Showing 1–50 of 1,171 results for author: Sun, P

Hierarchical Data Selection via Manifold Coverage and Sparse Feature Coverage in LLM Post-training

Authors: Peng Sun, Yi Yang, Antong Zhang, Chunxiao Li, Yanbo Wang, Dianbo Liu, xin chen, Kai Yu, Lu Chen, Tianfan Fu

Abstract: As supervised fine-tuning data continues to scale, selecting high-value subsets from large candidate pools is crucial for reducing training cost and improving model performance. Existing methods often measure diversity directly in the original embedding space, where geometric metrics entangle dominant semantic directions, fine-grained supervision differences, and local noise. We address this limit… ▽ More As supervised fine-tuning data continues to scale, selecting high-value subsets from large candidate pools is crucial for reducing training cost and improving model performance. Existing methods often measure diversity directly in the original embedding space, where geometric metrics entangle dominant semantic directions, fine-grained supervision differences, and local noise. We address this limitation by formulating data selection as a coarse-to-fine hierarchical coverage problem and propose MASS. MASS learns low-dimensional principal manifold coordinates with a dense autoencoder for coarse semantic grouping, and then performs quality-aware sparse feature coverage within each group using a TopK sparse autoencoder. Experiments on Vision Flan and LLaVA-CoT show that MASS consistently outperforms strong data selection baselines across multiple budgets, and in several settings matches or surpasses full data training with only a small subset of data. △ Less

Submitted 5 August, 2026; originally announced August 2026.

Data-DPO: Direct Preference Optimization for Target Model Data Selection in LLM Post-Training

Authors: Peng Sun, Yi Yang, Antong Zhang, Chunxiao Li, Yanbo Wang, Dianbo Liu, xin chen, Kai Yu, Lu Chen, Tianfan Fu

Abstract: Data selection in supervised fine-tuning aims to select a small set of effective samples from large-scale candidate data, reducing training cost while preserving model performance. However, existing methods usually treat data value as a relatively static property, and pay limited attention to the compatibility between data and the capability distribution of the target model. To address this issue,… ▽ More Data selection in supervised fine-tuning aims to select a small set of effective samples from large-scale candidate data, reducing training cost while preserving model performance. However, existing methods usually treat data value as a relatively static property, and pay limited attention to the compatibility between data and the capability distribution of the target model. To address this issue, we propose Data-DPO, a target model-oriented SFT data selection method. Data-DPO observes the local training feedback of the target model on different samples through one-step probing, transforms activation differences among samples into pairwise data preferences, and trains a lightweight reward model to learn target-model-aware data preferences. In the final selection stage, Data-DPO further combines target model preference, external quality scores, and marginal diversity to construct a more stable and effective training subset. Experimental results on Vision-Flan and LLaVA-CoT show that Data-DPO consistently outperforms existing data selection baselines under multiple data budgets and stably surpasses full data training performance. △ Less

Submitted 5 August, 2026; originally announced August 2026.

Carnot Meets Quantum Information: Thermal Machine Driven by Probabilistic Non-orthogonal State Discrimination

Authors: Tan-Ji Zhou, Yun-Qian Lin, Yu-Han Ma, C. P. Sun

Abstract: While the impossibility of perfectly identifying non-orthogonal states is a cornerstone of quantum information science, their probabilistic discrimination is nonetheless permissible. Here, we propose a two-reservoir quantum machine driven by this mechanism to map its functional boundaries across the parameter space of the state overlap $μ$ and the Carnot efficiency $η_C$. Within this $η_C$-$μ$ pla… ▽ More While the impossibility of perfectly identifying non-orthogonal states is a cornerstone of quantum information science, their probabilistic discrimination is nonetheless permissible. Here, we propose a two-reservoir quantum machine driven by this mechanism to map its functional boundaries across the parameter space of the state overlap $μ$ and the Carnot efficiency $η_C$. Within this $η_C$-$μ$ plane, the machine exhibits phase-transition-like functional switching among a pure heat-engine phase, a mixed phase, and a dissipative phase. We identify critical thresholds governing these transitions: strong thermal driving ($η_C \ge 0.5$) unconditionally guarantees positive work extraction, whereas weak driving ($η_C \lesssim 0.13$) induces an anomalous reentrant transition, where increasing $μ$ unexpectedly restores engine functionality after a purely dissipative regime. Our results explicitly demonstrate how quantum mechanics and thermodynamics jointly constrain information-to-energy conversion. △ Less

Submitted 17 August, 2026; originally announced August 2026.

Comments: 6 pages, 3 figures

Belayer: Efficient Fault Tolerance for LLM Agentic RL Training

Authors: Jiecheng Zhou, Qinghao Hu, Peng Sun, Xingcheng Zhang, Weiming Zhang

Abstract: Large language model (LLM) agents are increasingly trained with reinforcement learning in long-horizon, sandboxed environments. Unlike conventional RL, agentic RL couples GPU-intensive rollout engines with stateful environment containers whose actions may produce visible side effects, such as file edits, command execution, and dependency installation. A single trajectory can span many rounds of ge… ▽ More Large language model (LLM) agents are increasingly trained with reinforcement learning in long-horizon, sandboxed environments. Unlike conventional RL, agentic RL couples GPU-intensive rollout engines with stateful environment containers whose actions may produce visible side effects, such as file edits, command execution, and dependency installation. A single trajectory can span many rounds of gen- eration and environment interaction, so a component failure can discard completed work or expose the model to an environment state that is inconsistent with its context. However, existing systems lack efficient and correct recovery mechanisms for this distributed execution model. This paper presents Belayer, an efficient fault-tolerant system for LLM agentic RL training. Belayer handles failures in both rollout engines and environment execution while targeting low failure-free overhead. For scoped worker-local rollout failures, Belayer equips each pre-initialized shadow worker with a selective GPU-state reuse protocol that retains independently owned weights and raw KV-arena allocations after owner and GPU health checks, reinitializes worker-local state, and rebuilds request-specific KV contents from logged token prefixes. For environment failures, Belayer introduces full checkpoint and full restore to jointly capture and restore container file-system and runtime state, and coordinates the recovered environment with the LLM context to preserve prefix consistency. An adaptive policy opportunistically overlaps full-state checkpointing with natural LLM inference bubbles when the predicted interval is long enough. Empirical results show low measured overhead during failure-free training, a worker-recovery-time reduction of up to 42 times faster compared with a full engine cold start, and 1.5 to 3.5 times faster recovery from environment failures. △ Less

Submitted 17 August, 2026; v1 submitted 27 July, 2026; originally announced August 2026.

Evolve Vision-Language-Action Model into an Agent with On-the-fly Tool-use

Authors: Yi Ding, Yanzhao Yu, Xili Dai, Xianbiao Qi, Peiwen Sun, Xueqian Wang, Xiangyu Yue, Jianan Wang

Abstract: This paper integrates end-to-end Visual-Language-Action (VLA) models with agentic tool-use to propose Agentic Robot with Tool-use (ART). ART is a tool-injection framework that tunes any VLA model to leverage off-the-shelf tool modules for low-level vision, high-level affordance, and embodiment enhancement. Compared to vanilla VLA models with a whole continuous action solution space, ART reduces th… ▽ More This paper integrates end-to-end Visual-Language-Action (VLA) models with agentic tool-use to propose Agentic Robot with Tool-use (ART). ART is a tool-injection framework that tunes any VLA model to leverage off-the-shelf tool modules for low-level vision, high-level affordance, and embodiment enhancement. Compared to vanilla VLA models with a whole continuous action solution space, ART reduces the complexity of the action solution space through tool-use, which not only improves generalizability across different tasks but also reduces data dependency. To demonstrate the advantages (high generalizability and low data dependency) of this framework, we first built a dataset of 30K tool-use trajectories and action demonstrations, which is much smaller than those used by baseline methods. We then designed a training regimen for long-trajectory tool-use reasoning in challenging environments. Experiments show that ART achieves a 20% higher success rate than mainstream baselines on simulation and real-world tasks, such as pick-and-place in the dark at novel viewpoints. Empirical results highlight the benefits of an agent-based approach: modular tool utilization enables more efficient training, lightweight deployment, and scalable integration of new tools. This design fosters robustness, adaptability, and extensibility, paving the way for the practical deployment of VLA systems in complex real-world scenarios. △ Less

Submitted 14 August, 2026; originally announced August 2026.

Comments: 12 pages, 4 figures, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern (CVPR) Findings

Journal ref: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern (CVPR) Findings, 2026, pp. 1346-1357

Bridging Severe Cross-Modal Misalignment: End-to-End Visible-Infrared Object Detection via Explicit Feature-Domain Affine Registration

Authors: Qi Ming, Yuyang Wang, Mingjing Zhao, Yifan Xiao, Zhixin Guo, Zhiqiang Zhou, Peng Sun, Juan Fang, Fuqiang Yang, Xudong Zhao

Abstract: Visible-infrared object detection relies on complementary RGB and thermal cues, but its performance is often degraded by cross-modal spatial misalignment. Most existing methods rely on implicit feature adaptation to handle weakly misaligned scenarios, while large-offset geometric discrepancies remain insufficiently addressed. In this paper, we propose a Joint Feature-domain Registration and Detect… ▽ More Visible-infrared object detection relies on complementary RGB and thermal cues, but its performance is often degraded by cross-modal spatial misalignment. Most existing methods rely on implicit feature adaptation to handle weakly misaligned scenarios, while large-offset geometric discrepancies remain insufficiently addressed. In this paper, we propose a Joint Feature-domain Registration and Detection network (JFRDet), an end-to-end visible-infrared oriented object detector tailored for severely cross-modal geometric discrepancies. JFRDet introduces a Cross-Modal Affine Alignment (CMAA) module to estimate an image-level affine transformation for explicit multi-level feature alignment. Note that illumination changes directly affect the reliability of RGB cues, an Illumination-Guided Complementary Fusion (IGCF) module adaptively exploits modality reliability under varying illumination conditions for cross-modal fusion. Then, an Alignment Quality-Consistency Gating (AQCG) strategy stabilizes joint optimization by modulating detection supervision according to alignment reliability and gradient consistency. We further construct DroneVehicle Misaligned (DVMA), a benchmark for evaluating visible-infrared oriented object detection under severe cross-modal geometric misalignment. The proposed JFRDet achieves 69.7\% $\mathrm{mAP}_{50}$ on DVMA, which represents state-of-the-art (SOTA) performance. The code and dataset will be available on GitHub. △ Less

Submitted 11 August, 2026; originally announced August 2026.

Reliability-Safety Trade-off in AI Distillation: A Renormalization-Group Approach

Authors: Y. M. Du, Miao-Miao Yi, Tan-Ji Zhou, C. P. Sun

Abstract: Knowledge distillation transfers more than task competence: it also transmits response propensities, refusal policies, error boundaries, and latent safety biases. We formulate this behavioral inheritance as a coarse-graining model grounded in statistical mechanics, in which the student's answer and refusal decisions define two macrostates, while the teacher induces an effective field that reshapes… ▽ More Knowledge distillation transfers more than task competence: it also transmits response propensities, refusal policies, error boundaries, and latent safety biases. We formulate this behavioral inheritance as a coarse-graining model grounded in statistical mechanics, in which the student's answer and refusal decisions define two macrostates, while the teacher induces an effective field that reshapes the student's free-energy landscape. The model yields a reliability-safety trade-off relation controlled by a single parameter K, which we term the hazard discrimination capability. The predicted trade-off is consistent with refusal-token data [arXiv: 2412.06748]. In knowledge distillation, a teacher with strong hazard discrimination improves the student's attainable reliability and safety, whereas poor discrimination limits the attainable trade-off. Repeated distillation acts as an iterated renormalization-group-like transformation, under which K follows a flow across generations. The flow exhibits a tricritical structure separating regimes of K loss, stable transmission, and threshold-dependent inheritance, and yields testable scaling predictions for multigenerational distillation. △ Less

Submitted 9 August, 2026; originally announced August 2026.

MetaVideoAgent: Automated Video-Agent Evolution for Long-Form Video Understanding

Authors: Benlei Cui, Ruize Wang, Junjie Li, Jinhao Chen, Longtao Huang, Yinghao Chen, Yuwen Zhai, Jingqun Tang, Ruijian Jia, Weiwei Wu, Pengfei Sun, Haiwen Hong

Abstract: Long-form video understanding requires locating sparse, question-relevant evidence in long, multimodal videos. Real-world video distributions differ in modality-specific information density, content structure, and evidence patterns, causing fixed video-agent designs to incur redundant processing or fail when mismatched. Extending automated agent evolution from text to video is challenging because… ▽ More Long-form video understanding requires locating sparse, question-relevant evidence in long, multimodal videos. Real-world video distributions differ in modality-specific information density, content structure, and evidence patterns, causing fixed video-agent designs to incur redundant processing or fail when mismatched. Extending automated agent evolution from text to video is challenging because full long-video execution makes candidate validation expensive, failures propagate across coupled evidence-processing stages, and complex preprocessing, perception tools, and localization strategies make code-level updates difficult to implement reliably. We introduce MetaVideoAgent, a framework that automatically evolves a video agent for a target distribution. It profiles information density and evidence requirements from sparsely sampled frames and associated queries to guide initial design, then compresses localized failures into independently executable minimal validation tasks. It constructs evidence-grounded Gold Paths, audits Student trajectories, aggregates recurring failures across samples, and attributes them to responsible modules. A modular agent representation constrains each update to the primary responsible module and its necessary dependencies. We further introduce VA-EvoBench, covering eight video distributions with separate evolution and held-out splits. With four evolution iterations per distribution, MetaVideoAgent improves every initial agent and raises macro-average accuracy from 38.44% to 51.47%, at an average evolution cost of 3.54M tokens per distribution. The evolved agents outperform the strongest prior fixed-design video agent by 6.39 percentage points while using the fewest tokens and video frames per question among the compared video agents. We will release all code and data to support reproducible research. △ Less

Submitted 5 August, 2026; originally announced August 2026.

Comments: 16 pages, 7 figures. Code: https://github.com/Alibaba-VELLDEPTH/MetaVideoAgent

Latent Reward Registers for Diffusion Preference Alignment

Authors: Yuanshen Guan, Zipeng Feng, Chengru Song, Zhiwei Xiong, Peiqin Sun

Abstract: Aligning diffusion models with human preferences usually relies on a sparse terminal reward evaluated on the final generated samples, which creates a severe temporal credit-assignment problem across the denoising process. We propose Latent Reward Registers, a mechanism that estimates terminal preference directly from intermediate noisy latents. Learnable, position-free register tokens are appended… ▽ More Aligning diffusion models with human preferences usually relies on a sparse terminal reward evaluated on the final generated samples, which creates a severe temporal credit-assignment problem across the denoising process. We propose Latent Reward Registers, a mechanism that estimates terminal preference directly from intermediate noisy latents. Learnable, position-free register tokens are appended as an auxiliary read path to a frozen Diffusion Transformer (DiT), extracting preference signals without altering the generator's hidden states or velocity field. The resulting dense, differentiable reward field spans the full denoising trajectory and supports two alignment strategies. For training, Reward-Gradient On-Policy Distillation (RG-OPD) converts this dense reward field into per-step targets at states visited by the current generator, replacing rollout-intensive policy gradients with direct on-policy distillation. For inference, Reward-Guided Sampling (RGS) steers trajectories with magnitude-matched reward-gradient corrections and no parameter updates. Empirically, at high noise levels (t=0.8) the registers reach the highest pairwise accuracy among the evaluated latent reward models. RG-OPD outperforms online reinforcement learning baselines while reducing GPU hours by up to 33x. RGS achieves significant reward improvement with a favorable reward-quality balance against training-free baselines. Code and weights are to be available at https://github.com/Guanys-dar/latent-reward-register △ Less

Submitted 14 August, 2026; v1 submitted 4 August, 2026; originally announced August 2026.

CMuon: Accelerating and Stabilizing Diffusion Transformer Training via Chunked Momentum Orthogonalization

Authors: Chuyan Chen, Peng Sun, Kun Yuan

Abstract: Diffusion Transformers (DiTs) have achieved state-of-the-art (SOTA) performance in visual generative modeling, yet their training remains computationally prohibitive. While the recently proposed Momentum Orthogonalization (Muon) optimizer offers a promising alternative to AdamW, its direct application to DiTs yields suboptimal late-stage convergence. In this paper, we identify the root cause of th… ▽ More Diffusion Transformers (DiTs) have achieved state-of-the-art (SOTA) performance in visual generative modeling, yet their training remains computationally prohibitive. While the recently proposed Momentum Orthogonalization (Muon) optimizer offers a promising alternative to AdamW, its direct application to DiTs yields suboptimal late-stage convergence. In this paper, we identify the root cause of this bottleneck: standard DiT architectures fuse functionally distinct weights (e.g., within AdaLN and QKV layers) into unified tensors for computational efficiency. Applying Muon to these fused tensors inadvertently induces implicit subspace coupling, which distorts update directions and degrades global optimization. To address this, we introduce Chunked Muon (CMuon), a simple yet highly effective strategy that partitions these matrices into independent sub-components prior to orthogonalization. Extensive experiments demonstrate that a 675M-parameter DiT trained with CMuon achieves a FID of 1.18 on ImageNet 256 in just 200 epochs. This represents more than a 2x training speedup over AdamW, while effectively overcoming the late-stage convergence plateaus of vanilla Muon. △ Less

Submitted 3 August, 2026; originally announced August 2026.

Comments: ECCV 2026

Dynamic Resolution Routing for Efficient Egocentric Grounding

Authors: Huixin Sun, Wangbo Zhao, Fanyue Wei, Qiuxia Lin, Pengzhan Sun, Angela Yao

Abstract: Egocentric visual grounding requires high-resolution inputs to localize small objects. However, scaling Multimodal Large Language Models to this domain is constrained by the excessive cost of visual token processing. We identify that current efficient strategies based on token reduction are unreliable for selecting object-centric spatial evidence. To overcome this, we propose SmartRes, a framework… ▽ More Egocentric visual grounding requires high-resolution inputs to localize small objects. However, scaling Multimodal Large Language Models to this domain is constrained by the excessive cost of visual token processing. We identify that current efficient strategies based on token reduction are unreliable for selecting object-centric spatial evidence. To overcome this, we propose SmartRes, a framework that performs efficiency optimization in the pixel space via dynamic resolution routing. SmartRes first encodes a low-resolution view for global context and uses a lightweight router to activate high-resolution patches in object-centric regions and constructs an order-preserving visual sequence. To further enable robust routing under severe foreground-background imbalance, we introduce a margin-regularized routing objective that increases foreground-background logit separation and improves foreground recall. Experiments on Ego4D and EgoIntention show that SmartRes reduces visual tokens by up to 67% while retaining 86.4% of full-resolution performance, and achieves up to 1.66X faster inference than state-of-the-art token reduction methods with higher accuracy. Furthermore, strong performance on small object grounding indicates the effectiveness of SmartRes towards egocentric applications. Code will be publicly available. △ Less

Submitted 2 August, 2026; originally announced August 2026.

RayViT: Ray-Conditioned Visual Representations for Viewpoint-Robust Imitation Learning

Authors: Qian Wang, Longrui Chen, Peiran Sun, Aleksandar Taranovic, Niklas Freymuth, Ge Li, Weiran Liao, C. F. Maximilian Nagy, Yucheng Tan, Tao Chen, Gerhard Neumann

Abstract: Visual imitation learning enables robots to acquire visuomotor skills directly from images, yet RGB observations lack explicit geometric cues, making learned policies brittle to camera perturbations. To address this, we propose \textbf{Ray-conditioned Vision Transformer Encoder (RayViT)}, a lightweight architecture that injects camera geometry into pretrained ViT backbones. RayViT represents camer… ▽ More Visual imitation learning enables robots to acquire visuomotor skills directly from images, yet RGB observations lack explicit geometric cues, making learned policies brittle to camera perturbations. To address this, we propose \textbf{Ray-conditioned Vision Transformer Encoder (RayViT)}, a lightweight architecture that injects camera geometry into pretrained ViT backbones. RayViT represents camera geometry as a Plücker ray map, patchifies it into ray features, and uses gated cross-attention to produce a ray-conditioned class token. These ray features are added as dense positional embeddings, while the ray class token replaces the original ViT class token to provide a geometry-aware summary representation. We combine this approach with an auxiliary cosine similarity loss to consistently improve the performance and robustness for geometry-aware tokens. Experiments on sim- and real-robot tasks demonstrate that RayViT improves robustness by approximately 13 percentage points under camera perturbations in multi-task RoboCasa benchmark and by 1.78 average completed stages in real-world multi-task success rate compared to baselines. △ Less

Submitted 31 July, 2026; originally announced July 2026.

Pion Transition Form Factor in Lattice QCD

Authors: Shihao Su, Liuming Liu, Peng Sun

Abstract: We investigate the neutral pion transition form factor $F_{π^0γ^\astγ^\ast}(q_1^2,q_2^2)$ in lattice QCD and confirm that the connected and disconnected contributions have the same sign. We employ the recently proposed blending method, which supplies an unbiased and cheap estimators for the required all-to-all propagators. The external pion states are treated within the distillation framework, whi… ▽ More We investigate the neutral pion transition form factor $F_{π^0γ^\astγ^\ast}(q_1^2,q_2^2)$ in lattice QCD and confirm that the connected and disconnected contributions have the same sign. We employ the recently proposed blending method, which supplies an unbiased and cheap estimators for the required all-to-all propagators. The external pion states are treated within the distillation framework, while the electromagnetic currents are evaluated in the full blending space. Numerical tests are performed on an $N_f=2+1$ lattice ensemble. Our result shows that the contribution of the disconnected part is approximately $1\%$ of that of the connected part and enables constructive interference of probability amplitudes. △ Less

Submitted 30 July, 2026; originally announced July 2026.

CheckVLA: Execution-Time Verification with Action-Conditioned World Model for Long-Horizon Mobile Manipulation

Authors: Yushan Liu, Peibo Sun, Xintao Chao, Zhenyang Yang, Yifan Xie, Lingfeng Zhang, Shoujie Li, Chenyu Tang, Fang Chen, Xiao-Ping Zhang, Wenbo Ding

Abstract: Vision-language-action (VLA) policies commonly execute long-horizon mobile manipulation through open-loop action chunks, issuing multiple actions without receiving new high-level visual input. A committed chunk therefore implies how observations should evolve, but accidental deviations can violate this expectation while the remaining actions continue to propagate the error: commit-time policy conf… ▽ More Vision-language-action (VLA) policies commonly execute long-horizon mobile manipulation through open-loop action chunks, issuing multiple actions without receiving new high-level visual input. A committed chunk therefore implies how observations should evolve, but accidental deviations can violate this expectation while the remaining actions continue to propagate the error: commit-time policy confidence cannot react to a deviation that occurs after dispatch, and observation-only anomaly scores lack an action-conditioned reference for separating expected effects from unexplained changes. We propose CheckVLA, which verifies execution with a separately trained, frozen action-conditioned world model. A conformally calibrated risk threshold bounds the episode-level probability of an unnecessary first intervention and determines when to intervene, its exceedance controls how strongly the rewritten suffix retains the superseded chunk, latency-aware hard prefixing restricts replacement to actions that remain deployable, and an event-driven keyframe bank preserves evidence of prior progress across repairs. On RoboCasa365, under a common training recipe and a matched invocation budget, CheckVLA attains a 36.1% average success rate against 27.6% for periodic replanning (+8.5 points). At a matched 5% episode-level false-alarm target, action conditioning raises timely recall to 77.9%, against 48.6% for an observation-only control and 37.9% for an action-shuffled control. These simulation results support action-conditioned verification as a way to restore feedback during chunked execution while keeping the repair consistent with inference latency. △ Less

Submitted 29 July, 2026; originally announced July 2026.

Quantum advantage of nonlinear quantum battery and superconducting circuit implementation

Authors: Wei-Jun Han, Peng-Yu Sun, Guo-Feng Zhang

Abstract: A quantum battery is a novel energy storage device that operates on the principles of quantum mechanics. To enhance the charging performance of quantum batteries and further provide theoretical support for their physical implementation, we constructed an optical-field-dependent nonlinear quantum battery model. Meanwhile, we solved for the unbiased form of nonlinear interactions in this model, wher… ▽ More A quantum battery is a novel energy storage device that operates on the principles of quantum mechanics. To enhance the charging performance of quantum batteries and further provide theoretical support for their physical implementation, we constructed an optical-field-dependent nonlinear quantum battery model. Meanwhile, we solved for the unbiased form of nonlinear interactions in this model, where the charging power of the proposed model exhibits a superlinear quantum advantage, and the charging time saturates the quantum speed limit. Through theoretical analysis, we confirm that this quantum advantage arises from the quantum effect of multiphoton absorption. Subsequently, with the derived nonlinear function form, we further investigated other properties of this nonlinear quantum battery. Finally, an experimental design scheme for this nonlinear quantum battery in superconducting quantum circuits is presented. △ Less

Submitted 23 July, 2026; originally announced July 2026.

Comments: 26 pages, 14 figures

Journal ref: Frontiers of Physics,21,6,063203 (2026)

Memory-Augmented Multimodal Large Language Models for Small Object Understanding in Streaming Aerial Videos

Authors: Penglei Sun, Yehua Huang, Zhuoli Tao, Xiang Li, Runwei Guan, Yaoxian Song, Kaiyong Zhao, Henghui Ding, Bo Han, Yang Yang, Xiaowen Chu

Abstract: Language-guided aerial perception aims to understand user-specified tiny targets in complex unmanned aerial vehicle (UAV) scenes. In real UAV deployment, the UAV must respond while it flies, so such perception runs in an online streaming manner, where frames arrive sequentially and the model responds to each one without access to future frames. However, applying current Multimodal Large Language M… ▽ More Language-guided aerial perception aims to understand user-specified tiny targets in complex unmanned aerial vehicle (UAV) scenes. In real UAV deployment, the UAV must respond while it flies, so such perception runs in an online streaming manner, where frames arrive sequentially and the model responds to each one without access to future frames. However, applying current Multimodal Large Language Models (MLLMs) to this setting raises two challenges. First, targets viewed from the air are often tiny, yet the visual compression in existing MLLMs treats all regions equally and discards their fine-grained details. Second, understanding a continuous stream requires past-frame context, yet retaining the entire history is infeasible on resource-constrained onboard hardware, whereas discarding it causes the target to drift or disappear. We address the tiny object and streaming challenges from both data and method perspectives. From the data perspective, we present \textbf{DroneEyes}, the \textbf{first} pixel-level and open-vocabulary referring-segmentation dataset for tiny aerial targets, comprising $2,140$ high-definition videos and $176,623$ pairs across Object Description and Referring Expression tasks, with dense per-frame masks. From the method perspective, we propose \textbf{SkyAnchor}, an MLLM with two designs to the above challenges: a Semantics-Aware Token Router that preserves small-target under a reduced visual-token budget, and a Hierarchical Memory Bank that keeps the target consistently understood on streams. △ Less

Submitted 22 July, 2026; originally announced July 2026.

Three-Body Scattering for Generative Modeling

Authors: Peng Sun, Zhenglin Cheng, Deyuan Liu, Jun Xie, Xinyi Shang, Tao Lin

Abstract: Modern generative models typically rely on an adversarial critic, a prescribed noise-to-data path, or an autoregressive factorization. Instead, we show that a proper distributional energy can induce sample-level motion and provide direct regression supervision for a one-step generator. Three-Body Scattering Modeling (TBSM) for generation turns the energy distance into a constant-size per-projectil… ▽ More Modern generative models typically rely on an adversarial critic, a prescribed noise-to-data path, or an autoregressive factorization. Instead, we show that a proper distributional energy can induce sample-level motion and provide direct regression supervision for a one-step generator. Three-Body Scattering Modeling (TBSM) for generation turns the energy distance into a constant-size per-projectile interaction: each projectile is attracted toward one real source and repelled from one independently generated source. Conditioned on the projectile and its condition, its expectation equals the $2$-Wasserstein gradient-flow velocity of $\frac12D_E^2(P_θ,Q)$. A batch of $B$ frozen-target events yields $O(B)$ sample-level losses, each using one reference for its condition instead of the minibatch-wide all-pairs field used by methods such as Drifting Models. Tracking this conditional expectation online can reduce field noise. Using scattering in frozen image features, TBSM trains one-step generators on ImageNet-256, achieving FID${}=2.23$ with pixel-space PixelDiT-XL and FID${}=1.63$ with latent-space DiT-XL at NFE${}=1$. We provide a design map relating diffusion-related supervision, Drift-like dynamics, and GAN-like objectives. These results establish tracked scattering as a route to high-dimensional one-step generation. Code: https://github.com/sp12138/TBSM. △ Less

Submitted 20 July, 2026; originally announced July 2026.

Comments: 31 pages, 5 figures, and 4 tables. Code: https://github.com/sp12138/TBSM

Adverse Selection with Quality Variance: A Maximum-Entropy Approach

Authors: Zhi-Lei Zhang, Tan-Ji Zhou, C. P. Sun

Abstract: The adverse-selection mechanism in markets explains how asymmetric information between buyers and sellers can drive high-quality goods out of the market, thereby causing market deterioration. In its simplest formulation, only the mean quality is used to describe the market, and this is insufficient to determine how fast the market deteriorates or how the quality distribution evolves. To resolve th… ▽ More The adverse-selection mechanism in markets explains how asymmetric information between buyers and sellers can drive high-quality goods out of the market, thereby causing market deterioration. In its simplest formulation, only the mean quality is used to describe the market, and this is insufficient to determine how fast the market deteriorates or how the quality distribution evolves. To resolve the two problems, we describe the adverse selection as a dynamic truncation of the quality distribution: buyers set an upper bound proportional to the mean quality by a rate $ξ$ that is larger than unity, and sellers whose quality exceeds this upper bound reject an offer and exit the market. The retained market is then characterized by the conditional distribution obtained after this truncation, and the corresponding evolution process is iterated until market quality reaches a stable state. This statistical approach gives three results. (i) We identify a mechanism for preventing complete adverse selection, defined as the process where the quality of the market is driven down to the minimum quality floor. (ii) A larger quality variance or a smaller price premium, defined as the amount by which the payment upper bound exceeds the current mean quality, raises the upper bound on the deterioration in mean quality. (iii) A maximum-entropy benchmark shows numerically how quality variance and the payment rate jointly determine market deterioration and the final stable quality platform. This approach also clarifies how market interventions can slow adverse selection: they may raise buyers' payment rate, reduce quality variance, or increase the minimum quality floor. △ Less

Submitted 19 July, 2026; originally announced July 2026.

Robust Betatron-Tune Measurement from Schottky Spectra: Complementary Classical and Deep-Learning Paradigms

Authors: Peihan Sun, Manzhou Zhang, Renxian Yuan, Deming Li, Jian Dong

Abstract: Schottky spectra provide key beam diagnostics, with betatron sidebands encoding the fractional tune. Reliable tune measurement is particularly important for third-order resonance slow extraction in compact medical proton synchrotrons, where low signal-to-noise ratios and limited frequency resolution can compromise conventional peak-detection and curve-fitting methods. This work develops two comple… ▽ More Schottky spectra provide key beam diagnostics, with betatron sidebands encoding the fractional tune. Reliable tune measurement is particularly important for third-order resonance slow extraction in compact medical proton synchrotrons, where low signal-to-noise ratios and limited frequency resolution can compromise conventional peak-detection and curve-fitting methods. This work develops two complementary tune estimators with a shared spectral front-end but different temporal representations. The classical estimator coherently pools motion-compensated spectra, detects the sideband using a multi-width matched-filter bank, and performs sub-bin estimation through local argmax and an adaptive MAD-gated centroid. The deep-learning estimator converts each spectrum into a tune-likelihood map using a convolutional neural network with FFT-based global convolutions, then propagates the posterior with a discrete two-dimensional (q,v) Bayesian tracker under a Gaussian motion model while also reporting posterior uncertainty. On a synthetic dynamic-tune benchmark, the deep-learning estimator outperforms published baselines across the operating range, while the classical estimator exceeds the latency-compensated baseline and requires neither training data nor GPU acceleration. On near-stationary SAPT beam data, both methods operate end-to-end, with the deep-learning model requiring no retraining. Median per-frame latency remains below 1 ms on commodity hardware, supporting real-time-capable tune measurement in compact medical synchrotrons. △ Less

Submitted 20 July, 2026; v1 submitted 15 July, 2026; originally announced July 2026.

Giant magnetocaloric effect at low fields in triangular-lattice NdMgAl$_{11}$O$_{19}$

Authors: Yantao Cao, He Sun, Zhendong Fu, Zhaoming Tian, Huiqian Luo, Junsen Xiang, Peijie Sun, Jinkui Zhao, Hanjie Guo

Abstract: Magnetic refrigeration in the sub-Kelvin regime requires refrigerant materials to retain a large magnetic entropy at low temperatures by suppressing magnetic ordering. Quantum spin liquids (QSLs), which evade long-range magnetic ordering while retaining strong quantum fluctuations to the lowest temperatures, therefore provide a promising platform for realizing high-performance magnetic refrigerant… ▽ More Magnetic refrigeration in the sub-Kelvin regime requires refrigerant materials to retain a large magnetic entropy at low temperatures by suppressing magnetic ordering. Quantum spin liquids (QSLs), which evade long-range magnetic ordering while retaining strong quantum fluctuations to the lowest temperatures, therefore provide a promising platform for realizing high-performance magnetic refrigerants. Here, we investigate the magnetic ground state and the magnetocaloric effect of the hexaaluminate, NdMgAl$_{11}$O$_{19}$, in which the Nd$^{3+}$ ions form a network of triangular lattices. Magnetic susceptibility and specific heat measurements indicate a magnetically dynamic state down to 50~mK, consistent with a QSL state. Specific heat measurements further reveal substantial magnetic entropy retained below 50~mK. Quasi-adiabatic demagnetization measurements demonstrate a superior cooling performance of NdMgAl$_{11}$O$_{19}$, which can be cooled to 113~mK from 1.9~K by only a small magnetic field change of 2~T. The outstanding refrigeration performance is attributed to the persistent spin fluctuations associated with the QSL-like ground state, together with a large effective \textit{g} factor and the smallness of the exchange interactions along the easy-axis direction. This study demonstrates that frustration, combined with strong spin-orbit coupling and crystal-electric-field effect in the rare earth magnets provides a promising design principle for next-generation cryogenic magnetic refrigerants. △ Less

Submitted 13 July, 2026; originally announced July 2026.

Comments: 8 pages, 4 figures, 1 table

Condensing Large-Scale Datasets Directly with Minimal Information Loss

Authors: Xinyi Shang, Peng Sun, Bei Shi, Zixuan Wang, Tao Lin

Abstract: Recent advancements in scaling dataset distillation rely heavily on decoupled information extraction pipelines, comprising SQUEEZE, RECOVER, and RELABEL stages. Despite their scalability to large-scale datasets, these methods suffer from prohibitive computational overhead and poor cross-architecture generalization. In this paper, we reveal the root cause of these bottlenecks: the implicit dual-com… ▽ More Recent advancements in scaling dataset distillation rely heavily on decoupled information extraction pipelines, comprising SQUEEZE, RECOVER, and RELABEL stages. Despite their scalability to large-scale datasets, these methods suffer from prohibitive computational overhead and poor cross-architecture generalization. In this paper, we reveal the root cause of these bottlenecks: the implicit dual-compression process, from data to model and back to images, inherently induces severe information loss. Crucially, we empirically and theoretically demonstrate that this loss creates a distribution shift that fundamentally compromises the widely adopted RELABEL strategy, transforming the pre-trained model into an unreliable labeler that yields sub-optimal labels. To overcome these critical flaws, we propose CIM, a novel, metric-driven framework that abandons the flawed dual-compression paradigm. Instead, CIM explicitly quantifies and minimizes the information gap between the original and synthetic datasets. By directly aligning the data distributions, our approach ensures high-fidelity information condensation and inherently satisfies the prerequisites for effective relabeling. Extensive experiments demonstrate that CIM establishes a new state-of-the-art. Notably, it distills ImageNet-1K at an IPC=10 in merely 80 minutes on a single RTX-4090 GPU, achieving an unprecedented 48.7% Top-1 accuracy on ResNet-18 and significantly outperforming previous SOTA approaches, such as NRR-DD and DELT, by 2.6% and 2.9%, respectively. Our code is available at https://github.com/LINs-lab/CIM. △ Less

Submitted 1 July, 2026; originally announced July 2026.

Comments: Accepted by ECCV 2026

Yuvion LLM: An Adversarially-Aware Large Language Model for Content And AI Safety

Authors: Ting Ma, Xiufeng Huang, Benlei Cui, Xiaowen Xu, Shikai Qiu, Ruijie Jian, Hongxing Li, Guanghui Wang, Longtao Huang, Haiwen Hong, Haolei Xu, Wenjing Jiang, Ziwen Xu, Zhaoyu Fan, Shaoxuan He, Chuxi Xiao, Yujian Li, Xinyue Chen, Chunyang Chai, Wenxuan Liu, Ziheng Wang, Dongjie Zhang, Yangfan Zhou, Libin Dong, Yupeng Cao , et al. (21 additional authors not shown)

Abstract: As large language models are increasingly deployed in real-world systems, safety failures can still lead to harmful outputs and dangerous misuse. We argue that the essence of safety is adversarial: many failures arise not from natural inputs alone, but from strategic attempts to evade model policies and safeguards. However, existing general-purpose model development largely overlook this adversari… ▽ More As large language models are increasingly deployed in real-world systems, safety failures can still lead to harmful outputs and dangerous misuse. We argue that the essence of safety is adversarial: many failures arise not from natural inputs alone, but from strategic attempts to evade model policies and safeguards. However, existing general-purpose model development largely overlook this adversarial nature, and often remain insufficient for realistic safety scenarios involving planning, tool use, and multi-step reasoning, causing measured safety performance to overestimate real deployment robustness. To address this gap, we present Yuvion LLM, a large language model built for adversarially robust content safety and broader AI safety. Yuvion LLM treats adversarial robustness and agentic capability as first-class objectives. Its pipeline combines adversarially aware data construction, knowledge-enhanced continued pretraining, and policy-grounded multi-task safety post-training, including risk-aware supervised fine-tuning and reinforcement learning-based policy optimization, together with safety-aware agentic reinforcement learning for tool use and multi-step reasoning in complex safety scenarios. We further introduce the Yuvion LLM RiskEval (YLRE), a collection of 93 benchmarks across four evaluation categories, covering diverse open and internal evaluations with a focus on safety, adversarial robustness, and real-world capability requirements. Across these evaluations, Yuvion LLM demonstrates clear advantages on safety-focused benchmarks and particularly strong robustness under adversarial conditions, while maintaining solid overall capability. Notably, Yuvion-8B outperforms most state-of-the-art baselines, including substantially larger models such as GPT-5.4 and Qwen3-MAX, on several safety tasks. △ Less

Submitted 25 June, 2026; originally announced June 2026.

Yuvion VL: A Multimodal Foundation Model for Adversarial Content and AI Safety

Authors: Shikai Qiu, Xiaowen Xu, Benlei Cui, Ting Ma, Xiufeng Huang, Wenjing Jiang, Shaoxuan He, Haolei Xu, Chunyang Chai, Yujian Li, Yiliang Zhang, Guanghui Wang, Ziheng Wang, Ziwen Xu, Zhaoyu Fan, Jinhao Chen, Ruijie Jian, Hongxing Li, Chuxi Xiao, Xinyue Chen, Wenxuan Liu, Libin Dong, Yupeng Cao, Xiaoqian Xia, Jing Wang , et al. (33 additional authors not shown)

Abstract: General-purpose models often struggle to reliably identify and understand real-world multimodal risks, largely due to the inherent multimodal adversarial nature of content and AI safety. We present Yuvion VL, a family of multimodal large language models purpose-built for content and AI safety, with both instruction-tuned and reasoning-oriented variants. Yuvion VL addresses this gap by treating saf… ▽ More General-purpose models often struggle to reliably identify and understand real-world multimodal risks, largely due to the inherent multimodal adversarial nature of content and AI safety. We present Yuvion VL, a family of multimodal large language models purpose-built for content and AI safety, with both instruction-tuned and reasoning-oriented variants. Yuvion VL addresses this gap by treating safety as an inherently adversarial and multimodal problem and designing the entire pipeline around adversarial robustness. For data construction, we develop an automated pipeline integrating adversarial-aware data synthesis with multi-stage quality control, producing large-scale, high-quality multimodal samples augmented with domain knowledge and reasoning annotations. For training, we adopt a three-stage pipeline that includes continued pretraining for risk-concept cross-modal alignment, instruct post-training for production-grade safety tasks, and reasoning post-training for enhanced interpretability and performance in complex tasks. We further introduce Confuse-then-Contrast Fine-Tuning, a contrastive framework that mines model-specific confusions and constructs multi-image contrastive groups to enforce explicit discrimination of fine-grained visual-semantic elements, enabling the model to distinguish between visually similar cases with different safety implications in adversarial safety tasks. To support rigorous evaluation, we further introduce Yuvion VL RiskEval (YVRE), a collection of benchmarks covering diverse open and internal evaluations, with a focus on content and AI safety, adversarial robustness, and real-world capability requirements. Experiments show that Yuvion VL-32B achieves industry-leading safety performance, surpassing comparably sized open-source models and best closed-source commercial models, while maintaining comparable general capabilities. △ Less

Submitted 26 June, 2026; v1 submitted 23 June, 2026; originally announced June 2026.

Self-Supervised Mask-Aware Transformers for Fault-Tolerant FBG Force Sensing in Minimally Invasive Surgical Robotics

Authors: Peibo Sun, Shiyuan Dong, Shucheng Ye, Jianrong Cai, Yushan Liu, Hongen Liao, Tianqi Huang, Fang Chen

Abstract: In minimally invasive surgical robotics, catheter-scale Fiber Bragg Grating (FBG) sensors are promising due to their ability to estimate multi-dimensional forces by multiplexing several optical channels. However, deploying these compact multi-channel sensors introduces two critical engineering challenges: inherent nonlinear cross-axis coupling during complex deformations, and intermittent channel… ▽ More In minimally invasive surgical robotics, catheter-scale Fiber Bragg Grating (FBG) sensors are promising due to their ability to estimate multi-dimensional forces by multiplexing several optical channels. However, deploying these compact multi-channel sensors introduces two critical engineering challenges: inherent nonlinear cross-axis coupling during complex deformations, and intermittent channel dropouts caused by fiber fractures in constrained workspaces. These compounding issues severely degrade force estimation. Existing fault-tolerant approaches rely on combinatorial model banks, which scale exponentially with the channel count and demand prohibitively expensive per-pattern calibration. In this paper, we propose a unified, self-supervised mask-aware Transformer that explicitly models channel availability to enable graceful degradation under diverse and dynamic sensor failures. The encoder is pretrained via masked-channel reconstruction on unlabeled data streams and fine-tuned for force regression using a balanced clean-and-corrupted-view objective alongside a dynamic corruption curriculum. Furthermore, a parallel uncertainty head, trained via heteroscedastic Gaussian negative log-likelihood, predicts per-axis confidence in a single forward pass, circumventing the overhead of multi-pass ensembles. Evaluated on a catheter-scale 8-channel FBG dataset, our single unified model achieves a nominal Root Mean Square Error (RMSE) of 0.0066~N and degrades gracefully to 0.0126~N under severe 4-channel failures. This significantly outperforms a comprehensive model bank of 255 per-pattern neural networks (0.0154~N at 4-channel loss) while eliminating pattern-specific calibration. △ Less

Submitted 16 June, 2026; originally announced June 2026.

Curvature-Informed Potential Energy Surface for Protein-Ligand Binding Affinity Prediction

Authors: Peng-Fei Sun, Chuan-Xian Ren, Hong Yan

Abstract: Accurate prediction of protein-ligand binding affinity is essential for structure-based drug discovery. Recent geometric deep learning methods have achieved promising performance by representing protein-ligand complexes as three-dimensional graphs. However, most existing approaches mainly rely on static interaction geometry from a single bound conformation, while neglecting molecular flexibility a… ▽ More Accurate prediction of protein-ligand binding affinity is essential for structure-based drug discovery. Recent geometric deep learning methods have achieved promising performance by representing protein-ligand complexes as three-dimensional graphs. However, most existing approaches mainly rely on static interaction geometry from a single bound conformation, while neglecting molecular flexibility and binding-induced conformational changes. To address this limitation, we propose a curvature-informed potential energy surface (CPES) graph neural network for protein-ligand binding affinity prediction, which incorporates physics-informed curvature representations to model conformational flexibility. CPES first derives curvature spectral descriptors from the Hessian of the potential energy surface evaluated at equilibrium configurations, whose eigenvalues define the local principal curvatures of the potential energy surface. It then uses spectral cross-attention to compare the unbound ligand and protein with the bound complex, thereby capturing binding-induced changes in conformational dynamics. In parallel, hierarchical protein-ligand interaction representations are learned from static structural features through geometry-aware message passing, soft clustering, and bidirectional cross-attention. Finally, CPES fuses the curvature-informed dynamic representations with static interaction representations for affinity regression. Extensive evaluations on multiple benchmark datasets demonstrate that CPES achieves improved predictive performance and offers physical interpretability. △ Less

Submitted 12 June, 2026; originally announced June 2026.

Boosting Direct Preference Optimization with Penalization

Authors: Pengwei Sun

Abstract: Offline preference optimization has become a practical substitute for reinforcement learning from human feedback, but pairwise objectives such as Direct Preference Optimization (DPO) and its variants use only the chosen and rejected responses stored in a static dataset. This leaves a useful signal unused: the response that the reference model itself would generate for the same prompt. We propose D… ▽ More Offline preference optimization has become a practical substitute for reinforcement learning from human feedback, but pairwise objectives such as Direct Preference Optimization (DPO) and its variants use only the chosen and rejected responses stored in a static dataset. This leaves a useful signal unused: the response that the reference model itself would generate for the same prompt. We propose Direct Preference Optimization with Penalization (DPOP), a simple extension of DPO that augments the base preference loss with a gated penalty on reference-greedy responses. DPOP activates this penalty only when the current policy still assigns a lower likelihood to the preferred response than to the rejected response. On AlpacaEval 2.0, DPOP improves length-controlled win rate over DPO, SimPO, and AlphaDPO on both Llama-3-8b-it and Gemma-2-9b-it, achieving relative gains of 5.3\% and 4.4\% over baselines on the two models, respectively. Ablations further show that a SimNPO-style length-normalized penalty is stronger than NPO and token-level unlikelihood in this setting. △ Less

Submitted 10 June, 2026; originally announced June 2026.

Comments: Accepted at ICML 2026 Workshop on Decision-Making from Offline Datasets to Online Adaptation: Black-Box Optimization to Reinforcement Learning

Which Speech Representation Better Matches Text-Native Reasoning? A Study of Speech-Text Alignment on Frame Rate and Representation

Authors: Zhen Ye, Xu Tan, Yiming Li, Guangyan Zhang, Chimin Chan, Haohe Liu, Zhengxi Liu, Hongzhan Lin, Zheqi Dai, Xinshen Zhang, Peiwen Sun, Qiuqiang Kong, Wei Xue

Abstract: Spoken dialogue models typically start from text LLM backbones, yet reasoning often degrades when conditioning on speech instead of text. We attribute part of this modality gap to a temporal-granularity mismatch: speech tokens are temporally redundant and far longer than text under matched semantics, diluting per-token semantic density and weakening text-native reasoning dynamics. We study speech… ▽ More Spoken dialogue models typically start from text LLM backbones, yet reasoning often degrades when conditioning on speech instead of text. We attribute part of this modality gap to a temporal-granularity mismatch: speech tokens are temporally redundant and far longer than text under matched semantics, diluting per-token semantic density and weakening text-native reasoning dynamics. We study speech token design as a representation selection problem and sweep frame rates under a frozen LLM backbone with a fixed information rate. To make low frame rates feasible, we introduce factorized FSQ and a lightweight non-autoregressive audio LM head, scaling capacity to nearly 300\,bits/frame without sacrificing efficient prediction. With the bottleneck removed, we sweep frame rates (50$\rightarrow$2.08\,Hz) and alignment depth, and observe a consistent best regime for speech QA at 4.17\,Hz with intermediate-layer representation alignment. △ Less

Submitted 10 June, 2026; originally announced June 2026.

Comments: Accepted by Interspeech 2026 long paper

OmniGen-AR: AutoRegressive Any-to-Image Generation

Authors: Junke Wang, Xun Wang, Qiushan Guo, Peize Sun, Weilin Huang, Zuxuan Wu, Yu-Gang Jiang

Abstract: Autoregressive (AR) models have demonstrated strong potential in visual generation, offering superior performance with simple architectures and optimization objectives. However, existing methods are typically limited to single-modality conditions, e.g., text, restricting their applicability in real-world scenarios that demand image synthesis from diverse controls. In this work, we present OmniGen-… ▽ More Autoregressive (AR) models have demonstrated strong potential in visual generation, offering superior performance with simple architectures and optimization objectives. However, existing methods are typically limited to single-modality conditions, e.g., text, restricting their applicability in real-world scenarios that demand image synthesis from diverse controls. In this work, we present OmniGen-AR, a unified autoregressive framework for Any-to-Image generation. By discretizing various visual conditions through a shared visual tokenizer and text prompts with a text tokenizer, OmniGen-AR supports a broad spectrum of conditional inputs within a single model, including text (text-to-image generation), spatial signals (segmentation-to-image and depth-to-image), and visual context (image editing, frame prediction, and text-to-video generation). To mitigate the risk of information leakage from condition tokens to content tokens, we introduce Disentangled Causal Attention (DCA), which separates the full-sequence causal mask into condition causal attention and content causal attention. It serves as a training-time regularizer without affecting the standard next-token prediction during inference. With this design, OmniGen-AR achieves new state-of-the-art or at least competitive results across a range of benchmark, e.g., 0.63 on GenEval and 80.02 on VBench, demonstrating its effectiveness in flexible and high-fidelity visual generation. △ Less

Submitted 8 June, 2026; originally announced June 2026.

Comments: Accepted by NeurIPS

Hyperon-Nucleon Spectrometer

Authors: Xiaozhi Bai, Xu Cao, Zhe Cao, Jinhui Chen, Kai Chen, Qibo Chen, Shi Chen, Xin Chen, Yuquan Chen, Zhenyu Chen, Jianping Dai, Heng-Tong Ding, Dongshuo Du, Shuxian Du, Limin Duan, Zhe Duan, Anhui Feng, Jie Feng, Yicheng Feng, Jinlin Fu, Xiaofeng Fu, Chaosong Gao, Liang Ge, Wenwen Ge, Lisheng Geng , et al. (215 additional authors not shown)

Abstract: Chirality lies at the heart of low-energy QCD, governing the symmetry structure that shapes hadron masses and strong interaction dynamics. Among the most compelling open questions tied to chiral dynamics and spontaneous chiral symmetry breaking is the longstanding $Λ$ polarization puzzle, in which $Λ$ hyperons produced in unpolarized hadronic collisions exhibit a surprisingly large transverse pola… ▽ More Chirality lies at the heart of low-energy QCD, governing the symmetry structure that shapes hadron masses and strong interaction dynamics. Among the most compelling open questions tied to chiral dynamics and spontaneous chiral symmetry breaking is the longstanding $Λ$ polarization puzzle, in which $Λ$ hyperons produced in unpolarized hadronic collisions exhibit a surprisingly large transverse polarization that remains theoretically unexplained. This whitepaper presents the proposal for the Hyperon-Nucleon Spectrometer (H-NS) at the High-Intensity heavy-ion Accelerator Facility (HIAF). Leveraging the high energy and high intensity of HIAF's proton and heavy-ion beams, the H-NS experiment will perform systematic studies of hyperon polarization phenomena and their underlying mechanisms in proton-proton ($pp$), proton-nucleus ($pA$), and nucleus-nucleus ($AA$) collisions in the fixed target mode. A wide-range beam energy scan, including proton beams from 3 GeV up to 9.3 GeV (HIAF) and up to 32 GeV (upgraded HIAF), will be conducted to examine the dependence of polarization on collision energy. The spectrometer is designed with specialized detectors capable of high-precision reconstruction of final-state baryon polarizations. Among its many interesting and important measurements, H-NS will simultaneously measure hyperon and proton spin observables to explore the polarization mechanism in hadronic interactions and the spin structure of baryons. Furthermore, the use of $pA$ and $AA$ collisions will enable detailed investigations of cold and hot nuclear matter effects on spin polarization. Its physics program and detector development will significantly benefit the future Electron-ion Collider in China. △ Less

Submitted 4 June, 2026; originally announced June 2026.

Comments: 69 pages, Hyperon-Nucleon Spectrometer (H-NS) white paper

Benchmark Everything Everywhere All at Once

Authors: Shiyun Xiong, Dongming Wu, Peiwen Sun, Yuang Ai, Bokang Yang, Wencheng Han, Xiao-Hui Li, Xiangyu Yue

Abstract: Benchmarks are fundamental for evaluating and advancing LLMs and MLLMs by providing standardized and explicit measures of performance. However, their construction is labor-intensive and hard to reuse, raising concerns about sustainability and scalability. Moreover, existing benchmarks often quickly reach performance saturation after their release, resulting in insufficient discrimination among sta… ▽ More Benchmarks are fundamental for evaluating and advancing LLMs and MLLMs by providing standardized and explicit measures of performance. However, their construction is labor-intensive and hard to reuse, raising concerns about sustainability and scalability. Moreover, existing benchmarks often quickly reach performance saturation after their release, resulting in insufficient discrimination among state-of-the-art models. To address these challenges, we introduce Benchmark Agent, a fully autonomous agentic system designed for benchmark building. Our framework orchestrates the complete benchmark construction pipeline, from user query analysis and subtask design to data annotation and quality control. To assess Benchmark Agent, we implement it to produce 15 representative benchmarks, spanning diverse evaluation scenarios, including text understanding, multimodal understanding, and domain-specific reasoning. Extensive experiments, including human evaluation, LLM-as-a-judge assessment, and consistency checks, demonstrate Benchmark Agent can generate high-quality benchmark samples with minimal human involvement. More importantly, through continual evaluation, we observe several insightful findings, including that current models struggle with certain domain-specific reasoning tasks. We believe that rapidly evolving benchmarks can contribute significantly to the research community. The preview and code will be publicly available at the demo page and code repository. △ Less

Submitted 4 June, 2026; originally announced June 2026.

Comments: Project page: https://benchmarkagent.github.io/

LongSpace: Exploring Long-Horizon Spatial Memory from Perception to Recall in Video

Authors: Shiqiang Lang, Jing Liu, Haoyang He, Peiwen Sun, Yuanteng Chen, Tao Liu, Lan Yang, Longteng Guo, Honggang Zhang

Abstract: Multimodal Large Language Models (MLLMs) have advanced image and video understanding and can increasingly handle longer visual inputs. Long-horizon tasks such as autonomous driving and robotic navigation require more than recognizing the current view, as models must remember and retrieve previously observed spatial layouts, routes, viewpoint changes, and object states. To evaluate this capability,… ▽ More Multimodal Large Language Models (MLLMs) have advanced image and video understanding and can increasingly handle longer visual inputs. Long-horizon tasks such as autonomous driving and robotic navigation require more than recognizing the current view, as models must remember and retrieve previously observed spatial layouts, routes, viewpoint changes, and object states. To evaluate this capability, we introduce LongSpace-Bench, a room-tour video benchmark for long-horizon spatial memory, covering scene perception, spatial relations, and spatial memory. In this work, we further propose LongSpace, a memory framework for long-video spatial reasoning. LongSpace models long videos as sequential chunks, incorporates 3D structural cues into early decoder layers, and constructs layer-aware memory for question-guided retrieval. Experiments on multiple spatial reasoning benchmarks show that LongSpace improves long-video spatial understanding, further demonstrating explicit spatial memory as a key capability for long-horizon video MLLMs. △ Less

Submitted 4 June, 2026; originally announced June 2026.

X-Stream: Exploring MLLMs as Multiplexers for Multi-Stream Understanding

Authors: Peiwen Sun, Xudong Lu, Huadai Liu, Yang Bo, Dongming Wu, Huankang Guan, Minghong Cai, Jinpeng Chen, Xintong Guo, Shuhan Li, Fang Liu, Rui Liu, Xiangyu Yue

Abstract: While video streaming understanding has made significant strides, real-world applications, such as live sports broadcasting, autonomous driving, and multi-screen collaboration, inherently demand continuous, multi-stream interactions. However, existing benchmarks are confined to single-stream paradigms, leaving a critical gap in evaluating online, cross-stream reasoning. To bridge this, we introduc… ▽ More While video streaming understanding has made significant strides, real-world applications, such as live sports broadcasting, autonomous driving, and multi-screen collaboration, inherently demand continuous, multi-stream interactions. However, existing benchmarks are confined to single-stream paradigms, leaving a critical gap in evaluating online, cross-stream reasoning. To bridge this, we introduce X-Stream, the first benchmark dedicated to multi-stream streaming understanding. Comprising 4,220 rigorously curated QA pairs across 932 videos, X-Stream evaluates 11 subtasks across multi-window, multi-view, and multi-device scenarios. Crucially, our dataset is constructed using a novel dual-verification pipeline that prevents over-reliance on a single stream. Furthermore, we pioneer the conceptualization of multi-modal large language models (MLLMs) as naive multiplexers, systematically evaluating their performance through the lens of Signal Multiplexing Theory. Our extensive online inference experiments reveal a stark reality: state-of-the-art MLLMs struggle significantly with concurrent streams, achieving only about 50% score and exhibiting poor proactive ability. Ultimately, X-Stream exposes the trade-off of current multiplexing schemes, providing both a practical evaluation protocol and empirical guidance for next-generation multi-stream agents. △ Less

Submitted 29 June, 2026; v1 submitted 1 June, 2026; originally announced June 2026.

Comments: Project Page: https://peiwensun2000.github.io/xstream/

A 32-Channel 3.53-μW Per Channel Brain-Machine Interface SoC Featuring Dual-Threshold Delta-modulation, In-Memory Spike Detection and Bi-SNN Based Motor Decoding

Authors: Ye Ke, Zhengnan Fu, Pao-Sheng Vincent Sun, An Guo, Shuai Dong, Junyi Yang, Yahan Yang, Abdelrahman B. M. Eldaly, Xin Si, Leanne Chan, Arindam Basu

Abstract: With the scaling of sensor channel counts, systems confront challenges in frontend data sensing and on-implant data processing. This work presents a 32-channel fully event-based iBMI SoC in 65nm CMOS for an efficient neuromorphic signal processing pipeline. The SoC integrates a 32-channel dual-threshold delta modulation (DTDM) frontend array that provides up to 26x data compression at the frontend… ▽ More With the scaling of sensor channel counts, systems confront challenges in frontend data sensing and on-implant data processing. This work presents a 32-channel fully event-based iBMI SoC in 65nm CMOS for an efficient neuromorphic signal processing pipeline. The SoC integrates a 32-channel dual-threshold delta modulation (DTDM) frontend array that provides up to 26x data compression at the frontend, an in-memory computing (IMC) spike detector (SPD) for efficient in-pixel spike detection, and a bipolar LIF-based spiking neural network (Bi-SNN) decoder for on-chip motor intention decoding (MID). Consuming only 3.53 μW per channel and achieving ~0.62 decoding R2 with a compact 0.034 mm2 per-channel area, the chip enables high-efficiency signal recording, processing, and decoding for implantable devices. △ Less

Submitted 1 June, 2026; originally announced June 2026.

Crystallisation kinetics of supercooled liquid palladium

Authors: Zuzanna Kostera, Christian Bressler, Przemyslaw Dziegielewski, Wojciech Gawelda, Konstantinos Georgarakis, Dmitry Khakhulin, Oleksii I. Liubchenko, Adam Olczak, Angel Rodriguez-Fernandez, Ryszard Sobierajski, Klaus Sokolowski-Tinten, Peihao Sun, Robert W. E. van de Kruijs, Hazem Yousef, Peter Zalden, Jerzy Antonowicz

Abstract: In this study, we employ classical molecular dynamics (MD) simulations to investigate the crystallisation kinetics of supercooled liquid palladium and relate the results to time-resolved X-ray diffraction measurements on rapidly quenched Pd thin films. Crystal nucleation and growth rates are determined over the temperature range $700$--$1150~\mathrm{K}$ ($0.38$--$0.65 T_{\mathrm{m}}$) by analysing… ▽ More In this study, we employ classical molecular dynamics (MD) simulations to investigate the crystallisation kinetics of supercooled liquid palladium and relate the results to time-resolved X-ray diffraction measurements on rapidly quenched Pd thin films. Crystal nucleation and growth rates are determined over the temperature range $700$--$1150~\mathrm{K}$ ($0.38$--$0.65 T_{\mathrm{m}}$) by analysing the evolution of the microstructure during the liquid-to-crystal transition. The self-diffusion coefficient of Pd, obtained from the atomic mean-squared displacement, follows Arrhenius behaviour over the investigated temperature range, with an activation energy of $467(6)~\mathrm{meV/atom}$, consistent with available data for supercooled liquid metals. The steady-state homogeneous nucleation rate exhibits a maximum of approximately $4 \times 10^{35}~\mathrm{m^{-3} s^{-1}}$ near $0.5 T_{\mathrm{m}}$. Crystal growth occurs at velocities of the order of metres per second, with a temperature dependence consistent with diffusion-limited Wilson-Frenkel kinetics rather than the collision-limited regime. Based on multiple statistically independent simulations, a time-temperature-transformation (TTT) diagram for crystallisation onset is constructed. The TTT curve exhibits a nose near $0.5 T_{\mathrm{m}}$ and $100~\mathrm{ps}$, corresponding to a critical cooling rate for vitrification on the order of $10^{13}~\mathrm{K s^{-1}}.$ The simulations reproduce the crystallisation onset time and temperature observed in time-resolved X-ray diffraction experiments on optically molten Pd thin films quenched at $5 \times 10^{11}~\mathrm{K s^{-1}}.$ These results indicate that homogeneous, rather than heterogeneous, nucleation governs the achievable supercooling in the experimentally studied films. △ Less

Submitted 2 June, 2026; v1 submitted 29 May, 2026; originally announced May 2026.

Comments: 27 pages, 6 figures plus 4 pages of Supplementary Material

VIPER-MCP: Detecting and Exploiting Taint-Style Vulnerabilities in Model Context Protocol Servers

Authors: Pengyu Sun, Zifeng Kang, Qishu Jin, Enhao Huang, Xin Liu, Dakun Shen, Song Li

Abstract: Model Context Protocol (MCP) has emerged as a standard interface for connecting LLM agents to external tools. Because MCP servers expose privileged operations such as shell execution, network access, and file-system manipulation to agent-driven invocation, implementation flaws in tool handlers can create a direct path from natural-language input to security-sensitive sinks, potentially granting at… ▽ More Model Context Protocol (MCP) has emerged as a standard interface for connecting LLM agents to external tools. Because MCP servers expose privileged operations such as shell execution, network access, and file-system manipulation to agent-driven invocation, implementation flaws in tool handlers can create a direct path from natural-language input to security-sensitive sinks, potentially granting attackers remote code execution or full system compromise. Existing approaches either produce unconfirmed static alerts without dynamic validation, or rely on fixed template libraries that lack code-level guidance and fail to trigger vulnerabilities requiring specific parameter shapes or multi-step taint paths. In this paper, we present VIPER-MCP, the first end-to-end automated vulnerability auditing framework for MCP servers that not only detects taint-style vulnerabilities but also dynamically confirms their exploitability by producing concrete proof-of-concept prompts. VIPER-MCP introduces two novel techniques: (1) an anchor-query pass in a two-pass static analysis strategy that augments standard taint alerts with function-level structural context, resolving file-level static artifacts to specific MCP tool handlers and producing vulnerability-anchored call chains; and (2) a feedback-driven prompt evolution mechanism that employs dual-mutator scheduling that independently corrects tool-selection drift and deepens parameter penetration, together with fitness-scored seed selection to iteratively refine natural-language prompts toward vulnerable sinks. In a large-scale scan of 39,884 real-world open-source MCP server repositories, VIPER-MCP discovered 106 0-day vulnerabilities, all of which were confirmed through end-to-end exploit traces, with 67 CVE IDs assigned to date. △ Less

Submitted 12 August, 2026; v1 submitted 20 May, 2026; originally announced May 2026.

PlexRL: Cluster-Level Orchestration of Serviceized LLM Execution for RLVR

Authors: Yiqi Zhang, Fangzheng Jiao, Tian Tang, Boyu Tian, Hangyu Wang, Qiaoling Chen, Guoteng Wang, Zhen Jiang, Peng Sun, Ping Zhang, Xiaohe Hu, Ziming Liu, Menghao Zhang, Yanmin Jia, Yang You, Siyuan Feng

Abstract: Reinforcement learning with verifiable rewards (RLVR) has recently unlocked strong reasoning capabilities in large language models (LLMs), triggering rapid exploration of new algorithms and data. However, RLVR training is notoriously inefficient: long-tailed rollouts, tool-induced stalls, and asymmetric resource requirements between rollout and training introduce substantial idle time that cannot… ▽ More Reinforcement learning with verifiable rewards (RLVR) has recently unlocked strong reasoning capabilities in large language models (LLMs), triggering rapid exploration of new algorithms and data. However, RLVR training is notoriously inefficient: long-tailed rollouts, tool-induced stalls, and asymmetric resource requirements between rollout and training introduce substantial idle time that cannot be eliminated by job-local optimizations such as synchronous pipelining, asynchronous rollout, or colocated execution. We argue that this inefficiency is structural. While idle gaps are unavoidable within individual RLVR jobs, they are largely anti-correlated across jobs and therefore exploitable at the cluster level. Leveraging this observation, we present PlexRL, a cluster-level runtime for multiplexing unified LLM services across RLVR jobs. By centrally managing model placement, state transitions, and function-level scheduling under strict affinity constraints, PlexRL time-slices LLM execution across jobs to fill otherwise idle periods without expensive model migration. Our implementation and evaluations demonstrate that PlexRL significantly improves effective cluster capacity and reduces user GPU hour cost by maximum 37.58% while preserving algorithmic flexibility and introducing minimal per-job overhead. △ Less

Submitted 20 May, 2026; originally announced May 2026.

Two-nucleon systems at $m_π\approx292$ MeV from lattice QCD

Authors: Kuan Zhang, Kang Yu, Yiqi Geng, Chuan Liu, Liuming Liu, Peng Sun, Jia-Jun Wu, Ruilin Zhu

Abstract: Nucleon-nucleon systems in the $^3S_1$ and the $^1S_0$ channels are studied in lattice quantum chromodynamics at a pion mass of approximately $m_π\approx292$ MeV, employing three $N_f = 2+1$ ensembles with the same pion mass and lattice spacing $a=0.10530(18)$ fm but different spatial volumes. Finite-volume energies of the nucleon-nucleon systems are determined in both the rest frame and a moving… ▽ More Nucleon-nucleon systems in the $^3S_1$ and the $^1S_0$ channels are studied in lattice quantum chromodynamics at a pion mass of approximately $m_π\approx292$ MeV, employing three $N_f = 2+1$ ensembles with the same pion mass and lattice spacing $a=0.10530(18)$ fm but different spatial volumes. Finite-volume energies of the nucleon-nucleon systems are determined in both the rest frame and a moving frame. The distillation quark smearing method is applied to improve the precision and to ensure the symmetric correlators by using the same interpolating operators at sink and source. The scattering amplitudes are extracted from the finite-volume spectra using the Lüscher's finite-volume method. At the studied pion mass, both the $^3S_1$ (deuteron) and $^1S_0$(di-neutron) channels exhibit a virtual state pole, with binding energies of $6^{+5}_{-3}$ MeV and $11^{+6}_{-5}$ MeV, respectively. To investigate the effects of the left-hand cut, an alternative method -- the Non-Perturbative Hamiltonian framework (NPHF) -- is used for the scattering analysis and yields consistent results with those from the Lüscher method. △ Less

Submitted 22 July, 2026; v1 submitted 16 May, 2026; originally announced May 2026.

Journal ref: Phys. Rev. D 114, 034509 (2026)

Nonlinear Bipolar Compensation: Handling Outliers in Post-Training Quantization

Authors: Peilin Sun, Jianxin Wu

Abstract: Network quantization has emerged as one of the most practical model compression techniques, which significantly reduces a model's memory and compute consumption by mapping floating-point numbers to low-bit representations. However, existing quantization methods typically suffer from the speed-accuracy tradeoff and limited generalization. To address these issues, recent compensation-based methods o… ▽ More Network quantization has emerged as one of the most practical model compression techniques, which significantly reduces a model's memory and compute consumption by mapping floating-point numbers to low-bit representations. However, existing quantization methods typically suffer from the speed-accuracy tradeoff and limited generalization. To address these issues, recent compensation-based methods offer an efficient yet general solution by introducing additional lightweight linear layers into the quantized network. However, the accuracy of these methods suffers from their limited compensation capability and high sensitivity to outliers. In this paper, we propose Nonlinear Bipolar Compensation (NBC), a post-training quantization approach that introduces nonlinear compensation to reduce the effect of outliers. We further design Bipolar Logarithmic Transformation (BLT), which compresses outliers by mapping both the quantized input and the quantization error into a transformed space. A simple linear layer is then applied for compensation in the transformed space, preserving the efficiency of our method. Extensive experiments across various tasks, models, and quantization methods confirm the effectiveness, efficiency, robustness, and generality of our NBC approach. △ Less

Submitted 14 May, 2026; originally announced May 2026.

Wave packet landscape in open quantum systems

Authors: Kang Xu, Miao-Miao Yi, Zi-Hong Yan, C. P. Sun

Abstract: We formulate a landscape theory for the long-time wave packet spreading of free and harmonically trapped particles with quantum fluctuations and its related dissipation. We show that the diffusion, localization, and collapse of wave packets arise from symmetry structures of an underlying landscape in covariance space. The geometry of this landscape determines the asymptotic fate of the wave packet… ▽ More We formulate a landscape theory for the long-time wave packet spreading of free and harmonically trapped particles with quantum fluctuations and its related dissipation. We show that the diffusion, localization, and collapse of wave packets arise from symmetry structures of an underlying landscape in covariance space. The geometry of this landscape determines the asymptotic fate of the wave packet. In the quantum landscape description, the trapping potential and bath fluctuation break the landscape symmetry in distinct ways: the former lifts the valley-like landscape of a fluctuation-free free particle into a bowl-like landscape, leading to collapse, whereas the latter tilts the valley and turns localization into diffusion. The resulting landscape symmetry breaking accounts for the noncommuting long-time limits and abrupt changes in the asymptotic wave-packet width. This establishes landscape symmetry breaking as a unified geometric origin of wave-packet diffusion, localization, and collapse in quantum Brownian motion. △ Less

Submitted 17 June, 2026; v1 submitted 15 May, 2026; originally announced May 2026.

Ego2World: Compiling Egocentric Cooking Videos into Executable Worlds for Belief-State Planning

Authors: Qinchuan Cheng, Zhantao Gong, Pengzhan Sun, Angela Yao, Xulei Yang, Shijie Li

Abstract: Embodied agents in household environments must plan under partial observation: they need to remember objects, track state changes, and recover when actions fail. Existing benchmarks only partially test this ability. Egocentric video datasets capture realistic human activities but remain passive, while interactive simulators support execution but rely on synthetic scenes and hand-crafted dynamics,… ▽ More Embodied agents in household environments must plan under partial observation: they need to remember objects, track state changes, and recover when actions fail. Existing benchmarks only partially test this ability. Egocentric video datasets capture realistic human activities but remain passive, while interactive simulators support execution but rely on synthetic scenes and hand-crafted dynamics, introducing a sim-to-real gap and often assuming fully observable state. We introduce Ego2World, an executable benchmark that turns egocentric cooking videos into executable symbolic worlds governed by graph-transition rules. Built on HD-EPIC, Ego2World derives reusable transition rules from video annotations and executes them in a hidden symbolic world graph. During evaluation, the simulator maintains the hidden world graph, while the agent plans over its own partial belief graph using only local observations and execution feedback. This separation forces agents to update memory and replan without observing the true world state. Experiments show that action-overlap scores overestimate physical-state success, and that persistent belief memory improves task completion while reducing repeated visual exploration -- suggesting that belief maintenance should be a first-class target of embodied-agent evaluation. △ Less

Submitted 13 May, 2026; originally announced May 2026.

Comments: Project page: https://sj-li.com/PROJ/Ego2World/

PRISM: : Planning and Reasoning with Intent in Simulated Embodied Environments

Authors: Yunn Kang Lim, Pengzhan Sun, Ziyi Bai, Xun Xu, Angela Yao, Xulei Yang, Shijie Li

Abstract: When an LLM-based embodied agent fails at a household task, the culprit could be misidentified objects, forgotten sub-goals, or poor action sequencing -- yet existing benchmarks report only a single success rate, making it impossible to tell which cognitive module is responsible. We present PRISM, a diagnostic benchmark that reframes this problem: rather than asking only \textit{did the agent succ… ▽ More When an LLM-based embodied agent fails at a household task, the culprit could be misidentified objects, forgotten sub-goals, or poor action sequencing -- yet existing benchmarks report only a single success rate, making it impossible to tell which cognitive module is responsible. We present PRISM, a diagnostic benchmark that reframes this problem: rather than asking only \textit{did the agent succeed?}, PRISM asks \textit{which capability is most likely responsible for failure?} Built on five photorealistic multi-room apartments (4--8 rooms each), PRISM structures 300 human-verified tasks into three capability tiers -- \textit{Basic Ability}, \textit{Reasoning Ability}, and \textit{Long-horizon Ability} -- that isolate perception-to-action grounding, implicit intent resolution, and sustained multi-step coordination respectively. PRISM exposes an agent-agnostic executable action API that allows arbitrary agents: LLM agents, VLM agents, symbolic planners, RL policies, and hybrid systems, to be evaluated end-to-end under the same benchmark protocol. To support deeper diagnosis, optional probes for perception, memory, and planning can be adopted, replaced, or bypassed entirely, enabling controlled component-level analysis when desired. Experiments on seven contemporary LLMs establish a clear hierarchy: explicit spatial grounding is not the dominant failure source under oracle perception, implicit intent resolution is a significant bottleneck for all model families, and long-horizon coordination exposes a stark capability cliff -- lightweight models collapse to as low as 20.0\% success while simultaneously consuming more tokens than their frontier counterparts, a signature of compensatory over-reasoning rather than genuine planning capability. Project page: \href{https://sj-li.com/PROJ/PRISM}{link}. △ Less

Submitted 12 May, 2026; originally announced May 2026.

Permit: Permission-Aware Representation Intervention for Controlled Generation in Large Language Models

Authors: Pengcheng Sun, Lan Zhang, Zhaopeng Zhang, Jiewei Lai, Chen Tang

Abstract: Large language models (LLMs) are increasingly deployed in enterprise settings where they handle sensitive documents and user context, raising acute concerns over security and controllability. Conventional access control regulates whether information is accessible to the model, yet leaves how the model uses that information at generation time largely unconstrained: once sensitive content enters the… ▽ More Large language models (LLMs) are increasingly deployed in enterprise settings where they handle sensitive documents and user context, raising acute concerns over security and controllability. Conventional access control regulates whether information is accessible to the model, yet leaves how the model uses that information at generation time largely unconstrained: once sensitive content enters the context, outputs may still drift beyond a user's authorized scope. We present Permit, a novel permission-aware representation intervention framework that closes this gap by enforcing fine-grained control directly on the model's hidden states. Through exploratory analysis, we find that permission conditions induce hidden-state shifts that are (i) separable across permissions and (ii) concentrated in a small set of dominant directions. Permit exploits this geometry in two stages: it first identifies a permission-sensitive subspace from activation differences across permission conditions, and then performs lightweight interventions within this subspace to steer generation, with two concrete instantiations (offset-based and gated). Both operate atop a frozen backbone with only a handful of permission-specific parameters, achieving precise control with minimal overhead. Experimental results demonstrate that Permit performs better than the state-of-the-art method across multiple permission settings while driving information leakage to near zero, achieving over 18% F1-score improvement with >98% fewer trainable parameters. △ Less

Submitted 10 May, 2026; originally announced May 2026.

OA-WAM: Object-Addressable World Action Model for Robust Robot Manipulation

Authors: Yushan Liu, Peibo Sun, Shoujie Li, Yifan Xie, Lingfeng Zhang, Xintao Chao, Shiyuan Dong, Fang Chen, Xiao-Ping Zhang, Wenbo Ding

Abstract: World Action Models (WAMs) enhance Vision-Language-Action policies by jointly predicting scene evolution and robot actions, but existing methods usually represent the predicted world as holistic images, video tokens, or global latents. These representations are difficult for an action decoder to address when an instruction refers to a particular object, especially under scene shifts where object i… ▽ More World Action Models (WAMs) enhance Vision-Language-Action policies by jointly predicting scene evolution and robot actions, but existing methods usually represent the predicted world as holistic images, video tokens, or global latents. These representations are difficult for an action decoder to address when an instruction refers to a particular object, especially under scene shifts where object identity is entangled with context. We propose OA-WAM, an Object-Addressable World Action Model for robust robot manipulation. OA-WAM decomposes each frame into N+1 slot states, with one robot slot and N object slots. Each slot contains a persistent address vector and a time-varying content vector, and is fused with text, image, proprioception, and past-action tokens in a block-causal sequence. A world head predicts next-frame slot states, while a flow-matching action head decodes a 16-step continuous action chunk in the same forward pass. Addressability is enforced by routing cross-slot attention through address-only keys and resetting the address slice at every transformer layer, separating which object to act on from what that object currently is without adding extra tokens. OA-WAM matches strong VLA and WAM baselines on LIBERO (97.8%) and SimplerEnv (79.3%), reaches state-of-the-art performance on the most relevant LIBERO-Plus geometric axes, and remains competitive on the seven-axis aggregate. A causal slot-intervention test yields a swap-binding cosine of 0.87, versus at most 0.09 for holistic baselines. These results suggest that addressable object states provide an effective interface for robust world-action modeling under scene perturbations. △ Less

Submitted 7 May, 2026; originally announced May 2026.

Stateful Agent Backdoor

Authors: Zhengchunmin Dai, Jiaxiong Tang, Liantao Wu, Peng Sun, Honglong Chen

Abstract: Existing backdoor attacks on Large Language Model-based agents remain stateless, executing fixed behaviors confined to a single session. We propose a stateful agent backdoor that extends the attack lifecycle across multiple sessions under permission isolation. The attack maintains state through persistent components, enabling autonomous, incremental execution across sessions following a one-time t… ▽ More Existing backdoor attacks on Large Language Model-based agents remain stateless, executing fixed behaviors confined to a single session. We propose a stateful agent backdoor that extends the attack lifecycle across multiple sessions under permission isolation. The attack maintains state through persistent components, enabling autonomous, incremental execution across sessions following a one-time trigger injection. Formally, we model the attack as a Mealy machine and derive a decomposition framework that enables independent per-transition data construction. We instantiate this framework with a primary attack and two extensibility variants. The primary instantiation achieves an attack success rate of 80\%--95\% across four models, with per-transition analysis demonstrating the effectiveness of the decomposition. Extensibility variants with alternative topologies and persistent components demonstrate consistent effectiveness. Code and data are available at https://anonymous.4open.science/r/stateful_agent_backdoor-E89F. △ Less

Submitted 7 May, 2026; originally announced May 2026.

Safety Anchor: Defending Harmful Fine-tuning via Geometric Bottlenecks

Authors: Guoxin Lu, Letian Sha, Qing Wang, Peijie Sun, Hao Zhou, Hua Dai, Fu Xiao

Abstract: The safety alignment of Large Language Models (LLMs) remains vulnerable to Harmful Fine-tuning (HFT). While existing defenses impose constraints on parameters, gradients, or internal representations, we observe that they can be effectively circumvented under persistent HFT. Our analysis traces this failure to the inherent redundancy of the high-dimensional parameter space: attackers exploit optimi… ▽ More The safety alignment of Large Language Models (LLMs) remains vulnerable to Harmful Fine-tuning (HFT). While existing defenses impose constraints on parameters, gradients, or internal representations, we observe that they can be effectively circumvented under persistent HFT. Our analysis traces this failure to the inherent redundancy of the high-dimensional parameter space: attackers exploit optimization trajectories that are orthogonal to defense constraints to restore harmful capabilities while deceptively adhering to safety restrictions. To address this, we propose Safety Bottleneck Regularization (SBR). SBR shifts the defensive focus from the redundant parameter space to the unembedding layer, which serves as a geometric bottleneck. By anchoring the final hidden states of harmful queries to those of the safety-aligned model, SBR enables the model to maintain safe responses even under persistent HFT. Extensive experiments confirm SBR's effectiveness, demonstrating that utilizing just a single safety anchor is sufficient to reduce the Harmful Score to $<$10 while preserving competitive performance on benign downstream tasks. △ Less

Submitted 7 May, 2026; v1 submitted 7 May, 2026; originally announced May 2026.

Comments: Accepted to ICML 2026

CSGuard: Toward Forgery-Resistant Watermarking in Diffusion Models via Compressed Sensing Constraint

Authors: Jiewei Lai, Lan Zhang, Chen Tang, Pengcheng Sun, Zhaopeng Zhang, Yunhao Wang, Hui Jin

Abstract: Latent-based diffusion model watermarking embeds watermarks into generated images' latent space to enable content attribution, offering a training-free solution for intellectual property protection and digital forensics. However, these methods exhibit a critical vulnerability to the forgery attack, attackers can extract the watermark by inverting the watermarked image and re-generating it with an… ▽ More Latent-based diffusion model watermarking embeds watermarks into generated images' latent space to enable content attribution, offering a training-free solution for intellectual property protection and digital forensics. However, these methods exhibit a critical vulnerability to the forgery attack, attackers can extract the watermark by inverting the watermarked image and re-generating it with an arbitrary prompt, thereby enabling false attribution on malicious content. In this paper, we propose the CSGuard, the first forgery-resistant watermarking schema that leverages compressed sensing to bind the watermarked image generation and verification to a secret matrix. This ensures that only users possessing the secret matrix can correctly embed or verify the image watermark, prevents the illegal users from forgery without compromising generation quality and watermark integrity. Experimental results demonstrate that CSGuard achieves strong forgery resistance, reduces the attack success rate from 100.0\% to 28.12\%, and achieve 100\% detection rate on benign watermarked images without compromising watermarking effectiveness. △ Less

Submitted 2 May, 2026; originally announced May 2026.

Colinearity Decay: Training Quantization-Friendly ViTs with Outlier Decay

Authors: Jin Tong, Guang Liang, Peilin Sun, Jianxin Wu

Abstract: Low-bit quantization is a practical route for efficiently deploying vision Transformers, yet activation outliers complicate fully quantized deployment. Existing methods either handle quantization post-training or suppress large activations during training; however, aggressively restricting outliers in vision models can lead to a poorer trade-off between full-precision and quantized accuracy. We ar… ▽ More Low-bit quantization is a practical route for efficiently deploying vision Transformers, yet activation outliers complicate fully quantized deployment. Existing methods either handle quantization post-training or suppress large activations during training; however, aggressively restricting outliers in vision models can lead to a poorer trade-off between full-precision and quantized accuracy. We argue that rather than simply suppressing outliers, the training objective should control the structural amplification that makes them harmful. To this end, we introduce Colinearity-Decay (CD), a structural regularizer for ordered matrix pairs within Transformer blocks. CD penalizes detrimental cross-matrix alignment and mitigates extreme activations without altering the architecture or task loss. Applied as a decoupled update, CD is non-invasive and introduces minimal training overhead. Across ImageNet-1K pre-training, COCO detection, and downstream fine-tuning, CD consistently boosts quantized accuracy across multiple pipelines while preserving, or even improving, full-precision performance. Ultimately, our results demonstrate that structural regularization effectively prepares vision Transformers for low-bit deployment with zero inference-time overhead. △ Less

Submitted 2 May, 2026; originally announced May 2026.

Comments: 17 pages, 5 figures

Decoding Scientific Experimental Images: The SPUR Benchmark for Perception, Understanding, and Reasoning

Authors: Junpeng Ding, Zichen Tang, Haihong E, Mengyuan Ji, Yang Liu, Haolin Tian, Haiyang Sun, Pengqi Sun, Yang Xu, Yichen Liu, Haocheng Gao, Zijie Xi, Ruomeng Jiang, Peizhi Zhao, Rongjin Li, Yuanze Li, Jiacheng Liu, Zhongjun Yang, Jintong Chen, Siying Lin

Abstract: We introduce SPUR, a comprehensive benchmark for scientific experimental image perception, understanding, and reasoning, comprising 4,264 question-answering (QA) pairs derived from 1,084 expert-curated images. SPUR features three key innovations: (1) Panel-Level Fine-Grained Perception: evaluating the visual perception of multimodal large language models (MLLMs) across three dimensions (numerical,… ▽ More We introduce SPUR, a comprehensive benchmark for scientific experimental image perception, understanding, and reasoning, comprising 4,264 question-answering (QA) pairs derived from 1,084 expert-curated images. SPUR features three key innovations: (1) Panel-Level Fine-Grained Perception: evaluating the visual perception of multimodal large language models (MLLMs) across three dimensions (numerical, morphological, and information localization) on six fine-grained panel types; (2) Cross-Panel Relation Understanding: utilizing complex images with an average of 14.3 panels per sample to evaluate MLLMs' ability to decipher intricate cross-panel relations; (3) Expert-Level Reasoning: assessment of qualitative and quantitative reasoning across five experimental paradigms to determine if models can infer conclusions from evidence as human experts do. Comprehensive evaluation of 20 MLLMs and four multimodal Chain-of-Thought (MCoT) methods reveals that current models fall significantly short of the expert-level requirements for scientific image interpretation, underscoring a critical bottleneck in AI for Science (AI4S) research. △ Less

Submitted 26 May, 2026; v1 submitted 30 April, 2026; originally announced April 2026.

Comments: Accepted to ACL 2026 Main Conference

SpatialGrammar: A Domain-Specific Language for LLM-Based 3D Indoor Scene Generation

Authors: Song Tang, Kaiyong Zhao, Yuliang Li, Qingsong Yan, Penglei Sun, Junyi Zou, Qiang Wang, Xiaowen Chu

Abstract: Automatically generating interactive 3D indoor scenes from natural language is crucial for virtual reality, gaming, and embodied AI. However, existing LLM-based approaches often suffer from spatial errors and collisions, in part because common scene representations-raw coordinates or verbose code-are difficult for models to reason about 3D spatial relationships and physical constraints. We propose… ▽ More Automatically generating interactive 3D indoor scenes from natural language is crucial for virtual reality, gaming, and embodied AI. However, existing LLM-based approaches often suffer from spatial errors and collisions, in part because common scene representations-raw coordinates or verbose code-are difficult for models to reason about 3D spatial relationships and physical constraints. We propose SpatialGrammar, a domain-specific language that represents gravity-aligned indoor layouts as BEV grid placements with deterministic compilation to valid 3D geometry, enabling verifiable constraint checking. Building on this representation, we develop (1) SG-Agent, a closed-loop system that uses compiler feedback to iteratively refine scenes and enforce collision constraints, and (2) SG-Mini, a 104M-parameter model trained entirely on compiler-validated synthetic data. Across 159 test scenes spanning five scenarios of different complexity, SG-Agent improves spatial fidelity and physical plausibility over prior methods, while SG-Mini performs competitively against larger LLM-based baselines on single-shot generation scenarios. △ Less

Submitted 30 April, 2026; originally announced April 2026.

Diffusion Model as a Generalist Segmentation Learner

Authors: Haoxiao Wang, Antao Xiang, Haiyang Sun, Peilin Sun, Changhao Pan, Yifu Chen, Minjie Hong, Weijie Wang, Shuang Chen, Yue Chen, Zhou Zhao

Abstract: Diffusion models are primarily trained for image synthesis, yet their denoising trajectories encode rich, spatially aligned visual priors. In this paper, we demonstrate that these priors can be utilized for text-conditioned semantic and open-vocabulary segmentation, and this approach can be generalized to various downstream tasks to make a general-purpose diffusion segmentation framework. Concrete… ▽ More Diffusion models are primarily trained for image synthesis, yet their denoising trajectories encode rich, spatially aligned visual priors. In this paper, we demonstrate that these priors can be utilized for text-conditioned semantic and open-vocabulary segmentation, and this approach can be generalized to various downstream tasks to make a general-purpose diffusion segmentation framework. Concretely, we introduce DiGSeg (Diffusion Models as a Generalist Segmentation Learner), which repurposes a pretrained diffusion model into a unified segmentation framework. Our approach encodes the input image and ground-truth mask into the latent space and concatenates them as conditioning signals for the diffusion U-Net. A parallel CLIP-aligned text pathway injects language features across multiple scales, enabling the model to align textual queries with evolving visual representations. This design transforms an off-the-shelf diffusion backbone into a universal interface that produces structured segmentation masks conditioned on both appearance and arbitrary text prompts. Extensive experiments demonstrate state-of-the-art performance on standard semantic segmentation benchmarks, as well as strong open-vocabulary generalization and cross-domain transfer to medical, remote sensing, and agricultural scenarios-without domain-specific architectural customization. These results indicate that modern diffusion backbones can serve as generalist segmentation learners rather than pure generators, narrowing the gap between visual generation and visual understanding. △ Less

Submitted 27 April, 2026; originally announced April 2026.