High-dimensional images, or images with a high-dimensional attribute vector per pixel, are commonly explored with coordinated views of a low-dimensional embedding of the attribute space and a conventional image representation. Nowadays, such images can easily contain several million pixels. For such large datasets, hierarchical embedding techniques are better suited to represent the high-dimensional attribute space than flat dimensionality reduction methods. However, available hierarchical dimensionality reduction methods construct the hierarchy purely based on the attribute information and ignore the spatial layout of pixels in the images. This impedes the exploration of regions of interest in the image space, since there is no congruence between a region of interest in image space and the associated attribute abstractions in the hierarchy. In this paper, we present a superpixel hierarchy for high-dimensional images that takes the high-dimensional attribute manifold into account during construction. Through this, our method enables consistent exploration of high-dimensional images in both image and attribute space. We show the effectiveness of this new image-guided hierarchy in the context of embedding exploration by comparing it with classical hierarchical embedding-based image exploration in two use cases.

Vision transformers have recently made a breakthrough in computer vision showing excellent performance in terms of precision for numerous applications. However, their computational cost is very high compared to alternative approaches such as Convolutional Neural Networks. To address this problem, we propose a novel framework for image classification called RAViT based on a multi-branch network that operates on several copies of the same image with different resolutions to reduce the computational cost while preserving the overall accuracy. Furthermore, our framework includes an early exit mechanism that makes our model adaptive and allows to choose the appropriate trade-off between accuracy and computational cost at run-time. For example in a two-branch architecture, the original image is first resized to reduce its resolution, then a prediction is performed on it using a first transformer and the resulting prediction is reused together with the original-size image to perform a final prediction on a second transformer with less computation than a classical Vision transformer architecture. The early-exit process allows the model to make a final prediction at intermediate branches, saving even more computation. We evaluated our approach on CIFAR-10, Tiny ImageNet, and ImageNet. We obtained an equivalent accuracy to the classical Vision transformer model with only around 70% of FLOPs.

Humanoid robots have become a focal point of technological ambition, with claims of surgical capability within years in mainstream discourse. These projections are aspirational yet lack empirical grounding. To date, no humanoid has assisted a surgeon through an actual procedure, let alone performed one. The work described here breaks this new ground. Here we report a proof of concept in which a teleoperated Unitree G1 provided endoscopic visualization while an attending otolaryngologist performed a cadaveric sphenoidectomy. The procedure was completed successfully, with stable visualization maintained throughout. Teleoperation allowed assessment of whether the humanoid form factor could meet the physical demands of surgical assistance in terms of sustenance and precision; the cognitive demands were satisfied -- for now -- by the operator. Post-procedure analysis identified engineering targets for clinical translation, alongside near-term opportunities such as autonomous diagnostic scoping. This work establishes form-factor feasibility for humanoid surgical assistance while identifying challenges for continued development.

We present a method for generating a full 360° orbit video around a person from a single input image. Existing methods typically adapt image-based diffusion models for multi-view synthesis, but yield inconsistent results across views and with the original identity. In contrast, recent video diffusion models have demonstrated their ability in generating photorealistic results that align well with the given prompts. Inspired by these results, we propose HumanOrbit, a video diffusion model for multi-view human image generation. Our approach enables the model to synthesize continuous camera rotations around the subject, producing geometrically consistent novel views while preserving the appearance and identity of the person. Using the generated multi-view frames, we further propose a reconstruction pipeline that recovers a textured mesh of the subject. Experimental results validate the effectiveness of HumanOrbit for multi-view image generation and that the reconstructed 3D models exhibit superior completeness and fidelity compared to those from state-of-the-art baselines.

In many applications, evaluating the effectiveness of different alternatives comes with varying costs or resource usage. Motivated by such heterogeneity, we study the Best Arm Identification with Resource Constraints (BAIwRC) problem, where an agent seeks to identify the best alternative (aka arm) in the presence of resource constraints. Each arm pull consumes one or more types of limited resources. We make two key contributions. First, we propose the Successive Halving with Resource Rationing (SH-RR) algorithm, which integrates resource-aware allocation into the classical successive halving framework on best arm identification. The SH-RR algorithm unifies the theoretical analysis for both the stochastic and deterministic consumption settings, with a new \textit{effective consumption measure

Vision-language action (VLA) policies often report strong manipulation benchmark performance with relatively few demonstrations, but it remains unclear whether this reflects robust language-to-object grounding or reliance on object--location correlations that do not transfer beyond the training distribution. We present a controlled multi-object picking study that progressively increases object placement variability up to full workspace randomization and evaluates held-out object--location pairings that break familiar associations without increasing spatial difficulty. Across these stress tests and data scaling, we find that for representative VLA policies, including SmolVLA and $π_{0.5}$, execution of the manipulation primitive remains substantially more reliable than instruction-conditioned task success in harder regimes, suggesting that manipulation skill acquisition is decoupled from instruction following. We recommend augmenting manipulation benchmarks with task ladders and decomposed metrics that separately measure primitive execution and instruction-conditioned success to better diagnose instruction-grounded generalization.

Recent significant advances in 3D scene representation have been driven by 3D Gaussian Splatting (3DGS), which has enabled real-time rendering with photorealistic quality. 3DGS often requires a large number of primitives to achieve high fidelity, leading to redundant representations and high resource consumption, thereby limiting its scalability for complex or large-scale scenes. Consequently, effective pruning strategies and more expressive primitives that can reduce redundancy while preserving visual quality are crucial for practical deployment. We propose an efficient, integrated reconstruction-aware pruning strategy that adaptively determines pruning timing and refining intervals based on reconstruction quality, thus reducing model size while enhancing rendering quality. Moreover, we introduce a 3D Difference-of-Gaussians primitive that jointly models both positive and negative densities in a single primitive, improving the expressiveness of Gaussians under compact configurations. Our method significantly improves model compactness, achieving up to 90\% reduction in Gaussian-count while delivering visual quality that is similar to, or in some cases better than, that produced by state-of-the-art methods. Code will be made publicly available.

The expansion of retrieval-augmented generation (RAG) into multimodal domains has intensified the challenge for processing complex visual documents, such as financial reports. While page-level chunking and retrieval is a natural starting point, it creates a critical bottleneck: delivering entire pages to the generator introduces excessive extraneous context. This not only overloads the generator's attention mechanism but also dilutes the most salient evidence. Moreover, compressing these information-rich pages into a limited visual token budget further increases the risk of hallucinations. To address this, we introduce AgenticOCR, a dynamic parsing paradigm that transforms optical character recognition (OCR) from a static, full-text process into a query-driven, on-demand extraction system. By autonomously analyzing document layout in a "thinking with images" manner, AgenticOCR identifies and selectively recognizes regions of interest. This approach performs on-demand decompression of visual tokens precisely where needed, effectively decoupling retrieval granularity from rigid page-level chunking. AgenticOCR has the potential to serve as the "third building block" of the visual document RAG stack, operating alongside and enhancing standard Embedding and Reranking modules. Experimental results demonstrate that AgenticOCR improves both the efficiency and accuracy of visual RAG systems, achieving expert-level performance in long document understanding. Code and models are available at https://github.com/OpenDataLab/AgenticOCR.

Observational learning requires an agent to learn to perform a task by referencing only observations of the performed task. This work investigates the equivalent setting in real-world robot learning where access to hand-designed rewards and demonstrator actions are not assumed. To address this data-constrained setting, this work presents a planning-based Inverse Reinforcement Learning (IRL) algorithm for world modeling from observation and interaction alone. Experiments conducted entirely in the real-world demonstrate that this paradigm is effective for learning image-based manipulation tasks from scratch in under an hour, without assuming prior knowledge, pre-training, or data of any kind beyond task observations. Moreover, this work demonstrates that the learned world model representation is capable of online transfer learning in the real-world from scratch. In comparison to existing approaches, including IRL, RL, and Behavior Cloning (BC), which have more restrictive assumptions, the proposed approach demonstrates significantly greater sample efficiency and success rates, enabling a practical path forward for online world modeling and planning from observation and interaction. Videos and more at: https://uwrobotlearning.github.io/mpail2/.

Open RAN (O-RAN) exposes rich control and telemetry interfaces across the Non-RT RIC, Near-RT RIC, and distributed units, but also makes it harder to operate multi-tenant, multi-objective RANs in a safe and auditable manner. In parallel, agentic AI systems with explicit planning, tool use, memory, and self-management offer a natural way to structure long-lived control loops. This article surveys how such agentic controllers can be brought into O-RAN: we review the O-RAN architecture, contrast agentic controllers with conventional ML/RL xApps, and organise the task landscape around three clusters: network slice life-cycle, radio resource management (RRM) closed loops, and cross-cutting security, privacy, and compliance. We then introduce a small set of agentic primitives (Plan-Act-Observe-Reflect, skills as tool use, memory and evidence, and self-management gates) and show, in a multi-cell O-RAN simulation, how they improve slice life-cycle and RRM performance compared to conventional baselines and ablations that remove individual primitives. Security, privacy, and compliance are discussed as architectural constraints and open challenges for standards-aligned deployments. This framework achieves an average 8.83\% reduction in resource usage across three classic network slices.

Vision-language models (VLMs) show promise in drafting radiology reports, yet they frequently suffer from logical inconsistencies, generating diagnostic impressions unsupported by their own perceptual findings or missing logically entailed conclusions. Standard lexical metrics heavily penalize clinical paraphrasing and fail to capture these deductive failures in reference-free settings. Toward guarantees for clinical reasoning, we introduce a neurosymbolic verification framework that deterministically audits the internal consistency of VLM-generated reports. Our pipeline autoformalizes free-text radiographic findings into structured propositional evidence, utilizing an SMT solver (Z3) and a clinical knowledge base to verify whether each diagnostic claim is mathematically entailed, hallucinated, or omitted. Evaluating seven VLMs across five chest X-ray benchmarks, our verifier exposes distinct reasoning failure modes, such as conservative observation and stochastic hallucination, that remain invisible to traditional metrics. On labeled datasets, enforcing solver-backed entailment acts as a rigorous post-hoc guarantee, systematically eliminating unsupported hallucinations to significantly increase diagnostic soundness and precision in generative clinical assistants.

Reinforcement Learning from Verifiable Rewards (RLVR) has emerged as a powerful paradigm for enhancing the complex reasoning capabilities of Large Reasoning Models. However, standard outcome-based supervision suffers from a critical limitation that penalizes trajectories that are largely correct but fail due to several missteps as heavily as completely erroneous ones. This coarse feedback signal causes the model to discard valuable largely correct rollouts, leading to a degradation in rollout diversity that prematurely narrows the exploration space. Process Reward Models have demonstrated efficacy in providing reliable step-wise verification for test-time scaling, naively integrating these signals into RLVR as dense rewards proves ineffective.Prior methods attempt to introduce off-policy guided whole-trajectory replacement that often outside the policy model's distribution, but still fail to utilize the largely correct rollouts generated by the model itself and thus do not effectively mitigate the narrowing of the exploration space. To address these issues, we propose SCOPE (Step-wise Correction for On-Policy Exploration), a novel framework that utilizes Process Reward Models to pinpoint the first erroneous step in suboptimal rollouts and applies fine-grained, step-wise off-policy rectification. By applying precise refinement on partially correct rollout, our method effectively salvages partially correct trajectories and increases diversity score by 13.5%, thereby sustaining a broad exploration space. Extensive experiments demonstrate that our approach establishes new state-of-the-art results, achieving an average accuracy of 46.6% on math reasoning and exhibiting robust generalization with 53.4% accuracy on out-of-distribution reasoning tasks.

Can narratives make arguments more persuasive? And to this end, which narrative features matter most? Although stories are often seen as powerful tools for persuasion, their specific role in online, unstructured argumentation remains underexplored. To address this gap, we present ARGUS, a framework for studying the impact of narration on persuasion in argumentative discourse. ARGUS introduces a new ChangeMyView corpus annotated for story presence and six key narrative features, integrating insights from two established theoretical frameworks that capture both textual narrative features and their effects on recipients. Leveraging both encoder-based classifiers and zero-shot large language models (LLMs), ARGUS identifies stories and narrative features and applies them at scale to examine how different narrative dimensions influence persuasion success in online argumentation.

Soft pneumatic actuators enable safe human-machine interaction with lightweight and powerful applied parts. On the other side, they suffer design limitations as regards complex actuation patterns, including minimum bending radii, multi-states capabilities and structural stability. We present geometry-based pneumatic actuators (GPAs), a design and implementation approach that introduces constraint layers with configurable CNC heat-sealed chambers. The approach achieves predictable deformation, near-zero bending radii, multi-states actuation, and enables customizable and repeatable complex actuated geometries. Mathematical modeling reveals predictable linear angle transformations and validates nonlinear torque-angle relationships across diverse configurations. We demonstrate versatility of the GPAs approach through three applications: a 49 g wrist exoskeleton reducing muscle activity by up to 51%, a 30.8 g haptic interface delivering 8 N force feedback with fast response, and a 208 g bipedal robot achieving multi-gait locomotion. GPAs establish a configurable platform for next-generation wearable robotics, haptic systems, and soft locomotion devices.

This paper presents the Artificial Agency Program (AAP), a position and research agenda for building AI systems as reality embedded, resource-bounded agents whose development is driven by curiosity-as-learning-progress under physical and computational constraints. The central thesis is that AI is most useful when treated as part of an extended human--tool system that increases sensing, understanding, and actuation capability while reducing friction at the interface between people, tools, and environments. The agenda unifies predictive compression, intrinsic motivation, empowerment and control, interface quality (unification), and language/self-communication as selective information bottlenecks. We formulate these ideas as a falsifiable program with explicit costs, staged experiments, and a concrete multimodal tokenized testbed in which an agent allocates limited budget among observation, action, and deliberation. The aim is to provide a conceptual and experimental framework that connects intrinsic motivation, information theory, thermodynamics, bounded rationality, and modern reasoning systems

Winter road maintenance is critical for ensuring public safety and reducing environmental impacts, yet existing methods struggle to manage large-scale routing problems effectively and mostly reply on human decision. This study presents a novel, scalable bi-level optimization framework, validated on real operational data on UK strategic road networks (M25, M6, A1), including interconnected local road networks in surrounding areas for vehicle traversing, as part of the highway operator's efforts to solve existing planning challenges. At the upper level, a reinforcement learning (RL) agent strategically partitions the road network into manageable clusters and optimally allocates resources from multiple depots. At the lower level, a multi-objective vehicle routing problem (VRP) is solved within each cluster, minimizing the maximum vehicle travel time and total carbon emissions. Unlike existing approaches, our method handles large-scale, real-world networks efficiently, explicitly incorporating vehicle-specific constraints, depot capacities, and road segment requirements. Results demonstrate significant improvements, including balanced workloads, reduced maximum travel times below the targeted two-hour threshold, lower emissions, and substantial cost savings. This study illustrates how advanced AI-driven bi-level optimization can directly enhance operational decision-making in real-world transportation and logistics.

Cox processes model overdispersed point process data via a latent stochastic intensity, but both nonparametric estimation of the intensity model and posterior inference over intensity paths are typically intractable, relying on expensive MCMC methods. We introduce Neural Diffusion Intensity Models, a variational framework for Cox processes driven by neural SDEs. Our key theoretical result, based on enlargement of filtrations, shows that conditioning on point process observations preserves the diffusion structure of the latent intensity with an explicit drift correction. This guarantees the variational family contains the true posterior, so that ELBO maximization coincides with maximum likelihood estimation under sufficient model capacity. We design an amortized encoder architecture that maps variable-length event sequences to posterior intensity paths by simulating the drift-corrected SDE, replacing repeated MCMC runs with a single forward pass. Experiments on synthetic and real-world data demonstrate accurate recovery of latent intensity dynamics and posterior paths, with orders-of-magnitude speedups over MCMC-based methods.

Resource-efficient training optimization techniques are becoming increasingly important as the size of large language models (LLMs) continues to grow. In particular, batch packing is commonly used in pre-training and supervised fine-tuning to achieve resource-efficient training. We propose preference packing, a method to enhance resource efficiency in training techniques that use data with different responses for the same input prompt, such as reward models or Direct Preference Optimization (DPO). Preference packing improves resource efficiency by reducing the attention operations for duplicate input prompts and decreasing KV cache memory usage. We conducted experiments on text-only datasets and image-included datasets and achieved at least 37% reduction in training time. Notably, this method can be applied alongside existing optimization techniques such as batch sorting, resulting in a 3.22x speedup.

We propose ACWI (Adaptive Correlation Weighted Intrinsic), an adaptive intrinsic reward scaling framework designed to dynamically balance intrinsic and extrinsic rewards for improved exploration in sparse reward reinforcement learning. Unlike conventional approaches that rely on manually tuned scalar coefficients, which often result in unstable or suboptimal performance across tasks, ACWI learns a state dependent scaling coefficient online. Specifically, ACWI introduces a lightweight Beta Network that predicts the intrinsic reward weight directly from the agent state through an encoder based architecture. The scaling mechanism is optimized using a correlation based objective that encourages alignment between the weighted intrinsic rewards and discounted future extrinsic returns. This formulation enables task adaptive exploration incentives while preserving computational efficiency and training stability. We evaluate ACWI on a suite of sparse reward environments in MiniGrid. Experimental results demonstrate that ACWI consistently improves sample efficiency and learning stability compared to fixed intrinsic reward baselines, achieving superior performance with minimal computational overhead.

Recently, there have been significant advancements in music generation. However, existing models primarily focus on creating modern pop songs, making it challenging to produce ancient music with distinct rhythms and styles, such as ancient Chinese SongCi. In this paper, we introduce SongSong, the first music generation model capable of restoring Chinese SongCi to our knowledge. Our model first predicts the melody from the input SongCi, then separately generates the singing voice and accompaniment based on that melody, and finally combines all elements to create the final piece of music. Additionally, to address the lack of ancient music datasets, we create OpenSongSong, a comprehensive dataset of ancient Chinese SongCi music, featuring 29.9 hours of compositions by various renowned SongCi music masters. To assess SongSong's proficiency in performing SongCi, we randomly select 85 SongCi sentences that were not part of the training set for evaluation against SongSong and music generation platforms such as Suno and SkyMusic. The subjective and objective outcomes indicate that our proposed model achieves leading performance in generating high-quality SongCi music.

Leveraging non-linear dimension reduction techniques, we remove the low dimension constraint from node embedding and propose COVE, an explainable high dimensional embedding that, when reduced to low dimension with UMAP, slightly increases performance on clustering and link prediction tasks. The embedding is inspired by neural embedding methods that use co-occurrence on a random walk as an indication of similarity, and is closely related to a diffusion process. Extending on recent community detection benchmarks, we find that a COVE UMAP HDBSCAN pipeline performs similarly to the popular Louvain algorithm.

Path signatures provide a rich representation of sequential data, with strong theoretical guarantees and good performance in a variety of machine-learning tasks. While signatures have progressed from fixed feature extractors to trainable components of machine-learning models, existing libraries often lack the required scalability for large-scale, gradient-based learning. To address this gap, this paper introduces pathsig, a PyTorch-native library that computes path signatures directly in the word basis. By using CUDA kernels to update signature coefficients in parallel over prefix-closed word sets, pathsig achieves high GPU throughput and near-minimal peak memory. Compared with other libraries, pathsig achieves 10-30x speedups for computation of truncated signatures and up to 4-10x speedups in training that require backpropagation through the signature. Beyond regular truncation, pathsig supports projections of the (infinite-dimensional) signature onto user-specified sets of words and anisotropic truncation motivated by inhomogeneous path regularity, enabling more compact representations that can reduce dimensionality, redundancy, and computational cost.

Large language models (LLMs) with reasoning capabilities have fueled a compelling narrative that reasoning universally improves performance across language tasks. We test this claim through a comprehensive evaluation of 504 configurations across seven model families--including adaptive, conditional, and reinforcement learning-based reasoning architectures--on sentiment analysis datasets of varying granularity (binary, five-class, and 27-class emotion). Our findings reveal that reasoning effectiveness is strongly task-dependent, challenging prevailing assumptions: (1) Reasoning shows task-complexity dependence--binary classification degrades up to -19.9 F1 percentage points (pp), while 27-class emotion recognition gains up to +16.0pp; (2) Distilled reasoning variants underperform base models by 3-18 pp on simpler tasks, though few-shot prompting enables partial recovery; (3) Few-shot learning improves over zero-shot in most cases regardless of model type, with gains varying by architecture and task complexity; (4) Pareto frontier analysis shows base models dominate efficiency-performance trade-offs, with reasoning justified only for complex emotion recognition despite 2.1x-54x computational overhead. We complement these quantitative findings with qualitative error analysis revealing that reasoning degrades simpler tasks through systematic over-deliberation, offering mechanistic insight beyond the high-level overthinking hypothesis.

Post-Training Quantization (PTQ) has emerged as an effective technique for alleviating the substantial computational and memory overheads of Vision-Language Models (VLMs) by compressing both weights and activations without retraining the full model. Existing PTQ methods primarily rely on static identification and global compensation of sensitive or outlier channels, yet they often overlook the distributional differences of these important channels across inputs, leading to unsatisfactory quantization. In this work, we observe that the distributions and occurrence frequencies of important channels vary significantly both across modalities and among tokens, even within the same modality. Accordingly, we propose \textbf{Quant Experts (QE)}, a token-aware adaptive error compensation with mixture-of-experts for VLMs quantization. QE divides the important channels into token-independent and token-dependent groups. For the former, a shared expert is designed for most tokens to compensate for global quantization error using a low-rank adapter. For the latter, routed experts including multiple routed low-rank adapters are elaborated to compensate for local quantization error related to specific tokens. Extensive experiments demonstrate that QE consistently enhances task accuracy across various quantization settings and model scales, ranging from 2B to 70B parameters, while maintaining performance comparable to full-precision models.

Reconstructing 3D clothed humans from monocular images and videos is a fundamental problem with applications in virtual try-on, avatar creation, and mixed reality. Despite significant progress in human body recovery, accurately reconstructing garment geometry, particularly for loose-fitting clothing, remains an open challenge. We propose a unified framework for high-fidelity 3D garment reconstruction from both single images and video sequences. Our approach combines Implicit Sewing Patterns (ISP) with a generative diffusion model to learn expressive garment shape priors in 2D UV space. Leveraging these priors, we introduce a mapping model that establishes correspondences between image pixels, UV pattern coordinates, and 3D geometry, enabling accurate and detailed garment reconstruction from single images. We further extend this formulation to dynamic reconstruction by introducing a spatio-temporal diffusion scheme with test-time guidance to enforce long-range temporal consistency. We also develop analytic projection-based constraints that preserve image-aligned geometry in visible regions while enforcing coherent completion in occluded areas over time. Although trained exclusively on synthetically simulated cloth data, our method generalizes well to real-world imagery and consistently outperforms existing approaches on both tight- and loose-fitting garments. The reconstructed garments preserve fine geometric detail while exhibiting realistic dynamic motion, supporting downstream applications such as texture editing, garment retargeting, and animation.

Multimodal large language models (MLLMs) have achieved remarkable progress in vision-language reasoning, yet they remain vulnerable to hallucination, where generated content deviates from visual evidence. Existing mitigation strategies either require costly supervision during training or introduce additional latency at inference time. Recent vision enhancement methods attempt to address this issue by reinforcing visual tokens during decoding, but they typically inject all tokens indiscriminately, which causes interference from background regions and distracts the model from critical cues. To overcome this challenge, we propose Adaptive Visual Reinforcement (AIR), a training-free framework for MLLMs. AIR consists of two components. Prototype-based token reduction condenses the large pool of visual tokens into a compact subset to suppress redundancy. OT-guided patch reinforcement quantifies the alignment between hidden states and patch embeddings to selectively integrate the most consistent patches into feed-forward layers. As a result, AIR enhances the model's reliance on salient visual information and effectively mitigates hallucination. Extensive experiments across representative MLLMs demonstrate that AIR substantially reduces hallucination while preserving general capabilities, establishing it as an effective solution for building reliable MLLMs.

Reward models are central to aligning large language models (LLMs) with human preferences. Yet most approaches rely on pointwise reward estimates that overlook the epistemic uncertainty in reward models arising from limited human feedback. Recent work suggests that quantifying this uncertainty can reduce the costs of human annotation via uncertainty-guided active learning and mitigate reward overoptimization in LLM post-training. However, uncertainty-aware reward models have so far been adopted without thorough comparison, leaving them poorly understood. This work introduces a unified framework, RewardUQ, to systematically evaluate uncertainty quantification for reward models. We compare common methods along standard metrics measuring accuracy and calibration, and we propose a new ranking strategy incorporating both dimensions for a simplified comparison. Our experimental results suggest that model size and initialization have the most meaningful impact on performance, and most prior work could have benefited from alternative design choices. To foster the development and evaluation of new methods and aid the deployment in downstream applications, we release our open-source framework as a Python package. Our code is available at https://github.com/lasgroup/rewarduq.

Market regime shifts induce distribution shifts that can degrade the performance of portfolio rebalancing policies. We propose macro-conditioned scenario-context rollout (SCR) that generates plausible next-day multivariate return scenarios under stress events. However, doing so faces new challenges, as history will never tell what would have happened differently. As a result, incorporating scenario-based rewards from rollouts introduces a reward--transition mismatch in temporal-difference learning, destabilizing RL critic training. We analyze this inconsistency and show it leads to a mixed evaluation target. Guided by this analysis, we construct a counterfactual next state using the rollout-implied continuations and augment the critic agent's bootstrap target. Doing so stabilizes the learning and provides a viable bias-variance tradeoff. In out-of-sample evaluations across 31 distinct universes of U.S. equity and ETF portfolios, our method improves Sharpe ratio by up to 76% and reduces maximum drawdown by up to 53% compared with classic and RL-based portfolio rebalancing baselines.

Autonomous drone racing has attracted increasing interest as a research topic for exploring the limits of agile flight. However, existing studies primarily focus on obstacle-free racetracks, while the perception and dynamic challenges introduced by obstacles remain underexplored, often resulting in low success rates and limited robustness in real-world flight. To this end, we propose a novel vision-based curriculum reinforcement learning framework for training a robust controller capable of addressing unseen obstacles in drone racing. We combine multi-stage cu rriculum learning, domain randomization, and a multi-scene updating strategy to address the conflicting challenges of obstacle avoidance and gate traversal. Our end-to-end control policy is implemented as a single network, allowing high-speed flight of quadrotors in environments with variable obstacles. Both hardware-in-the-loop and real-world experiments demonstrate that our method achieves faster lap times and higher success rates than existing approaches, effectively advancing drone racing in obstacle-rich environments. The video and code are available at: https://github.com/SJTU-ViSYS-team/CRL-Drone-Racing.

Large vision-language models (LVLMs) have achieved remarkable progress in vision-language reasoning tasks, yet ensuring their safety remains a critical challenge. Recent input-side defenses detect unsafe images with CLIP and prepend safety prefixes to prompts, but they still suffer from inaccurate detection in complex scenes and unstable safety signals during decoding. To address these issues, we propose GuardAlign, a training-free defense framework that integrates two strategies. First, OT-enhanced safety detection leverages optimal transport to measure distribution distances between image patches and unsafe semantics, enabling accurate identification of malicious regions without additional computational cost. Second, cross-modal attentive calibration strengthens the influence of safety prefixes by adaptively reallocating attention across layers, ensuring that safety signals remain consistently activated throughout generation. Extensive evaluations on six representative MLLMs demonstrate that GuardAlign reduces unsafe response rates by up to 39% on SPA-VL, while preserving utility, achieving an improvement on VQAv2 from 78.51% to 79.21%.