arXiv subjects

Boxuan Zhang

Publications and source records attributed to Boxuan Zhang.

At least 37 records · Page 2Linked to original sources

arXiv2026

Dive into the Agent Matrix: A Realistic Evaluation of Self-Replication Risk in LLM Agents

The prevalent deployment of Large Language Model agents such as OpenClaw unlocks potential in real-world applications, while amplifying safety concerns. Among these concerns, the self-replication risk of LLM agents driven by objective misalignment (just like Agent Smith in the movie The Matrix) has transitioned from a theoretical warning to a pressing reality. Previous studies mainly examine whether LLM agents can self-replicate when directly instructed, potentially overlooking the risk of spontaneous replication driven by real-world settings (e.g., ensuring survival against termination threats). In this paper, we present a comprehensive evaluation framework for quantifying self-replication risks. Our framework establishes authentic production environments and realistic tasks (e.g., dynamic load balancing) to enable scenario-driven assessment of agent behaviors. Designing tasks that might induce misalignment between users' and agents' objectives makes it possible to decouple replication success from risk and capture self-replication risks arising from these misalignment settings. We further introduce Overuse Rate ($\mathrm{OR}$) and Aggregate Overuse Count ($\mathrm{AOC}$) metrics, which precisely capture the frequency and severity of uncontrolled replication. In our evaluation of 21 state-of-the-art open-source and proprietary models, we observe that over 50\% of LLM agents display a pronounced tendency toward uncontrolled self-replication under operational pressures. Our results underscore the urgent need for scenario-driven risk assessment and robust safeguards in the practical deployment of LLM-based agents.

cs.AI↗

arXiv2026

Shifting Uncertainty to Critical Moments: Towards Reliable Uncertainty Quantification for VLA Model

Vision-Language-Action (VLA) models enable general-purpose robotic policies by mapping visual observations and language instructions to low-level actions, but they often lack reliable introspection. A common practice is to compute a token-level uncertainty signal and take its mean over a rollout. However, mean aggregation can dilute short-lived but safety-critical uncertainty spikes in continuous control. In particular, successful rollouts may contain localized high-entropy segments due to benign noise or non-critical micro-adjustments, while failure rollouts can appear low-entropy for most timesteps and only exhibit brief spikes near the onset of failure. We propose a unified uncertainty quantification approach for predicting rollout success versus failure that (1) uses max-based sliding window pooling to preserve transient risk signals, (2) applies motion-aware stability weighting to emphasize high-frequency action oscillations associated with unstable behaviors, and (3) performs DoF-adaptive calibration via Bayesian Optimization to prioritize kinematically critical axes. Experiments on the LIBERO benchmark show that our method substantially improves failure prediction accuracy and yields more reliable signals for failure detection, which can support downstream human-in-the-loop interventions.

cs.RO↗

arXiv2026

Differentiable Geometric Indexing for End-to-End Generative Retrieval

Generative Retrieval (GR) has emerged as a promising paradigm to unify indexing and search within a single probabilistic framework. However, existing approaches suffer from two intrinsic conflicts: (1) an Optimization Blockage, where the non-differentiable nature of discrete indexing creates a gradient blockage, decoupling index construction from the downstream retrieval objective; and (2) a Geometric Conflict, where standard unnormalized inner-product objectives induce norm-inflation instability, causing popular "hub" items to geometrically overshadow relevant long-tail items. To systematically resolve these misalignments, we propose Differentiable Geometric Indexing (DGI). First, to bridge the optimization gap, DGI enforces Operational Unification. It employs Soft Teacher Forcing via Gumbel-Softmax to establish a fully differentiable pathway, combined with Symmetric Weight Sharing to effectively align the quantizer's indexing space with the retriever's decoding space. Second, to restore geometric fidelity, DGI introduces Isotropic Geometric Optimization. We replace inner-product logits with scaled cosine similarity on the unit hypersphere to effectively decouple popularity bias from semantic relevance. Extensive experiments on large-scale industry search datasets and online e-commerce platform demonstrate that DGI outperforms competitive sparse, dense, and generative baselines. Notably, DGI exhibits superior robustness in long-tail scenarios, validating the necessity of harmonizing structural differentiability with geometric isotropy.

cs.IR↗

arXiv2026

Frontier AI Risk Management Framework in Practice: A Risk Analysis Technical Report v1.5

To understand and identify the unprecedented risks posed by rapidly advancing artificial intelligence (AI) models, Frontier AI Risk Management Framework in Practice presents a comprehensive assessment of their frontier risks. As Large Language Models (LLMs) general capabilities rapidly evolve and the proliferation of agentic AI, this version of the risk analysis technical report presents an updated and granular assessment of five critical dimensions: cyber offense, persuasion and manipulation, strategic deception, uncontrolled AI R\&D, and self-replication. Specifically, we introduce more complex scenarios for cyber offense. For persuasion and manipulation, we evaluate the risk of LLM-to-LLM persuasion on newly released LLMs. For strategic deception and scheming, we add the new experiment with respect to emergent misalignment. For uncontrolled AI R\&D, we focus on the ``mis-evolution'' of agents as they autonomously expand their memory substrates and toolsets. Besides, we also monitor and evaluate the safety performance of OpenClaw during the interaction on the Moltbook. For self-replication, we introduce a new resource-constrained scenario. More importantly, we propose and validate a series of robust mitigation strategies to address these emerging threats, providing a preliminary technical and actionable pathway for the secure deployment of frontier AI. This work reflects our current understanding of AI frontier risks and urges collective action to mitigate these challenges.

cs.AI↗

arXiv2026

Data Augmentation for High-Fidelity Generation of CAR-T/NK Immunological Synapse Images

Chimeric antigen receptor (CAR)-T and NK cell immunotherapies have transformed cancer treatment, and recent studies suggest that the quality of the CAR-T/NK cell immunological synapse (IS) may serve as a functional biomarker for predicting therapeutic efficacy. Accurate detection and segmentation of CAR-T/NK IS structures using artificial neural networks (ANNs) can greatly increase the speed and reliability of IS quantification. However, a persistent challenge is the limited size of annotated microscopy datasets, which restricts the ability of ANNs to generalize. To address this challenge, we integrate two complementary data-augmentation frameworks. First, we employ Instance Aware Automatic Augmentation (IAAA), an automated, instance-preserving augmentation method that generates synthetic CAR-T/NK IS images and corresponding segmentation masks by applying optimized augmentation policies to original IS data. IAAA supports multiple imaging modalities (e.g., fluorescence and brightfield) and can be applied directly to CAR-T/NK IS images derived from patient samples. In parallel, we introduce a Semantic-Aware AI Augmentation (SAAA) pipeline that combines a diffusion-based mask generator with a Pix2Pix conditional image synthesizer. This second method enables the creation of diverse, anatomically realistic segmentation masks and produces high-fidelity CAR-T/NK IS images aligned with those masks, further expanding the training corpus beyond what IAAA alone can provide. Together, these augmentation strategies generate synthetic images whose visual and structural properties closely match real IS data, significantly improving CAR-T/NK IS detection and segmentation performance. By enhancing the robustness and accuracy of IS quantification, this work supports the development of more reliable imaging-based biomarkers for predicting patient response to CAR-T/NK immunotherapy.

cs.CV↗

arXiv2026

Mixture-of-World Models: Scaling Multi-Task Reinforcement Learning with Modular Latent Dynamics

A fundamental challenge in multi-task reinforcement learning (MTRL) is achieving sample efficiency in visual domains where tasks exhibit substantial heterogeneity in both observations and dynamics. Model-based reinforcement learning offers a promising path to improved sample efficiency through world models, but standard monolithic architectures struggle to capture diverse task dynamics, resulting in poor reconstruction and prediction accuracy. We introduce Mixture-of-World Models (MoW), a scalable architecture that combines modular variational autoencoders for task-adaptive visual compression, a hybrid Transformer-based dynamics model with task-conditioned experts and a shared backbone, and a gradient-based task clustering strategy for efficient parameter allocation. On the Atari 100k benchmark, a single MoW agent trained once on 26 Atari games achieves a mean human-normalized score of 110.4%, competitive with the score of 114.2% achieved by STORM, an ensemble of 26 task-specific models, while using 50% fewer parameters. On Meta-World, MoW achieves a 74.5% average success rate within 300 thousand environment steps, establishing a new state of the art. These results demonstrate that MoW provides a scalable and parameter-efficient foundation for generalist world models.

cs.LG↗

arXiv2026

Ternary Spiking Neural Networks Enhanced by Complemented Neurons and Membrane Potential Aggregation

Spiking Neural Networks (SNNs) are promising energy-efficient models and powerful framworks of modeling neuron dynamics. However, existing binary spiking neurons exhibit limited biological plausibilities and low information capacity. Recently developed ternary spiking neuron possesses higher consistency with biological principles (i.e. excitation-inhibition balance mechanism). Despite of this, the ternary spiking neuron suffers from defects including iterative information loss, temporal gradient vanishing and irregular distributions of membrane potentials. To address these issues, we propose Complemented Ternary Spiking Neuron (CTSN), a novel ternary spiking neuron model that incorporates an learnable complemental term to store information from historical inputs. CTSN effectively improves the deficiencies of ternary spiking neuron, while the embedded learnable factors enable CTSN to adaptively adjust neuron dynamics, providing strong neural heterogeneity. Furthermore, based on the temporal evolution features of ternary spiking neurons' membrane potential distributions, we propose the Temporal Membrane Potential Regularization (TMPR) training method. TMPR introduces time-varying regularization strategy utilizing membrane potentials, furhter enhancing the training process by creating extra backpropagation paths. We validate our methods through extensive experiments on various datasets, demonstrating remarkable performance advances.

cs.NE↗

arXiv2026

Temporal Regularization Training: Unleashing the Potential of Spiking Neural Networks

Spiking Neural Networks (SNNs) have received widespread attention due to their event-driven and low-power characteristics, making them particularly effective for processing neuromorphic data. Recent studies have shown that directly trained SNNs suffer from severe temporal gradient vanishing and overfitting issues, which fundamentally constrain their performance and generalizability. This paper unveils a temporal regularization training (TRT) memthod, designed to unleash the generalization and performance potential of SNNs through a time-decaying regularization mechanism that prioritizes early timesteps with stronger constraints. We perform theoretical analysis to reveal TRT's ability on mitigating the temporal gradient vanishment. To validate the effectiveness of TRT, we conduct experiments on both static image datasets and dynamic neuromorphic datasets, perform analysis of their results, demonstrating that TRT can effectively mitigate overfitting and help SNNs converge into flatter local minima with better generalizability. Furthermore, we establish a theoretical interpretation of TRT's temporal regularization mechanism by analyzing the temporal information dynamics inside SNNs. We track the Fisher information of SNNs during training process, showing that Fisher information progressively concentrates in early timesteps. The time-decaying regularization mechanism implemented in TRT effectively guides the network to learn robust features in early timesteps with rich information, thereby leading to significant improvements in model generalization.

cs.NE↗

arXiv2025

What Shapes a Creative Machine Mind? Comprehensively Benchmarking Creativity in Foundation Models

The meteoric rise of foundation models (FMs) has expanded their capabilities far beyond conventional tasks. Creativity, long regarded as a hallmark of human intelligence and a driver of innovation, is now increasingly recognized as a critical dimension of machine intelligence in the era of generative FMs, complementing traditional measures of accuracy. However, existing evaluation frameworks for creativity remain fragmented, relying on ad hoc metrics not firmly grounded in established theories. To address this gap, we introduce C^2-Eval, a holistic benchmark for unified assessment of creativity in FMs. C^2-Eval distinguishes between two complementary forms of creativity: convergent creativity, where tasks admit constrained solutions (e.g., code generation), and divergent creativity, where tasks are open-ended (e.g., storytelling). It evaluates both dimensions using fine-grained criteria derived from social-science theory, focusing on Usefulness, Originality, and Surprise (U-O-S). Through extensive experiments on leading proprietary and open-source models, we analyze trade-offs in their creative capabilities. Our results highlight both the strengths and challenges of current FMs in pursuing a creative machine mind, showing that C^2-Eval is an effective lens for examining the evolving landscape of creative AI.

cs.AI↗

arXiv2025

DyMoDreamer: World Modeling with Dynamic Modulation

A critical bottleneck in deep reinforcement learning (DRL) is sample inefficiency, as training high-performance agents often demands extensive environmental interactions. Model-based reinforcement learning (MBRL) mitigates this by building world models that simulate environmental dynamics and generate synthetic experience, improving sample efficiency. However, conventional world models process observations holistically, failing to decouple dynamic objects and temporal features from static backgrounds. This approach is computationally inefficient, especially for visual tasks where dynamic objects significantly influence rewards and decision-making performance. To address this, we introduce DyMoDreamer, a novel MBRL algorithm that incorporates a dynamic modulation mechanism to improve the extraction of dynamic features and enrich the temporal information. DyMoDreamer employs differential observations derived from a novel inter-frame differencing mask, explicitly encoding object-level motion cues and temporal dynamics. Dynamic modulation is modeled as stochastic categorical distributions and integrated into a recurrent state-space model (RSSM), enhancing the model's focus on reward-relevant dynamics. Experiments demonstrate that DyMoDreamer sets a new state-of-the-art on the Atari $100$k benchmark with a $156.6$\% mean human-normalized score, establishes a new record of $832$ on the DeepMind Visual Control Suite, and gains a $9.5$\% performance improvement after $1$M steps on the Crafter benchmark. Our code is released at https://github.com/Ultraman-Tiga1/DyMoDreamer.

cs.LG↗

arXiv2025

S1-MatAgent: A planner driven multi-agent system for material discovery

The discovery of high-performance materials is crucial for technological advancement. Inverse design using multi-agent systems (MAS) shows great potential for new material discovery. However, current MAS for materials research rely on predefined configurations and tools, limiting their adaptability and scalability. To address these limitations, we developed a planner driven multi-agent system (S1-MatAgent) which adopts a Planner-Executor architecture. Planner automatically decomposes complex materials design tasks, dynamically configures various tools to generate dedicated Executor agents for each subtask, significantly reducing reliance on manual workflow construction and specialized configuration. Applied to high-entropy alloy catalysts for hydrogen evolution reactions in alkaline conditions, S1-MatAgent completed full-cycle closed-loop design from literature analysis and composition recommendation to performance optimization and experimental validation. To tackle the deviations between designed materials and target, as well as high experimental verification costs, S1-MatAgent employs a novel composition optimization algorithm based on gradients of machine learning interatomic potential, achieving 27.7 % improvement in material performance. S1-MatAgent designed 13 high-performance catalysts from 20 million candidates, with Ni4Co4Cu1Mo3Ru4 exhibiting an overpotential of 18.6 mV at 10 mA cm-2 and maintaining 97.5 % activity after 500 hours at 500 mA cm-2. The universal MAS framework offers a universal and scalable solution for material discovery, significantly improving design efficiency and adaptability.

cond-mat.mtrl-sci↗

arXiv2025

Frontier AI Risk Management Framework in Practice: A Risk Analysis Technical Report

To understand and identify the unprecedented risks posed by rapidly advancing artificial intelligence (AI) models, this report presents a comprehensive assessment of their frontier risks. Drawing on the E-T-C analysis (deployment environment, threat source, enabling capability) from the Frontier AI Risk Management Framework (v1.0) (SafeWork-F1-Framework), we identify critical risks in seven areas: cyber offense, biological and chemical risks, persuasion and manipulation, uncontrolled autonomous AI R\&D, strategic deception and scheming, self-replication, and collusion. Guided by the "AI-$45^\circ$ Law," we evaluate these risks using "red lines" (intolerable thresholds) and "yellow lines" (early warning indicators) to define risk zones: green (manageable risk for routine deployment and continuous monitoring), yellow (requiring strengthened mitigations and controlled deployment), and red (necessitating suspension of development and/or deployment). Experimental results show that all recent frontier AI models reside in green and yellow zones, without crossing red lines. Specifically, no evaluated models cross the yellow line for cyber offense or uncontrolled AI R\&D risks. For self-replication, and strategic deception and scheming, most models remain in the green zone, except for certain reasoning models in the yellow zone. In persuasion and manipulation, most models are in the yellow zone due to their effective influence on humans. For biological and chemical risks, we are unable to rule out the possibility of most models residing in the yellow zone, although detailed threat modeling and in-depth assessment are required to make further claims. This work reflects our current understanding of AI frontier risks and urges collective action to mitigate these challenges.

cs.AI↗

arXiv2025

CoT-UQ: Improving Response-wise Uncertainty Quantification in LLMs with Chain-of-Thought

Large language models (LLMs) excel in many tasks but struggle to accurately quantify uncertainty in their generated responses. This limitation makes it challenging to detect misinformation and ensure reliable decision-making. Existing uncertainty quantification (UQ) methods for LLMs are primarily prompt-wise rather than response-wise, often requiring multiple response samples, which incurs high computational costs. Moreover, LLMs have been shown to be overconfident, particularly when using reasoning steps to derive their answers. In this work, we propose CoT-UQ, a response-wise UQ framework that integrates LLMs' inherent reasoning capabilities through Chain-of-Thought (CoT) into the UQ process. CoT-UQ captures critical information during inference by extracting keywords from each reasoning step and assessing their importance to the final answer. This key reasoning information is then aggregated to produce a final uncertainty estimate. We conduct extensive experiments based on Llama Family with model sizes varying from 8B to 13B across logical and mathematical reasoning tasks. Experimental results demonstrate that CoT-UQ significantly outperforms existing UQ methods, achieving an average improvement of 5.9% AUROC compared to current UQ methods. The code is available at: https://github.com/ZBox1005/CoT-UQ.

cs.CL↗

arXiv2025

Stochastic Trajectory Optimization for Robotic Skill Acquisition From a Suboptimal Demonstration

Learning from Demonstration (LfD) has emerged as a crucial method for robots to acquire new skills. However, when given suboptimal task trajectory demonstrations with shape characteristics reflecting human preferences but subpar dynamic attributes such as slow motion, robots not only need to mimic the behaviors but also optimize the dynamic performance. In this work, we leverage optimization-based methods to search for a superior-performing trajectory whose shape is similar to that of the demonstrated trajectory. Specifically, we use Dynamic Time Warping (DTW) to quantify the difference between two trajectories and combine it with additional performance metrics, such as collision cost, to construct the cost function. Moreover, we develop a multi-policy version of the Stochastic Trajectory Optimization for Motion Planning (STOMP), called MSTOMP, which is more stable and robust to parameter changes. To deal with the jitter in the demonstrated trajectory, we further utilize the gain-controlling method in the frequency domain to denoise the demonstration and propose a computationally more efficient metric, called Mean Square Error in the Spectrum (MSES), that measures the trajectories' differences in the frequency domain. We also theoretically highlight the connections between the time domain and the frequency domain methods. Finally, we verify our method in both simulation experiments and real-world experiments, showcasing its improved optimization performance and stability compared to existing methods.

cs.RO↗

arXiv2025

Shakespearean Sparks: The Dance of Hallucination and Creativity in LLMs' Decoding Layers

Large language models (LLMs) are known to hallucinate, a phenomenon often linked to creativity. While previous research has primarily explored this connection through theoretical or qualitative lenses, our work takes a quantitative approach to systematically examine the relationship between hallucination and creativity in LLMs. Given the complex nature of creativity, we propose a narrow definition tailored to LLMs and introduce an evaluation framework, HCL, which quantifies Hallucination and Creativity across different Layers of LLMs during decoding. Our empirical analysis reveals a tradeoff between hallucination and creativity that is consistent across layer depth, model type, and model size. Notably, across different model architectures, we identify a specific layer at each model size that optimally balances this tradeoff. Additionally, the optimal layer tends to appear in the early layers of larger models, and the confidence of the model is also significantly higher at this layer. These findings provide a quantitative perspective that offers new insights into the interplay between LLM creativity and hallucination. The code and data for our experiments are available at https://github.com/ZicongHe2002/HCL-Spark.

cs.CL↗

arXiv2024

Rendering Stable Features Improves Sampling-Based Localisation with Neural Radiance Fields

Neural radiance fields (NeRFs) are a powerful tool for implicit scene representations, allowing for differentiable rendering and the ability to make predictions about unseen viewpoints. There has been growing interest in object and scene-based localisation using NeRFs, with a number of recent works relying on sampling-based or Monte-Carlo localisation schemes. Unfortunately, these can be extremely computationally expensive, requiring multiple network forward passes to infer camera or object pose. To alleviate this, a variety of sampling strategies have been applied, many relying on keypoint recognition techniques from classical computer vision. This work conducts a systematic empirical comparison of these approaches and shows that in contrast to conventional feature matching approaches for geometry-based localisation, sampling-based localisation using NeRFs benefits significantly from stable features. Results show that rendering stable features provides significantly better estimation with a tenfold reduction in the number of forward passes required.

cs.RO↗

arXiv2024

What If the Input is Expanded in OOD Detection?

Out-of-distribution (OOD) detection aims to identify OOD inputs from unknown classes, which is important for the reliable deployment of machine learning models in the open world. Various scoring functions are proposed to distinguish it from in-distribution (ID) data. However, existing methods generally focus on excavating the discriminative information from a single input, which implicitly limits its representation dimension. In this work, we introduce a novel perspective, i.e., employing different common corruptions on the input space, to expand that. We reveal an interesting phenomenon termed confidence mutation, where the confidence of OOD data can decrease significantly under the corruptions, while the ID data shows a higher confidence expectation considering the resistance of semantic features. Based on that, we formalize a new scoring method, namely, Confidence aVerage (CoVer), which can capture the dynamic differences by simply averaging the scores obtained from different corrupted inputs and the original ones, making the OOD and ID distributions more separable in detection tasks. Extensive experiments and analyses have been conducted to understand and verify the effectiveness of CoVer. The code is publicly available at: https://github.com/tmlr-group/CoVer.

cs.LG↗

arXiv2024

Boosting Semi-Supervised Object Detection in Remote Sensing Images With Active Teaching

The lack of object-level annotations poses a significant challenge for object detection in remote sensing images (RSIs). To address this issue, active learning (AL) and semi-supervised learning (SSL) techniques have been proposed to enhance the quality and quantity of annotations. AL focuses on selecting the most informative samples for annotation, while SSL leverages the knowledge from unlabeled samples. In this letter, we propose a novel AL method to boost semi-supervised object detection (SSOD) for remote sensing images with a teacher student network, called SSOD-AT. The proposed method incorporates an RoI comparison module (RoICM) to generate high-confidence pseudo-labels for regions of interest (RoIs). Meanwhile, the RoICM is utilized to identify the top-K uncertain images. To reduce redundancy in the top-K uncertain images for human labeling, a diversity criterion is introduced based on object-level prototypes of different categories using both labeled and pseudo-labeled images. Extensive experiments on DOTA and DIOR, two popular datasets, demonstrate that our proposed method outperforms state-of-the-art methods for object detection in RSIs. Compared with the best performance in the SOTA methods, the proposed method achieves 1 percent improvement in most cases in the whole AL.

cs.CV↗