Tendon-driven anthropomorphic robotic hands often lack direct joint angle sensing, as the integration of joint encoders can compromise mechanical compactness and dexterity. This paper presents a computational method for estimating joint positions from measured tendon displacements and tensions. An efficient kinematic modeling framework for anthropomorphic hands is first introduced based on the Denavit-Hartenberg convention. Using a simplified tendon model, a system of nonlinear equations relating tendon states to joint positions is derived and solved via a nonlinear optimization approach. The estimated joint angles are then employed for closed-loop control through a Jacobian-based proportional-integral (PI) controller augmented with a feedforward term, enabling gesture tracking without direct joint sensing. The effectiveness and limitations of the proposed estimation and control framework are demonstrated in the MuJoCo simulation environment using the Anatomically Correct Biomechatronic Hand, featuring five degrees of freedom for each long finger and six degrees of freedom for the thumb.

Large language models (LLMs) have achieved remarkable progress in code generation, yet their potential for software protection remains largely untapped. Reverse engineering continues to threaten software security, while traditional virtual machine protection (VMP) relies on rigid, rule-based transformations that are costly to design and vulnerable to automated analysis. In this work, we present the first protection-aware framework that learns robust representations of VMP-protected code. Our approach builds large-scale paired datasets of source code and normalized VM implementations, and introduces hierarchical dependency modeling at intra-, preceding-, and inter-instruction levels. We jointly optimize language modeling with functionality-aware and protection-aware contrastive objectives to capture both semantic equivalence and protection strength. To further assess resilience, we propose a protection effectiveness optimization task that quantifies and ranks different VM variants derived from the same source. Coupled with a two-stage continual pre-training and fine-tuning pipeline, our method enables models to generate, compare, and reason over protected code. Extensive experiments show that our framework significantly improves robustness across diverse protection levels, opening a new research direction for learning-based software defense. In this work, we present ShieldedCode, the first protection-aware framework that learns robust representations of VMP-protected code. Our method achieves 26.95% Pass@1 on L0 VM code generation compared to 22.58% for GPT-4o., and improves binary similarity detection Recall@1 by 10% over state of art methods like jTrans.

Visual Question-Answering (VQA) is a challenging multimodal task that requires integrating visual and textual information to generate accurate responses. While multimodal Retrieval-Augmented Generation (mRAG) has shown promise in enhancing VQA systems by providing more evidence on both image and text sides, the default procedure that addresses VQA queries, especially the knowledge-intensive ones, often relies on multi-stage pipelines of mRAG with inherent dependencies. To mitigate the inefficiency limitations while maintaining VQA task performance, this paper proposes a method that trains a multimodal planning agent, dynamically decomposing the mRAG pipeline to solve the VQA task. Our method optimizes the trade-off between efficiency and effectiveness by training the agent to intelligently determine the necessity of each mRAG step. In our experiments, the agent can help reduce redundant computations, cutting search time by over 60\% compared to existing methods and decreasing costly tool calls. Meanwhile, experiments demonstrate that our method outperforms all baselines, including a Deep Research agent and a carefully designed prompt-based method, on average over six various datasets. Code will be released.

Prompt learning (PL) has emerged as an effective strategy to adapt vision-language models (VLMs), such as CLIP, for downstream tasks under limited supervision. While PL has demonstrated strong generalization on natural image datasets, its transferability to remote sensing (RS) imagery remains underexplored. RS data present unique challenges, including multi-label scenes, high intra-class variability, and diverse spatial resolutions, that hinder the direct applicability of existing PL methods. In particular, current prompt-based approaches often struggle to identify dominant semantic cues and fail to generalize to novel classes in RS scenarios. To address these challenges, we propose BiMoRS, a lightweight bi-modal prompt learning framework tailored for RS tasks. BiMoRS employs a frozen image captioning model (e.g., BLIP-2) to extract textual semantic summaries from RS images. These captions are tokenized using a BERT tokenizer and fused with high-level visual features from the CLIP encoder. A lightweight cross-attention module then conditions a learnable query prompt on the fused textual-visual representation, yielding contextualized prompts without altering the CLIP backbone. We evaluate BiMoRS on four RS datasets across three domain generalization (DG) tasks and observe consistent performance gains, outperforming strong baselines by up to 2% on average. Codes are available at https://github.com/ipankhi/BiMoRS.

This study applies Large Language Models (LLMs) to two foundational Electronic Health Record (EHR) data science tasks: structured data querying (using programmatic languages, Python/Pandas) and information extraction from unstructured clinical text via a Retrieval Augmented Generation (RAG) pipeline. We test the ability of LLMs to interact accurately with large structured datasets for analytics and the reliability of LLMs in extracting semantically correct information from free text health records when supported by RAG. To this end, we presented a flexible evaluation framework that automatically generates synthetic question and answer pairs tailored to the characteristics of each dataset or task. Experiments were conducted on a curated subset of MIMIC III, (four structured tables and one clinical note type), using a mix of locally hosted and API-based LLMs. Evaluation combined exact-match metrics, semantic similarity, and human judgment. Our findings demonstrate the potential of LLMs to support precise querying and accurate information extraction in clinical workflows.

Training reinforcement learning (RL) policies for legged robots remains challenging due to high-dimensional continuous actions, hardware constraints, and limited exploration. Existing methods for locomotion and whole-body control work well for position-based control with environment-specific heuristics (e.g., reward shaping, curriculum design, and manual initialization), but are less effective for torque-based control, where sufficiently exploring the action space and obtaining informative gradient signals for training is significantly more difficult. We introduce Growing Policy Optimization (GPO), a training framework that applies a time-varying action transformation to restrict the effective action space in the early stage, thereby encouraging more effective data collection and policy learning, and then progressively expands it to enhance exploration and achieve higher expected return. We prove that this transformation preserves the PPO update rule and introduces only bounded, vanishing gradient distortion, thereby ensuring stable training. We evaluate GPO on both quadruped and hexapod robots, including zero-shot deployment of simulation-trained policies on hardware. Policies trained with GPO consistently achieve better performance. These results suggest that GPO provides a general, environment-agnostic optimization framework for learning legged locomotion.

Skill assessment in procedural videos is crucial for the objective evaluation of human performance in settings such as manufacturing and procedural daily tasks. Current research on skill assessment has predominantly focused on sports and lacks large-scale datasets for complex procedural activities. Existing studies typically involve only a limited number of actions, focus on either pairwise assessments (e.g., A is better than B) or on binary labels (e.g., good execution vs needs improvement). In response to these shortcomings, we introduce ProSkill, the first benchmark dataset for action-level skill assessment in procedural tasks. ProSkill provides absolute skill assessment annotations, along with pairwise ones. This is enabled by a novel and scalable annotation protocol that allows for the creation of an absolute skill assessment ranking starting from pairwise assessments. This protocol leverages a Swiss Tournament scheme for efficient pairwise comparisons, which are then aggregated into consistent, continuous global scores using an ELO-based rating system. We use our dataset to benchmark the main state-of-the-art skill assessment algorithms, including both ranking-based and pairwise paradigms. The suboptimal results achieved by the current state-of-the-art highlight the challenges and thus the value of ProSkill in the context of skill assessment for procedural videos. All data and code are available at https://fpv-iplab.github.io/ProSkill/

Assessing ways in which Language Models can reduce their hallucinations and improve the outputs' quality is crucial to ensure their large-scale use. However, methods such as fine-tuning on domain-specific data or the training of a separate \textit{ad hoc} verifier require demanding computational resources (not feasible for many user applications) and constrain the models to specific fields of knowledge. In this thesis, we propose a dialectic pipeline that preserves LLMs' generalization abilities while improving the quality of its answer via self-dialogue, enabling it to reflect upon and correct tentative wrong answers. We experimented with different pipeline settings, testing our proposed method on different datasets and on different families of models. All the pipeline stages are enriched with the relevant context (in an oracle-RAG setting) and a study on the impact of its summarization or its filtering is conducted. We find that our proposed dialectic pipeline is able to outperform by significative margins the standard model answers and that it consistently achieves higher performances than Chain-of-Thought only prompting.

Face morphing attacks present a significant threat to face recognition systems used in electronic identity enrolment and border control, particularly in single-image morphing attack detection (S-MAD) scenarios where no trusted reference is available. In spite of the vast amount of research on this problem, morph detection systems struggle in cross-dataset scenarios. To address this problem, we introduce a region-aware frequency-based morph detection strategy that drastically improves over strong baseline methods in challenging cross-dataset and cross-morph settings using a lightweight approach. Having observed the separability of bona fide and morph samples in the frequency domain of different facial parts, our approach 1) introduces the concept of residual frequency domain, where the frequency of the signal is decoupled from the natural spectral decay to easily discriminate between morph and bona fide data; 2) additionally, we reason in a global and local manner by combining the evidence from different facial regions in a Markov Random Field, which infers a globally consistent decision. The proposed method, trained exclusively on the synthetic morphing attack detection development dataset (SMDD), is evaluated in challenging cross-dataset and cross-morph settings on FRLL-Morph and MAD22 sets. Our approach achieves an average equal error rate (EER) of 1.85\% on FRLL-Morph and ranks second on MAD22 with an average EER of 6.12\%, while also obtaining a good bona fide presentation classification error rate (BPCER) at a low attack presentation classification error rate (APCER) using only spectral features. These findings indicate that Fourier-domain residual modeling with structured regional fusion offers a competitive alternative to deep S-MAD architectures.

While reinforcement learning with verifiable rewards (RLVR) has advanced LLM reasoning in structured domains like mathematics and programming, its application to general-domain reasoning tasks remains challenging due to the absence of verifiable reward signals. To this end, methods like Reinforcement Learning with Reference Probability Reward (RLPR) have emerged, leveraging the probability of generating the final answer as a reward signal. However, these outcome-focused approaches neglect crucial step-by-step supervision of the reasoning process itself. To address this gap, we introduce Probabilistic Process Supervision (P2S), a novel self-supervision framework that provides fine-grained process rewards without requiring a separate reward model or human-annotated reasoning steps. During reinforcement learning, P2S synthesizes and filters a high-quality reference reasoning chain (gold-CoT). The core of our method is to calculate a Path Faithfulness Reward (PFR) for each reasoning step, which is derived from the conditional probability of generating the gold-CoT's suffix, given the model's current reasoning prefix. Crucially, this PFR can be flexibly integrated with any outcome-based reward, directly tackling the reward sparsity problem by providing dense guidance. Extensive experiments on reading comprehension and medical Question Answering benchmarks show that P2S significantly outperforms strong baselines.

We investigate whether \emph{LLM-based agents} can develop task-oriented communication protocols that differ from standard natural language in collaborative reasoning tasks. Our focus is on two core properties such task-oriented protocols may exhibit: Efficiency -- conveying task-relevant information more concisely than natural language, and Covertness -- becoming difficult for external observers to interpret, raising concerns about transparency and control. To investigate these aspects, we use a referential-game framework in which vision-language model (VLM) agents communicate, providing a controlled, measurable setting for evaluating language variants. Experiments show that VLMs can develop effective, task-adapted communication patterns. At the same time, they can develop covert protocols that are difficult for humans and external agents to interpret. We also observe spontaneous coordination between similar models without explicitly shared protocols. These findings highlight both the potential and the risks of task-oriented communication, and position referential games as a valuable testbed for future work in this area.

Accurately modelling the dynamics of complex systems and discovering their governing differential equations are critical tasks for accelerating scientific discovery. Using noisy, synthetic data from two damped oscillatory systems, we explore the extrapolation capabilities of Neural Ordinary Differential Equations (NODEs) and the ability of Symbolic Regression (SR) to recover the underlying equations. Our study yields three key insights. First, we demonstrate that NODEs can extrapolate effectively to new boundary conditions, provided the resulting trajectories share dynamic similarity with the training data. Second, SR successfully recovers the equations from noisy ground-truth data, though its performance is contingent on the correct selection of input variables. Finally, we find that SR recovers two out of the three governing equations, along with a good approximation for the third, when using data generated by a NODE trained on just 10% of the full simulation. While this last finding highlights an area for future work, our results suggest that using NODEs to enrich limited data and enable symbolic regression to infer physical laws represents a promising new approach for scientific discovery.

Virtual sensors use machine learning to predict target signals from available measurements, replacing expensive physical sensors in critical applications. Existing virtual sensor approaches require application-specific models with hand-selected inputs for each sensor, cannot leverage task synergies, and lack consistent benchmarks. At the same time, emerging time series foundation models are computationally expensive and limited to predicting their input signals, making them incompatible with virtual sensors. We introduce the first foundation model for virtual sensors addressing both limitations. Our unified model can simultaneously predict diverse virtual sensors exploiting synergies while maintaining computational efficiency. It learns relevant input signals for each virtual sensor, eliminating expert knowledge requirements while adding explainability. In our large-scale evaluation on a standard benchmark and an application-specific dataset with over 18 billion samples, our architecture achieves 415x reduction in computation time and 951x reduction in memory requirements, while maintaining or even improving predictive quality compared to baselines. Our model scales gracefully to hundreds of virtual sensors with nearly constant parameter count, enabling practical deployment in large-scale sensor networks.

We propose DIVERSE, a framework for systematically exploring the Rashomon set of deep neural networks, the collection of models that match a reference model's accuracy while differing in their predictive behavior. DIVERSE augments a pretrained model with Feature-wise Linear Modulation (FiLM) layers and uses Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to search a latent modulation space, generating diverse model variants without retraining or gradient access. Across MNIST, PneumoniaMNIST, and CIFAR-10, DIVERSE uncovers multiple high-performing yet functionally distinct models. Our experiments show that DIVERSE offers a competitive and efficient exploration of the Rashomon set, making it feasible to construct diverse sets that maintain robustness and performance while supporting well-balanced model multiplicity. While retraining remains the baseline to generate Rashomon sets, DIVERSE achieves comparable diversity at reduced computational cost.

Multimodal sarcasm detection (MSD) aims to identify sarcasm within image-text pairs by modeling semantic incongruities across modalities. Existing methods often exploit cross-modal embedding misalignment to detect inconsistency but struggle when visual and textual content are loosely related or semantically indirect. While recent approaches leverage large language models (LLMs) to generate sarcastic cues, the inherent diversity and subjectivity of these generations often introduce noise. To address these limitations, we propose the Generative Discrepancy Comparison Network (GDCNet). This framework captures cross-modal conflicts by utilizing descriptive, factually grounded image captions generated by Multimodal LLMs (MLLMs) as stable semantic anchors. Specifically, GDCNet computes semantic and sentiment discrepancies between the generated objective description and the original text, alongside measuring visual-textual fidelity. These discrepancy features are then fused with visual and textual representations via a gated module to adaptively balance modality contributions. Extensive experiments on MSD benchmarks demonstrate GDCNet's superior accuracy and robustness, establishing a new state-of-the-art on the MMSD2.0 benchmark.

Reinforcement Learning with Verifiable Rewards (RLVR) offers a robust mechanism for enhancing mathematical reasoning in large models. However, we identify a systematic lack of emphasis on more challenging questions in existing methods from both algorithmic and data perspectives, despite their importance for refining underdeveloped capabilities. Algorithmically, widely used Group Relative Policy Optimization (GRPO) suffers from an implicit imbalance where the magnitude of policy updates is lower for harder questions. Data-wise, augmentation approaches primarily rephrase questions to enhance diversity without systematically increasing intrinsic difficulty. To address these issues, we propose a two-dual MathForge framework to improve mathematical reasoning by targeting harder questions from both perspectives, which comprises a Difficulty-Aware Group Policy Optimization (DGPO) algorithm and a Multi-Aspect Question Reformulation (MQR) strategy. Specifically, DGPO first rectifies the implicit imbalance in GRPO via difficulty-balanced group advantage estimation, and further prioritizes harder questions by difficulty-aware question-level weighting. Meanwhile, MQR reformulates questions across multiple aspects to increase difficulty while maintaining the original gold answer. Overall, MathForge forms a synergistic loop: MQR expands the data frontier, and DGPO effectively learns from the augmented data. Extensive experiments show that MathForge significantly outperforms existing methods on various mathematical reasoning tasks. The code and augmented data are all available at https://github.com/AMAP-ML/MathForge.

Time series forecasting (TSF) faces challenges in modeling complex intra-channel temporal dependencies and inter-channel correlations. Although recent research has highlighted the efficiency of linear architectures in capturing global trends, these models often struggle with non-linear signals. To address this gap, we conducted a systematic receptive field analysis of convolutional neural network (CNN) TSF models. We introduce the "individual receptive field" to uncover granular structural dependencies, revealing that convolutional layers act as feature extractors that mirror channel-wise attention while exhibiting superior robustness to non-linear fluctuations. Based on these insights, we propose ACFormer, an architecture designed to reconcile the efficiency of linear projections with the non-linear feature-extraction power of convolutions. ACFormer captures fine-grained information through a shared compression module, preserves temporal locality via gated attention, and reconstructs variable-specific temporal patterns using an independent patch expansion layer. Extensive experiments on multiple benchmark datasets demonstrate that ACFormer consistently achieves state-of-the-art performance, effectively mitigating the inherent drawbacks of linear models in capturing high-frequency components.

Reconstructing dynamical evolution from limited observations is a fundamental challenge in single-cell biology, where dynamic unbalanced optimal transport provides a principled framework for modeling coupled transport and mass variation. However, existing approaches rely on trajectory simulation at inference time, making inference a key bottleneck for scalable applications. In this work, we propose a mean-flow framework for unbalanced flow matching that summarizes both transport and mass-growth dynamics over arbitrary time intervals using mean velocity and mass-growth fields, enabling fast one-step generation without trajectory simulation. To solve dynamic unbalanced optimal transport under the Wasserstein-Fisher-Rao geometry, we further build on this framework to develop Wasserstein-Fisher-Rao Mean Flow Matching (WFR-MFM). Across synthetic and real single-cell RNA sequencing datasets, WFR-MFM achieves orders-of-magnitude faster inference than a range of existing baselines while maintaining high predictive accuracy, and enables efficient perturbation response prediction on large synthetic datasets with thousands of conditions.

Uncrewed Aerial Vehicles (UAVs) are increasingly used in civilian and industrial applications, making secure low-altitude operations crucial. In dense mmWave environments, accurately classifying low-altitude UAVs as either inside authorized or restricted airspaces remains challenging, requiring models that handle complex propagation and signal variability. This paper proposes a deep learning model, referred to as CoBA, which stands for integrated Convolutional Neural Network (CNN), Bidirectional Long Short-Term Memory (BiLSTM), and Attention which leverages Fifth Generation (5G) millimeter-wave (mmWave) radio measurements to classify UAV operations in authorized and restricted airspaces at low altitude. The proposed CoBA model integrates convolutional, bidirectional recurrent, and attention layers to capture both spatial and temporal patterns in UAV radio measurements. To validate the model, a dedicated dataset is collected using the 5G mmWave network at TalTech, with controlled low altitude UAV flights in authorized and restricted scenarios. The model is evaluated against conventional ML models and a fingerprinting-based benchmark. Experimental results show that CoBA achieves superior accuracy, significantly outperforming all baseline models and demonstrating its potential for reliable and regulated UAV airspace monitoring.

This paper introduces a methodological framework for empirically testing AI alignment strategies through structured multi-model dialogue. Drawing on Peace Studies traditions - particularly interest-based negotiation, conflict transformation, and commons governance - we operationalize Viral Collaborative Wisdom (VCW), an approach that reframes alignment from a control problem to a relationship problem developed through dialogical reasoning. Our experimental design assigns four distinct roles (Proposer, Responder, Monitor, Translator) to different AI systems across six conditions, testing whether current large language models can engage substantively with complex alignment frameworks. Using Claude, Gemini, and GPT-4o, we conducted 72 dialogue turns totaling 576,822 characters of structured exchange. Results demonstrate that AI systems can engage meaningfully with Peace Studies concepts, surface complementary objections from different architectural perspectives, and generate emergent insights not present in initial framings - including the novel synthesis of "VCW as transitional framework." Cross-architecture patterns reveal that different models foreground different concerns: Claude emphasized verification challenges, Gemini focused on bias and scalability, and GPT-4o highlighted implementation barriers. The framework provides researchers with replicable methods for stress-testing alignment proposals before implementation, while the findings offer preliminary evidence about AI capacity for the kind of dialogical reasoning VCW proposes. We discuss limitations, including the observation that dialogues engaged more with process elements than with foundational claims about AI nature, and outline directions for future research including human-AI hybrid protocols and extended dialogue studies.

Person Re-Identification (ReID) remains a challenging problem in computer vision. This work reviews various training paradigm and evaluates the robustness of state-of-the-art ReID models in cross-domain applications and examines the role of foundation models in improving generalization through richer, more transferable visual representations. We compare three training paradigms, supervised, self-supervised, and language-aligned models. Through the study the aim is to answer the following questions: Can supervised models generalize in cross-domain scenarios? How does foundation models like SigLIP2 perform for the ReID tasks? What are the weaknesses of current supervised and foundational models for ReID? We have conducted the analysis across 11 models and 9 datasets. Our results show a clear split: supervised models dominate their training domain but crumble on cross-domain data. Language-aligned models, however, show surprising robustness cross-domain for ReID tasks, even though they are not explicitly trained to do so. Code and data available at: https://github.com/moiiai-tech/object-reid-benchmark.

We investigate structural traces of language contact in the intermediate representations of a monolingual language model. Focusing on Persian (Farsi) as a historically contact-rich language, we probe the representations of a Persian-trained model when exposed to languages with varying degrees and types of contact with Persian. Our methodology quantifies the amount of linguistic information encoded in intermediate representations and assesses how this information is distributed across model components for different morphosyntactic features. The results show that universal syntactic information is largely insensitive to historical contact, whereas morphological features such as Case and Gender are strongly shaped by language-specific structure, suggesting that contact effects in monolingual language models are selective and structurally constrained.

Ordinal regression and ranking are challenging due to inherent ordinal dependencies that conventional methods struggle to model. We propose Ranking-Aware Reinforcement Learning (RARL), a novel RL framework that explicitly learns these relationships. At its core, RARL features a unified objective that synergistically integrates regression and Learning-to-Rank (L2R), enabling mutual improvement between the two tasks. This is driven by a ranking-aware verifiable reward that jointly assesses regression precision and ranking accuracy, facilitating direct model updates via policy optimization. To further enhance training, we introduce Response Mutation Operations (RMO), which inject controlled noise to improve exploration and prevent stagnation at saddle points. The effectiveness of RARL is validated through extensive experiments on three distinct benchmarks.

Speech emotion recognition (SER) is crucial in speech understanding and generation. Most approaches are based on either classification models or large language models. Different from previous methods, we propose Gen-SER, a novel approach that reformulates SER as a distribution shift problem via generative models. We propose to project discrete class labels into a continuous space, and obtain the terminal distribution via sinusoidal taxonomy encoding. The target-matching-based generative model is adopted to transform the initial distribution into the terminal distribution efficiently. The classification is achieved by calculating the similarity of the generated terminal distribution and ground truth terminal distribution. The experimental results confirm the efficacy of the proposed method, demonstrating its extensibility to various speech-understanding tasks and suggesting its potential applicability to a broader range of classification tasks.

The practical deployment of diffusion-based Neural Video Compression (NVC) faces critical challenges, including severe information loss, prohibitive inference latency, and poor temporal consistency. To bridge this gap, we propose DiffVC-RT, the first framework designed to achieve real-time diffusion-based perceptual NVC. First, we introduce an Efficient and Informative Model Architecture. Through strategic module replacements and pruning, this architecture significantly reduces computational complexity while mitigating structural information loss. Second, to address generative flickering artifacts, we propose Explicit and Implicit Consistency Modeling. We enhance temporal consistency by explicitly incorporating a zero-cost Online Temporal Shift Module within the U-Net, complemented by hybrid implicit consistency constraints. Finally, we present an Asynchronous and Parallel Decoding Pipeline incorporating Mixed Half Precision, which enables asynchronous latent decoding and parallel frame reconstruction via a Batch-dimension Temporal Shift design. Experiments show that DiffVC-RT achieves 80.1% bitrate savings in terms of LPIPS over VTM-17.0 on HEVC dataset with real-time encoding and decoding speeds of 206 / 30 fps for 720p videos on an NVIDIA H800 GPU, marking a significant milestone in diffusion-based video compression.

Unsupervised ensemble learning emerged to address the challenge of combining multiple learners' predictions without access to ground truth labels or additional data. This paradigm is crucial in scenarios where evaluating individual classifier performance or understanding their strengths is challenging due to limited information. We propose a novel deep energy-based method for constructing an accurate meta-learner using only the predictions of individual learners, potentially capable of capturing complex dependence structures between them. Our approach requires no labeled data, learner features, or problem-specific information, and has theoretical guarantees for when learners are conditionally independent. We demonstrate superior performance across diverse ensemble scenarios, including challenging mixture of experts settings. Our experiments span standard ensemble datasets and curated datasets designed to test how the model fuses expertise from multiple sources. These results highlight the potential of unsupervised ensemble learning to harness collective intelligence, especially in data-scarce or privacy-sensitive environments.

Rich contact perception is crucial for robotic manipulation, yet traditional tactile skins remain expensive and complex to integrate. This paper presents a scalable alternative: high-accuracy whole-body touch localization via vibro-acoustic sensing. By equipping a robotic hand with seven low-cost piezoelectric microphones and leveraging an Audio Spectrogram Transformer, we decode the vibrational signatures generated during physical interaction. Extensive evaluation across stationary and dynamic tasks reveals a localization error of under 5 mm in static conditions. Furthermore, our analysis highlights the distinct influence of material properties: stiff materials (e.g., metal) excel in impulse response localization due to sharp, high-bandwidth responses, whereas textured materials (e.g., wood) provide superior friction-based features for trajectory tracking. The system demonstrates robustness to the robot's own motion, maintaining effective tracking even during active operation. Our primary contribution is demonstrating that complex physical contact dynamics can be effectively decoded from simple vibrational signals, offering a viable pathway to widespread, affordable contact perception in robotics. To accelerate research, we provide our full datasets, models, and experimental setups as open-source resources.

We study risk-sensitive planning under partial observability using the dynamic risk measure Iterated Conditional Value-at-Risk (ICVaR). A policy evaluation algorithm for ICVaR is developed with finite-time performance guarantees that do not depend on the cardinality of the action space. Building on this foundation, three widely used online planning algorithms--Sparse Sampling, Particle Filter Trees with Double Progressive Widening (PFT-DPW), and Partially Observable Monte Carlo Planning with Observation Widening (POMCPOW)--are extended to optimize the ICVaR value function rather than the expectation of the return. Our formulations introduce a risk parameter $α$, where $α= 1$ recovers standard expectation-based planning and $α< 1$ induces increasing risk aversion. For ICVaR Sparse Sampling, we establish finite-time performance guarantees under the risk-sensitive objective, which further enable a novel exploration strategy tailored to ICVaR. Experiments on benchmark POMDP domains demonstrate that the proposed ICVaR planners achieve lower tail risk compared to their risk-neutral counterparts.

We present DeepSeek-OCR 2 to investigate the feasibility of a novel encoder-DeepEncoder V2-capable of dynamically reordering visual tokens upon image semantics. Conventional vision-language models (VLMs) invariably process visual tokens in a rigid raster-scan order (top-left to bottom-right) with fixed positional encoding when fed into LLMs. However, this contradicts human visual perception, which follows flexible yet semantically coherent scanning patterns driven by inherent logical structures. Particularly for images with complex layouts, human vision exhibits causally-informed sequential processing. Inspired by this cognitive mechanism, DeepEncoder V2 is designed to endow the encoder with causal reasoning capabilities, enabling it to intelligently reorder visual tokens prior to LLM-based content interpretation. This work explores a novel paradigm: whether 2D image understanding can be effectively achieved through two-cascaded 1D causal reasoning structures, thereby offering a new architectural approach with the potential to achieve genuine 2D reasoning. Codes and model weights are publicly accessible at http://github.com/deepseek-ai/DeepSeek-OCR-2.

Large language models (LLMs) are increasingly used in verbal creative tasks. However, previous assessments of the creative capabilities of LLMs remain weakly grounded in human creativity theory and are thus hard to interpret. The widely used Divergent Association Task (DAT) focuses on novelty, ignoring appropriateness, a core component of creativity. We evaluate a range of state-of-the-art LLMs on DAT and show that their scores on the task are lower than those of two baselines that do not possess any creative abilities, undermining its validity for model evaluation. Grounded in human creativity theory, which defines creativity as the combination of novelty and appropriateness, we introduce Conditional Divergent Association Task (CDAT). CDAT evaluates novelty conditional on contextual appropriateness, separating noise from creativity better than DAT, while remaining simple and objective. Under CDAT, smaller model families often show the most creativity, whereas advanced families favor appropriateness at lower novelty. We hypothesize that training and alignment likely shift models along this frontier, making outputs more appropriate but less creative. We release the dataset and code.