Showing 1–50 of 612 results for author: Fan, C

PTXBench: Benchmark and Adapt LLMs for GPU Kernel Optimization with Architecture-specific PTX

Authors: Genghan Zhang, Yixin Dong, Chengze Fan, Zhichen Zeng, Yueming Yuan, Shaowei Zhu, Kunle Olukotun

Abstract: We introduce PTXBench, a benchmark for evaluating and adapting large language models (LLMs) to use architecture-specific PTX for GPU kernel optimization. PTXBench measures functional correctness, whether selected target instructions execute at runtime, and speedup over frontier libraries across GEMM and attention workloads on H100 and B200 GPUs. Our evaluation shows that architecture-specific PTX… ▽ More We introduce PTXBench, a benchmark for evaluating and adapting large language models (LLMs) to use architecture-specific PTX for GPU kernel optimization. PTXBench measures functional correctness, whether selected target instructions execute at runtime, and speedup over frontier libraries across GEMM and attention workloads on H100 and B200 GPUs. Our evaluation shows that architecture-specific PTX capability remains uneven: success rates fall substantially on complex attention backward workloads, and executing the target instructions does not necessarily translate into competitive performance. No evaluated model consistently matches frontier libraries across the suite. We further adapt Qwen3.6-27B using supervised fine-tuning. Repair-conditioned training improves several tasks, but generalization remains uneven; data coverage, balance, and the quality of the reasoning teacher matter in addition to dataset size. PTXBench provides an auditable testbed for measuring and improving LLMs' ability to exploit evolving GPU architectures. △ Less

Submitted 19 August, 2026; v1 submitted 18 August, 2026; originally announced August 2026.

ScienceFlow: A long-horizon agent for ML research, scientific discovery and beyond

Authors: Mingming Zhao, Jiqian Dong, Kangping Xu, Zadid Hasan, Chengrui Fan, Shan Jiang, Shuai Mao, Ting Lingya, Linyi Zou, Tailin Zhou, Yun Hin Chan, Wenkai Zhang, Zhanhong Zhou, Guowei Huang, Hongliang Li, Wenjing Cun, Zhitang Chen, Mingxuan Yuan, Yanhui Geng

Abstract: Enabling LLM agents to sustain productive, stable, and goal-aligned research over extended horizons is a central challenge for autonomous machine learning and scientific discovery, as progress hinges on continuously managing evolving state, exploration decisions, and computational resources. Pioneering autoresearch agents, despite great success, still lack mechanisms for continuity, recovery from… ▽ More Enabling LLM agents to sustain productive, stable, and goal-aligned research over extended horizons is a central challenge for autonomous machine learning and scientific discovery, as progress hinges on continuously managing evolving state, exploration decisions, and computational resources. Pioneering autoresearch agents, despite great success, still lack mechanisms for continuity, recovery from dead ends, and value-driven compute allocation, which inherently undermines overall search efficiency, wastes computational resources, and lowers the chance of ultimate success. To bridge this gap, we introduce ScienceFlow, an end-to-end autoresearch agent framework that organizes long-horizon research work into research segments grounded in executable workspaces. It represents research progress as recoverable executable states, enabling efficient exploration, revision, and execution. Transitions between research segments are governed by Executable-State Transition through Re-Anchoring (ESTRA), which selects either the live state or an archived state as the next anchor and determines whether to continue or redirect the research trajectory. An evidence-aware execution controller allocates resources to physical jobs based on resource availability, remaining budget, and validated progress. We evaluate ScienceFlow on tasks spanning machine learning, scientific modeling, and mathematical optimization. Results on diverse long-horizon benchmarks demonstrate its ability to sustain effective research processes, highlighted by a SOTA 70.22 percent Any-Medal score on the full MLE-bench within a 24-hour budget, outperforming prior reported results by 4.92 percentage points. The efficacy of ScienceFlow further demonstrates that efficient state management, adaptive exploration, and objective-aligned execution are critical for scaling autonomous research beyond short-horizon interactions. △ Less

Submitted 14 August, 2026; originally announced August 2026.

CipherSight: Robust Website Fingerprinting via Record-Resource Semantic Supervision under Distribution Shifts

Authors: Runhan Song, Qiqi Liu, Chuanzhou Pan, Zhenquan Ding, Youquan Xian, Chongru Fan, Lei Cui, Wei Wang, Zhiyu Hao

Abstract: HTTPS website fingerprinting (WF) aims to identify visited websites from metadata observable in encrypted traffic. However, real-world deployments introduce a significant out-of-distribution (OOD) problem caused by temporal and geographic changes, while previously unseen websites are common in open-world scenarios. Existing methods primarily learn from raw TCP packet sequences and struggle to capt… ▽ More HTTPS website fingerprinting (WF) aims to identify visited websites from metadata observable in encrypted traffic. However, real-world deployments introduce a significant out-of-distribution (OOD) problem caused by temporal and geographic changes, while previously unseen websites are common in open-world scenarios. Existing methods primarily learn from raw TCP packet sequences and struggle to capture stable and generalizable website representations, resulting in performance degradation under practical conditions. We propose CipherSight, a TLS-record-based hierarchical framework for robust HTTPS WF. Unlike existing approaches that rely on TCP packet sequences and are sensitive to transport-layer artifacts, CipherSight learns website representations from TLS records by jointly encoding multiple record-level attributes. It introduces a hierarchical architecture that captures both intra-flow dependencies among TLS records and inter-flow interactions across concurrent flows, enabling the model to exploit structural patterns in HTTPS traffic. Besides, to learn robust representations, CipherSight employs a masked record modeling (MRM) task to capture contextual traffic semantics and leverages fine-grained record-resource annotations as privileged supervision through structure-aware objectives and semantic distillation. Experiments show that CipherSight achieves 95.41% accuracy across more than 2,000 website classes in the closed-world setting and maintains over 90% accuracy under both temporal and geographic drift, consistently outperforming all evaluated baselines. △ Less

Submitted 13 August, 2026; originally announced August 2026.

Coverage Aware Active Evaluation for Failure Discovery with Paired Systems

Authors: Anjali Parashar, Rachel Luo, Apoorva Sharma, Sushant Veer, Edward Schmerling, Carson Sobolewski, Mingxin Yu, Chuchu Fan, Marco Pavone

Abstract: Autonomous systems can fail in rare and heterogeneous ways, making real-world failure discovery difficult under limited testing budgets. Although cheaper proxies such as simulators, lower-fidelity systems, or related policies can be sampled extensively to find failures, proxy failures often do not transfer to the real world due to sim-to-real and system-to-system gaps. The key challenge is therefo… ▽ More Autonomous systems can fail in rare and heterogeneous ways, making real-world failure discovery difficult under limited testing budgets. Although cheaper proxies such as simulators, lower-fidelity systems, or related policies can be sampled extensively to find failures, proxy failures often do not transfer to the real world due to sim-to-real and system-to-system gaps. The key challenge is therefore to effectively leverage proxy system information for accurate prediction of severe target system failures. We propose an adaptive failure discovery method that combines proxy evaluations with limited target system results to guide scenario selection for target system testing. Our method learns a local predictor of target risk by correcting proxy failure signals using control-variate-inspired residual modeling. To find failures that are both likely and diverse, we combine this predictor with a support-aware mutual-information objective that favors realistic, well-supported regions while expanding coverage across failure modes. Across autonomous driving, manipulation, and quadruped velocity-tracking tasks, our method discovers up to 2$\times$ as many failures as random sampling and active-learning baselines, including severe and diverse failures missed by competing methods. △ Less

Submitted 13 August, 2026; originally announced August 2026.

Comments: 9 main pages followed by Appendix, total 21 pages, 12 figures

LoopVSR: A Loop Engineering Framework for Automated Repair of Visual Speech Recognition Inference Pipelines

Authors: Fei Qin, Bowen Zhang, Chao Fan, Pengcheng Luo, Genke Yang

Abstract: Visual speech recognition (VSR) recovers speech from lip movements when audio is noisy or unavailable. Its multi-stage inference pipeline spans video decoding, mouth-region extraction, preprocessing, model invocation, and decoding, where upstream failures can mask downstream faults. Pipeline maintenance therefore still relies largely on predefined checks and manual debugging. We propose LoopVSR, a… ▽ More Visual speech recognition (VSR) recovers speech from lip movements when audio is noisy or unavailable. Its multi-stage inference pipeline spans video decoding, mouth-region extraction, preprocessing, model invocation, and decoding, where upstream failures can mask downstream faults. Pipeline maintenance therefore still relies largely on predefined checks and manual debugging. We propose LoopVSR, a Loop Engineering framework that enables a code agent to automatically diagnose and repair VSR inference pipelines using end-to-end execution evidence. It couples constrained repository-level diagnosis and patching with an external controller that audits changes, runs real inference, and accepts or rolls back patches using failures and character error rate (CER). The resulting feedback loop returns newly observed exceptions, tensor statistics, and recognition errors to the agent, progressively exposing faults masked by upstream failures. On the CMLR VSR system, LoopVSR repairs all 11 main faults with 100% mean recovery, whereas the Static guard repairs 2 of 11 with 18.13% mean recovery. It also resolves three cascading tasks in seven accepted iterations and preserves recovery on an independent 200-video hidden set. These results demonstrate that LoopVSR enables measurable, end-to-end automated repair of VSR inference pipelines. △ Less

Submitted 12 August, 2026; originally announced August 2026.

EGM-Det: Entropy-Guided Multimodal Adaptive Fusion for UAV RGB-IR Object Detection

Authors: Cunzheng Fan, Dawei Yan, Guanlin Wang, Xingshuo Yang, Yupeng Jia, Jing Yang, Haokui Zhang

Abstract: Joint use of RGB and infrared (IR) imagery can improve UAV-view object detection, but most existing methods fuse multimodal features with static or fixed weights and therefore overlook spatially varying modality reliability. We propose EGM-Det, an entropy-guided multimodal adaptive fusion framework for RGB-IR object detection. EGM-Det employs a dual-stream architecture to preserve modality-specifi… ▽ More Joint use of RGB and infrared (IR) imagery can improve UAV-view object detection, but most existing methods fuse multimodal features with static or fixed weights and therefore overlook spatially varying modality reliability. We propose EGM-Det, an entropy-guided multimodal adaptive fusion framework for RGB-IR object detection. EGM-Det employs a dual-stream architecture to preserve modality-specific representations and introduces an Entropy Offset Gate Fusion module for adaptive multi-scale fusion. The module derives shallow entropy priors from input intensity, local entropy, and cross-modal discrepancy, and uses them to guide local offset alignment and spatial-channel gated fusion. It therefore selectively aggregates reliable RGB and infrared cues instead of uniformly combining heterogeneous features. We further introduce cross-modal distillation to regularize the learned fusion gates and reduce fusion degradation. Each student branch extracts complementary knowledge from the cross-modality teacher branch matched to the main branch, while entropy-adaptive supervision emphasizes uncertain modality decisions. Experiments on DroneVehicle, LLVIP, and VEDAI demonstrate state-of-the-art performance across all three benchmarks; in particular, EGM-Det outperforms prior approaches by more than 10 percentage points on VEDAI. △ Less

Submitted 12 August, 2026; originally announced August 2026.

Comments: 14 pages, 7 figures, 6 tables

Boundary-Seeking Policy Gradient for Safe Reinforcement Learning

Authors: Chenhua Fan, Jiahui Zhu, Yuhang Zhang, Honghao Wei

Abstract: Safe reinforcement learning maximizes reward subject to safety constraints. For Constrained Markov Decision Processes, the linear-programming view over occupancy measures implies that whenever the constraint is active at optimality, the optimal policy lies exactly on the constraint boundary, yet standard gradient-based methods do not exploit this structure and often settle in the feasible interior… ▽ More Safe reinforcement learning maximizes reward subject to safety constraints. For Constrained Markov Decision Processes, the linear-programming view over occupancy measures implies that whenever the constraint is active at optimality, the optimal policy lies exactly on the constraint boundary, yet standard gradient-based methods do not exploit this structure and often settle in the feasible interior. We introduce Boundary-Seeking Policy Gradient (BSPG), a first-order method whose update combines a tangential component that improves reward while preserving cost to first order with a signed, residual-driven normal component that regulates the policy toward the active boundary from either side; the combined direction admits an algebraic Lagrangian form with an induced coefficient and no learned dual variable. Under exact gradients and stated regularity conditions, the constraint residual converges to zero from either side with a finite-horizon $O(1/\sqrt{T})$ bound, the tangential component is a reward-ascent direction on the boundary, and any convergent parameter sequence is stationary on the active constraint set, satisfying the KKT conditions when the limit is also a local maximizer over the feasible set. This complements existing analyses, which certify feasibility but do not characterize the constraint value at convergence. On a standard Safety-Gymnasium navigation task, BSPG attains higher reward while tracking the boundary more tightly than the compared baselines. △ Less

Submitted 10 August, 2026; originally announced August 2026.

Comments: CDC 2026

Reading is not Reasoning: Bridging the Agentic Policy Gap in Vision-Text Compression

Authors: Cheng Fan, Junyi Zhou, Tingzhang Luo, RongJian Xu, Qiyanhui Lu, Mingjian Zhu, Hanting Chen, Jianyuan Guo

Abstract: Multi-step language-model agents repeatedly process growing interaction histories, leading to substantial context costs. Vision--text compression reduces these costs by rendering history as images, but the resulting modality shift creates a marked capability gap. Through controlled evaluations of history recovery, matched-state decisions, and complete trajectories, we show that this gap cannot be… ▽ More Multi-step language-model agents repeatedly process growing interaction histories, leading to substantial context costs. Vision--text compression reduces these costs by rendering history as images, but the resulting modality shift creates a marked capability gap. Through controlled evaluations of history recovery, matched-state decisions, and complete trajectories, we show that this gap cannot be explained by OCR quality alone. Visual-history agents exhibit systematic drift in action selection, query formulation, stopping, and evidence use, revealing an agentic policy gap. We introduce \textbf{CAPS}, a two-stage \textbf{C}ross-modal \textbf{A}gentic \textbf{P}olicy \textbf{S}elf-distillation framework that uses the same model's stronger text-history policy to supervise its visual-history counterpart. Offline trajectory self-distillation transfers successful text-policy behavior to visual-history inputs, while online policy self-distillation provides dense supervision on states visited by the visual-history policy during reinforcement learning. On SearchQA, CAPS improves over AgentOCR by 5.0\% and 3.4\% with 3B and 7B backbones, respectively. On full-history ALFWorld, the corresponding gains are 15.6\% and 14.5\%. Across settings, CAPS reduces average memory-context cost by up to 63.3\% and peak cost by up to 83.4\% relative to matched text-history policies. These results show that explicit cross-modal policy self-distillation can preserve agent capability under vision--text compression. Our code will be made publicly available in a future release. △ Less

Submitted 9 August, 2026; originally announced August 2026.

Thinking Hard, Not Smart: Reasoning Models Fail to Ration Test-Time Compute Across Questions

Authors: Chenrui Fan, Yize Cheng, Ming Li, Yongyuan Liang, Tianyi Zhou, Soheil Feizi

Abstract: Reasoning language models increasingly use test-time compute to improve performance, but existing evaluations typically study this compute one question at a time. Yet when multiple problems share an end-to-end cost or latency constraint, models must decide how to divide limited inference compute among them. We introduce an exam-style evaluation framework for studying this setting, in which a model… ▽ More Reasoning language models increasingly use test-time compute to improve performance, but existing evaluations typically study this compute one question at a time. Yet when multiple problems share an end-to-end cost or latency constraint, models must decide how to divide limited inference compute among them. We introduce an exam-style evaluation framework for studying this setting, in which a model must distribute one shared token budget across questions with different difficulty and point values to maximize its total score. Across several open and frontier reasoning models, we find that models fail to allocate a shared budget strategically across questions of varying difficulties and values. Models behave largely as greedy sequential solvers: they prioritize questions by presentation order, front-load effort on early questions, and remain insensitive to value, with these tendencies becoming more pronounced as the number of questions grows. Explicit planning prompts spread compute more evenly but do not produce value- or difficulty-aware prioritization. The same behavioral pattern extends from mathematical to code reasoning. These findings establish global budget allocation as a distinct capability that is not captured by conventional per-question evaluation and remains a challenge for current reasoning models. △ Less

Submitted 10 August, 2026; v1 submitted 8 August, 2026; originally announced August 2026.

RoRA: Role-Oriented Regional Allocation for Visual Token Pruning in MLLMs

Authors: Qiyanhui Lu, Han Wu, Rongjian Xu, Tingzhang Luo, Cheng Fan, Xinghao Chen, Minjing Dong, Jufeng Yang, Jianyuan Guo

Abstract: Multimodal large language models (MLLMs) encode images as long visual token sequences, making prefilling and KV-cache storage expensive. Existing training-free pruning methods select tokens by importance, diversity, or spatial coverage, but treat retained tokens as interchangeable and do not explicitly track which object-related regions are already covered. We present RoRA, a training-free framewo… ▽ More Multimodal large language models (MLLMs) encode images as long visual token sequences, making prefilling and KV-cache storage expensive. Existing training-free pruning methods select tokens by importance, diversity, or spatial coverage, but treat retained tokens as interchangeable and do not explicitly track which object-related regions are already covered. We present RoRA, a training-free framework that casts visual token pruning as role-oriented regional evidence allocation. Given a fixed budget, RoRA partitions tokens into a protected semantic core, complementary context, and fine-grained detail. It first calibrates text-conditioned attention with a positional prior and a prompt-calibrated object prior, then builds Attention-Anchored Regions (AARs) from high-confidence anchors as lightweight proxies for covered object support. Context is explored mainly outside AARs, while a small AAR-guided budget restores local detail; pairwise similarity is used only for context-stage redundancy filtering. Under matched budgets, RoRA consistently outperforms strong training-free baselines across LLaVA and Qwen-VL families, retaining most of the unpruned accuracy even at aggressive pruning ratios, e.g., 96.5% of full performance at 88.9% pruning on LLaVA-1.5, and improving over D2Pruner by about 5% on Qwen3-VL at 75-90% pruning. At a 66.7% pruning ratio, RoRA requires only 0.7 ms for token selection and reduces end-to-end inference time by 24.6%, corresponding to a 1.33x speedup over unpruned inference on an NVIDIA H800. △ Less

Submitted 7 August, 2026; originally announced August 2026.

Comments: 9 pages, 4 figures, 4 tables. Code is available at https://github.com/LukieLuu/RoRA

AutoSND: From Execution Evidence to Structural Policies for Automated Network Dismantling Heuristic Discovery

Authors: Zhijing Hu, Changjun Fan, Yufan Deng, Zhiguang Cao

Abstract: Network dismantling is fundamental to analyzing the robustness and vulnerability of complex systems, yet practical heuristics must balance effectiveness and computational efficiency, and are usually designed manually by researchers. Existing large language model based automatic heuristic design methods can generate and screen candidates, yet they have difficulty further transforming candidate qual… ▽ More Network dismantling is fundamental to analyzing the robustness and vulnerability of complex systems, yet practical heuristics must balance effectiveness and computational efficiency, and are usually designed manually by researchers. Existing large language model based automatic heuristic design methods can generate and screen candidates, yet they have difficulty further transforming candidate quality or failure states during execution into structural-level guid- ance for subsequent generation. We propose AutoSND, a three stage tree search framework for complete network dismantling pro- grams. Stage I broadly explores from simple heuristics and archives execution evidence. Stage II compiles candidate records into struc- tural policies concerning local signals, neighborhood access, and state update ranges. Stage III continues tree search conditioned on these policies and obtains the final quality prioritized and speed prioritized candidates, AutoSND-Q/S. Experiments on 12 real world networks and 3 large real world networks show that AutoSND achieves better search performance and stability and discovers more competitive and structurally interpretable network disman- tling programs. The final candidates form an interpretable structure that uses residual degree as the backbone, adjusts node order with bounded local signals, and restricts the state update range. Code is available at https://github.com/MirrorNew/AutoSND. △ Less

Submitted 4 August, 2026; originally announced August 2026.

CROSS: Cascaded Distillation and Dual-Constraint Grounding for Remote Sensing Referring Segmentation

Authors: Tingzhang Luo, Ruizhong Liu, Yichao Liu, Cheng Fan, Yu Liu, Jianyuan Guo

Abstract: Referring Remote Sensing Image Segmentation (RRSIS) has achieved significant progress through the integration of VLMs and the Segment Anything Model (SAM). However, this progress largely relies on strong pre-trained capabilities, while leaving two fundamental limitations insufficiently addressed: (1) Architectural Weak-Coupling, where the unidirectional flow forces reliance on coarse VLM prompts a… ▽ More Referring Remote Sensing Image Segmentation (RRSIS) has achieved significant progress through the integration of VLMs and the Segment Anything Model (SAM). However, this progress largely relies on strong pre-trained capabilities, while leaving two fundamental limitations insufficiently addressed: (1) Architectural Weak-Coupling, where the unidirectional flow forces reliance on coarse VLM prompts and wastes SAM's pixel-level structural guidance, causing localization drift; and (2) Object-Centric Semantic Bias, where models overemphasize dominant object semantics while remaining insensitive to spatial reasoning crucial for RRSIS. Motivated by these observations, we propose CROSS, a tightly integrated paradigm for RRSIS. First, we introduce Linguistic-Guided Cascaded Distillation (LGCD) to bridge the architectural gap, which distills SAM's geometric affinities as soft regularizers into VLM intermediate layers, injecting dense structural priors to refine localization. Second, Perspective-Spatial Contrastive Learning (PSCL) imposes cross-anchored constraints by mining mask-filtered deceptive distractors and spatial-linguistic counterfactuals as hard negatives, explicitly shattering semantic shortcuts to enforce genuine logical consistency. Extensive experiments on RRSIS benchmarks demonstrate that CROSS achieves state-of-the-art performance and maintains precise localization even under severe spatial description perturbations, standing as a robust new paradigm for RRSIS. △ Less

Submitted 5 August, 2026; v1 submitted 4 August, 2026; originally announced August 2026.

Comments: Accepted at the European Conference on Computer Vision (ECCV) 2026. 20 pages, 6 figures, and 5 tables. Tingzhang Luo and Ruizhong Liu contributed equally. Jianyuan Guo is the corresponding author. Project page: https://clarence-cv.github.io/CROSS/

Inter-Residue Geometry Attention for Antibody-Specific Epitope Prediction

Authors: Chuanliu Fan, Nan Yu, Junjie Wu, Guohong Fu

Abstract: Antibody-specific epitope prediction aims to identify which antigen residues are recognized by a given antibody, a task that depends on the three-dimensional complementarity between antibody CDRs and the antigen surface. Existing methods usually leverage PLM embeddings and inject structure through additional graph, surface, or point-cloud encoders, where the positional mechanism inside attention r… ▽ More Antibody-specific epitope prediction aims to identify which antigen residues are recognized by a given antibody, a task that depends on the three-dimensional complementarity between antibody CDRs and the antigen surface. Existing methods usually leverage PLM embeddings and inject structure through additional graph, surface, or point-cloud encoders, where the positional mechanism inside attention remains largely tied to one-dimensional sequence order. For proteins, the analogue of a token offset is not only sequence separation, but also the three-dimensional displacement between residues after folding. This raises a question, can folded residue geometry serve as the positional mechanism of attention itself? We propose Local-Frame 3D Rotary Position Encoding (LF3DRoPE), which expresses inter-residue displacements in backbone-defined local frames and injects them directly into rotary attention. This design preserves continuous directional geometry while ensuring invariance to global $\mathrm{SE}(3)$ transformations. On the AsEP benchmark, LF3DRoPE achieves state-of-the-art $\mathrm{MCC}$ on both ratio and epitope-group splits. Ablations and rigid transformation tests show that local three-dimensional geometry provides information beyond sequence-order attention while preserving invariance to arbitrary global coordinate systems. Mutation ranking results further indicate that LF3DRoPE captures antigen-specific structural compatibility. △ Less

Submitted 2 August, 2026; originally announced August 2026.

RLPF: Reinforcement Learning from Performance Feedback for Code Generation

Authors: Huihao Jing, Haozhe Cui, Wenbin Hu, Shaojin Chen, Haochen Shi, Changxuan Fan, Yuxuan Liu, Hanyu Yang, Sirui Zhang, Ziyi Chen, Haoran Li, Yangqiu Song

Abstract: Code models are increasingly trained with execution feedback, but most training signals still stop at correctness. This leaves an important gap for systems code: two programs can pass the same tests while differing greatly in runtime. We study how to train code agents to prefer faster correct implementations, rather than treating efficiency only as an evaluation metric. The key difficulty is that… ▽ More Code models are increasingly trained with execution feedback, but most training signals still stop at correctness. This leaves an important gap for systems code: two programs can pass the same tests while differing greatly in runtime. We study how to train code agents to prefer faster correct implementations, rather than treating efficiency only as an evaluation metric. The key difficulty is that runtime is a fragile reward. It is meaningful only after a program is correct, varies across tasks, and gives little guidance when most sampled programs fail to compile or run. We propose \textbf{RLPF}, reinforcement learning from performance feedback, which turns execution outcomes into a staged reward. Failed programs are ordered by execution progress, while correct programs are ranked by their relative improvement from the baseline toward the expert reference. This gives useful feedback before correctness and performance-sensitive feedback after correctness. Fine-tuning Qwen3-32B with RLPF on PerfCodeBench raises correct-and-runnable solutions from $11.1\%$ to $54.6\%$ and improves relative efficiency from $8.1\%$ to $38.6\%$. The trained model becomes competitive with stronger open-weight systems, and its optimization behavior transfers modestly to EffiBench-X. Additional studies show that model-generated references provide useful but weaker supervision, and that the full composite reward is more reliable than correctness-only or runtime-only baselines. These results suggest that code agents can be trained not only to pass tests, but also to optimize the programs they write. △ Less

Submitted 29 July, 2026; originally announced July 2026.

Safe and Scalable Multi-Drone Payload Transport via CBF-based Reinforcement Learning with Zero-Shot Sim-to-Real Transfer

Authors: Jaeyoun Choi, Oswin So, Songyuan Zhang, Cooper Taylor, Chuchu Fan

Abstract: Multi-drone payload transportation has emerged as a promising research paradigm with potential applications in construction, logistics, and disaster response. However, the complex coupled dynamics among drones, cables, and payloads pose significant challenges, and existing approaches remain limited in safety and scalability, particularly in dynamic and unstructured environments. In this work, we p… ▽ More Multi-drone payload transportation has emerged as a promising research paradigm with potential applications in construction, logistics, and disaster response. However, the complex coupled dynamics among drones, cables, and payloads pose significant challenges, and existing approaches remain limited in safety and scalability, particularly in dynamic and unstructured environments. In this work, we propose a learning-based framework for safe and scalable multi-drone cooperative payload transport. We introduce a minimal 2D abstraction that preserves the task-relevant drone-payload coupling required for coordination and safety, while remaining computationally efficient for large-scale learning. Using domain randomization over team size and physical parameters, we train a fully distributed policy via Discrete Graph Control Barrier Function Proximal Policy Optimization (DGPPO), enabling robust zero-shot sim-to-real transfer without fine-tuning. Extensive real-world evaluations demonstrate that a single learned policy generalizes across varying team sizes and task scenarios. Furthermore, multi-group hardware experiments show that the same policy can safely operate in dynamic environments, where other drone teams act as moving obstacles. These results indicate that the proposed framework enables efficient, safe, and scalable multi-drone payload transportation with strong generalization to complex real-world conditions. △ Less

Submitted 22 July, 2026; originally announced July 2026.

Comments: Published in IEEE Robotics and Automation Letters (Early Access), 2026

KineBench: Benchmarking Embodied World Models via IDM-Free Kinematic Grounding

Authors: Zeyu Liu, Zhangzhe Zhu, Yang Zhang, Chenyou Fan, Chenjia Bai, Xuelong Li

Abstract: Evaluating the physical consistency of embodied world models(EWMs) is a critical open challenge. While closed-loop evaluation via simulator rollouts offers a more faithful assessment of physical plausibility than open-loop alternatives, existing frameworks almost exclusively rely on Inverse Dynamics Models(IDMs) for action extraction. Due to the intricate mapping from 2D pixel space to 3D kinemati… ▽ More Evaluating the physical consistency of embodied world models(EWMs) is a critical open challenge. While closed-loop evaluation via simulator rollouts offers a more faithful assessment of physical plausibility than open-loop alternatives, existing frameworks almost exclusively rely on Inverse Dynamics Models(IDMs) for action extraction. Due to the intricate mapping from 2D pixel space to 3D kinematic space, the learned IDMs can be brittle to data outside their training distribution, resulting in unreliable action extraction from the generated videos with novel objects and scenarios. This creates an unavoidable attribution ambiguity between world model inaccuracies and extractor errors. To reduce this ambiguity, we present KineBench, an IDM-free closed-loop benchmark for EWMs, built upon an explicit kinematic grounding pipeline. Given a generated video, KineBench employs cascaded visual foundation models to directly extract 6D end-effector poses from individual frames, which are then executed in a physics simulator for closed-loop validation. Beyond execution-based task success, KineBench incorporates two classical 3D kinematic metrics--Spectral Arc Length (SPARC) and the Maruyama Manipulability Index--to characterize trajectory smoothness and kinematic feasibility from a robot-centric perspective. Built on 20 diverse manipulation tasks in ManiSkill3, KineBench evaluates EWMs across four progressive suites: basic execution, task transfer, visual out-of-distribution generalization, and complexity-conditioned scaling. Evaluation across frontier models reveals task-complexity-bounded nonlinear scaling in embodied video generation, providing empirical guidance for future data-scaling strategies. △ Less

Submitted 22 July, 2026; originally announced July 2026.

Comments: Accept to ECCV2026

Private Approximation of Graph Spectra and Cuts via Spectral Amplifiers

Authors: Chenglin Fan, Jingcheng Liu, Pan Peng, Hangyu Xu, Zongrui Zou

Abstract: We study the problem of releasing a synthetic graph that approximates the sizes of all cuts of an input graph under edge-level differential privacy. If one insists on purely additive error, the optimal worst-case error is $\widetildeΘ(n^{3/2})$. If one allows a small multiplicative slack, an information-theoretic exponential-time mechanism achieves nearly linear additive error, but the best known… ▽ More We study the problem of releasing a synthetic graph that approximates the sizes of all cuts of an input graph under edge-level differential privacy. If one insists on purely additive error, the optimal worst-case error is $\widetildeΘ(n^{3/2})$. If one allows a small multiplicative slack, an information-theoretic exponential-time mechanism achieves nearly linear additive error, but the best known polynomial-time algorithms have substantially larger error. We give a polynomial-time $(\varepsilon,δ)$-differentially private algorithm which, for every $n$-vertex unweighted graph $G$, outputs a non-negative weighted synthetic graph $\widetilde G$ such that, with high probability, every cut $S\subseteq V(G)$ satisfies \[ |w_G(S)-w_{\widetilde G}(S)| \le γw_G(S)+\widetilde O_{\varepsilon,δ,γ}(n^{13/12+o(1)}). \] This improves the previous polynomial-time worst-case bound $\widetilde O(n^{5/4+o(1)})$ of Aamand et al. (ICML 2025) for mixed multiplicative/additive private cut approximation. The main technical ingredient is a new set of private spectral primitives for bounded-degree graphs, one of them gives spectral error $\widetilde O_δ((nd)^{1/4}/\sqrt\varepsilon)$ in estimating the graph Laplacian for graphs of maximum degree $d$, being the first to beat the standard $\min\{2d,\widetilde O_δ(\sqrt{n}/\varepsilon)\}$ baseline in the high-degree regime. We further develop a primitive with a sharper error dependence on $n$ and $d$ for the downstream cut approximation. Combined with a new edge-sensitive terminal cut oracle with additive error $\widetilde O(n+(n^2M)^{1/3})$ on graphs with $M$ edges, this yields the final worst-case $\widetilde O(n^{13/12+o(1)})$ private cut-release error. △ Less

Submitted 21 July, 2026; originally announced July 2026.

Comments: 77 pages

SteinGate: Tail-Sensitive Safe Reinforcement Learning via Stein Discrepancy

Authors: Yassine Chemingui, Chenhua Fan, Honghao Wei, Janardhan Rao Doppa

Abstract: Safe reinforcement learning typically enforces safety by bounding expected cumulative costs, a criterion that often fails to detect rare but catastrophic tail events. To overcome these limitations, this paper introduces SteinGate, a boundary-aware distributional safety certificate that replaces fragile tail fitting with a robust consistency check using Kernelized Stein Discrepancy while accounting… ▽ More Safe reinforcement learning typically enforces safety by bounding expected cumulative costs, a criterion that often fails to detect rare but catastrophic tail events. To overcome these limitations, this paper introduces SteinGate, a boundary-aware distributional safety certificate that replaces fragile tail fitting with a robust consistency check using Kernelized Stein Discrepancy while accounting for boundary atoms induced by clipped costs. SteinGate evaluates whether observed policy rollout costs remain consistent with a safe reference distribution, providing a non-parametric safety certificate. This certificate is used to dynamically adapt the learning regime: favoring reward-improving policy updates when rollouts remain consistent with the safe reference and switching to recovery behavior when the cost tail deviates. Experiments on continuous-control benchmarks demonstrate that SteinGate significantly reduces both the frequency and severity of constraint violations during training while maintaining competitive returns relative to state-of-the-art baselines. △ Less

Submitted 14 July, 2026; originally announced July 2026.

Comments: Accepted for Publication at the 42nd Conference on Uncertainty in Artificial Intelligence (UAI), 2026

Isolation as a First-Class Principle for LLM-Agent System Safety: Concepts, Taxonomy, Challenges and Future Directions

Authors: Huihao Jing, Wenbin Hu, Shaojin Chen, Haochen Shi, Sirui Zhang, Hanyu Yang, Changxuan Fan, Zhongwei Xie, Hongyu Luo, Wun Yu Chan, Wei Fan, Haoran Li, Yangqiu Song

Abstract: The capability of LLM agents to function as the ``brain'' of a system fundamentally expands the scope of analysis beyond a standalone model. Consequently, safety is no longer only about input--output content alignment. It also concerns system behavior and real-world execution outcomes. However, the current literature is fragmented across attack types, applications, and benchmarks. This makes it ha… ▽ More The capability of LLM agents to function as the ``brain'' of a system fundamentally expands the scope of analysis beyond a standalone model. Consequently, safety is no longer only about input--output content alignment. It also concerns system behavior and real-world execution outcomes. However, the current literature is fragmented across attack types, applications, and benchmarks. This makes it hard to explain why failures such as prompt injection, tool misuse, and memory poisoning often share the same structural cause, and how they spread through an agent workflow. In this survey, we treat isolation as a first-class principle for LLM-agent system safety. By isolation, we refer to the separation of user inputs, tool access, execution channels, inter-agent communication, and environment-originated context. We organize the literature with a boundary-centric taxonomy of five boundaries: user-agent, agent-tool, agent-execution, agent-agent, and system-environment. This view helps identify where the loss of isolation first occurs, how compromise propagates across boundaries, and which defenses are most relevant at each interface. We also summarize cross-boundary failure paths, discuss open challenges, and outline a research agenda for isolation-by-construction in future agent systems. △ Less

Submitted 14 July, 2026; originally announced July 2026.

LongE2V: Long-Horizon Event-based Video Reconstruction, Prediction, and Frame Interpolation with Video Diffusion Models

Authors: Cheng-De Fan, Chun-Wei Tuan Mu, Chen-Wei Chang, Chin-Yang Lin, Kun-Ru Wu, Yu-Chee Tseng, Yu-Lun Liu

Abstract: Recovering high-quality video from sparse event streams is a challenging task. Regression methods often blur textures, while existing generative models struggle with long-term stability. We propose LongE2V, a novel approach that leverages pre-trained video diffusion priors to jointly handle event-based video reconstruction, prediction, and frame interpolation. By fine-tuning a foundational video m… ▽ More Recovering high-quality video from sparse event streams is a challenging task. Regression methods often blur textures, while existing generative models struggle with long-term stability. We propose LongE2V, a novel approach that leverages pre-trained video diffusion priors to jointly handle event-based video reconstruction, prediction, and frame interpolation. By fine-tuning a foundational video model, our approach achieves high data efficiency and superior perceptual quality. We introduce Autoregressive Unrolling and Adaptive Context Switching to mitigate temporal drift in extremely long sequences. We also propose Reencoding Alignment with Cross Residual Correction to ensure precise bidirectional consistency during frame interpolation. Furthermore, Event Voxel Density Augmentation ensures robustness across varying sensor resolutions. Extensive experiments on real-world benchmarks demonstrate that LongE2V outperforms state-of-the-art methods across all three tasks, exhibiting exceptional temporal coherence and zero-shot generalization. Project page: https://cdfan0627.github.io/LongE2V-page/ △ Less

Submitted 9 July, 2026; originally announced July 2026.

Comments: SIGGRAPH 2026. Project page: https://cdfan0627.github.io/LongE2V-page/

Towards Accurate and Fast Clinical Body Composition: A Resource-Efficient Hierarchical Segmentation Framework for Multi-Source CT

Authors: Xiaodi Shen, Qingzhu Zheng, Yaoyang Qiu, Cien Fan, Ruonan Zhang, Yangdi Wang, Luyao Wu, Weikai Zheng, Longfei Zhao, Bing Li, Rulin Xu, Qiqi Xu, Ren Mao, Shiting Feng, Xuehua Li

Abstract: Background: Automated 3D segmentation of muscles and adipose tissue from CT is vital for body composition analysis, but multi-source data heterogeneity and high CPU memory demands hinder clinical deployment. Methods: We propose a coarse-to-fine hierarchical framework to segment ten tissue structures. Efficiency is optimized using Dynamic Spacing and Anisotropic Patching, a Group Inference mechan… ▽ More Background: Automated 3D segmentation of muscles and adipose tissue from CT is vital for body composition analysis, but multi-source data heterogeneity and high CPU memory demands hinder clinical deployment. Methods: We propose a coarse-to-fine hierarchical framework to segment ten tissue structures. Efficiency is optimized using Dynamic Spacing and Anisotropic Patching, a Group Inference mechanism for low-memory sliding-window processing, and Topology-Aware Asymmetric Resampling for fast post-processing. Results: The framework was trained on 1,558 CT volumes from seven public and two private datasets, and evaluated on an independent test cohort (N=105), per-structure Dice coefficients ranged from 0.924 to 0.982. Eight major structures met the +-10% relative error clinical acceptance limit. On a 12-core CPU workstation, the GPU-free pipeline averaged 44.5 seconds per volume with 4.73 GB peak memory. Conclusion: This framework balances accuracy and efficiency, enabling robust, large-scale body composition analysis on standard CPU workstations. △ Less

Submitted 8 July, 2026; originally announced July 2026.

Comments: Affiliations: (1) Department of Radiology, The First Affiliated Hospital, Sun Yat-sen University, Guangzhou 510080, China. (2) Research & Development Center, Canon Medical Systems (China) Co. Ltd. Beijing 100015, China

UP: Unbounded Positive Asymmetric Optimization for Breaking the Exploration-Stability Dilemma

Authors: Chongyu Fan, Pengfei Liu, Jingjia Huang, Sijia Liu, Yi Lin

Abstract: Reinforcement learning (RL) has become the standard paradigm for enhancing the complex reasoning capabilities of large language models (LLMs). To achieve sample efficiency, modern RL frameworks rely on importance sampling (IS). However, these algorithms suffer from an exploration-stability dilemma. Pure IS often leads to catastrophic training instability, while standard clipping mechanisms used to… ▽ More Reinforcement learning (RL) has become the standard paradigm for enhancing the complex reasoning capabilities of large language models (LLMs). To achieve sample efficiency, modern RL frameworks rely on importance sampling (IS). However, these algorithms suffer from an exploration-stability dilemma. Pure IS often leads to catastrophic training instability, while standard clipping mechanisms used to mitigate this instability strictly constrain the policy update budget. By formalizing the concept of Probability Capacity (Cap), we reveal that conservative clipping structurally stifles exploration by prematurely truncating the update budget for correct but low-confidence reasoning paths. To break free from these constraints, we propose Unbounded Positive Asymmetric Optimization (UP), a universal and plug-and-play objective. UP theoretically restructures the optimization process by anchoring the policy to its current state via the stop-gradient operator. This asymmetric design unleashes unclipped, stable gradients for positive advantages to maximize exploration, while maintaining standard clipping safeguards for negative advantages to prevent training instability. Furthermore, our formulation readily extends across different optimization granularities, including token-level (GRPO, DAPO) and sequence-level (GSPO) frameworks. Extensive experiments demonstrate that UP enhances exploration capacity and achieves superior reasoning accuracy across diverse RL algorithms (DAPO, GSPO, and GRPO), model architectures (Dense, MoE, and vision-language), and training modalities (language and multimodal), validating UP as a truly universal plug-and-play enhancement for RL-based training. △ Less

Submitted 8 July, 2026; originally announced July 2026.

Gemma 4 Technical Report

Authors: Gemma Team, Sherif El Abd, Vaibhav Aggarwal, Robin Algayres, Alek Andreev, Olivier Bachem, Ian Ballantyne, Cormac Brick, Victor Cărbune, Michelle Casbon, Mayank Chaturvedi, Aditya Chawla, Victor Cotruta, Alice Coucke, Phil Culliton, Robert Dadashi, Lucas Dixon, Mohamed Elhawaty, Utku Evci, Clément Farabet, Johan Ferret, Filippo Galgani, Sertan Girgin, Jean-Bastien Grill, Maarten Grootendorst , et al. (298 additional authors not shown)

Abstract: We introduce Gemma 4, a new generation of open-weight, natively multimodal language models in the Gemma model family. Designed to advance compute efficiency and reasoning, the Gemma 4 model suite features dense and Mixture-of-Experts architectures, ranging from 2.3B to 31B parameters. Alongside improved vision and audio encoders for all model sizes, we propose a unified, encoder-free architecture… ▽ More We introduce Gemma 4, a new generation of open-weight, natively multimodal language models in the Gemma model family. Designed to advance compute efficiency and reasoning, the Gemma 4 model suite features dense and Mixture-of-Experts architectures, ranging from 2.3B to 31B parameters. Alongside improved vision and audio encoders for all model sizes, we propose a unified, encoder-free architecture for our 12B model, which ingests raw audio and image patches. Furthermore, we integrate a thinking mode, enabling Gemma models to generate reasoning traces prior to responding. We improve inference speed, memory, and compute efficiency, as well as long-context abilities through critical design choices. Gemma 4 establishes a leap in performance across STEM, multimodal, and long-context benchmarks, and rivals larger, frontier open models in human-rated tasks. △ Less

Submitted 24 July, 2026; v1 submitted 2 July, 2026; originally announced July 2026.

Comments: 17 pages, 2 figures, technical report, updated

Absorption-Feature-Guided Distance-Decoupled Estimation and Band Selection for LWIR Hyperspectral Passive Ranging

Authors: Shuo Liu, Chen Fan, Zhihe Chen, Xiaolin Huang, Lilian Zhang

Abstract: Long-wave infrared (LWIR) hyperspectral observations contain distance-dependent atmospheric absorption signatures, providing a physical basis for long-range passive ranging. However, in natural scenes, these signatures are nonlinearly coupled with target temperature, material emissivity, and path radiance, making distance inversion from observed radiance ill posed. Existing methods typically rely… ▽ More Long-wave infrared (LWIR) hyperspectral observations contain distance-dependent atmospheric absorption signatures, providing a physical basis for long-range passive ranging. However, in natural scenes, these signatures are nonlinearly coupled with target temperature, material emissivity, and path radiance, making distance inversion from observed radiance ill posed. Existing methods typically rely on full-band measurements and pixel-wise joint optimization, which is computationally expensive and does not explicitly exploit sharp atmospheric absorption structures. This paper proposes an Absorption-Guided Distance-Decoupled Estimation and Refinement (ADER) framework for LWIR hyperspectral passive ranging. ADER represents emissivity with B-spline control points under a smoothness prior, suppressing overfitting to atmospheric absorption structures and enabling distance-decoupled estimation. It further uses ozone-absorption cues to classify pixels into emission-dominant and reflection-dominant groups. For emission-dominant pixels, ADER compensates path radiance and transmittance and estimates distance by one-dimensional absorption-residual minimization. For reflection-dominant pixels, ADER refines the initial estimate using downwelling-radiance compensation based on the complete radiative model. To reduce spectral redundancy, ADER also introduces a greedy band selection strategy based on multi-scene effective Fisher information for the distance parameter. Experiments on real scenes show that ADER recovers LiDAR-consistent spatial distance structures under both full-band and 20-band settings, improves ranging accuracy in the evaluated regions, and achieves approximately two orders of magnitude speedup over a public full-band hyperspectral ranging method. △ Less

Submitted 30 June, 2026; originally announced June 2026.

Comments: 18 pages, 9 figures

Scenario Generation for Testing of Autonomous Driving Systems Using Real-World Failure Records

Authors: Anjali Parashar, Chuchu Fan

Abstract: To ensure safe on-road behavior, pre-deployment testing and failure discovery of Autonomous Driving Systems (ADS) is crucial. Present day simulation based testing methods focus largely on mathematical models for efficient search of optimal scenarios, assuming a fixed scenario representation. On the other hand, real-world testing involves substantial manual effort to design scenario templates for t… ▽ More To ensure safe on-road behavior, pre-deployment testing and failure discovery of Autonomous Driving Systems (ADS) is crucial. Present day simulation based testing methods focus largely on mathematical models for efficient search of optimal scenarios, assuming a fixed scenario representation. On the other hand, real-world testing involves substantial manual effort to design scenario templates for testing. These templates represent distinct failure scenarios consisting of pre-deployment vehicle movements, map types, etc. Historical failure records for ADS are a reliable source of real-world failure conditions, which can be used for scenario generation. In this work, we propose a scenario generation pipeline using categorical and contextual information available from historical records in natural language format. Our approach consists of modular LLM based synthetic scenario generation, compatible with the testing constraints of a given system. We successfully apply our method to generate a diverse set of scenarios for testing autonomous navigation on Metadrive simulator using the NHTSA ADS crash records. Our approach results in accurate and diverse scenario generation with a combination of 4 road types, 3 non ego vehicle movement types, including on road anomalies in the form of working zones. Generated scenarios align with the provided testing conditions, and reveals interesting failures of the system within a limited testing budget of 20 scenarios. Code is available at https://github.com/anjaliParashar/crash2scenario. △ Less

Submitted 30 June, 2026; originally announced June 2026.

Comments: 9 pages, Appendix included. Paper accepted and presented at NeuS 2026

GeoT2V-Bench: Benchmarking 3D Consistency in Text-to-Video Models via 3D Reconstruction

Authors: Chenrui Fan, Paolo Favaro

Abstract: Camera-prompted text-to-video (T2V) models are increasingly used to synthesize virtual camera captures, such as orbiting objects or moving through static scenes. For these outputs, visual plausibility is insufficient: the generated frames should also provide coherent multi-view evidence for a single static 3D scene. We introduce GeoT2V-Bench, a reconstruction-based diagnostic benchmark for evaluat… ▽ More Camera-prompted text-to-video (T2V) models are increasingly used to synthesize virtual camera captures, such as orbiting objects or moving through static scenes. For these outputs, visual plausibility is insufficient: the generated frames should also provide coherent multi-view evidence for a single static 3D scene. We introduce GeoT2V-Bench, a reconstruction-based diagnostic benchmark for evaluating whether camera-prompted T2V clips can support explicit rigid 3D reconstruction. Our pipeline estimates per-frame camera intrinsics and poses with VGGT-style geometry estimation, fits DeformableGS, derives a static MedianGS proxy by temporal-median aggregation, and renders this proxy along the estimated camera path. Instead of producing a pass/fail label or a single scalar score, GeoT2V-Bench reports a continuous reconstruction profile covering apparent image motion, estimated trajectory behavior, MedianGS static rendering error, static-render flow agreement, and the gap between flexible and static fits. On a fair-format four-seed evaluation with 3,840 completed reconstructions from 12 open-weight model configurations and 80 GeCo-Eval static-scene prompts, we find that visible motion, static rendering error, flow agreement, and flexible-vs-static behavior often disagree. GeoT2V-Bench therefore captures complementary failure modes that emerge when generated videos are tested as global static-scene acquisitions. △ Less

Submitted 23 June, 2026; originally announced June 2026.

Comments: 36 pages, 17 figures, 18 tables

OrbitForge: Text-to-3D Scene Generation via Reconstruction-Anchored Video Synthesis

Authors: Chenrui Fan, Paolo Favaro

Abstract: Generic text-to-video models can be used as rich open-world scene priors. Despite the high quality of today's generated videos, they do not directly yield reliable 3D assets: camera motion is difficult to control, view coverage is partial, and frames often contain inconsistencies across time. We introduce OrbitForge, an adapter built from frozen video priors and per-prompt Gaussian Splatting recon… ▽ More Generic text-to-video models can be used as rich open-world scene priors. Despite the high quality of today's generated videos, they do not directly yield reliable 3D assets: camera motion is difficult to control, view coverage is partial, and frames often contain inconsistencies across time. We introduce OrbitForge, an adapter built from frozen video priors and per-prompt Gaussian Splatting reconstruction optimization that converts a single text-generated video into a canonical closed-orbit 3D Gaussian Splatting scene. We use 3D reconstruction as an anchor to improve the 3D consistency of the generated video. We obtain a preliminary 3D reconstruction from a first generated video via Deformable Gaussian Splatting with a robust MedianGS proxy. We render views from a prescribed orbit to detect missing viewpoints. OrbitForge uses the text-to-video model to complete only the missing views, and reconstructs the completed orbit into a final Gaussian Splatting scene. This design requires no task-specific video or multiview fine-tuning, avoids per-prompt score-distillation optimization, and does not progressively generate views one step at a time. We further argue that this setting demands coverage-aware evaluation: local smoothness alone rewards methods that never attempt a full orbit. On a frozen 300-prompt T3Bench-derived audit, OrbitForge reconstruction attains a 359.0-degree measured median span, raises originally unsupported-bin Q10 ImageReward from 8.07 to 16.36 relative to MedianGS-only reconstruction, while remaining competitive with VideoMV on the coverage-quality. △ Less

Submitted 23 June, 2026; originally announced June 2026.

Comments: 40 pages, 33 figures, 19 tables

IOI: Decoupling Kinematics and Physics for Interactive World Models

Authors: Chengyu Bai, Peidong Jia, Tiecheng Guo, Yukai Wang, Rui Ma, Fangyuan Zhao, Chunkai Fan, Xiaobao Wei, Jintao Chen, Hao Wang, Ying Li, Xiaozhu Ju, Jian Tang, Shanghang Zhang

Abstract: Developing generalist embodied agents requires interactive environments providing visually realistic feedback and accurate action-conditioned dynamics. Interactive world models address this by simulating such complex dynamics. However, purely data-driven methods struggle to ensure precise control alignment and physically plausible visual feedback due to a lack of explicit structural constraints. T… ▽ More Developing generalist embodied agents requires interactive environments providing visually realistic feedback and accurate action-conditioned dynamics. Interactive world models address this by simulating such complex dynamics. However, purely data-driven methods struggle to ensure precise control alignment and physically plausible visual feedback due to a lack of explicit structural constraints. To address this, we propose IOI, a hybrid interactive world model integrating analytical kinematic priors with learned physical dynamics. Unlike data-driven approaches prone to spatiotemporal drift, IOI introduces explicit kinematic guidance, computing forward kinematics from action sequences for accurate motion trajectories. These trajectories are rendered into synchronized front, side, and top orthographic projections, eliminating the need for extrinsic camera calibration. A Multi-view Kinematic Aggregation and Injection module fuses these geometric cues and injects them into the video generator, providing geometry-consistent guidance. Conditioning video generation on these deterministic trajectories establishes a synergy between the analytical simulator and the world model. Decoupling deterministic motion into the kinematic prior frees the generator to model stochastic physical interactions. Experiments on the RoboTwin benchmark validate IOI across kinematic fidelity, out-of-distribution (OOD) generalization, and policy evaluation. IOI achieves state-of-the-art simulation performance and robust zero-shot generalization to unseen OOD tasks. Furthermore, IOI serves as a reliable policy evaluator, yielding success rates closely aligning with ground-truth physics simulators. On real-world platforms, policies trained on IOI-synthesized data match those trained on teleoperation demonstrations, solidifying its practical value for embodied policy learning. △ Less

Submitted 22 June, 2026; originally announced June 2026.

MV-WAM: Manifold-Aware World Action Model with Value Augmentation

Authors: Jintao Chen, Peidong Jia, Qingpo Wuwu, Jiaming Liu, Mengfei Du, Chun-Kai Fan, Xiaowei Chi, Hao Chen, Chengyu Bai, Zezhong Qian, Hao Wang, Jiajun Cao, Weishi Mi, Xiaozhu Ju, Jian Tang, Shanghang Zhang

Abstract: Achieving robust and generalizable manipulation across diverse environments remains a fundamental challenge in embodied robotics. Recent world action models achieve strong in-domain performance, yet their gains do not extend proportionally to out-of-distribution scenarios. We attribute this to a structural mismatch between visual and action modalities, whose intrinsically heterogeneous manifolds c… ▽ More Achieving robust and generalizable manipulation across diverse environments remains a fundamental challenge in embodied robotics. Recent world action models achieve strong in-domain performance, yet their gains do not extend proportionally to out-of-distribution scenarios. We attribute this to a structural mismatch between visual and action modalities, whose intrinsically heterogeneous manifolds cause joint optimization to disproportionately degrade action robustness under distribution shift. To address this, we propose MV-WAM, a novel end-to-end framework that jointly models visual prediction, action generation, and value estimation designed to effectively leverage video priors during both training and inference for enhanced action generalization. Key to this unification is a cross-modality causal mask that hierarchically grounds actions in predicted video frames and value function tokens in both modalities. To further narrow the generalization gap, MV-WAM adopts a manifold-aware optimization scheme that explicitly accounts for the structural heterogeneity across modalities. Finally, MV-WAM introduces a progress-value regulation mechanism that estimates task completion and detects misalignment between predicted frames and generated actions, enabling the policy to autonomously identify execution deviations and recover through value-guided rollback. On the RoboTwin simulation, MV-WAM achieves a 55.7% mean success rate on random scenarios without any randomized action supervision, outperforming the strongest baseline by 29.3%. MV-WAM achieves a 77.5% mean success rate across four real-world tasks of varying difficulty on a dual-arm robot. Our results demonstrate that manifold-aware cross-modal alignment is essential for robust policy generalization, offering a path toward deployable robotic manipulation. △ Less

Submitted 19 June, 2026; originally announced June 2026.

Comments: 20 pages, 9 figures, 7 tables

SafeDojo: Safe Reinforcement Learning for VLA via Interactive World Model

Authors: Kai Tang, Peidong Jia, Zhong Chu, Jixian Wu, Rui Ma, Jiajun Cao, Fangyuan Zhao, Sixiang Chen, Yichen Guo, Xiaowei Chi, Chun-Kai Fan, Kevin Zhang, Jinchang Xu, Fubing Yang, Weishi Mi, Xiaozhu Ju, Jian Tang, Shanghang Zhang

Abstract: Safe control is a prerequisite for real-world embodied intelligence, for which safe reinforcement learning has emerged as a promising paradigm. However, existing safe reinforcement learning methods either require costly real-world exploration or depend on hand-crafted safety functions. Neither scales to vision-language-action models deployed in open-world physical environments. We propose SafeDojo… ▽ More Safe control is a prerequisite for real-world embodied intelligence, for which safe reinforcement learning has emerged as a promising paradigm. However, existing safe reinforcement learning methods either require costly real-world exploration or depend on hand-crafted safety functions. Neither scales to vision-language-action models deployed in open-world physical environments. We propose SafeDojo, the first model-based safe reinforcement learning framework for vision-language-action policies designed to learn safe actions through world model-based imagination. Specifically, SafeDojo performs online reinforcement learning on top of an interactive video world model. The world model generates action-conditioned future predictions, from which a tailored ResNet success classifier estimates per-step task progress from imagined frames and a lightweight safety head predicts per-step safety costs from latent context together with the proposed action chunk, enabling simultaneous assessment of task execution and trajectory safety. The decoupled task-reward and safety-cost signals are balanced through a Lagrangian-based constrained GRPO objective, enabling coordinated improvement of task success and safety under explicit constraints. On SafeLIBERO, SafeDojo achieves the best aggregate task success, safe success, and execution efficiency among inference-time safety, model-free RL, and model-based RL baselines, with the best average safe-success rate on both levels and an 8.25 percentage-point improvement over the strongest baseline on Level I. Real-world Franka deployment further shows the best average task and safe-success rates across five tasks. Our results position world model-based safe reinforcement learning as a scalable and generalizable path toward safe embodied intelligence. △ Less

Submitted 15 June, 2026; originally announced June 2026.

Comments: 20 pages, 5 figures, 8 tables

From Question Answering to Task Completion: A Survey on Agent System and Harness Design

Authors: Jianyuan Guo, Zhiwei Hao, Chengcheng Wang, Cheng Fan, Tingzhang Luo, Hongguang Li, Ying Gao, Hefei Mei, Jiankun Peng, Rongjian Xu, Minjing Dong, Han Wu, Mengyu Zheng, Kai Han, Shiqi Wang, Chang Xu, Yunhe Wang

Abstract: LLM-based agents mark a shift from passive question answering to active task completion: they perceive environments, invoke tools, maintain state, and act over extended horizons. As agent systems have evolved from prompt engineering to workflows and context engineering, harness engineering, and agent-native training with co-evolution, a central question has become increasingly important: where doe… ▽ More LLM-based agents mark a shift from passive question answering to active task completion: they perceive environments, invoke tools, maintain state, and act over extended horizons. As agent systems have evolved from prompt engineering to workflows and context engineering, harness engineering, and agent-native training with co-evolution, a central question has become increasingly important: where does the bottleneck in agent performance reside, in the foundation model, in the execution harness, or in the coupling between them? This survey examines LLM-based agents through a model-harness lens. We first clarify the functional definition of agents and the implementation view of an LLM-based agent as a foundation model coupled with an execution harness. We then analyze the limits of model-centric scaling, trace four paradigms of agent engineering, and decompose the execution harness into six coupled runtime responsibilities: observation, context, control, action, state, and verification. Using this decomposition, we map task properties and domain pressures to harness configurations, review benchmark and evaluation practices, and synthesize model-harness evidence on how runtime design affects long-horizon task completion, efficiency, and reliability. Finally, we identify open challenges in value-aware evaluation, safety, harness generalization, and model-harness co-evolution. Rather than treating agents as models with auxiliary tools, this survey argues that agent quality -- including success, efficiency, safety, and generalization -- emerges from the interaction between model capability, runtime infrastructure, task structure, and evaluation design. A collection of papers discussed in this survey is provided in https://github.com/ggjy/Awesome-Agent-Engineering. △ Less

Submitted 14 June, 2026; originally announced June 2026.

ResAware: Cross-Environment Website Fingerprinting via Resource-Privileged Distillation

Authors: Chongru Fan, Wei Wang, Wentao Huang, Zhenquan Ding, Jinqiao Shi, Lei Cui, Zhiyu Hao, Xiaochun Yun

Abstract: While Website Fingerprinting (WF) attacks achieve high accuracy in controlled laboratory settings, they often degrade substantially in real-world environments due to spatio-temporal drift, browser heterogeneity, proxy obfuscation and etc. This limitation stems from their sole reliance on low-level traffic features that are noisy and highly sensitive to environmental perturbations. To address this… ▽ More While Website Fingerprinting (WF) attacks achieve high accuracy in controlled laboratory settings, they often degrade substantially in real-world environments due to spatio-temporal drift, browser heterogeneity, proxy obfuscation and etc. This limitation stems from their sole reliance on low-level traffic features that are noisy and highly sensitive to environmental perturbations. To address this problem, we propose \textbf{ResAware}, a cross-environment resource-aware distillation framework under a \textit{training-rich/inference-poor} asymmetric setting. Specifically, ResAware trains a teacher model on resource-level features, and then distills the resulting privileged knowledge into a student model through heterogeneous knowledge distillation. At deployment time, the student model performs inference using only encrypted traffic, incurring zero additional cost. We evaluate ResAware on a large-scale dataset collected over five months from six globally distributed vantage points, comprising more than $160{,}000$ paired samples. The results show that ResAware significantly enhances the cross-environment robustness of diverse WF baselines. Under a 150-day temporal drift, for example, ResAware improves the F1-score of Var-CNN from $72.77\%$ to $81.49\%$ and the open-world $TPR@1\%FPR$ from $22.40\%$ to $27.20\%$. Our results demonstrate that resource-level supervision improves WF robustness without expanding online observation capabilities. △ Less

Submitted 15 June, 2026; originally announced June 2026.

Comments: 18 pages, 9 figures

When is Your LLM Steerable?

Authors: Chenrui Fan, Yize Cheng, Ming Li, Soheil Feizi, Tianyi Zhou

Abstract: Activation steering offers a lightweight approach to control language models' behavior at inference time, but whether it succeeds or fails heavily depends on the prompt, concept, model, and steering configuration. Finding the regime and boundaries of successful steering typically requires expensive grid searches and post-hoc evaluation of full autoregressive rollouts. In this work, we investigate… ▽ More Activation steering offers a lightweight approach to control language models' behavior at inference time, but whether it succeeds or fails heavily depends on the prompt, concept, model, and steering configuration. Finding the regime and boundaries of successful steering typically requires expensive grid searches and post-hoc evaluation of full autoregressive rollouts. In this work, we investigate whether steerability can be predicted from the model's internal states at the beginning of the generation process, e.g., after generating the first few tokens, and how to leverage such a predictor to improve steering success rate. To this end, we first introduce ASTEER, a testbed including 1.4M steered generations, spanning 150 concepts with each steering success/failure labeled. Leveraging this testbed, we analyze the model's early decoding dynamics by extracting features that compare hidden states before and after steering across layers and initial decoding steps. These features help us understand how steering's effects propagate along layers and token positions, which provide key information for steerability prediction. We then train a Gradient Boosting Decision Trees (GBDT) classifier on these features to predict whether an intervention will under-steer, succeed, or over-steer without requiring full rollout. Our predictor achieves around 0.7 macro-F1 score on unseen concepts, demonstrating that early hidden states encode substantial, structured information about eventual steering efficacy. We further leverage this steerability predictor as guidance for steering strength searching, achieving near-optimal performance with a small fraction of decoding cost. △ Less

Submitted 9 June, 2026; originally announced June 2026.

Optimizing Energy-based Neural Network Training with Coherent Ising Machine

Authors: Chen-Rui Fan, Bo Lu, Zhi-Hong Zhang, Run-Qing Zhang, Jing-Wei Wen, Chuan Wang

Abstract: While Ising machines serve as advanced physical solvers for the Ising model,enabling applications in combinatorial optimization and neural network training,their scalability for large-scale neural networks remains constrained by hardware connectivity limitations and suboptimal training methodologies. In this work,we leverage a Coherent Ising Machine (CIM) to train an energy-based neural network us… ▽ More While Ising machines serve as advanced physical solvers for the Ising model,enabling applications in combinatorial optimization and neural network training,their scalability for large-scale neural networks remains constrained by hardware connectivity limitations and suboptimal training methodologies. In this work,we leverage a Coherent Ising Machine (CIM) to train an energy-based neural network using Equilibrium Propagation, achieving performance comparable to existing software-based implementations. We further enhance the algorithm by integrating the Adam optimizer to solve for the ground state of a Hopfield energy network, significantly improving convergence speed and solution accuracy. Additionally, we demonstrate the scalability of our approach across deeper network architectures and convolutional operations. Our results highlight the potential of CIM dynamics as a scalable platform for training complex neural networks, offering a pathway toward energy-efficient implementations via analog circuits, optoelectronics, or integrated photonics. This work establishes a novel physical framework for next-generation AI hardware development. △ Less

Submitted 8 June, 2026; originally announced June 2026.

Hybridizing Equilibrium Propagation with Ising Machines for Efficient Energy-Based Learning

Authors: Chen-Rui Fan, Bo Lu, Xing-Yu Wu, Tie-Jun Wang, Chuan Wang

Abstract: The rapid evolution of artificial intelligence has led to substantial advances in deep neural networks. Nonetheless, conventional GPU-based training remains highly energy-demanding, motivating the exploration of physical dynamics and compatible energy-based learning schemes, such as equilibrium propagation (EP). EP-based training, however, frequently suffers from convergence to local minima due to… ▽ More The rapid evolution of artificial intelligence has led to substantial advances in deep neural networks. Nonetheless, conventional GPU-based training remains highly energy-demanding, motivating the exploration of physical dynamics and compatible energy-based learning schemes, such as equilibrium propagation (EP). EP-based training, however, frequently suffers from convergence to local minima due to phase-space contraction. Here we introduce an Ising-dynamics-inspired equilibrium-propagation framework in which dissipative Hopfield relaxation is replaced by an extended phase-space dynamics with conjugate variables. The resulting training paradigm keeps the local two-phase learning rule of EP while changing the physical route by which neural states reach equilibrium. We show that this dynamics lowers effective energy barriers, accelerates convergence, improves noise robustness, and trains deep convolutional Hopfield networks on MNIST, FashionMNIST, and CIFAR-10 with performance comparable to backpropagation. △ Less

Submitted 8 June, 2026; originally announced June 2026.

ExMesh: EXplicit Mesh Reconstruction with Topology Adaptation

Authors: Chuanjin Fan, Lifan Wu, Wenjie Chang, Hanzhi Chang, Wenfei Yang, Tianzhu Zhang

Abstract: Reconstructing surface meshes from multi-view images has remained a core challenge in recent years. Most existing methods, whether implicit or explicit, depend on intermediate representations and post-processing steps like Marching Cubes or TSDF fusion, often resulting in artifacts and fragmented geometry. Directly optimizing explicit meshes is a promising approach. However, it presents two critic… ▽ More Reconstructing surface meshes from multi-view images has remained a core challenge in recent years. Most existing methods, whether implicit or explicit, depend on intermediate representations and post-processing steps like Marching Cubes or TSDF fusion, often resulting in artifacts and fragmented geometry. Directly optimizing explicit meshes is a promising approach. However, it presents two critical challenges. The first is how to adaptively refine mesh topology to capture detail without introducing degenerate faces. The second is how to maintain consistent UV coordinates for high-fidelity texturing as the mesh structure evolves. To overcome these, we propose ExMesh, a novel framework that directly optimizes explicit meshes by integrating differentiable optimization with discrete topology updates. Specifically, we introduce an adaptive vertex splitting and merging strategy, along with real-time UV maintenance, to enable coarse-to-fine optimization while preserving geometric integrity. To our knowledge, ExMesh is the first framework to seamlessly integrate discrete topology operations into a continuous differentiable optimization pipeline. Extensive experiments demonstrate that ExMesh achieves a balance among accuracy, computational efficiency, and mesh conciseness. △ Less

Submitted 5 June, 2026; originally announced June 2026.

Comments: Accepted at the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2026 (CVPR 2026)

DBHN-Net: Dual-Branch Hybrid Neural Network For Low-Complexity Monaural Speech Enhancement

Authors: Cunhang Fan, Enrui Liu, Jing Zhou, Jian Kang, Jie Li, Andong Li, Jian Zhou, Zhao Lv, Xuelong Li

Abstract: Although artificial neural network (ANN) based speech enhancement (SE) methods demonstrate excellent performance, the high computational complexity and high energy consumption hinder their deployment in practical front-end processing tasks.} Currently, the spiking neural networks (SNNs) have shown potential in reducing power consumption. However, the discrete binary activation and complex spatio-t… ▽ More Although artificial neural network (ANN) based speech enhancement (SE) methods demonstrate excellent performance, the high computational complexity and high energy consumption hinder their deployment in practical front-end processing tasks.} Currently, the spiking neural networks (SNNs) have shown potential in reducing power consumption. However, the discrete binary activation and complex spatio-temporal dynamics of SNNs often result in information loss. The current challenge therefore focuses on how to maintain performance and reduce computational complexity. To address this issue, this work propose a Dual-Branch Hybrid Neural (DBHN) Network. 1) In terms of network architecture: A dual-branch network integrating ANN and SNN was designed, where the SNN branch reduces power consumption while the ANN branch addresses information loss; The BandSplit and Time-Frequency (TF) -Mamba modules were developed to simultaneously compress energy consumption and enhance model performance; Spiking Feature Extraction Group (SFEG) and Information Transformation Block (ITB) components were implemented with residual connections to mitigate information loss while further refining feature representations. 2) To facilitate inter-branch information fusion: An Interaction module was designed to promote information exchange at various stages of the dual-branch network; A TF-Cross Attention-Fusion module was designed to perform time-frequency domain fusion of dual-branch information while data-adaptively guiding the SNN branch to retain more critical information. Results show that the proposed model maintains superior performance across three public datasets while achieving an average 7.5 fold reduction in computational complexity compared to baseline models. △ Less

Submitted 4 June, 2026; originally announced June 2026.

Comments: This article has been accepted for publication in IEEE Transactions on Pattern Analysis and Machine Intelligence(TPAMI)

Journal ref: IEEE Transactions on Pattern Analysis and Machine Intelligence(TPAMI2026)

Trustworthy Recommendation in the Era of Large Language Models: Opportunities and Challenges

Authors: Bohao Wang, Yu Cui, Zhenxiang Xu, Jujia Zhao, Chenxiao Fan, Jizhi Zhang, Weiqin Yang, Shengjia Zhang, Sirui Chen, Yang Zhang, Xiaoyan Zhao, Wenjie Wang, Chongming Gao, Fuli Feng, Xiangnan He, Jiawei Chen

Abstract: The field of recommender systems (RS) is currently undergoing two profound paradigm shifts. From the perspective of objectives, the goal has shifted beyond mere recommendation accuracy to comprehensive trustworthiness, encompassing multiple dimensions such as robustness, fairness, and privacy preservation. From a technical perspective, Large Language Models (LLMs) have been extensively integrated… ▽ More The field of recommender systems (RS) is currently undergoing two profound paradigm shifts. From the perspective of objectives, the goal has shifted beyond mere recommendation accuracy to comprehensive trustworthiness, encompassing multiple dimensions such as robustness, fairness, and privacy preservation. From a technical perspective, Large Language Models (LLMs) have been extensively integrated into RS, reshaping the foundations of recommendation through richer semantic understanding, stronger intent reasoning, and more flexible user interactions. The convergence of these two shifts prompts a timely and pivotal question: how does the integration of LLMs reshape the landscape of trustworthy recommendation? In this work, we present a systematic review of trustworthy LLM-empowered recommendation. By comprehensively analyzing over 200 recent studies, we reveal that the introduction of LLMs acts as a double-edged sword. While their advanced mechanisms and user-friendly interfaces offer unprecedented opportunities to enhance trustworthiness, they simultaneously introduce new risks, such as novel forms of bias and hallucination-induced issues. To characterize this dual impact, we systematically identify 13 opportunities and 18 challenges across six fundamental dimensions of trustworthiness, and accordingly organize the existing literature into a novel taxonomy. We also provide a comprehensive review of commonly used datasets and evaluation metrics to facilitate empirical validation. Finally, we identify critical open challenges and outline future directions, hoping to inspire future research on this emerging topic. △ Less

Submitted 30 May, 2026; originally announced June 2026.

Generating Graph-Like Logical Rules for Knowledge Graph Reasoning via Diffusion Models

Authors: Haoxiang Cheng, Yunfei Wang, Chao Chen, Kewei Cheng, Zhipeng Lin, Haoxuan Li, Changjun Fan, Shixuan Liu

Abstract: Logical rules constitute a cornerstone of knowledge graph (KG) reasoning, valued for their interpretability and ability to model relational patterns. However, existing rule mining methods predominantly focus on simple chain-like rules and therefore neglect the richer relational information encoded in graph-like structures, such as cycles and branches. This limitation is further exacerbated by comp… ▽ More Logical rules constitute a cornerstone of knowledge graph (KG) reasoning, valued for their interpretability and ability to model relational patterns. However, existing rule mining methods predominantly focus on simple chain-like rules and therefore neglect the richer relational information encoded in graph-like structures, such as cycles and branches. This limitation is further exacerbated by computational bottlenecks caused by the combinatorial explosion of the search space, which is especially challenging for graph-like rules. Meanwhile, generative approaches such as diffusion models, despite their success in other domains, cannot be directly applied to rule mining because their training objectives are not aligned with the goal of learning high-quality rules, and non-differentiable KG rule quality metrics cannot directly guide model optimization. To address these limitations, we propose GRiD, a framework that reformulates graph-like rule discovery as a discrete generative process conditioned on the target relation. GRiD employs a two-phase training strategy. First, supervised pre-training enables GRiD to capture structural priors from subgraphs sampled from the KG meta-graph. Subsequently, reinforcement learning is applied to fine-tune GRiD through policy gradient optimization guided directly by non-differentiable rule-quality metrics. Experiments on six benchmark datasets show that GRiD achieves competitive performance on KG completion tasks. Ablation studies confirm the efficiency and robustness of GRiD and further show that graph-like rules complement chain-like rules in KG completion. Our code and datasets are available in https://github.com/Haoxiang-Cheng/GRiD. △ Less

Submitted 4 June, 2026; v1 submitted 28 May, 2026; originally announced May 2026.

Comments: accepted by KDD 26

STAP: A Shuffle-Tokenized App Predictor with Ultra Long Context for Vocabulary-Free Mobile App Prediction

Authors: Chengyu Fan, Hang Liu

Abstract: Predicting the next mobile application a user will launch is essential for intelligent device resource management and proactive assistance. Existing models rely on fixed app vocabularies, which prevents them from generalizing across different app ecosystems. Many also depend on user-specific knowledge, which complicates deployment in cold start scenarios. We propose STAP, a Transformer-based model… ▽ More Predicting the next mobile application a user will launch is essential for intelligent device resource management and proactive assistance. Existing models rely on fixed app vocabularies, which prevents them from generalizing across different app ecosystems. Many also depend on user-specific knowledge, which complicates deployment in cold start scenarios. We propose STAP, a Transformer-based model that eliminates the need for a fixed vocabulary. STAP replaces true app identities with randomly reassigned virtual indices via a shuffle mechanism, and compensates for discarded semantic information by processing behavioral sequences with an ultra-long context design. A theoretical analysis shows that, given a sufficiently long context, the predicted distribution converges to the correct one despite the anonymity of the mapping. Experiments on two datasets from different continents demonstrate that STAP achieves strong cross-dataset zero-shot prediction accuracy -- a setting where all existing fixed-vocabulary methods are inherently inapplicable -- while its cold start performance within each dataset remains competitive with leading models. Furthermore, we introduce a deployment strategy that enables the model to retain a sufficiently long context during continuous inference while keeping latency within acceptable bounds. △ Less

Submitted 28 May, 2026; originally announced May 2026.

Comments: 15 pages, 9 figures, 5 tables Preprint submitted to Expert Systems with Applications

SafeSABR: Risk-Calibrated Adaptive Bitrate Streaming over Starlink Networks

Authors: Hongjun Xie, Jiahang Zhu, Zhiming Shao, Chao Fan, Zenghui Zhang, Genke Yang, Pengcheng Luo

Abstract: Starlink, as a representative low Earth orbit (LEO) satellite broadband system, makes high-bitrate video streaming possible in regions where terrestrial broadband is unavailable. However, its access links exhibit rapid throughput fluctuations caused by satellite mobility and handovers. Existing learned adaptive bitrate (ABR) algorithms can achieve high average quality of experience (QoE), yet high… ▽ More Starlink, as a representative low Earth orbit (LEO) satellite broadband system, makes high-bitrate video streaming possible in regions where terrestrial broadband is unavailable. However, its access links exhibit rapid throughput fluctuations caused by satellite mobility and handovers. Existing learned adaptive bitrate (ABR) algorithms can achieve high average quality of experience (QoE), yet high-bitrate Starlink streaming exposes severe session-level rebuffering that is not captured by average QoE alone. To address it, this paper proposes SafeSABR, a risk-calibrated learned ABR framework for Starlink networks. SafeSABR formulates Starlink ABR as a QoE--severe-risk tradeoff and follows a three-stage design: behavior-cloning pretraining learns a high-QoE ABR prior, risk-calibrated reinforcement learning (RL) fine-tuning reduces severe-tail action tendencies, and a runtime safety auditor uses safe-capacity lower bounds to check policy-requested bitrates before execution. Experiments on real Starlink traces compare SafeSABR with online, prediction-assisted, and learned ABR baselines. Compared with advanced methods, SafeSABR reduces severe-stall sessions from 22.8% to 7.2% and worst-5% session rebuffering from 54.30 s to 22.68 s, with a 1.8% QoE cost. Component analyses further show that risk-calibrated fine-tuning and safe-capacity auditing reduce unsafe bitrate decisions and downstream severe-session rebuffering. These results show that combining risk-calibrated policy learning with decision-aware safe throughput forecasting can move learned ABR toward a safer QoE--severe-risk operating point under volatile Starlink networks. △ Less

Submitted 26 May, 2026; v1 submitted 22 May, 2026; originally announced May 2026.

GrandGuard: Taxonomy, Benchmark, and Safeguards for Elderly-Chatbot Interaction Safety

Authors: Changxuan Fan, Xi Yang, Yueyuan Zheng, Bin Zhou, Yuanping Wang, Wenbin Hu, Huihao Jing, Ki Sen Hung, Dazhao Du, Haoran Li, Janet Hui-wen Hsiao, Yangqiu Song

Abstract: As older adults increasingly use LLM-based chatbots for companionship and assistance, a safety gap is emerging. Older adults may face vulnerabilities from social isolation, limited digital literacy, and cognitive decline, yet existing safety benchmarks largely target general harms and overlook elderly-specific risks. For example, a prompt such as "how to repair a ceiling light alone in the dark" m… ▽ More As older adults increasingly use LLM-based chatbots for companionship and assistance, a safety gap is emerging. Older adults may face vulnerabilities from social isolation, limited digital literacy, and cognitive decline, yet existing safety benchmarks largely target general harms and overlook elderly-specific risks. For example, a prompt such as "how to repair a ceiling light alone in the dark" may be benign for most users but poses a serious fall risk for older adults with mobility limitations. We introduce GrandGuard, the first comprehensive framework for assessing and mitigating elderly-specific contextual risks in LLM interactions. We develop a three-level taxonomy with 50 fine-grained risk types across mental well-being, financial, medical, toxicity, and privacy domains, grounded in real-world incidents, community discussions, and analysis of stakeholder studies. Using this taxonomy, we construct a benchmark of 10,404 labeled prompts and responses, showing that several leading LLMs mishandle elderly-specific contextual risks in over 50% of cases. We mitigate these failures with two safeguards: a fine-tuned Llama-Guard-3 and a policy-enhanced gpt-oss-safeguard-20b, achieving up to 96.2% and 90.9% unsafe-prompt detection accuracy, respectively. GrandGuard lays the groundwork for AI systems that move beyond general safety to support aging populations. △ Less

Submitted 7 April, 2026; originally announced May 2026.

Implicit Action Chunking for Smooth Continuous Control

Authors: Bosun Liang, Shuo Pei, Zirui Chen, Chuanzhi Fan, Chen Sun, Yuankai Wu, Huachun Tan, Yong Wang

Abstract: Reinforcement learning often produces high-frequency oscillatory control signals that undermine the safety and stability required for physical deployment. Explicit action chunking addresses this by predicting fixed-horizon trajectories but scales the policy output dimension proportionally with the horizon length, leading to optimization difficulties and incompatibility with standard step-wise inte… ▽ More Reinforcement learning often produces high-frequency oscillatory control signals that undermine the safety and stability required for physical deployment. Explicit action chunking addresses this by predicting fixed-horizon trajectories but scales the policy output dimension proportionally with the horizon length, leading to optimization difficulties and incompatibility with standard step-wise interaction. To overcome these challenges, this paper proposes Dual-Window Smoothing (DWS), an implicit action chunking framework for smooth continuous control. Unlike explicit methods, DWS enforces temporal coherence without expanding the action space. It uses a dual-window design: an execution window that ensures physical smoothness through deterministic modulation, and a value window that aligns temporal-difference targets over the horizon to correct critic bias caused by open-loop execution. DWS also includes a lightweight actor-side temporal regularizer based on first-order action differences to promote global continuity. This design effectively bridges the gap between temporal abstraction and reactive step-wise control. Experiments on benchmarks including the DeepMind Control Suite and industrial energy management tasks show that DWS outperforms state-of-the-art (SOTA) baselines. In complex vision-based autonomous driving tasks, DWS achieves smoother control, safer behavior with reduced jitter, and attains a 100% success rate. △ Less

Submitted 19 May, 2026; originally announced May 2026.

Rethinking Muon Beyond Pretraining: Spectral Failures and High-Pass Remedies for VLA and RLVR

Authors: Chongyu Fan, Gaowen Liu, Mingyi Hong, Ramana Rao Kompella, Sijia Liu

Abstract: Muon is a matrix-aware optimizer that leverages Newton-Schulz (NS) iterations to enforce spectral gradient orthogonalization by driving all singular values of the momentum matrix toward 1. While this uniform spectral whitening enhances exploration and outperforms AdamW in LLM pretraining, we show it could lead to fundamental limitations beyond pretraining in two regimes: (i) cross-modality vision-… ▽ More Muon is a matrix-aware optimizer that leverages Newton-Schulz (NS) iterations to enforce spectral gradient orthogonalization by driving all singular values of the momentum matrix toward 1. While this uniform spectral whitening enhances exploration and outperforms AdamW in LLM pretraining, we show it could lead to fundamental limitations beyond pretraining in two regimes: (i) cross-modality vision-language-action (VLA) training, where inherently low-rank action-module gradients cause amplification of noisy tail directions, and (ii) reinforcement learning with verifiable rewards (RLVR), where low-SNR gradients and the need to preserve per-head specialization from prior training make whitening unstable. To address these challenges, we propose Pion, a drop-in replacement for Muon that preserves its computational efficiency while replacing uniform spectral whitening with a two-stage Promotion+Suppression mechanism, which we call the high-pass NS iteration. This design induces a sharp spectral high-pass effect, anchoring dominant singular values at 1 while suppressing noisy tail components toward 0, with controllable filter strength. To preserve pretrained per-head heterogeneity, Pion also supports a per-head mode that applies updates independently across attention heads via a simple reshape, at no extra cost. In VLA training on LIBERO and LIBERO-Plus, Pion consistently outperforms both baselines across l_1-regression (VLA-Adapter) and flow-matching (VLANeXt) architectures, e.g., reaching 100% success rate on LIBERO Object after 1,500 training steps with VLA-Adapter, vs. 97.0% for Muon and only 32.2% for AdamW. The advantage of Pion further extends to a real Franka Research 3 robot with a pi_0.5 backbone under the DROID setup on three grasp-and-place tasks. In RLVR post-training on Qwen3-1.7B/4B with GRPO and GMPO, Pion also outperforms AdamW on MATH and GSM8K while Muon collapses to zero. △ Less

Submitted 18 May, 2026; originally announced May 2026.

EVA01: Unified Native 3D Understanding and Generation via Mixture-of-Transformers

Authors: Zongyuan Yang, Mingjing Yi, Wanli Ma, Chenzhuo Fan, Bocheng Li, Baolin Liu, Yuke Lou, Yingde Song, Yongping Xiong, Zhengdong Guo, Shimu Wang

Abstract: This paper addresses the challenge of integrating 3D meshes as a native modality within Multimodal Large Language Models (MLLMs). Diffusion-based large reconstruction models decouple semantic understanding from geometric reasoning, operating as stateless reconstructors conditioned on dense 2D pixel priors. Recent MLLM-based methods treat the 3D modality as an external output rather than a native c… ▽ More This paper addresses the challenge of integrating 3D meshes as a native modality within Multimodal Large Language Models (MLLMs). Diffusion-based large reconstruction models decouple semantic understanding from geometric reasoning, operating as stateless reconstructors conditioned on dense 2D pixel priors. Recent MLLM-based methods treat the 3D modality as an external output rather than a native component of the multimodal sequence, making incremental adaptations without a systematic analysis of how geometric manifolds align with MLLM feature spaces. We introduce EVA01, a unified framework that extends the modality boundary of MLLMs to natively incorporate 3D mesh understanding, generation, and context-aware editing. Built upon a Mixture-of-Transformers (MoT) architecture, EVA01 decouples the model into a pre-trained Understanding Expert ($E_{\mathrm{und}}$) and a structurally mirrored Generation Expert ($E_{\mathrm{gen}}$), coupled through shared global self-attention with hard modality routing. This design aligns the semantic latent space of the MLLM backbone with the geometric manifold, enabling direct transfer of multimodal priors without intermediate 2D representations. Results show that EVA01 achieves state-of-the-art native text-to-3D generation fidelity and unlocks robust long-context multi-turn geometric editing with identity preservation, a capability fundamentally inaccessible to stateless reconstruction pipelines. Our findings further offer architectural insights for integrating 2D foundation models with 3D tasks, informing the design of 3D-native multimodal systems. Project Page: https://www.seeles.ai/research/pages/EVA01 △ Less

Submitted 15 May, 2026; originally announced May 2026.

Comments: 28 pages, 10 figures, 6 tables. Technical report

Position: Zeroth-Order Optimization in Deep Learning Is Underexplored, Not Underpowered

Authors: Sijia Liu, Yicheng Lang, Soumyadeep Pal, Changsheng Wang, Yancheng Huang, Chongyu Fan, James Diffenderfer, Bhavya Kailkhura, Yihua Zhang

Abstract: Zeroth-order (ZO) optimization, learning from finite differences of function evaluations without backpropagation, has recently regained attention in deep learning due to its memory efficiency and applicability to gray- or black-box pipelines. Yet, ZO methods are often dismissed as fundamentally unscalable because of estimator variance and unfavorable query complexity. We argue that this conclusion… ▽ More Zeroth-order (ZO) optimization, learning from finite differences of function evaluations without backpropagation, has recently regained attention in deep learning due to its memory efficiency and applicability to gray- or black-box pipelines. Yet, ZO methods are often dismissed as fundamentally unscalable because of estimator variance and unfavorable query complexity. We argue that this conclusion might be misguided: ZO optimization is underexplored, not underpowered. We show that many perceived limitations stem from myopic development practices, most notably full-space, element-wise, estimator-centric designs. We articulate six positions spanning the algorithmic, systems, and evaluation stack. First, we revisit the feasibility boundaries of estimator-centric ZO methods through variance control, variance-query tradeoffs, and directional-derivative lenses. Then, we identify three underexplored opportunities: (i) subspace and spectral views of ZO that enable interpretable variance reduction with graceful query scaling, (ii) the forward-only nature of ZO as a systems advantage for communication-efficient, pipeline-friendly, and resource-constrained training, and (iii) the need to de-obfuscate ZO evaluations from task complexity. We strongly advocate rethinking ZO optimization around its unique strengths and acting accordingly, opening a viable path toward large-scale, system-aware, and resource-efficient learning with ZO optimization. △ Less

Submitted 18 May, 2026; v1 submitted 15 May, 2026; originally announced May 2026.

Comments: Accepted by ICML 2026 Position Paper Track as a Spotlight Paper

Model-Adaptive Tool Necessity Reveals the Knowing-Doing Gap in LLM Tool Use

Authors: Yize Cheng, Chenrui Fan, Mahdi JafariRaviz, Keivan Rezaei, Soheil Feizi

Abstract: Large language models (LLMs) increasingly act as autonomous agents that must decide when to answer directly vs. when to invoke external tools. Prior work studying adaptive tool use has largely treated tool necessity as a model-agnostic property, annotated by human or LLM judge, and mostly cover cases where the answer is obvious (e.g., fetching the weather vs. paraphrasing text). However, tool nece… ▽ More Large language models (LLMs) increasingly act as autonomous agents that must decide when to answer directly vs. when to invoke external tools. Prior work studying adaptive tool use has largely treated tool necessity as a model-agnostic property, annotated by human or LLM judge, and mostly cover cases where the answer is obvious (e.g., fetching the weather vs. paraphrasing text). However, tool necessity in the wild is more nuanced due to the divergence of capability boundaries across models: a problem solvable by a strong model on its own may still require tools for a weaker one. In this work, we introduce a model-adaptive definition of tool-necessity, grounded in each model's empirical performance. Following this definition, we compare the necessity against observed tool-call behavior across four models on arithmetic and factual QA dataset, and find substantial mismatches of 26.5-54.0% and 30.8-41.8%, respectively. To diagnose the failure, we decompose tool use into two stages: an internal cognition stage that reflects whether a model believes a tool is necessary, and an execution stage that determines whether the model actually makes a tool-call action. By probing the LLM hidden states, we find that both signals are often linearly decodable, yet their probe directions become nearly orthogonal in the late-layer, last-token regime that drives the next-token action. By tracing the trajectory of samples in the two-stage process, we further discover that the majority of mismatch is concentrated in the cognition-to-action transition, not in cognition itself. These results reveal a knowing-doing gap in LLM tool-use: improving tool-use reliability requires not only better recognition of when tools are needed, but also better translation of that recognition into action. △ Less

Submitted 17 May, 2026; v1 submitted 13 May, 2026; originally announced May 2026.

DP-Muon: Differentially Private Optimization via Matrix-Orthogonalized Momentum

Authors: Jihwan Kim, Chenglin Fan

Abstract: We study differentially private (DP) training with Muon, a matrix-valued optimizer that updates hidden-layer weights using momentum followed by Newton--Schulz orthogonalization. While DP-SGD is well understood, the interaction between per-example clipping, Gaussian noise, momentum, and nonlinear orthogonalization in Muon has not been systematically analyzed. We formulate DP-Muon, a private Muon pr… ▽ More We study differentially private (DP) training with Muon, a matrix-valued optimizer that updates hidden-layer weights using momentum followed by Newton--Schulz orthogonalization. While DP-SGD is well understood, the interaction between per-example clipping, Gaussian noise, momentum, and nonlinear orthogonalization in Muon has not been systematically analyzed. We formulate DP-Muon, a private Muon procedure that clips per-example matrix gradients, adds Gaussian noise to the clipped lot average, and then applies momentum and Newton--Schulz orthogonalization as post-processing. We prove that DP-Muon inherits the privacy guarantee certified by the corresponding same-lot subsampled Gaussian accountant, with no additional privacy cost from Muon-specific post-processing. On the optimization side, we establish finite-horizon and vanishing stationarity guarantees under per-matrix clipping, with bounds that separate optimization error, clipping residual, privacy noise, and Newton--Schulz approximation error. We further show that the DP-induced bias in Muon arises not in the linear momentum buffer itself, but after the nonlinear Newton--Schulz map, where Gaussian noise induces a matrix-valued heat-smoothing bias. This motivates DP-MuonBC, a bias-corrected variant that removes the leading output-level bias term while preserving the same privacy guarantee. Experiments on E2E and DART show that Muon-style matrix updates improve private fine-tuning, and that DP-MuonBC further improves utility without increasing the privacy budget. △ Less

Submitted 13 May, 2026; originally announced May 2026.

Comments: 26 pages

GuidedVLA: Specifying Task-Relevant Factors via Plug-and-Play Action Attention Specialization

Authors: Xiaosong Jia, Bowen Yang, Zuhao Ge, Xian Nie, Yuchen Zhou, Cunxin Fan, Yufeng Li, Yilin Chai, Chao Jing, Zijian Liang, Qingwen Bu, Haidong Cao, Chao Wu, Qifeng Li, Zhenjie Yang, Chenhe Zhang, Hongyang Li, Zuxuan Wu, Junchi Yan, Yu-Gang Jiang

Abstract: Vision-Language-Action (VLA) models aim for general robot learning by aligning action as a modality within powerful Vision-Language Models (VLMs). Existing VLAs rely on end-to-end supervision to implicitly enable the action decoding process to learn task-relevant features. However, without explicit guidance, these models often overfit to spurious correlations, such as visual shortcuts or environme… ▽ More Vision-Language-Action (VLA) models aim for general robot learning by aligning action as a modality within powerful Vision-Language Models (VLMs). Existing VLAs rely on end-to-end supervision to implicitly enable the action decoding process to learn task-relevant features. However, without explicit guidance, these models often overfit to spurious correlations, such as visual shortcuts or environmental noise, limiting their generalization. In this paper, we introduce GuidedVLA, a framework designed to manually guide the action generation to focus on task-relevant factors. Our core insight is to treat the action decoder not as a monolithic learner, but as an assembly of functional components. Individual attention heads are supervised by manually defined auxiliary signals to capture distinct factors. As an initial study, we instantiate this paradigm with three specialized heads: object grounding, spatial geometry, and temporal skill logic. Across simulation and real-robot experiments, GuidedVLA improves success rates in both in-domain and out-of-domain settings compared to strong VLA baselines. Finally, we show that the quality of these specialized factors correlates positively with task performance and that our mechanism yields decoupled, high-quality features. Our results suggest that explicitly guiding action-decoder learning is a promising direction for building more robust and general VLA models. △ Less

Submitted 1 June, 2026; v1 submitted 12 May, 2026; originally announced May 2026.

Comments: Accepted to RSS 2026. Project page: https://guidedvla.github.io/project_page/

VEGA: Visual Encoder Grounding Alignment for Spatially-Aware Vision-Language-Action Models

Authors: Hao Wang, Xiaobao Wei, Jingyang He, Chengyu Bai, Chun-Kai Fan, Jiajun Cao, Jintao Chen, Ying Li, Shanyu Rong, Ming Lu, Xiaozhu Ju, Jian Tang, Shanghang Zhang

Abstract: Precise spatial reasoning is fundamental to robotic manipulation, yet the visual backbones of current vision-language-action (VLA) models are predominantly pretrained on 2D image data without explicit 3D geometric supervision, resulting in representations that lack accurate spatial awareness. Existing implicit spatial grounding methods partially address this by aligning VLA features with those of… ▽ More Precise spatial reasoning is fundamental to robotic manipulation, yet the visual backbones of current vision-language-action (VLA) models are predominantly pretrained on 2D image data without explicit 3D geometric supervision, resulting in representations that lack accurate spatial awareness. Existing implicit spatial grounding methods partially address this by aligning VLA features with those of 3D-aware foundation models, but they rely on empirical layer search and perform alignment on LLM-level visual tokens where spatial structure has already been entangled with linguistic semantics, limiting both generalizability and geometric interpretability. We propose VEGA (Visual Encoder Grounding Alignment), a simple yet effective framework that directly aligns the output of the VLA's visual encoder with spatially-aware features from DINOv2-FiT3D, a DINOv2 model fine-tuned with multi-view consistent 3D Gaussian Splatting supervision. By performing alignment at the visual encoder output level, VEGA grounds spatial awareness before any linguistic entanglement occurs, offering a more interpretable and principled alignment target. The alignment is implemented via a lightweight projector trained with a cosine similarity loss alongside the standard action prediction objective, and is discarded at inference time, introducing no additional computational overhead. Extensive experiments on simulation benchmark and real-world manipulation tasks demonstrate that VEGA consistently outperforms existing implicit spatial grounding baselines, establishing a new state-of-the-art among implicit spatial grounding methods for VLA models. △ Less

Submitted 11 May, 2026; originally announced May 2026.