Existing VLMs have achieved strong performance in video understanding, yet they struggle with long-video spatiotemporal reasoning when target objects become invisible, often mistaking "invisible" for "unknown". We define this challenge as hidden-state spatiotemporal reasoning: inferring object states during prolonged invisible intervals from context interactions. To address this, we propose StateTrace, a novel object-centric framework that endows VideoLLMs with an explicit mechanism for hidden state reasoning in long videos. StateTrace builds a reusable spatiotemporal state memory that organizes object trajectories, inter-object relations, and state-transition events into a structured reasoning substrate. At inference time, it retrieves question-relevant state-evolution trajectories and converts them into compact reasoning cues, enabling the model to explicitly reason about why an object disappears, how its state evolves while invisible, and whether that state should persist at query time. We further build HSR-Bench, a diagnostic benchmark for hidden-state reasoning, containing 1,427 video-QA samples from 1,384 unique videos. Extensive experiments across multiple VideoLLMs show that StateTrace consistently improves performance on both public benchmarks and HSR-Bench (e.g., improving VideoLLaMA3 from 39.6 to 64.2 on HSR-Bench).

Leveraging multiple specialized LLMs can combine complementary strengths, but existing approaches trade adaptability for stability: routing commits prematurely, heuristic ensembling depends on fragile proxies, and parameter merging introduces interference. We propose DLLG (Dynamic Logit-Level Gating), a dynamic logit-level ensembling framework that learns token-level expert fusion from sparse response-level supervision. A lightweight gating module predicts step-wise fusion weights, linking trajectory-level correctness to generation without token-level labels or expert retraining. Across diverse reasoning and code benchmarks, DLLG consistently outperforms strong routing, heuristic ensembling, and parameter-merging baselines across model scales, highlighting learned logit-level fusion as a robust and scalable paradigm for integrating specialized experts.

Vision-Language-Action (VLA) models have demonstrated remarkable generalization capabilities in robotic manipulation tasks, yet their substantial computational overhead remains a critical obstacle to real-world deployment. Improving inference efficiency is therefore essential for practical robotic applications. Existing acceleration methods often rely on heuristic or static strategies–such as rule-based token caching or pruning–that are decoupled from task objectives and fail to adapt to dynamic scene changes. In this work, we reformulate inference acceleration as a learnable policy optimization problem and propose a novel framework that integrates a dynamic, task-aware decision-making process directly into the VLA model. At its core are two lightweight, cooperative modules: a Cached Token Selector, which determines which tokens should be reused, and a Cache Ratio Predictor, which controls how many tokens to reuse. Training these modules is non-trivial due to their discrete decisions. We address this by adopting a differentiable relaxation that allows gradient-based end-to-end optimization. Extensive experiments on the LIBERO and SIMPLER benchmarks, as well as real-robot evaluations, show that our method achieves a 1.76x wall-clock inference speedup while simultaneously improving the average success rate by 1.9 percentage points (from 75.0

Video multimodal large language models have made strong progress on open-ended video understanding, but they still lack precise local spatiotemporal perception. When two videos share almost the same global semantics and differ only in a short time span or a small region, current models often fail to find the change and provide reliable evidence. We propose DELTAVID, a verifiable proxy-task framework that enhances fine-grained spatiotemporal perception with cross-video differences. The key idea is to turn cross-video spot-the-difference into a trainable perception signal, where a model identifies local changes, judges temporal boundaries, and organizes spatial evidence by comparing similar videos. To make this signal scalable to train and reliable to evaluate, we further introduce DELTAVID-10K and DELTAVID-Bench, which convert controllable local differences in real videos into evidence-labeled training and test samples. Experiments show that DELTAVID substantially improves performance on cross-video difference understanding and transfers the learned local evidence ability to general video understanding benchmarks, including MMVU, MLVU, Video-MME, VideoHolmes, VideoMMMU, LVBench, TempCompass, and LongVideoBench. These results show that cross-video differences are not only an effective way to diagnose fine-grained perception failures, but also a scalable proxy supervision that moves Video MLLMs from coarse semantic understanding toward fine-grained spatiotemporal evidence reasoning.

Large language model (LLM)–based multi-agent systems (MAS) show strong promise for complex reasoning, planning, and tool-augmented tasks, but designing effective MAS architectures remains labor-intensive, brittle, and hard to generalize. Existing automatic MAS generation methods either rely on code generation, which often leads to executability and robustness failures, or impose rigid architectural templates that limit expressiveness and adaptability. We propose Evolutionary Generation of Multi-Agent Systems (EvoMAS), which formulates MAS generation as structured configuration generation. EvoMAS performs evolutionary generation in configuration space. Specifically, EvoMAS selects initial configurations from a pool, applies feedbackconditioned mutation and crossover guided by execution traces, and iteratively refines both the candidate pool and an experience memory. We evaluate EvoMAS on diverse benchmarks, including BBEH, SWE-Bench, and WorkBench, covering reasoning, software engineering, and tool-use tasks. EvoMAS consistently improves task performance over both human-designed MAS and prior automatic MAS generation methods, while producing generated systems with higher executability and runtime robustness. EvoMAS outperforms the agent evolution method EvoAgent by +10.5 points on BBEH reasoning and +7.1 points on WorkBench. With Claude-4.5-Sonnet, EvoMAS also reaches 79.1% on SWE-BenchVerified, matching the top of the leaderboard.

We introduce StaminaBench, a benchmark that measures the stamina of coding agents: how many consecutive interaction turns (change requests) they can handle before failing. Unlike the prevailing fraction-of-tasks-solved metric, this matches real vibe-coding where sessions run dozens or hundreds of turns. In StaminaBench, agents implement a REST API server and modify it across a tunable number of procedurally generated follow-up change requests - 100 in our experiments, resulting in codebases of up to 6,000 lines. Tests are generated fully programmatically without LLM involvement, ensuring reproducibility and reliability; change sequences are drawn from either a hardcoded or LLM-driven sampler, both constrained to a structured action space to ensure changes are valid. The agent and the server run in an isolated environment and communicate with the benchmark through HTTP, making testing fully black-box and language-agnostic. We evaluate six agent harnesses paired with seven open-source LLMs across 20 scenarios of 100 turns each and find that: (1) all the tested models fail within 5-6 turns, confirming that vibe-coding-style programming without thorough testing produces bugs; (2) passing test feedback back to the agent and allowing it to retry improves passed turn count by up to 12x; and (3) a good harness is required for strong performance: stronger models exhibit up to a 6x gap between their best and worst harness, while weaker models fail with any harness. We release the benchmark and the generated tasks to enable further research into multi-turn coding agent behavior. Benchmark code and data: github.com/amazon-science/StaminaBench.

RL post-training strategies are dataset-dependent and reveal a recurring empirical pattern: capacity parameters accumulate monotonically across stages, while regularization parameters predominantly oscillate in response to shifting training dynamics. This distinction matters because fixed schedules commit all parameters to fixed trajectories and therefore cannot express the non-stationary exploration-exploitation tradeoffs that regularization must track; the principle provides actionable design rules for multi-stage training. We discover this through LLMZero, a system where LLM agents search over training trajectories via tree search, diagnosing pathologies at each checkpoint and proposing coordinated multi-parameter transitions. Across 4 diverse GRPO tasks, LLMZero discovers strategies that improve over the base model by 9

Reinforcement learning (RL) post-training has recently driven major gains in long chain-of-thought reasoning large language models (LLMs), but the high inference cost of such models motivates distillation into smaller students. Most existing knowledge distillation (KD) methods are designed for supervised fine-tuning (SFT), relying on fixed teacher traces or teacher-student Kullback–Leibler (KL) divergence based regularization. When combined with RL, these approaches often suffer from distribution mismatch and objective interference: teacher supervision may not align with the student’s evolving rollout distribution, and the KL regularizer can compete with reward maximization and require careful loss balancing. To address these issues, we propose \emph{RL-aware distillation} (RLAD), which performs selective imitation during RL---guiding the student toward the teacher only when it improves the current policy update. Our core component,Trust Region Ratio Distillation (TRRD), replaces the teacher-student KL regularizer with a PPO/GRPO-style likelihood-ratio objective anchored to a teacher--old-policy mixture, yielding advantage-aware, trust-region-bounded distillation on student rollouts and naturally balancing exploration, exploitation, and imitation. Across diverse logic reasoning and math benchmarks, RLAD}consistently outperforms offline distillation, standard GRPO, and KL-based on-policy teacher–student knowledge distillation.

We introduce Talk2Move, a reinforcement learning (RL) based diffusion framework for text-instructed spatial transformation of objects within scenes. Spatially manipulating objects in a scene through natural language poses a challenge for multimodal generation systems. While existing text-based manipulation methods can adjust appearance or style, they struggle to perform object-level geometric transformations-such as translating, rotating, or resizing objects-due to scarce paired supervision and pixel-level optimization limits. Talk2Move employs Group Relative Policy Optimization (GRPO) to explore geometric actions through diverse rollouts generated from input images and lightweight textual variations, removing the need for costly paired data. A spatial reward guided model aligns geometric transformations with linguistic description, while off-policy step evaluation and active step sampling improve learning efficiency by focusing on informative transformation stages. Furthermore, we design object-centric spatial rewards that evaluate displacement, rotation, and scaling behaviors directly, enabling interpretable and coherent transformations. Experiments on curated benchmarks demonstrate that Talk2Move achieves precise, consistent, and semantically faithful object transformations, outperforming existing text-guided editing approaches in both spatial accuracy and scene coherence.

Interfacial adhesion between selective laser-melted (SLM) AlSi10Mg and polyimide (PI) insulating coatings is often limited by mismatched physicochemical properties. To improve adhesion, Al-rich and Si-rich microstructured surfaces were fabricated on the XY plane (perpendicular to the build direction) and the Z plane (parallel to the build direction) by acidic and alkaline etching, exploiting the characteristic microstructure of SLM AlSi10Mg. Surface topography, chemical composition, and wettability were characterized, and interfacial mechanical performance was evaluated by shear and pull-off tests. The microstructures increased surface roughness and improved wettability. The shear strength rose from 2.6 ± 1.5 MPa for the polished surface to 43.2 ± 8.6 MPa. The polished surface showed a pull-off strength of 2.2 ± 0.25 MPa. In pull-off tests, failure mainly occurred within the dolly/adhesive/PI system, indicating that the interfacial tensile strength exceeded the strength of the adhesive system; the maximum measured pull-off strength was 29.0 ± 1.3 MPa. Fractography predominantly showed cohesive failure in PI on Al-rich microstructures. Si-rich microstructures exhibited mixed failure, including fracture of the Si skeleton and tearing of PI, together with interfacial microcracks.

Existing embodied control research demonstrates remarkable performance improvements by scaling training data and model size. We instead explore inference-time strategy as an alternative axis. Non-deterministic generative models, such as diffusion and autoregressive models, have been widely adopted in the field of embodied control. However, the single-shot inference paradigm limits their performance. In this paper, we propose \textbf{TapSampling}, a plug-and-play framework for inference-time sampling. First, we introduce an Action-VAE to represent actions in a low-dimensional latent space. The Action-VAE maps initial actions from policies into a compressed posterior distribution, from which an arbitrary number of latent samples can be drawn and decoded into candidate actions that approximately follow the true action distribution. Second, we formulate action verification as task-progress outcome prediction and train the verifier by leveraging the intrinsic sequential information of robotic datasets. The predicted scores have clear semantic grounding, enabling interpretable action selection. Furthermore, TapSampling is a policy-agnostic framework. Extensive experiments in both simulated and real-world environments demonstrate that our method effectively improves multiple generalist policies substantially without further finetuning the policy models.

A fully digital workflow incorporating autonomous robotic surgery is described. A prefabricated interim prosthesis offers the potential to streamline the process and reduce chairside time. Adopting this digital workflow can simplify the treatment procedure and help minimize the overall time required for the provision of implant-supported prostheses.

Harnessing the power of diffusion models to synthesize auxiliary training data based on latent space features has proven effective in enhancing out-of-distribution (OOD) detection performance. However, extracting effective features outside the in-distribution (ID) boundary in latent space remains challenging due to the difficulty of identifying decision boundaries between classes. This paper proposes a novel framework called Boundary-based Out-Of-Distribution data generation (BOOD), which synthesizes high-quality OOD features and generates human-compatible outlier images using diffusion models. BOOD first learns a text-conditioned latent feature space from the ID dataset, selects ID features closest to the decision boundary, and perturbs them to cross the decision boundary to form OOD features. These synthetic OOD features are then decoded into images in pixel space by a diffusion model. Compared to previous works, BOOD provides a more training efficient strategy for synthesizing informative OOD features, facilitating clearer distinctions between ID and OOD data. Extensive experimental results on common benchmarks demonstrate that BOOD surpasses the state-of-the-art method significantly, achieving a 29.64\% decrease in average FPR95 (40.31\% vs. 10.67\%) and a 7.27\% improvement in average AUROC (90.15\% vs. 97.42\%) on the Cifar-100 dataset.

Real-world data often contains intrinsic ambiguity that the common single-hard-label annotation paradigm ignores. Standard training using ambiguous data with these hard labels may produce overly confident models and thus leading to poor generalization. In this paper, we propose a novel framework called Quantized Label Learning (QLL) to alleviate this issue. First, we formulate QLL as learning from (very) ambiguous data with hard labels: ideally, each ambiguous instance should be associated with a ground-truth soft-label distribution describing its corresponding probabilistic weight in each class, however, this is usually not accessible; in practice, we can only observe a quantized label, i.e., a hard label sampled (quantized) from the corresponding ground-truth soft-label distribution, of each instance, which can be seen as a biased approximation of the ground-truth soft-label. Second, we propose a Class-wise Positive-Unlabeled (CPU) risk estimator that allows us to train accurate classifiers from only ambiguous data with quantized labels. Third, to simulate ambiguous datasets with quantized labels in the real world, we design a mixing-based ambiguous data generation procedure for empirical evaluation. Experiments demonstrate that our CPU method can significantly improve model generalization performance and outperform the baselines.

STUDY DESIGN:Retrospective cohort study. OBJECTIVE:This study aimed to analyze the correlation between vertebral bone quality (VBQ) scores and bone turnover markers (BTMs) in lumbar spine disorders. BACKGROUND:VBQ score is increasingly used in the assessment of bone mineral density (BMD), and bone quality profiles are closely related to bone metabolism. However, the level of bone turnover is often overlooked in clinical practice. PATIENTS AND METHODS:We retrospectively analyzed the data from 234 patients who underwent lumbar spine surgery. VBQ scores were evaluated using preoperative lumbar T1-weighted magnetic resonance imaging, with patients classified into high (>3.3), middle (2.7 to 3.3), and low (<2.7) VBQ groups. The data of computed tomography images and dual-energy x-ray absorptiometry were collected to obtain the Hounsfield unit (HU) and T value. Correlation analysis, linear regression, and 1-way ANOVA were used to analyze the relationship between BTMs, including parathyroid hormone (ng/L), 25-hydroxyvitamin D3 (nmol/L), osteocalcin (ng/mL), β-CrossLaps (ng/mL), and procollagen type I N-propeptide (ng/mL) and bone quality. P <0.05 was considered statistically different. RESULTS:Comparative analysis showed BTMs varied markedly across VBQ categories ( P < 0.0001 to P = 0.0158), with osteoblast-related markers [25-hydroxyvitamin D3, OC] decreasing and osteoclast-related markers (β-CrossLaps) increasing with higher VBQ scores. Multivariate analysis confirmed age, sex, BMI, and specific BTMs (except for PINP) as independent predictors of VBQ scores ( P = 0.0075 to 0.0256). VBQ demonstrated superior correlations with BTMs ( r = 0.52 to 0.63) compared with T-scores and HU values, highlighting its enhanced sensitivity to dynamic bone metabolism. Notably, patients with normal BMD/high HU but intermediate VBQ scores showed suppressed osteoblastic activity ( P = 0.0009 to 0.0036), while those with osteopenia-level BMD/HU and elevated VBQ scores exhibited exacerbated bone resorption ( P < 0.001). CONCLUSION:This is the first study to link VBQ scores with BTMs in lumbar spine patients. Preoperative VBQ assessment through magnetic resonance imaging can initially evaluate bone metabolism without radiation exposure, guiding osteoporosis treatment post-surgery to optimize bone health. LEVEL OF EVIDENCE:Level III.

Vehicle logo detection plays a crucial role in various computer vision applications, such as vehicle classification and detection. In this research, we propose an improved vehicle logo detection method leveraging the MAMBA structure. The MAMBA structure integrates multiple attention mechanisms and bidirectional feature aggregation to enhance the discriminative power of the detection model. Specifically, we introduce the Multi-Head Attention for Multi-Scale Feature Fusion (MHAMFF) module to capture multi-scale contextual information effectively. Moreover, we incorporate the Bidirectional Aggregation Mechanism (BAM) to facilitate information exchange between different layers of the detection network. Experimental results on a benchmark dataset (VLD-45 dataset) demonstrate that our proposed method outperforms baseline models in terms of both detection accuracy and efficiency. Furthermore, extensive ablation studies validate the effectiveness of each component in the MAMBA structure. Overall, the proposed MAMBA-based vehicle logo detection approach shows promising potential for real-world applications in intelligent transportation systems.

Promoting green innovation capability is of great significance to China's realization of green economic power and green innovation power. Based on the perspective of economic development, this paper establishes a comprehensive evaluation system for improving regional green innovation capability from 19 indicators in five aspects, including green innovation input, green innovation output, green development, talent development and utilization, and economic benefits, and uses the entropy weight TOPSIS method to estimate the green innovation capability of Shandong Province from 2012 to 2022. On this basis, the European distance method is used to identify the key influencing factors. The results show that the overall green innovation capability of Shandong Province is on the rise, but the speed of green innovation capability improvement is slightly lower than the speed of green innovation input; Under the average level of green innovation during the sample period, green innovation input, green innovation output, green development, and talent development and utilization have promoting effects on the improvement of green innovation ability, but the economic benefits fail to meet the requirements of promoting green innovation ability.

Traditional studies emphasize the significance of context information in improving matting performance. Consequently, deep learning-based matting methods delve into designing pooling or affinity-based context aggregation modules to achieve superior results. However, these modules cannot well handle the context scale shift caused by the difference in image size during training and inference, resulting in matting performance degradation. In this paper, we revisit the context aggregation mechanisms of matting networks and find that a basic encoder-decoder network without any context aggregation modules can actually learn more universal context aggregation, thereby achieving higher matting performance compared to existing methods. Building on this insight, we present AEMatter, a matting network that is straightforward yet very effective. AEMatter adopts a Hybrid-Transformer backbone with appearance-enhanced axis-wise learning (AEAL) blocks to build a basic network with strong context aggregation learning capability. Furthermore, AEMatter leverages a large image training strategy to assist the network in learning context aggregation from data. Extensive experiments on five popular matting datasets demonstrate that the proposed AEMatter outperforms state-of-the-art matting methods by a large margin. The source code is available at https://github.com/aipixel/AEMatter.