Robotics
Learning Distance-Conditioned Object Transport for Humanoid Loco-Manipulation from a Single Motion Clip
Yuhyeon Hwang, Daniel Sungho Jung, YongHyeok Seo et al. (+4 more)
Motion tracking can reproduce
humanoid loco-manipulation from a single retargeted motion clip, but a policy trained on a fixed reference primarily reproduces its demonstrated transport outcome. Although the source trajectory visits intermediate object displacements, transport termination is demonstrated only at its endpoint. We identify this mismatch as the termination-versus-passage gap: intermediate displacements are observed as passage states rather than termination-complete outcomes. We introduce Distance-Conditioned Reference Recomposition (DCRR), which relocates the demonstrated termination segment to intermediate transport states. A frozen tracking teacher replays the recomposed references under closed-loop dynamics, and the retained trajectories are relabeled by their achieved object placements and distilled into a reference-free policy. This procedure constructs distance-conditioned supervision from the interaction behavior encoded in the source motion. Across Carry, Kick-Push, Crouch-Push, and Drag, DCRR-BC produces command-dependent transport with an overall normalized distance mean absolute error (MAE) of 0.15, compared with 0.28 for source-only behavior cloning. RL fine-tuning further improves the command response and execution robustness in the training simulator and under sim-to-sim transfer. Finally, hardware experiments demonstrate transport-distance modulation across all four interaction modes.
Read the full abstract ▾
All authors Yuhyeon Hwang, Daniel Sungho Jung, YongHyeok Seo, Mingi Jung, Chang Nho Cho, Jung-Hoon Hwang, Dongin Shin
Information Theory
Orbital Detection: On Maximum-Entropy Priors
Kuranage Roche Rayan Ranasinghe, Takumi Takahashi and Giuseppe Thadeu Freitas de Abreu
Soft-input detection over a discrete constellation \(\mathcal{M}\) of cardinality \(M\) requires computing a posterior whose mean and mode are respectively given by the minimum mean square error (MMSE) and maximum a posteriori (MAP) estimates, both of which incur a computational cost of order \(\mathcal{O}(M)\) per symbol. We show that this cost is reduced to \(\mathcal{O}(L)\), where \(L \le M\) is the number of distinct amplitudes (rings), once the discrete prior is replaced by its maximum-
entropy counterpart subject to the same radial marginal. This orbital prior, which is a mixture of uniform circular shells, is obtained by maximizing a mixed discrete-continuous
entropy. We prove in this paper that such a distribution is the only distribution on \(\mathbb{C}\) that preserves the amplitude statistics of \(\mathcal{M}\) exactly while remaining maximally noncommittal in phase. Under the additive white Gaussian noise (AWGN) channel, the orbital prior induces a closed-form posterior that factors into a softmax over the \(L\) rings and a von Mises phase distribution whose concentration is supplied entirely by the observation, yielding closed-form orbital MMSE and MAP detectors of the discrete symbol at \(\mathcal{O}(L)\) cost. The resulting hierarchical rule selects the ring by posterior mass and the phase by conditional mode. We compare the pairwise ring boundary with that of the joint posterior-density and quantify the leading-order outward shift at high signal-to-noise ratio (SNR). Numerical results using standard constellations confirm that the orbital detectors maintain similar symbol error rate (SER) performance to optimal detectors, at a fraction of the complexity.
Read the full abstract ▾
Artificial Intelligence · Cross-listed from eess.AS
OmniVChat: Synthesizing, Benchmarking, and Training for Native Audio-Visual Dialogue
Haolin He, Yunfei Chu, Qi Chen et al. (+15 more)
We define OmniVChat (Omni
Video Chat) as the task of native audio-visual dialogue between a user and an omni model. In OmniVChat, omni models directly and simultaneously receive audio and
video from a user and return text. The user's query is embedded in the audio and
video, without a separate text question, external captioning, or
speech recognition. Direct audio-visual input reduces external latency and computation while preserving perceptual cues. However, research on OmniVChat faces two constraints: data availability and evaluation. Recordings of people using their own devices are scarce. Furthermore, good replies often depend on
multimodal context and can be phrased in many ways, making keyword matching unreliable for evaluation. Recent progress in
agent systems and
video generation makes generation for comprehension viable, which means using synthesized dialogues for training and evaluation. Therefore, we present OmniVChat-Studio, a multi-
agent data engine for synthesizing single- and multi-turn audio-visual dialogues. We use synthesized dialogues to build OmniVChat-Bench, an evaluation
benchmark that evaluates omni models' basic dialogue abilities across five ability categories. Replies are judged by a large
language model based on explicit scoring criteria. We also present OmniVChat-RL, a reinforcement learning reward design that jointly targets reply correctness, efficiency, and style in OmniVChat. Training Qwen3-Omni-Instruct with OmniVChat-RL on synthesized dialogues improves its performance on both OmniVChat-Bench and the human-recorded OmniVChat-Bench-Human. These gains validate the reward design and show transfer to real-world dialogues in training and evaluation.
Read the full abstract ▾
All authors Haolin He, Yunfei Chu, Qi Chen, Wen Huang, Yuan Feng, Muzhi Zhu, Zheqi Dai, Haoning Xu, Dongchao Yang, Chunyat Wu, Zining Liang, Zhengxi Liu, Xiquan Li, Xie Chen, Xize Cheng, Qize Yang, Jin Xu, Qiuqiang Kong
Computer Vision
PSEE: Progressive Sensor Event Expansion for Point-Supervised Temporal Action Localization
Jiaxi Yin, Ge Wang, Han Ding et al. (+1 more)
Temporal action localization (TAL) in wearable sensor streams identifies action classes and temporal boundaries, enabling finer-grained activity understanding than conventional action recognition. However, training typically requires costly start–end annotations for every action instance. To reduce this burden, we study point-supervised TAL, where each instance is labeled with only one timestamp and its class. We propose Progressive Sensor Event Expansion (PSEE), which combines semantic activations, sensor-specific transition evidence, and adaptive temporal ownership to recover point-supervised pseudo segments. These segments supervise standard TAL detectors without modifying their inference procedures. Cross-subject experiments on four inertial-sensing
benchmarks demonstrate improved pseudo-boundary quality over adapted point-supervised baselines, compatibility with different TAL detectors, and robustness to point sampling. Code is available at https://github.com/joeeeeyin/PSEE.
Read the full abstract ▾
All authors Jiaxi Yin, Ge Wang, Han Ding, Fei Wang
Robotics
AtomEgo: Exploring Ego-Robot Integration for Embodied Foundation Model Pretraining
Di Wu, Dongchen Zheng, Junhe Sheng et al. (+8 more)
Embodied foundation models are constrained by the limited scale and diversity of
robot demonstrations, motivating the use of large-scale
egocentric human interaction data. However, how to effectively incorporate such data into
embodied-model pre-training remains unclear because of substantial embodiment and action-space gaps between humans and
robots. We present AtomEgo, a systematic study of ego–
robot co-training supported by a curated corpus of approximately 2,659 hours and a scalable data processing pipeline. Across
vision–
language–action and world–action model architectures, we investigate three representative paradigms: joint co-training with domain-specific action heads, progressive ego-to-
robot transfer through embodiment
alignment, and joint
video–action modeling. We evaluate these paradigms through multi-task real-
robot experiments and
language-conditioned cross-embodiment representation analysis. Our results reveal a simple principle: Data Scale *
Alignment Quality –> Capability Gain;
egocentric data can improve generalization, but their value depends on how effectively they are aligned and utilized. This principle can provide practical guidance for scalable ego–
robot pre-training.
Read the full abstract ▾
All authors Di Wu, Dongchen Zheng, Junhe Sheng, Zhongxing Wei, Songxin Zhang, Zejian Xie, Xiaoquan Sun, Junyang Zheng, Zhuoyang Song, Jiaxing Zhang, Jiayu Chen
Information Theory
Sharp High-Entropy Bounds for Sums of Independent Discrete Random Variables
Haoran Wang
Sharp high-
entropy lower bounds for the
entropy of a sum were known for identically distributed summands in torsion-free abelian groups and in
prime cyclic groups. For arbitrary independent summands, Gavalakis, Goh and Kontoyiannis obtained an additive constant of $1/8$ and conjectured that the sharp constant is $1/2$. We prove that independent discrete random variables $X,Y$ with finite Shannon
entropies satisfy $H(X+Y)\ge (H(X)+H(Y))/2+1/2-o(1)$ in every torsion-free abelian group as $\max{H(X),H(Y)}\to\infty$. The same conclusion holds in the
prime cyclic group $\mathbb F_p$ when both $\max{H(X),H(Y)}$ and $\log_2 p-\max{H(X),H(Y)}$ tend to infinity. We give explicit error bounds in both settings. The proof extracts a component with paired point probabilities while controlling the
entropy of the remainder independently of its support. Discrete rearrangement and uniform perturbation then transfer the continuous
entropy power inequality to this component. In
prime cyclic groups, an additional estimate controls the
entropy lost under modular reduction. Binomial distributions show that the constant $1/2$ is optimal.
Read the full abstract ▾
Machine Learning
Efficient Architecture Search under Leave-One-Subject-Out Evaluation
Heinke Hihn and Friedhelm Schwenker
Deep neural architectures are widely used for signal processing in automated pain assessment systems. However, architecture design has remained largely a manual task despite the potential efficiency benefits of Neural Architecture Search (NAS).
Embedding NAS in a Leave-One-Subject-Out (LOSO) evaluation is computationally demanding because a fully nested implementation requires $N$ independent architecture searches and, assuming approximately linear training cost, scales as $\mathcal{O}(N^2)$. We propose a block-based, leakage-controlled approach that shares NAS runs between subjects, reducing the number of searches from $N$ to $B$, where $B \ll N$, dubbed PainNAS. On the BioVid Heat Pain
dataset, PainNAS yields comparable subject-level accuracy with substantially fewer parameters and FLOPs.
Read the full abstract ▾
Computer Vision
CompAdapt: Adaptable Composite Motion Modeling for Physics-Consistent Text-to-Video Generation
Haoran Qin, Renlong Wu, Tianyu Huang et al. (+3 more)
While
diffusion-based text-to-
video (T2V) models have demonstrated impressive capability in generating realistic and temporally coherent
videos, they often fail to respect fundamental physical dynamics. Although recent physics-constrained methods incorporate explicit dynamics priors to improve physical plausibility, they remain limited to simple single-type motions, depend on manually specified parameters, and struggle to generalize to unseen physical laws. In this work, we propose CompAdapt, a physics-consistent T2V framework for adaptable generation across complex real-world scenarios. It extends neural dynamics modeling beyond single-type motions to encompass composite physical behaviors, including coupled motions, multi-stage transitions, and multi-object collisions. Furthermore, CompAdapt translates natural
language prompts into structured physical semantics, enabling end-to-end specification of motion types, temporal relations, and initial physical parameters. To generalize to novel physical environments, CompAdapt introduces dynamics-aware prior matching, achieving one-shot adaptation without retraining the core dynamics module. In addition, a physics-aware latent feature fusion module improves visual fidelity under fast and complex motion. Experiments on physics-focused T2V
benchmarks demonstrate that CompAdapt improves physical consistency over both general T2V models and physics-constrained baselines, while preserving high visual quality and adaptability to unseen dynamics. The project page is available at https://makapic.github.io/CompAdapt/ .
Read the full abstract ▾
All authors Haoran Qin, Renlong Wu, Tianyu Huang, Yukang Ding, Hui Li, Wangmeng Zuo
Machine Learning · Cross-listed from stat.ML
Improving the Predictive Performance of Bootstrap Aggregating by Dirichlet Resampling
Quoc Viet Le and Joonha Park
We revisit Breiman's observation that reducing inter-tree correlation without weakening individual trees can improve random forests. Building on this principle, we introduce two variants: Dirichlet-Multinomial Bagging Random Forest (DM) and Dirichlet-Weighted Random Forest (DW). Both modulate sample reweighting via a concentration parameter $\alpha>0$. We provide a simple theoretical criterion that clarifies when these variants behave indistinguishably from standard random forests, and we use it to guide a lightweight tuning strategy. In a controlled evaluation on public classification
benchmarks, DM and DW are consistently competitive and often stronger than other random-forest (RF) baselines, with negligible additional runtime.
Read the full abstract ▾
Machine Learning
Understanding LLM Quantization through Activation-Guided Compensation and Orthogonal Residuals
Yamato Narita and Issei Sato
Post-training weight-activation quantization reduces the
memory and inference costs of large
language models, but aggressive W4A4 quantization remains difficult because activation outliers degrade effective quantization resolution. Although weight optimization, channel-wise scaling, and orthogonal rotation mitigate this problem, the error components they address and their relationship remain unclear. Using an exact decomposition of local weight-activation quantization error into an activation-guided weight compensation term and an orthogonal residual, we bound the residual using persistent channel-wise outlier and regular activation quantities. This decomposition clarifies which error components can be addressed by weight compensation and which require transformation design. We then use the residual bounds to derive practical guidelines for applying randomized Hadamard rotation, sign selection, and channel scaling. In particular, the analysis explains how random signs suppress constructive interference among persistent outlier channels, how sampling multiple sign patterns can improve transformation selection, and how second-moment balancing leads to an $L_2$ scaling rule while a further relaxation recovers SmoothQuant-style $L_\infty$ scaling. We evaluate these guidelines through backpropagation-free configurations across eight Llama and Mistral models, obtaining performance competitive with gradient-trained SpinQuant.
Read the full abstract ▾
Computer Vision
ME-Dex 1.0: Bringing Heterogeneous Tactile Sensing into World Action Modeling
Xuancheng Zhang, Xuetao Liu, Qianying Tang et al. (+8 more)
World Action Models bring the predictive capabilities of
video models into
robot action generation, providing a rich foundation for modeling future visual states. Tactile sensing complements this foundation with direct measurements of physical interaction. Some existing methods use tactile features as conditioning inputs without jointly predicting future tactile states, visual observations, and actions. Our key insight is that tactile signals, like
video, provide observations of the evolving world state and should be modeled as future observations alongside
video. We present ME-Dex-1.0 (MachEmbodied-Dex-1.0), a unified World Action Tactile Model for joint visual, tactile, and action learning. ME-Dex-1.0 adopts a Mixture-of-
Transformers architecture comprising a
Video Expert, a Tactile Expert, and an Action Expert, all trained with flow matching. We use shared
attention connects the experts in intermediate layers, allowing action generation to draw on learned representations of visual and tactile dynamics during joint denoising. To support multi-source heterogeneous tactile inputs, a Canonical Hand Model and a Unified Tactile Autoencoder map tactile observations from different embodiments and sensing layouts into shared spatial and latent spaces. To address the limited availability of paired visual, tactile, and action data, we develop the Agentic Tactile Data Engine, an
agent-based data production platform. It supplements RoboTwin and DexJoCo with tactile data recorded directly from force sensors during trajectory replay in simulation. Experiments on the RoboTwin, DexJoCo, and ManiFeel simulation platforms, together with real
robot evaluations, demonstrate improved manipulation performance using both grippers and dexterous hands equipped with tactile sensing.
Read the full abstract ▾
All authors Xuancheng Zhang, Xuetao Liu, Qianying Tang, Jizhe Wang, Runsheng Wang, Zhijing Cheng, Bochen Lin, Haoran Wen, Ming Li, Kun Zhan, Yu Liu
Robotics
Robotic Multiphase Interaction: Manipulating Coupled Liquid and Solid Dynamics with a World Model
Yixuan Feng and Peng Wang
This work presents Robotic Multiphase Interaction (RMI), a setting in which
liquid enters a porous material and interacts mechanically with its deforming solid skeleton. Manipulation can therefore change pore volume, expel or redistribute retained
liquid, and alter grasp stability at the same time. Spilled
liquid can also create safety risks in domestic and manufacturing settings. This differs from most manipulation of solid objects and from tasks that involve both
liquid and solid while keeping the phases spatially separate. We study a sponge filled with water as the first RMI example. We use implicit incompressible porous flow with smoothed particle hydrodynamics as the dynamics engine and enable robotic manipulation by adding Coulomb contact
memory, hybrid velocity and force regulation, and a stability gate for lifting. The resulting environment connects
robot commands to changes in the coupled
liquid and solid state. A world model conditioned on actions predicts how this state evolves under candidate commands, while a temporal UNet generates actions using either
Diffusion Policy or rectified flow matching. Our world model reduces retained water prediction error by more than $60\%$ compared with the baseline. The best action sequence selected by the world model from policy proposals further reduces the predicted terminal water error by about half. These improvements show that modelling the coupled
liquid and solid state helps the
robot predict how its actions affect both the porous object and the
liquid held inside.
Read the full abstract ▾
Robotics
FootQuery: Future-Touchdown-Guided Retrieval from Depth History for Perceptive Humanoid Locomotion
Tao Dong, Jia Yu, Yuxuan Fan et al. (+5 more)
Humanoid locomotion over complex terrain requires anticipating footholds that may no longer be visible at touchdown. Limited camera coverage and self-occlusion make it necessary to retrieve relevant terrain information from earlier observations. We present FootQuery, a perceptive locomotion framework that queries depth history using each foot's predicted next touchdown. The policy predicts touchdown locations and uncertainty from proprioception and uses these distributions, together with per-foot features, to query sparsely sampled historical depth frames. During training, realized contacts are projected into historical images to supervise
retrieval at the regions where those contacts were visible. The retrieved per-foot features are fused with global visual
memory to generate control actions. A progressive force-assistance curriculum supports early exploration, while event-consistent tread-midline shaping encourages coordinated stair contacts. Deployment requires only proprioception and onboard depth images. In simulation, the complete framework outperforms its component ablations on the most challenging tested stairs, gaps, and platforms. Real-world experiments on a Unitree G1 demonstrate continuous traversal with a single policy across outdoor stairs and indoor routes combining stair ascent and descent, platforms, and gaps. These results support organizing visual history around anticipated contacts for perceptive
humanoid locomotion.
Read the full abstract ▾
All authors Tao Dong, Jia Yu, Yuxuan Fan, Linna Zhao, Jiaqi Gong, Andong Yang, Chao Gao, Guyue Zhou
Machine Learning
Optimal Randomized Proper Online Learning
Zachary Chase and Idan Mehalel
We prove that the optimal expected mistake bound of online learning a function class $\mathcal{H}$ by a randomized proper learning
algorithm is $O(\mathtt{L}(\mathcal{H}) \log T)$, where $\mathtt{L}(\mathcal{H})$ is the Littlestone dimension of $\mathcal{H}$ and $T$ is the time horizon. Our result improves upon the previously best known bound of $O(\mathtt{L}(\mathcal{H}) \log^6 T)$ given by Daskalakis and Golowich (STOC 2022), and is optimal up to a universal constant for worst-case classes.
Read the full abstract ▾
Information Theory
Interference-Driven Clustered Optimisation for FM Spectrum Coordination
Federica Mangiatordi and Emiliano Pallotti
Cross-border FM spectrum coordination involves protecting foreign broadcasting services while preserving domestic coverage, amid increasingly large
radio-planning
datasets containing thousands of transmitters and millions of transmitter-pixel relationships. In such scenarios, conventional optimisation approaches become computationally demanding due to the high dimensionality of the associated power-control problem. This paper proposes an interference-driven clustered optimisation framework for large-scale FM spectrum coordination. The proposed method exploits the observation that violations of foreign-service protection are typically dominated by a limited subset of transmitters. Protected services are therefore analysed to identify dominant interferers and quantify their impact on interference. These relationships are represented through an interference graph from which optimisation-oriented transmitter clusters are extracted. The clusters decompose the global power-control problem into smaller optimisation tasks solved with clustered simulated annealing, followed by a global refinement that captures residual inter-cluster interactions. Coverage and interference are evaluated using frequency-dependent protection criteria and a dynamic strongest-service assignment model. To enable operational-scale planning, the framework uses sparse
matrices and GPU-accelerated computations. Tests on realistic cross-border FM coordination scenarios show that the clustering strategy greatly reduces optimisation complexity and runtime while maintaining foreign-service protection and domestic coverage. The method also yields an interpretable ranking of transmitters that contribute most to harmful interference, supporting optimisation and spectrum planning.
Read the full abstract ▾
Human-Computer Interaction
People escalate against a competitor labelled human and hold back against one labelled an optimising machine
Vinicius Ferraz and Leon Houf
People increasingly compete against AI
agents rather than other human opponents. We distinguish two channels: an opponent effect and an information effect. These are different elements with different consequences: the opponent effect is specific to a given computational system, the information effect a property of the information environment that an organisation or policymaker can control. We separate them in a preregistered experiment (N = 1,395) using a dynamic all-pay auction, a repeated contest in which escalation of commitment arises from the incentives. What participants are told about the opponent (human, an AI trained to imitate people, or an AI trained to compete well) is varied and crossed with who they actually face, in a deception-free design. What people are told influences escalation: the median price rises by 6.7 points when a human might be the opponent and falls by 8.8 when an optimising machine might be, a spread of about 15% of the prize value of the competition, produced by information alone. Competing against the AI
agents lowers prices, yet reduces the chance that both sides finish with positive earnings, showing distinct effects of the opponent channel. The information effect is not explained by articulated strategy, or individual differences, and is consistent with a competitive response engaged when a human is a live possibility. This shows that describing an AI competitor is not behaviourally neutral.
Read the full abstract ▾
Computer Vision
Think Locally, Refine Globally for Memory-Efficient 3D Reconstruction
Jingke Zhou, Chenhang Ma, Zhizhou Zhong et al. (+6 more)
We propose LoG-VGGT, a
memory-efficient framework for long-sequence 3D reconstruction that balances local temporal modeling with global camera consistency. Instead of relying on full global
attention, our method introduces cross-window
attention at a small subset of
transformer blocks, enabling effective information propagation across adjacent temporal windows while keeping
memory usage bounded. To mitigate long-term pose drift, we further design a global camera consistency refinement module, where camera
tokens interact with compact register
tokens via cross-
attention to enforce scene-level constraints across the entire sequence. This design enables joint optimization of camera representations and significantly improves long-horizon pose stability without incurring the high cost of sequence-wide
attention. Extensive experiments demonstrate that LoG-VGGT achieves improved depth accuracy and robust camera pose estimation across multiple long-sequence
benchmarks, while delivering competitive streaming reconstruction performance.
Read the full abstract ▾
All authors Jingke Zhou, Chenhang Ma, Zhizhou Zhong, Mingkai Liu, Zhuang Zhou, Yicheng ji, Binghua Su, Bo Cai, Xianliang Huang
Human-Computer Interaction
AESSI: An Around-Ear Silent Speech Interface for Cross-Day Online Reuse without Test-Day Calibration
Xiran Xu, Mochu Dong, Yujie Yan et al. (+3 more)
Silent
speech interfaces (SSIs) enable private communication without audible
speech and may support people with post-stroke dysarthria. Everyday reuse requires articulation-related representations that generalize across days despite sensor repositioning and physiological changes. We present AESSI, an around-ear SSI using masked-context representation pretraining (MCRP): a student predicts teacher representations from masked time-frequency inputs to encourage robustness to recording variability. We collected 44 electrophysiological recordings from 24 participants for 25 everyday Mandarin sentences. With six participants' complete final recordings held out, AESSI achieved 92.24 percent mean accuracy without test-day calibration. AESSI exceeded the best adapted baseline by 50.57 percentage points. At least 21 days after each participant's last recording, five participants each completed 50 independently randomized online tests without calibration, achieving 98.0 percent overall accuracy. Median preprocessing and inference time in CPU replay was 31.70 ms. These results demonstrate end-to-end system operation and support online reuse without test-day calibration. Demo is included with the paper.
Read the full abstract ▾
All authors Xiran Xu, Mochu Dong, Yujie Yan, Chenxi Wang, Yu Jiao, Jing Chen
Systems and Control
Iterative Single-Loop Coordination of Transmission and Distribution Systems and DER Aggregators with Limited Information Sharing
Robert Sosnowski, Yury Dvorkin, Marcin Baszynski et al. (+1 more)
The growing integration of DERs underscores the importance of unified coordination across TSO, DSO, and aggregator levels. This paper proposes a coordination method between TSO, DSOs and DER aggregators, based on a transparent, well-scalable, and decentralized single-loop iterative mechanism with limited information sharing. The approach enables iterative price-based coordination between adjacent levels, where the higher-level entities provide price signals and the lower-level ones respond with desired power schedules. The proposed method uses the proximal-gradient-based iterative co-optimization
algorithm extended to a three-level case, which enjoys convergence characteristic of proximal
algorithms. The performance of the proposed approach is compared with both a centralized formulation and a sequential reference formulation based on fixed transmission-level energy prices. The results demonstrate that the proposed coordination achieves outcomes very close to those obtained with the centralized formulation, while significantly improving performance relative to the sequential reference, particularly at higher levels of DER flexibility.
Read the full abstract ▾
All authors Robert Sosnowski, Yury Dvorkin, Marcin Baszynski, Charalambos Konstantinou
Artificial Intelligence
GVPO++: Group Variance Policy Optimization for LLM Post-Training and On-Policy Distillation
Kaichen Zhang, Yuzhong Hong, Junwei Bao et al. (+4 more)
Post-training plays a pivotal role in enhancing the
reasoning capabilities and task-specific expertise of large
language models (LLMs). Despite recent advances in post-training methods, such as Group Relative Policy Optimization (GRPO), their practical deployment remains impeded by training instability arising from the reliance on importance sampling. We introduce Group Variance Policy Optimization (GVPO), a novel post-training method that integrates the analytical solution of KL-constrained reward maximization into its gradient weighting scheme. This formulation provides an intuitive interpretation: GVPO's gradient corresponds to the mean squared error between the central distance of implicit rewards and that of actual rewards. GVPO offers two key advantages: (1) it guarantees a unique optimal solution, exactly to the KL-constrained reward maximization objective, and (2) it enables flexible sampling distributions without requiring importance sampling. Beyond general post-training, we show that GVPO naturally extends to on-policy
distillation (OPD). Furthermore, GVPO enables the optimization of a broad family of extended OPD objectives, providing a principled foundation for diverse objective design. By unifying theoretical guarantees with practical adaptability, GVPO establishes a new paradigm for reliable and versatile LLM post-training and on-policy
distillation.
Read the full abstract ▾
All authors Kaichen Zhang, Yuzhong Hong, Junwei Bao, Hongfei Jiang, Yang Song, Dingqian Hong, Hui Xiong
Machine Learning
Decision-Focused Learning for Mean-Variance Portfolio Optimization via KKT-Based Reformulation
Kensei Nosaka, Shunnosuke Ikeda and Yuichi Takano
Mean-variance portfolio optimization (MVO) is a central framework in data-driven asset management. A widely adopted approach is a two-stage framework that first predicts expected returns and then solves the optimization problem based on these predictions, with the predictive models trained by minimizing prediction errors. However, this objective of prediction is not aligned with the quality of the downstream portfolio decision. Decision-focused learning (DFL), which directly minimizes the downstream decision loss within the learning process, has thus emerged as a promising direction. However, existing DFL approaches to MVO rely on surrogate losses or constraint relaxations for tractability, creating a structural mismatch between predictive model training and the constrained MVO solved at evaluation. We propose a single-level optimization formulation that incorporates the Karush-Kuhn-Tucker (KKT) optimality conditions of the lower-level MVO into the upper-level learning problem. This formulation explicitly preserves the budget and short-sale constraints while remaining tractable for standard nonlinear optimization solvers. Rolling-window experiments on real-world ETF (Exchange Traded Funds) data across two asset universes with different correlation structures show that our method achieved the best performance on multiple investment metrics and also demonstrated performance improvement due to the proposed regularization.
Read the full abstract ▾
Cryptography and Security
DEFEAT: Stitching Fragmented File I/O Contexts for Early Ransomware Detection
Muhammad Ejaz Ahmed, Hyoungshick Kim, Mohsen Ali Alawami et al. (+4 more)
Ransomware increasingly fragments its file operations across temporary and intermediate files, scattering the semantic context that links individual I/O events to an overarching encryption campaign. This fragmentation defeats existing detectors that reason over isolated file streams – whether pattern-based methods that match rigid event sequences or learning-based methods that require accumulating statistical evidence across many files. We present DEFEAT, a framework that reconstructs this fragmented, scattered context by grouping causally related file events into File Event Gadgets (FEGs), semantically coherent units that capture the full intent behind sequences of file operations spanning multiple dynamically created files. Unlike provenance graphs (system-wide causal graphs that record relationships among all OS entities, such as processes, files, sockets, and registry keys, across the entire system), FEGs are scoped to the file-operation context of a single user asset, enabling lightweight, targeted analysis without whole-system instrumentation. Each FEG is modelled as an attributed control flow graph (ACFG) and embedded via a graph neural network for unsupervised clustering, enabling analysts to label entire behavioural clusters rather than individual samples, reducing annotation effort by 94%. Evaluated on a corpus of 97,816,471 file I/O events spanning 67 ransomware families, DEFEAT achieves 99.2% detection accuracy and outperforms state-of-the-art methods including UNVEIL, RWGuard, and Peeler by 6.57 to 7.56%. The framework operates at the granularity of a single file encryption: because each ACFG represents exactly one FEG (one user asset context), a cluster label can be assigned as soon as the first file operation completes, enabling detection at the first encrypted file.
Read the full abstract ▾
All authors Muhammad Ejaz Ahmed, Hyoungshick Kim, Mohsen Ali Alawami, Alsharif Abuadbba, Seyit Camtepe, Surya Nepal, Junaid Qadir
Machine Learning
Tracing the Evidence Behind Zero-Shot Time-Series Forecasting: A Source-First Taxonomy and Audit Framework
Delun Kong, Wanyun Ling, Chenxi Liu et al. (+1 more)
Zero-shot time-series
forecasting (TSF) is often described as
forecasting without target-specific parameter updates, but that training-status condition does not specify what evidence the system may use. A frozen
language model prompted with serialized values, a time-series model pretrained on broad
forecasting corpora, and a
retrieval-augmented forecaster may all satisfy the no-update condition while drawing on different transferable evidence. This paper argues that zero-shot TSF should therefore be governed as an evidence-access claim. We propose a source-first taxonomy that separates three primary evidence sources—frozen LLM prior reuse, parametric time-series pretraining, and
retrieval-augmented external
memory—from the architectures that implement them. After the source is identified, four additional audit questions remain: task interface, forecast object and scoring, prediction-time context, and resource budget. The resulting agenda is to make zero-shot leaderboards auditable by reporting evidence boundaries and interface assumptions alongside scores, so that
benchmark progress reflects transferable
forecasting capability rather than undisclosed changes in context,
memory, or budget.
Read the full abstract ▾
All authors Delun Kong, Wanyun Ling, Chenxi Liu, Ziyue Li
Computer Vision
P$^3$-SAM: SAM with Perceptual Parallel Prompt for Few-Shot Strip Steel Surface Defect Segmentation
Qian Xu, Hang Xiong, Anpeng Wang et al. (+3 more)
Few-shot semantic
segmentation (FSS) of strip steel surface defects (S$^3$D) has posed significant challenges distinct from natural scenes. Unlike natural images, S$^3$D task exhibits unique characteristics including low local contrast, uneven illumination, and complex fine-grained texture patterns. Although recent methods based on Segment Anything Model (SAM) have shown promise in FSS on natural images by leveraging SAM's powerful pre-trained representations, these unique industrial characteristics of S$^3$D images lead to performance drop when directly applying SAM to industrial defect scenarios. In this paper, we propose a novel Perceptual Parallel
Prompt (P$^3$) framework that empowers SAM, creating the P$^3$-SAM model to address these challenges through two core strategies. First, we develop a Perceptual-Optimized Encoding (POE) strategy that enhances local contrast and preserves critical texture details for S$^3$D
segmentation. Second, we introduce the Parallel
Prompt Generator (PPG) strategy that simultaneously generates both semantic and spatial
prompts, enabling comprehensive guidance for SAM's decoder across varying images. Extensive experiments on three few-shot S$^3$D
benchmarks demonstrate that P$^3$-SAM achieves state-of-the-art performance, with particularly notable improvements of 12.00% in mIoU on Surface Defects-4i
dataset.
Read the full abstract ▾
All authors Qian Xu, Hang Xiong, Anpeng Wang, Sam Kwong, Cong Zhang, Runmin Cong
Artificial Intelligence
DENSE: Distilling Agent Trajectories into Evidence-Grounded Shortcut Trees for Self-Refinement
Siyuan Liu, Fan Yu, Dongyu Ru et al. (+5 more)
Online
agent deployments accumulate execution trajectories at massive scale and behavioral diversity, for which predefined annotation criteria hardly exist. Extracting useful evidence therefore demands costly manual annotation or verifier signals that fail to scale, leaving valuable evidence buried among redundant, incomplete, and failed executions. This raises a question: without post-execution rewards or correctness labels, how can reusable experience be distilled from the trajectories themselves? To address this challenge, we introduce DENSE (Distilling Evidence from Nested Subtask Executions), which organizes trajectory-derived evidence into nested shortcut trees. By consolidating redundant attempts, identifying resolved subtasks, and retaining useful steps alongside outstanding requirements, DENSE transforms noisy execution traces into structured and reusable task-solving feedback. To evaluate whether such feedback helps
agents retry the same task, we design REFIT, which measures success-rate changes between the initial attempt and feedback-guided retries. Among feedback methods without external outcome supervision, DENSE achieves the highest strict pass rate across four
agent models on Terminal-Bench 2.1, improving over initial attempts by 7.12-21.81 percentage points with 19.0-43.6% fewer
agent tokens on retries. In addition, on hard tasks DENSE consistently outperforms self-reflection in cumulative pass rate across multiple feedback iterations on all four models, demonstrating its strong potential for continual
agent self-improvement.
Read the full abstract ▾
All authors Siyuan Liu, Fan Yu, Dongyu Ru, Yizhu Liu, Yifan Yang, Xuezhi Cao, Xunliang Cai, Yixin Cao
Machine Learning · Cross-listed from stat.ML
Brownian Heads for Deep ReLU Representations: Activation Mass and the Cost of Same-Sample Selection
Mahdi Mohammadigohari and Nicole Mücke
Deep representation learning often selects hidden features and fits the final predictor on the same sample, so fixed-feature analysis performed after selection can omit selection cost. We study the conditional empirical Rademacher complexity of deep ReLU representations followed by bounded-norm predictors in additive or Lévy-Brownian RKHSs, termed Brownian heads. For a fixed representation, we derive an exact dual identity and sharp bounds in terms of activation mass, the average norm of the observed hidden vectors. Under same-sample selection, the representation supremum induces a quadratic Rademacher process. Brownian layer-cake and Gaussian-projection identities reduce it to coordinatewise or signed projected threshold traces, separating realized scale from selection complexity. For samples with pairwise-distinct inputs, explicit scalar ReLU families match the finite-trace and VC rates up to universal constants at the realized trace-and-envelope level. Induced-norm contraction also yields architecture-level bounds for rectangular, rank-deficient ReLU networks. Experiments verify the sharp bounds and rates, exhibit a selection gap at fixed activation mass, and assess the predictive feasibility of Brownian heads.
Read the full abstract ▾
Cryptography and Security
TrustBOM: A Scalable Architecture for Confidentiality-Preserving SBOMs Across Organizations
Van Thang Nguyen, Frederic Rupprecht, Tom Lawrence et al. (+6 more)
Software Bills of Materials (SBOMs) have emerged as a key mechanism for software supply chain governance in enterprise architectures. However, their adoption across organizations remains limited due to concerns about exposing sensitive dependency information. To address this limitation, we propose TrustBOM, a scalable architecture for confidentiality-preserving SBOMs integrated into enterprise CI/CD workflows. TrustBOM enables software providers to attest that specific vulnerabilities or restricted licenses are absent from their software without revealing the underlying dependency graph. This is achieved using zero-knowledge non-membership proofs, which are applied selectively based on consumer-defined policy constraints. The architecture ensures that proof generation scales linearly with the number of asserted constraints rather than with the size of the SBOM, enabling efficient operation in large-scale enterprise environments. Empirical evaluation demonstrates linear performance, with an average proof generation time of 0.9 seconds per constraint on commodity hardware, indicating the feasibility of deployment in enterprise platform ecosystems.
Read the full abstract ▾
All authors Van Thang Nguyen, Frederic Rupprecht, Tom Lawrence, Lucca Di Benedetto, Sören Schubert, Amor Rezgui, Sebastian Werner, Maria C. Borges, Stefan Tai
Robotics
A Unified Dynamic Force Guidance Framework for Performance-Optimized Kinesthetic Teaching
Chunxin Li, Jianhua Wu, Zhenhua Xiong et al. (+1 more)
Collaborative
robots are increasingly deployed in industrial scenarios characterized by frequent product changeovers. As an intuitive programming method, kinesthetic teaching facilitates rapid
robot deployment. However, users may overlook the configuration of the
robot during kinesthetic teaching, leading to degradation in operational performance. Operational performance refers to the capability of the
robot to generate motion and can be quantified by the Minimum Singular Value of the Jacobian
matrix. To address this issue, this paper proposes an online dynamic force guidance method that integrates performance constraint and optimization mechanisms. Specifically, variable admittance control maintains the operational performance of the
robot above a predefined threshold, while a virtual force actively guides the user to drag the
robot towards configurations with improved performance. Experiments are conducted on a 6-DOF collaborative
robot, comparing three typical paths in the task space. To evaluate the quality of the taught trajectories, trajectory playback experiments are conducted to analyze the relationship between the operational performance of the
robot and the work efficiency. The results demonstrate that the proposed method effectively enhances the operational performance of the
robot and consequently improves the work efficiency, holding significant value for reducing production takt time in industrial deployment.
Read the full abstract ▾
All authors Chunxin Li, Jianhua Wu, Zhenhua Xiong, Xiangyang Zhu
Computational Engineering
Synthetic Human Mobility Data Generation: A Structured Review of Representations, Methods, and Practical Capabilities
Yanbo Pang, Chen Zhong, Song Gao et al. (+1 more)
Human mobility data has become an increasingly important component of urban analytics. Although the range of available mobility data sources has expanded substantially, access remains highly constrained by commercial restrictions,
privacy concerns, and institutional barriers. Data protection procedures also often reduce the analytical value of released
datasets. Synthetic mobility data has emerged as a promising solution, but existing methods differ substantially in their underlying mechanisms, the information they preserve, the outputs they generate, and the analytical questions they can support. Their comparative strengths and trade-offs remain insufficiently understood for urban analytics. This paper presents a structured review of synthetic human mobility data generation from an urban analytics perspective. We review the literature by methodological family and index it by the mobility outputs each family generates natively and the analytical capabilities those outputs enable. We first provide a taxonomy of synthetic data products, including population and persona representations, activity schedules, trip and tour records, trajectories, and aggregate mobility patterns. We then review the major methodological families, spanning mechanistic models, survey-driven population synthesis, activity- and
agent-based simulation, deep generative models,
transformer-based mobility
language models, and LLM-agentic systems. Building on this synthesis, we introduce a Meaning-Population-Autonomy framework that characterises these methods along three dimensions: behavioural meaning, population grounding and scale, and generation autonomy. We consider these dimensions the principal requirements for downstream urban analytics. Few methods deliver behavioural meaning, population grounding and autonomous generation at once, and fewer still with generation constrained to feasible trajectories.
Read the full abstract ▾
All authors Yanbo Pang, Chen Zhong, Song Gao, Yoshihide Sekimoto
Computer Vision
When Online Adaptation Hurts: Parameter-Frozen Test-Time Ensembling for Continual Medical Image Segmentation
Ruijie Huang
Medical image segmenters often get worse when sites, scanner vendors, or protocols change. Continual test-time adaptation (CTTA) addresses this problem without target labels, but it can be impossible to update a model on a non-stationary stream and can lead to a lot of errors. We examine a more reasonable and meaningful alternative: parameter-frozen inference enhancement(PIE). We use a source-trained segmenter that learns about anatomy-preserving scale and flip views, maps their predictions back to the native location, and averages the probabilities. We do not modify the weights of the model or the normalization statistics. On a cardiac MRI stream from M&Ms, which is trained on vendor A and evaluated sequentially on vendors B, C, and D, PIE has 0.7786 mean Dice, compared to 0.7680 for source-only inference and 0.7388–0.7416 for five other online-adaptation baselines. The controlled ablations show that performance saturates at 28 views, and confidence weighting, class-prior correction, connected-component filtering, morphological refinement, and inter-slice smoothing have no effect or cause negative transfer. Qualitative results on cardiac MRI and fundus images are also consistent with the frozen ensemble keeping thinner and nested anatomical structures. These results provide a strong, stable baseline for medical CTTA and expose an important failure mode: adaptation and handcrafted refinement can be less reliable than carefully designed inference.
Read the full abstract ▾