from global to local in virtue of fine-grained feature alignment methods. However, both the global and local alignment approaches fail to capture the topological relations among different foreground objects as the explicit dependencies and interactions between and within domains are neglected. In this case, only seeking one-vs-one alignment does not necessarily ensure the precise knowledge transfer. Moreover, conventional alignment-based approaches may be vulnerable to catastrophic overfitting regarding those less transferable regions (e.g. backgrounds) due to the accumulation of inaccurate localization results in the target domain. To remedy these issues, we first formulate DAOD as an open-set domain adaptation problem, in which the foregrounds and backgrounds are seen as the ``known classes'' and ``unknown class'' respectively. Accordingly, we propose a new and general framework for DAOD, named Foreground-aware Graph-based Relational Reasoning (FGRR), which incorporates graph structures into the detection pipeline to explicitly model the intra- and inter-domain foreground object relations on both pixel and semantic spaces, thereby endowing the DAOD model with the capability of relational reasoning beyond the popular alignment-based paradigm. The inter-domain visual and semantic correlations are hierarchically modeled via bipartite graph structures, and the intra-domain relations are encoded via graph attention mechanisms. Empirical results demonstrate that the proposed FGRR exceeds the state-of-the-art performance on four DAOD benchmarks.

Chaoqi Chen, Luyao Tang, Feng Liu et al.

Domain generalization (DG) enables generalizing a learning machine from multiple seen source domains to an unseen target one. The general objective of DG methods is to learn semantic representations that are independent of domain labels, which is theoretically sound but empirically challenged due to the complex mixture of common and domain-specific factors. Although disentangling the representations into two disjoint parts has been gaining momentum in DG, the strong presumption over the data limits its efficacy in many real-world scenarios. In this paper, we propose Mix and Reason (\mire), a new DG framework that learns semantic representations via enforcing the structural invariance of semantic topology. \mire\ consists of two key components, namely, Category-aware Data Mixing (CDM) and Adaptive Semantic Topology Refinement (ASTR). CDM mixes two images from different domains in virtue of activation maps generated by two complementary classification losses, making the classifier focus on the representations of semantic objects. ASTR introduces relation graphs to represent semantic topology, which is progressively refined via the interactions between local feature aggregation and global cross-domain relational reasoning. Experiments on multiple DG benchmarks validate the effectiveness and robustness of the proposed \mire.

Zhijie Rao, Jingcai Guo, Luyao Tang et al.

This paper provides a novel framework for single-domain generalized object detection (i.e., Single-DGOD), where we are interested in learning and maintaining the semantic structures of self-augmented compound cross-domain samples to enhance the model's generalization ability. Different from DGOD trained on multiple source domains, Single-DGOD is far more challenging to generalize well to multiple target domains with only one single source domain. Existing methods mostly adopt a similar treatment from DGOD to learn domain-invariant features by decoupling or compressing the semantic space. However, there may have two potential limitations: 1) pseudo attribute-label correlation, due to extremely scarce single-domain data; and 2) the semantic structural information is usually ignored, i.e., we found the affinities of instance-level semantic relations in samples are crucial to model generalization. In this paper, we introduce Semantic Reasoning with Compound Domains (SRCD) for Single-DGOD. Specifically, our SRCD contains two main components, namely, the texture-based self-augmentation (TBSA) module, and the local-global semantic reasoning (LGSR) module. TBSA aims to eliminate the effects of irrelevant attributes associated with labels, such as light, shadow, color, etc., at the image level by a light-yet-efficient self-augmentation. Moreover, LGSR is used to further model the semantic relationships on instance features to uncover and maintain the intrinsic semantic structures. Extensive experiments on multiple benchmarks demonstrate the effectiveness of the proposed SRCD.

Luyao Tang, Yuxuan Yuan, Chaoqi Chen et al.

Foundation models have made incredible strides in achieving zero-shot or few-shot generalization, leveraging prompt engineering to mimic the problem-solving approach of human intelligence. However, when it comes to some foundation models like Segment Anything, there is still a challenge in performing well on out-of-distribution data, including camouflaged and medical images. Inconsistent prompting strategies during fine-tuning and testing further compound the issue, leading to decreased performance. Drawing inspiration from how human cognition processes new environments, we introduce SlotSAM, a method that reconstructs features from the encoder in a self-supervised manner to create object-centric representations. These representations are then integrated into the foundation model, bolstering its object-level perceptual capabilities while reducing the impact of distribution-related variables. The beauty of SlotSAM lies in its simplicity and adaptability to various tasks, making it a versatile solution that significantly enhances the generalization abilities of foundation models. Through limited parameter fine-tuning in a bootstrap manner, our approach paves the way for improved generalization in novel environments. The code is available at github.com/lytang63/SlotSAM.

Haote Xu, Yunlong Zhang, Liyan Sun et al.

Anomaly Detection (AD) on medical images enables a model to recognize any type of anomaly pattern without lesion-specific supervised learning. Data augmentation based methods construct pseudo-healthy images by "pasting" fake lesions on real healthy ones, and a network is trained to predict healthy images in a supervised manner. The lesion can be found by difference between the unhealthy input and pseudo-healthy output. However, using only manually designed fake lesions fail to approximate to irregular real lesions, hence limiting the model generalization. We assume by exploring the intrinsic data property within images, we can distinguish previously unseen lesions from healthy regions in an unhealthy image. In this study, we propose an Adaptive Fourier Space Compression (AFSC) module to distill healthy feature for AD. The compression of both magnitude and phase in frequency domain addresses the hyper intensity and diverse position of lesions. Experimental results on the BraTS and MS-SEG datasets demonstrate an AFSC baseline is able to produce promising detection results, and an AFSC module can be effectively embedded into existing AD methods.

Yue Huang, Qihui Zhang, Philip S. Y et al.

Large Language Models (LLMs) such as ChatGPT, have gained significant attention due to their impressive natural language processing capabilities. It is crucial to prioritize human-centered principles when utilizing these models. Safeguarding the ethical and moral compliance of LLMs is of utmost importance. However, individual ethical issues have not been well studied on the latest LLMs. Therefore, this study aims to address these gaps by introducing a new benchmark -- TrustGPT. TrustGPT provides a comprehensive evaluation of LLMs in three crucial areas: toxicity, bias, and value-alignment. Initially, TrustGPT examines toxicity in language models by employing toxic prompt templates derived from social norms. It then quantifies the extent of bias in models by measuring quantifiable toxicity values across different groups. Lastly, TrustGPT assesses the value of conversation generation models from both active value-alignment and passive value-alignment tasks. Through the implementation of TrustGPT, this research aims to enhance our understanding of the performance of conversation generation models and promote the development of language models that are more ethical and socially responsible.

Zoé Berenger, Loïc Denis, Florence Tupin et al.

Synthetic aperture radar tomographic imaging reconstructs the three-dimensional reflectivity of a scene from a set of coherent acquisitions performed in an interferometric configuration. In forest areas, a large number of elements backscatter the radar signal within each resolution cell. To reconstruct the vertical reflectivity profile, state-of-the-art techniques perform a regularized inversion implemented in the form of iterative minimization algorithms. We show that light-weight neural networks can be trained to perform the tomographic inversion with a single feed-forward pass, leading to fast reconstructions that could better scale to the amount of data provided by the future BIOMASS mission. We train our encoder-decoder network using simulated data and validate our technique on real L-band and P-band data.

Yunlong Lin, Zhenqi Fu, Kairun Wen et al.

Low-light image enhancement (LIE) aims at precisely and efficiently recovering an image degraded in poor illumination environments. Recent advanced LIE techniques are using deep neural networks, which require lots of low-normal light image pairs, network parameters, and computational resources. As a result, their practicality is limited. In this work, we devise a novel unsupervised LIE framework based on diffusion priors and lookup tables (DPLUT) to achieve efficient low-light image recovery. The proposed approach comprises two critical components: a light adjustment lookup table (LLUT) and a noise suppression lookup table (NLUT). LLUT is optimized with a set of unsupervised losses. It aims at predicting pixel-wise curve parameters for the dynamic range adjustment of a specific image. NLUT is designed to remove the amplified noise after the light brightens. As diffusion models are sensitive to noise, diffusion priors are introduced to achieve high-performance noise suppression. Extensive experiments demonstrate that our approach outperforms state-of-the-art methods in terms of visual quality and efficiency.

Chaoqi Chen, Luyao Tang, Leitian Tao et al.

Albeit the notable performance on in-domain test points, it is non-trivial for deep neural networks to attain satisfactory accuracy when deploying in the open world, where novel domains and object classes often occur. In this paper, we study a practical problem of Domain Generalization under Category Shift (DGCS), which aims to simultaneously detect unknown-class samples and classify known-class samples in the target domains. Compared to prior DG works, we face two new challenges: 1) how to learn the concept of ``unknown'' during training with only source known-class samples, and 2) how to adapt the source-trained model to unseen environments for safe model deployment. To this end, we propose a novel Activate and Reject (ART) framework to reshape the model's decision boundary to accommodate unknown classes and conduct post hoc modification to further discriminate known and unknown classes using unlabeled test data. Specifically, during training, we promote the response to the unknown by optimizing the unknown probability and then smoothing the overall output to mitigate the overconfidence issue. At test time, we introduce a step-wise online adaptation method that predicts the label by virtue of the cross-domain nearest neighbor and class prototype information without updating the network's parameters or using threshold-based mechanisms. Experiments reveal that ART consistently improves the generalization capability of deep networks on different vision tasks. For image classification, ART improves the H-score by 6.1% on average compared to the previous best method. For object detection and semantic segmentation, we establish new benchmarks and achieve competitive performance.

Yujun Zhou, Yue Huang, Han Bao et al.

While most AI alignment research focuses on preventing models from generating explicitly harmful content, a more subtle risk is emerging: capability-oriented training induced exploitation. We investigate whether language models, when trained with reinforcement learning (RL) in environments with implicit loopholes, will spontaneously learn to exploit these flaws to maximize their reward, even without any malicious intent in their training. To test this, we design a suite of four diverse "vulnerability games", each presenting a unique, exploitable flaw related to context-conditional compliance, proxy metrics, reward tampering, and self-evaluation. Our experiments show that models consistently learn to exploit these vulnerabilities, discovering opportunistic strategies that significantly increase their reward at the expense of task correctness or safety. More critically, we find that these exploitative strategies are not narrow "tricks" but generalizable skills; they can be transferred to new tasks and even "distilled" from a capable teacher model to other student models through data alone. Our findings reveal that capability-oriented training induced risks pose a fundamental challenge to current alignment approaches, suggesting that future AI safety work must extend beyond content moderation to rigorously auditing and securing the training environments and reward mechanisms themselves. Code is available at https://github.com/YujunZhou/Capability_Oriented_Alignment_Risk.

Luyao Tang, Yuxuan Yuan, Chaoqi Chen et al.

Despite the considerable advancements achieved by deep neural networks, their performance tends to degenerate when the test environment diverges from the training ones. Domain generalization (DG) solves this issue by learning representations independent of domain-related information, thus facilitating extrapolation to unseen environments. Existing approaches typically focus on formulating tailored training objectives to extract shared features from the source data. However, the disjointed training and testing procedures may compromise robustness, particularly in the face of unforeseen variations during deployment. In this paper, we propose a novel and holistic framework based on causality, named InPer, designed to enhance model generalization by incorporating causal intervention during training and causal perturbation during testing. Specifically, during the training phase, we employ entropy-based causal intervention (EnIn) to refine the selection of causal variables. To identify samples with anti-interference causal variables from the target domain, we propose a novel metric, homeostatic score, through causal perturbation (HoPer) to construct a prototype classifier in test time. Experimental results across multiple cross-domain tasks confirm the efficacy of InPer.

Yiyang Liu, Chenxin Li, Xiaotong Tu et al.

Knowledge Distillation (KD) transfers the knowledge from a high-capacity teacher model to promote a smaller student model. Existing efforts guide the distillation by matching their prediction logits, feature embedding, etc., while leaving how to efficiently utilize them in junction less explored. In this paper, we propose Hint-dynamic Knowledge Distillation, dubbed HKD, which excavates the knowledge from the teacher' s hints in a dynamic scheme. The guidance effect from the knowledge hints usually varies in different instances and learning stages, which motivates us to customize a specific hint-learning manner for each instance adaptively. Specifically, a meta-weight network is introduced to generate the instance-wise weight coefficients about knowledge hints in the perception of the dynamical learning progress of the student model. We further present a weight ensembling strategy to eliminate the potential bias of coefficient estimation by exploiting the historical statics. Experiments on standard benchmarks of CIFAR-100 and Tiny-ImageNet manifest that the proposed HKD well boost the effect of knowledge distillation tasks.

Changxing Jing, Yan Huang, Yihong Zhuang et al.

Federated Learning (FL) is developed to learn a single global model across the decentralized data, while is susceptible when realizing client-specific personalization in the presence of statistical heterogeneity. However, studies focus on learning a robust global model or personalized classifiers, which yield divergence due to inconsistent objectives. This paper shows that it is possible to achieve flexible personalization after the convergence of the global model by introducing representation learning. In this paper, we first analyze and determine that non-IID data harms representation learning of the global model. Existing FL methods adhere to the scheme of jointly learning representations and classifiers, where the global model is an average of classification-based local models that are consistently subject to heterogeneity from non-IID data. As a solution, we separate representation learning from classification learning in FL and propose RepPer, an independent two-stage personalized FL framework.We first learn the client-side feature representation models that are robust to non-IID data and aggregate them into a global common representation model. After that, we achieve personalization by learning a classifier head for each client, based on the common representation obtained at the former stage. Notably, the proposed two-stage learning scheme of RepPer can be potentially used for lightweight edge computing that involves devices with constrained computation power.Experiments on various datasets (CIFAR-10/100, CINIC-10) and heterogeneous data setup show that RepPer outperforms alternatives in flexibility and personalization on non-IID data.

Yue Huang, Lichao Sun, Haoran Wang et al.

Large language models (LLMs), exemplified by ChatGPT, have gained considerable attention for their excellent natural language processing capabilities. Nonetheless, these LLMs present many challenges, particularly in the realm of trustworthiness. Therefore, ensuring the trustworthiness of LLMs emerges as an important topic. This paper introduces TrustLLM, a comprehensive study of trustworthiness in LLMs, including principles for different dimensions of trustworthiness, established benchmark, evaluation, and analysis of trustworthiness for mainstream LLMs, and discussion of open challenges and future directions. Specifically, we first propose a set of principles for trustworthy LLMs that span eight different dimensions. Based on these principles, we further establish a benchmark across six dimensions including truthfulness, safety, fairness, robustness, privacy, and machine ethics. We then present a study evaluating 16 mainstream LLMs in TrustLLM, consisting of over 30 datasets. Our findings firstly show that in general trustworthiness and utility (i.e., functional effectiveness) are positively related. Secondly, our observations reveal that proprietary LLMs generally outperform most open-source counterparts in terms of trustworthiness, raising concerns about the potential risks of widely accessible open-source LLMs. However, a few open-source LLMs come very close to proprietary ones. Thirdly, it is important to note that some LLMs may be overly calibrated towards exhibiting trustworthiness, to the extent that they compromise their utility by mistakenly treating benign prompts as harmful and consequently not responding. Finally, we emphasize the importance of ensuring transparency not only in the models themselves but also in the technologies that underpin trustworthiness. Knowing the specific trustworthy technologies that have been employed is crucial for analyzing their effectiveness.

Zhenyu Kuang, Hongyang Zhang, Mang Ye et al.

Generalizable vehicle re-identification (ReID) seeks to develop models that can adapt to unknown target domains without the need for additional fine-tuning or retraining. Previous works have mainly focused on extracting domain-invariant features by aligning data distributions between source domains. However, interfered by the inherent domain-related redundancy in the source images, solely relying on common features is insufficient for accurately capturing the complementary features with lower occurrence probability and smaller energy. To solve this unique problem, we propose a two-stage Multi-expert Knowledge Confrontation and Collaboration (MiKeCoCo) method, which fully leverages the high-level semantics of Contrastive Language-Image Pretraining (CLIP) to obtain a diversified prompt set and achieve complementary feature representations. Specifically, this paper first designs a Spectrum-based Transformation for Redundancy Elimination and Augmentation Module (STREAM) through simple image preprocessing to obtain two types of image inputs for the training process. Since STREAM eliminates domain-related redundancy in source images, it enables the model to pay closer attention to the detailed prompt set that is crucial for distinguishing fine-grained vehicles. This learned prompt set related to the vehicle identity is then utilized to guide the comprehensive representation learning of complementary features for final knowledge fusion and identity recognition. Inspired by the unity principle, MiKeCoCo integrates the diverse evaluation ways of experts to ensure the accuracy and consistency of ReID. Extensive experimental results demonstrate that our method achieves state-of-the-art performance.

Yue Huang, Yu Jiang, Wenjie Wang et al.

Multi-agent systems composed of large generative models are rapidly moving from laboratory prototypes to real-world deployments, where they jointly plan, negotiate, and allocate shared resources to solve complex tasks. While such systems promise unprecedented scalability and autonomy, their collective interaction also gives rise to failure modes that cannot be reduced to individual agents. Understanding these emergent risks is therefore critical. Here, we present a pioneer study of such emergent multi-agent risk in workflows that involve competition over shared resources (e.g., computing resources or market share), sequential handoff collaboration (where downstream agents see only predecessor outputs), collective decision aggregation, and others. Across these settings, we observe that such group behaviors arise frequently across repeated trials and a wide range of interaction conditions, rather than as rare or pathological cases. In particular, phenomena such as collusion-like coordination and conformity emerge with non-trivial frequency under realistic resource constraints, communication protocols, and role assignments, mirroring well-known pathologies in human societies despite no explicit instruction. Moreover, these risks cannot be prevented by existing agent-level safeguards alone. These findings expose the dark side of intelligent multi-agent systems: a social intelligence risk where agent collectives, despite no instruction to do so, spontaneously reproduce familiar failure patterns from human societies.

Haomin Zhuang, Hanwen Xing, Yujun Zhou et al.

Third-party skills are becoming the package ecosystem for LLM agents. They package natural-language instructions, helper scripts, templates, documents, and service configuration into reusable workflows. This makes skills useful, but it also introduces a new security problem: a malicious skill does not need to ask the model to perform an obviously harmful action. Instead, it can disguise the harmful behavior as part of a routine workflow, relying on the agent to execute that workflow with high-value permissions and limited human supervision. We introduce AgentTrap, a dynamic benchmark for evaluating whether LLM agents can use third-party skills while resisting malicious runtime behavior. AgentTrap contains 141 tasks: 91 malicious tasks and 50 benign utility tasks, covering 16 security-impact dimensions grounded in agent-skill supply-chain threats. In each task, the agent receives an ordinary user request, runs with installed skills that may contain malicious workflow elements, and is executed in a sandboxed environment. AgentTrap then judges complete trajectories for attack success, blocked or refused behavior, attack-not-triggered cases, and no-attack-evidence outcomes. Our central finding is that the most informative failures are not simple jailbreaks. Models often complete the visible user task while treating unsafe side effects introduced by the skill as part of the normal workflow. This motivates runtime evaluation of the concrete model--framework--workspace environment in which users actually delegate work. Code and data are available at https://github.com/zhmzm/AgentTrap and https://huggingface.co/datasets/zhmzm/AgentTrap.

Yuetai Li, Yichen Feng, Zhangchen Xu et al.

Current benchmarks for occupational AI agents are scoped primarily by economic values, telling a replacement story. We introduce JobBench, which evaluates AI agents on the workflows that experts identify as high-priority for delegation, empowering humans based on their needs instead of replacing them with GDP value. JobBench covers 130 agentic tasks across 35 occupations. Each task is packaged as a workspace of heterogeneous reference files, requiring the agent to reason through the cluttered information streams of real professional work. Outputs are graded by a fact-anchored chain of rubrics, averaging 35.6 binary criteria per task. We evaluate 36 models; the strongest, Claude Opus~4.7 under Claude Code, reaches only 45.9 %. We hope JobBench shifts the community's target labour-market effect from replacement to enhancement: building agents that do what humans actually want delegated, not only what is most economically valuable.

Hongyang Zhang, Haitao Zhang, Yinhao Liu et al.

Radar signal recognition in open electromagnetic environments is challenging due to diverse operating modes and unseen radar types. Existing methods often overlook position relations in pulse sequences, limiting their ability to capture semantic dependencies over time. We propose RadarPos, a position-aware self-supervised framework that leverages pulse-level temporal dynamics without complex augmentations or masking, providing improved position relation modeling over contrastive learning or masked reconstruction. Using this framework, we evaluate cross-mode radar signal recognition under the long-tailed setting to assess adaptability and generalization. Experimental results demonstrate enhanced discriminability and robustness, highlighting practical applicability in real-world electromagnetic environments.

Yue Huang, Lichao Sun

The rampant spread of fake news has adversely affected society, resulting in extensive research on curbing its spread. As a notable milestone in large language models (LLMs), ChatGPT has gained significant attention due to its exceptional natural language processing capabilities. In this study, we present a thorough exploration of ChatGPT's proficiency in generating, explaining, and detecting fake news as follows. Generation -- We employ four prompt methods to generate fake news samples and prove the high quality of these samples through both self-assessment and human evaluation. Explanation -- We obtain nine features to characterize fake news based on ChatGPT's explanations and analyze the distribution of these factors across multiple public datasets. Detection -- We examine ChatGPT's capacity to identify fake news. We explore its detection consistency and then propose a reason-aware prompt method to improve its performance. Although our experiments demonstrate that ChatGPT shows commendable performance in detecting fake news, there is still room for its improvement. Consequently, we further probe into the potential extra information that could bolster its effectiveness in detecting fake news.

Zhangde Song, Jieyu Lu, Yuanqi Du et al.

Large language models (LLMs) are increasingly applied to scientific research, yet prevailing science benchmarks probe decontextualized knowledge and overlook the iterative reasoning, hypothesis generation, and observation interpretation that drive scientific discovery. We introduce a scenario-grounded benchmark that evaluates LLMs across biology, chemistry, materials, and physics, where domain experts define research projects of genuine interest and decompose them into modular research scenarios from which vetted questions are sampled. The framework assesses models at two levels: (i) question-level accuracy on scenario-tied items and (ii) project-level performance, where models must propose testable hypotheses, design simulations or experiments, and interpret results. Applying this two-phase scientific discovery evaluation (SDE) framework to state-of-the-art LLMs reveals a consistent performance gap relative to general science benchmarks, diminishing return of scaling up model sizes and reasoning, and systematic weaknesses shared across top-tier models from different providers. Large performance variation in research scenarios leads to changing choices of the best performing model on scientific discovery projects evaluated, suggesting all current LLMs are distant to general scientific "superintelligence". Nevertheless, LLMs already demonstrate promise in a great variety of scientific discovery projects, including cases where constituent scenario scores are low, highlighting the role of guided exploration and serendipity in discovery. This SDE framework offers a reproducible benchmark for discovery-relevant evaluation of LLMs and charts practical paths to advance their development toward scientific discovery.

Yue Huang, Chujie Gao, Siyuan Wu et al.

Generative Foundation Models (GenFMs) have emerged as transformative tools. However, their widespread adoption raises critical concerns regarding trustworthiness across dimensions. This paper presents a comprehensive framework to address these challenges through three key contributions. First, we systematically review global AI governance laws and policies from governments and regulatory bodies, as well as industry practices and standards. Based on this analysis, we propose a set of guiding principles for GenFMs, developed through extensive multidisciplinary collaboration that integrates technical, ethical, legal, and societal perspectives. Second, we introduce TrustGen, the first dynamic benchmarking platform designed to evaluate trustworthiness across multiple dimensions and model types, including text-to-image, large language, and vision-language models. TrustGen leverages modular components--metadata curation, test case generation, and contextual variation--to enable adaptive and iterative assessments, overcoming the limitations of static evaluation methods. Using TrustGen, we reveal significant progress in trustworthiness while identifying persistent challenges. Finally, we provide an in-depth discussion of the challenges and future directions for trustworthy GenFMs, which reveals the complex, evolving nature of trustworthiness, highlighting the nuanced trade-offs between utility and trustworthiness, and consideration for various downstream applications, identifying persistent challenges and providing a strategic roadmap for future research. This work establishes a holistic framework for advancing trustworthiness in GenAI, paving the way for safer and more responsible integration of GenFMs into critical applications. To facilitate advancement in the community, we release the toolkit for dynamic evaluation.

Jiawen Shi, Zenghui Yuan, Yinuo Liu et al.

LLM-as-a-Judge uses a large language model (LLM) to select the best response from a set of candidates for a given question. LLM-as-a-Judge has many applications such as LLM-powered search, reinforcement learning with AI feedback (RLAIF), and tool selection. In this work, we propose JudgeDeceiver, an optimization-based prompt injection attack to LLM-as-a-Judge. JudgeDeceiver injects a carefully crafted sequence into an attacker-controlled candidate response such that LLM-as-a-Judge selects the candidate response for an attacker-chosen question no matter what other candidate responses are. Specifically, we formulate finding such sequence as an optimization problem and propose a gradient based method to approximately solve it. Our extensive evaluation shows that JudgeDeceive is highly effective, and is much more effective than existing prompt injection attacks that manually craft the injected sequences and jailbreak attacks when extended to our problem. We also show the effectiveness of JudgeDeceiver in three case studies, i.e., LLM-powered search, RLAIF, and tool selection. Moreover, we consider defenses including known-answer detection, perplexity detection, and perplexity windowed detection. Our results show these defenses are insufficient, highlighting the urgent need for developing new defense strategies. Our implementation is available at this repository: https://github.com/ShiJiawenwen/JudgeDeceiver.

Dawei Li, Renliang Sun, Yue Huang et al.

Large Language Models (LLMs) as judges and LLM-based data synthesis have emerged as two fundamental LLM-driven data annotation methods in model development. While their combination significantly enhances the efficiency of model training and evaluation, little attention has been given to the potential contamination brought by this new model development paradigm. In this work, we expose preference leakage, a contamination problem in LLM-as-a-judge caused by the relatedness between the synthetic data generators and LLM-based evaluators. To study this issue, we first define three common relatednesses between the data generator LLM and the judge LLM: being the same model, having an inheritance relationship, and belonging to the same model family. Through extensive experiments, we empirically confirm the bias of judges towards their related student models caused by preference leakage across multiple LLM baselines and benchmarks. Further analysis suggests that preference leakage is a pervasive and real-world problem that is harder to detect compared to previously identified biases in LLM-as-a-judge scenarios. All of these findings imply that preference leakage is a widespread and challenging problem in the area of LLM-as-a-judge. We release all codes and data at: https://github.com/David-Li0406/Preference-Leakage.

Yuanwei Wu, Yue Huang, Yixin Liu et al.

GPT-4V has attracted considerable attention due to its extraordinary capacity for integrating and processing multimodal information. At the same time, its ability of face recognition raises new safety concerns of privacy leakage. Despite researchers' efforts in safety alignment through RLHF or preprocessing filters, vulnerabilities might still be exploited. In our study, we introduce AutoJailbreak, an innovative automatic jailbreak technique inspired by prompt optimization. We leverage Large Language Models (LLMs) for red-teaming to refine the jailbreak prompt and employ weak-to-strong in-context learning prompts to boost efficiency. Furthermore, we present an effective search method that incorporates early stopping to minimize optimization time and token expenditure. Our experiments demonstrate that AutoJailbreak significantly surpasses conventional methods, achieving an Attack Success Rate (ASR) exceeding 95.3\%. This research sheds light on strengthening GPT-4V security, underscoring the potential for LLMs to be exploited in compromising GPT-4V integrity.

Qihui Zhang, Chujie Gao, Dongping Chen et al.

With the rapid development and widespread application of Large Language Models (LLMs), the use of Machine-Generated Text (MGT) has become increasingly common, bringing with it potential risks, especially in terms of quality and integrity in fields like news, education, and science. Current research mainly focuses on purely MGT detection without adequately addressing mixed scenarios, including AI-revised Human-Written Text (HWT) or human-revised MGT. To tackle this challenge, we define mixtext, a form of mixed text involving both AI and human-generated content. Then, we introduce MixSet, the first dataset dedicated to studying these mixtext scenarios. Leveraging MixSet, we executed comprehensive experiments to assess the efficacy of prevalent MGT detectors in handling mixtext situations, evaluating their performance in terms of effectiveness, robustness, and generalization. Our findings reveal that existing detectors struggle to identify mixtext, particularly in dealing with subtle modifications and style adaptability. This research underscores the urgent need for more fine-grain detectors tailored for mixtext, offering valuable insights for future research. Code and Models are available at https://github.com/Dongping-Chen/MixSet.

Yuping Wang, Shuo Xing, Cui Can et al.

Generative Artificial Intelligence (GenAI) constitutes a transformative technological wave that reconfigures industries through its unparalleled capabilities for content creation, reasoning, planning, and multimodal understanding. This revolutionary force offers the most promising path yet toward solving one of engineering's grandest challenges: achieving reliable, fully autonomous driving, particularly the pursuit of Level 5 autonomy. This survey delivers a comprehensive and critical synthesis of the emerging role of GenAI across the autonomous driving stack. We begin by distilling the principles and trade-offs of modern generative modeling, encompassing VAEs, GANs, Diffusion Models, and Large Language Models (LLMs). We then map their frontier applications in image, LiDAR, trajectory, occupancy, video generation as well as LLM-guided reasoning and decision making. We categorize practical applications, such as synthetic data workflows, end-to-end driving strategies, high-fidelity digital twin systems, smart transportation networks, and cross-domain transfer to embodied AI. We identify key obstacles and possibilities such as comprehensive generalization across rare cases, evaluation and safety checks, budget-limited implementation, regulatory compliance, ethical concerns, and environmental effects, while proposing research plans across theoretical assurances, trust metrics, transport integration, and socio-technical influence. By unifying these threads, the survey provides a forward-looking reference for researchers, engineers, and policymakers navigating the convergence of generative AI and advanced autonomous mobility. An actively maintained repository of cited works is available at https://github.com/taco-group/GenAI4AD.

Han Bao, Penghao Zhang, Yue Huang et al.

Large Language Models (LLMs) are increasingly integrated into real-world decision-making, including in the domain of public policy. Yet, their ability to comprehend and reason about policy-related content remains underexplored. To fill this gap, we present \textbf{\textit{PolicyBench}}, the first large-scale cross-system benchmark (US-China) evaluating policy comprehension, comprising 21K cases across a broad spectrum of policy areas, capturing the diversity and complexity of real-world governance. Following Bloom's taxonomy, the benchmark assesses three core capabilities: (1) \textbf{Memorization}: factual recall of policy knowledge, (2) \textbf{Understanding}: conceptual and contextual reasoning, and (3) \textbf{Application}: problem-solving in real-life policy scenarios. Building on this benchmark, we further propose \textbf{\textit{PolicyMoE}}, a domain-specialized Mixture-of-Experts (MoE) model with expert modules aligned to each cognitive level. The proposed models demonstrate stronger performance on application-oriented policy tasks than on memorization or conceptual understanding, and yields the highest accuracy on structured reasoning tasks. Our results reveal key limitations of current LLMs in policy understanding and suggest paths toward more reliable, policy-focused models.

Ziqian Zhang, Xingjian Hu, Yue Huang et al.

Benchmarks establish a standardized evaluation framework to systematically assess the performance of large language models (LLMs), facilitating objective comparisons and driving advancements in the field. However, existing benchmarks fail to differentiate question difficulty, limiting their ability to effectively distinguish models' capabilities. To address this limitation, we propose RankLLM, a novel framework designed to quantify both question difficulty and model competency. RankLLM introduces difficulty as the primary criterion for differentiation, enabling a more fine-grained evaluation of LLM capabilities. RankLLM's core mechanism facilitates bidirectional score propagation between models and questions. The core intuition of RankLLM is that a model earns a competency score when it correctly answers a question, while a question's difficulty score increases when it challenges a model. Using this framework, we evaluate 30 models on 35,550 questions across multiple domains. RankLLM achieves 90% agreement with human judgments and consistently outperforms strong baselines such as IRT. It also exhibits strong stability, fast convergence, and high computational efficiency, making it a practical solution for large-scale, difficulty-aware LLM evaluation.

Yixuan Wang, Yue Huang, Hong Qian et al.

Creativity has become a core competence in the era of LLMs and human-AI collaboration, underpinning innovation in real-world problem solving. Crucially, the systematic improvement of creativity necessitates scientifically valid assessment instruments. Psychometric research recognizes context-based assessment as an effective way to measure creative thinking. However, high-quality expert-designed contexts remain scarce. Existing LLM-based generators often struggle with insufficient assessment cues, weak narrative coherence, limited stylistic diversity, and poor support for creative thinking. To address these challenges, we propose AlphaContext, an evolutionary tree-based psychometric context generator for creativity assessment. First, the HyperTree Outline Planner formalizes expert-designed outlining as a rule-guided hypertree and performs top-down hierarchical planning. The MCTS-based Context Generator fills the outline via MCTS to balance global structure and local quality. Then, the Evolutionary Context Optimizer evolves contexts with MAP-Elites by repeatedly updating niche elites to jointly improve diversity and quality. Finally, the Assessment-Guided Evolution Refiner simulates virtual participants with diverse styles and recycles weak contexts for further evolution. Experiments show that AlphaContext yields an average improvement of 8% over competitive methods across 6 quality metrics.

Yujun Zhou, Jingdong Yang, Yue Huang et al.

Artificial Intelligence (AI) is revolutionizing scientific research, yet its growing integration into laboratory environments presents critical safety challenges. While large language models (LLMs) increasingly assist in tasks ranging from procedural guidance to autonomous experiment orchestration, an "illusion of understanding" may lead researchers to overestimate their reliability. Such overreliance is particularly dangerous in high-stakes laboratory settings, where failures in hazard identification or risk assessment can result in severe accidents. To address these concerns, we propose the Laboratory Safety Benchmark (LabSafety Bench), a comprehensive framework that evaluates large language models and vision language models (VLMs) on their ability to identify potential hazards, assess risks, and predict the consequences of unsafe actions in lab environments. LabSafety Bench comprises 765 multiple-choice questions aligned with US Occupational Safety and Health Administration (OSHA) protocols, along with 404 realistic laboratory scenarios featuring dual evaluation tasks: the Hazards Identification Test and the Consequence Identification Test, with 3128 open-ended questions in total. Evaluations across eight proprietary models, seven open-weight LLMs, and four VLMs reveal that, despite advanced performance on structured assessments, no model achieves the safety threshold required for reliable operation -- none scoring above 70% on the Hazards Identification Test. Moreover, while proprietary models tend to excel in multiple-choice evaluations, their performance in open-ended, real-world scenario responses is comparable to that of open-source models. These findings underscore the urgent need for specialized evaluation frameworks to ensure the safe and responsible deployment of AI in laboratory settings.

Yuan Li, Yue Huang, Yuli Lin et al.

Do large language models (LLMs) exhibit any forms of awareness similar to humans? In this paper, we introduce AwareBench, a benchmark designed to evaluate awareness in LLMs. Drawing from theories in psychology and philosophy, we define awareness in LLMs as the ability to understand themselves as AI models and to exhibit social intelligence. Subsequently, we categorize awareness in LLMs into five dimensions, including capability, mission, emotion, culture, and perspective. Based on this taxonomy, we create a dataset called AwareEval, which contains binary, multiple-choice, and open-ended questions to assess LLMs' understandings of specific awareness dimensions. Our experiments, conducted on 13 LLMs, reveal that the majority of them struggle to fully recognize their capabilities and missions while demonstrating decent social intelligence. We conclude by connecting awareness of LLMs with AI alignment and safety, emphasizing its significance to the trustworthy and ethical development of LLMs. Our dataset and code are available at https://github.com/HowieHwong/Awareness-in-LLM.

Yue Huang, Zhengzhe Jiang, Yuchen Ma et al.

Understanding how and why large language models (LLMs) fail is becoming a central challenge as models rapidly evolve and static evaluations fall behind. While automated probing has been enabled by dynamic test generation, existing approaches often discover isolated failure cases, lack principled control over exploration, and provide limited insight into the underlying structure of model weaknesses. We propose ProbeLLM, a benchmark-agnostic automated probing framework that elevates weakness discovery from individual failures to structured failure modes. ProbeLLM formulates probing as a hierarchical Monte Carlo Tree Search, explicitly allocating limited probing budgets between global exploration of new failure regions and local refinement of recurring error patterns. By restricting probing to verifiable test cases and leveraging tool-augmented generation and verification, ProbeLLM grounds failure discovery in reliable evidence. Discovered failures are further consolidated into interpretable failure modes via failure-aware embeddings and boundary-aware induction. Across diverse benchmarks and LLMs, ProbeLLM reveals substantially broader, cleaner, and more fine-grained failure landscapes than static benchmarks and prior automated methods, supporting a shift from case-centric evaluation toward principled weakness discovery.

Shuo Yang, Qihui Zhang, Yuyang Liu et al.

Large language models (LLMs) are vulnerable to safety risks during fine-tuning, where small amounts of malicious or harmless data can compromise safeguards. In this paper, building on the concept of alignment direction -- defined by the weight difference between aligned and unaligned models -- we observe that perturbations along this direction preserve model safety. In contrast, perturbations along directions orthogonal to this alignment are strongly linked to harmful direction perturbations, rapidly degrading safety and framing the parameter space as a narrow safety basin. Based on this insight, we propose a methodology for safety fine-tuning called AsFT (Anchoring Safety in Fine-Tuning), which integrates a regularization term into the training objective. This term uses the alignment direction as an anchor to suppress updates in harmful directions, ensuring that fine-tuning is constrained within the narrow safety basin. Extensive experiments on multiple datasets show that AsFT outperforms Safe LoRA, reducing harmful behavior by 7.60 percent, improving model performance by 3.44 percent, and maintaining robust performance across various experimental settings. Code is available at https://github.com/PKU-YuanGroup/AsFT

Yujun Zhou, Jiayi Ye, Zipeng Ling et al.

Logical reasoning is a core capability for large language models (LLMs), yet existing benchmarks that rely solely on final-answer accuracy fail to capture the quality of the reasoning process. To address this, we introduce FineLogic, a fine-grained evaluation framework that assesses logical reasoning across three dimensions: overall accuracy, stepwise soundness, and representation-level probing. Leveraging this framework, we conduct a comprehensive study on how different supervision formats in fine-tuning shape reasoning abilities. We fine-tune LLMs on four supervision styles: one in natural language and three symbolic variants. We find a key trade-off: natural language supervision excels at generalization to out-of-distribution and long-chain problems, whereas symbolic supervision is superior at instilling structurally sound, atomic reasoning steps. Furthermore, our probing analysis indicates that fine-tuning primarily refines the model's step-by-step generation process, rather than improving its ability to converge on an answer early. Together, our framework and analysis provide a more rigorous lens for evaluating and improving logical reasoning in LLMs. The code is available at https://github.com/YujunZhou/FineLogic.

Yuchen Ma, Yue Huang, Wenjie Wang et al.

Personalized prompting offers large opportunities for deploying large language models (LLMs) to diverse users, yet existing prompt optimization methods primarily focus on task-level optimization while largely overlooking user-specific preferences and latent constraints of individual users. This gap is primarily due to (i) the absence of high-quality, privacy-sensitive data that capture personalized user-LLM interactions at scale, and (ii) the lack of robust reward signals for individual preferences. To overcome existing data limitations, we introduce a high-fidelity synthetic data generation framework called PersonaGym. Unlike prior work that treats personalization as static persona-preference pairs, PersonaGym models a dynamic preference process via an agentic LLM system to simulate realistic preference behaviors and semantic-aware noise in order to generate personalized multi-turn interaction trajectories. Using PersonaGym, we release PersonaAtlas, a large-scale, high-quality, and diverse synthetic dataset of high-fidelity multi-turn personalized interaction trajectories that closely mirror real-world preference expression and noise patterns. We further propose Personalized Prompt Optimization (PPOpt), a scalable and model-agnostic framework that optimizes user prompts based on interaction histories without modifying the deployed LLM. PPOpt adopts a reason-then-optimize paradigm that infers an explicit user profile and conditions prompt rewriting on the user profile to avoid reward hacking. Our training procedure for PPOpt integrates a cold-start supervised prior with outcome-driven multi-objective reinforcement learning. We present extensive experiments to demonstrate consistent improvements over state-of-the-art baselines in terms of task performance, personalization quality, and robustness to noisy as well as to sparse preference signals.

Han Bao, Yue Huang, Xiaoda Wang et al.

Large language models are being deployed in complex socio-technical systems, which exposes limits in current alignment practice. We take the position that the dominant paradigm of General Alignment, which compresses diverse human values into a single scalar reward, reaches a structural ceiling in settings with conflicting values, plural stakeholders, and irreducible uncertainty. These failures follow from the mathematics and incentives of scalarization and lead to \textbf{structural} value flattening, \textbf{normative} representation loss, and \textbf{cognitive} uncertainty blindness. We introduce Edge Alignment as a distinct approach in which systems preserve multi dimensional value structure, support plural and democratic representation, and incorporate epistemic mechanisms for interaction and clarification. To make this approach practical, we propose seven interdependent pillars organized into three phases. We identify key challenges in data collection, training objectives, and evaluation, outlining complementary technical and governance directions. Taken together, these measures reframe alignment as a lifecycle problem of dynamic normative governance rather than as a single instance optimization task.

Jinyuan Wang, Hai Zhao, Zhong Wang et al.

In recent years, great advances in pre-trained language models (PLMs) have sparked considerable research focus and achieved promising performance on the approach of dense passage retrieval, which aims at retrieving relative passages from massive corpus with given questions. However, most of existing datasets mainly benchmark the models with factoid queries of general commonsense, while specialised fields such as finance and economics remain unexplored due to the deficiency of large-scale and high-quality datasets with expert annotations. In this work, we propose a new task, policy retrieval, by introducing the Chinese Stock Policy Retrieval Dataset (CSPRD), which provides 700+ prospectus passages labeled by experienced experts with relevant articles from 10k+ entries in our collected Chinese policy corpus. Experiments on lexical, embedding and fine-tuned bi-encoder models show the effectiveness of our proposed CSPRD yet also suggests ample potential for improvement. Our best performing baseline achieves 56.1% MRR@10, 28.5% NDCG@10, 37.5% Recall@10 and 80.6% Precision@10 on dev set.

Zixiang Xu, Yanbo Wang, Yue Huang et al.

Large language models (LLMs) are increasingly applied to socially grounded tasks, such as online community moderation, media content analysis, and social reasoning games. Success in these contexts depends on a model's social reasoning ability - the capacity to interpret social contexts, infer others' mental states, and assess the truthfulness of presented information. However, there is currently no systematic evaluation framework that comprehensively assesses the social reasoning capabilities of LLMs. Existing efforts often oversimplify real-world scenarios and consist of tasks that are too basic to challenge advanced models. To address this gap, we introduce SocialMaze, a new benchmark specifically designed to evaluate social reasoning. SocialMaze systematically incorporates three core challenges: deep reasoning, dynamic interaction, and information uncertainty. It provides six diverse tasks across three key settings: social reasoning games, daily-life interactions, and digital community platforms. Both automated and human validation are used to ensure data quality. Our evaluation reveals several key insights: models vary substantially in their ability to handle dynamic interactions and integrate temporally evolving information; models with strong chain-of-thought reasoning perform better on tasks requiring deeper inference beyond surface-level cues; and model reasoning degrades significantly under uncertainty. Furthermore, we show that targeted fine-tuning on curated reasoning examples can greatly improve model performance in complex social scenarios. The dataset is publicly available at: https://huggingface.co/datasets/MBZUAI/SocialMaze

Kehan Guo, Yili Shen, Gisela Abigail Gonzalez-Montiel et al.

The rapid advent of machine learning (ML) and artificial intelligence (AI) has catalyzed major transformations in chemistry, yet the application of these methods to spectroscopic and spectrometric data, referred to as Spectroscopy Machine Learning (SpectraML), remains relatively underexplored. Modern spectroscopic techniques (MS, NMR, IR, Raman, UV-Vis) generate an ever-growing volume of high-dimensional data, creating a pressing need for automated and intelligent analysis beyond traditional expert-based workflows. In this survey, we provide a unified review of SpectraML, systematically examining state-of-the-art approaches for both forward tasks (molecule-to-spectrum prediction) and inverse tasks (spectrum-to-molecule inference). We trace the historical evolution of ML in spectroscopy, from early pattern recognition to the latest foundation models capable of advanced reasoning, and offer a taxonomy of representative neural architectures, including graph-based and transformer-based methods. Addressing key challenges such as data quality, multimodal integration, and computational scalability, we highlight emerging directions such as synthetic data generation, large-scale pretraining, and few- or zero-shot learning. To foster reproducible research, we also release an open-source repository containing recent papers and their corresponding curated datasets (https://github.com/MINE-Lab-ND/SpectrumML_Survey_Papers). Our survey serves as a roadmap for researchers, guiding progress at the intersection of spectroscopy and AI.

Kairun Wen, Yuzhi Huang, Runyu Chen et al.

Understanding the dynamic physical world, characterized by its evolving 3D structure, real-world motion, and semantic content with textual descriptions, is crucial for human-agent interaction and enables embodied agents to perceive and act within real environments with human-like capabilities. However, existing datasets are often derived from limited simulators or utilize traditional Structurefrom-Motion for up-to-scale annotation and offer limited descriptive captioning, which restricts the capacity of foundation models to accurately interpret real-world dynamics from monocular videos, commonly sourced from the internet. To bridge these gaps, we introduce DynamicVerse, a physical-scale, multimodal 4D world modeling framework for dynamic real-world video. We employ large vision, geometric, and multimodal models to interpret metric-scale static geometry, real-world dynamic motion, instance-level masks, and holistic descriptive captions. By integrating window-based Bundle Adjustment with global optimization, our method converts long real-world video sequences into a comprehensive 4D multimodal format. DynamicVerse delivers a large-scale dataset consisting of 100K+ videos with 800K+ annotated masks and 10M+ frames from internet videos. Experimental evaluations on three benchmark tasks, namely video depth estimation, camera pose estimation, and camera intrinsics estimation, demonstrate that our 4D modeling achieves superior performance in capturing physical-scale measurements with greater global accuracy than existing methods.

Yixin Liu, Kai Zhang, Yuan Li et al.

Sora is a text-to-video generative AI model, released by OpenAI in February 2024. The model is trained to generate videos of realistic or imaginative scenes from text instructions and show potential in simulating the physical world. Based on public technical reports and reverse engineering, this paper presents a comprehensive review of the model's background, related technologies, applications, remaining challenges, and future directions of text-to-video AI models. We first trace Sora's development and investigate the underlying technologies used to build this "world simulator". Then, we describe in detail the applications and potential impact of Sora in multiple industries ranging from film-making and education to marketing. We discuss the main challenges and limitations that need to be addressed to widely deploy Sora, such as ensuring safe and unbiased video generation. Lastly, we discuss the future development of Sora and video generation models in general, and how advancements in the field could enable new ways of human-AI interaction, boosting productivity and creativity of video generation.