As multimodal LLM-driven agents advance in autonomy and generalization, traditional static datasets face inherent scalability limitations and are insufficient for fully assessing their capabilities in increasingly complex and diverse tasks. Existing studies have attempted to generate agent tasks using LLMs, but due to the inherent hallucinations of LLMs and the lack of internal data relationship modeling, these tasks often exhibit semantic inconsistencies and solvability issues. To address these challenges, we introduce Graph2Eval, a knowledge-graph-driven framework for automated, scalable, and semantically grounded agent task generation. At its core, Graph2Eval leverages a knowledge graph built from heterogeneous external data sources as a structured task space, generating multimodal agent tasks through subgraph sampling and task construction guided by task templates and meta-path strategies. To further ensure task reliability, a multi-stage filtering pipeline based on node reachability analysis, LLM scoring, and similarity analysis ensures the diversity and solvability of the generated tasks. By unifying both RAG Agent and Web Agent scenarios, Graph2Eval enables efficient generation of multimodal document understanding tasks and multi-step web interaction tasks. We instantiate the framework with Graph2Eval-Bench, a curated dataset of 1,319 tasks spanning document understanding and web interaction scenarios. Extensive experiments show that, on average, Graph2Eval improves task semantic consistency by 20% and solvability by 17% over baselines, while Graph2Eval-Bench effectively distinguishes agent performance, offering a new perspective on agent evaluation.

Reliable trajectory prediction from an ego-centric perspective is crucial for robotic navigation in human-centric environments. However, existing methods typically assume noiseless observation histories, failing to account for the perceptual artifacts inherent in first-person vision, such as occlusions, ID switches, and tracking drift. This discrepancy between training assumptions and deployment reality severely limits model robustness. To bridge this gap, we introduce EgoTraj-Bench, built upon TBD dataset, which is the first real-world benchmark that aligns noisy, first-person visual histories with clean, bird's-eye-view future trajectories, enabling robust learning under realistic perceptual constraints. Building on this benchmark, we propose BiFlow, a dual-stream flow matching model that concurrently denoises historical observations and forecasts future motion. To better model agent intent, BiFlow incorporates our EgoAnchor mechanism, which conditions the prediction decoder on distilled historical features via feature modulation. Extensive experiments show that BiFlow achieves state-of-the-art performance, reducing minADE and minFDE by 10-15% on average and demonstrating superior robustness. We anticipate that our benchmark and model will provide a critical foundation for robust real-world ego-centric trajectory prediction. The benchmark library is available at: https://github.com/zoeyliu1999/EgoTraj-Bench.

Current large language model (LLM) development treats task-solving and preference-alignment as separate challenges, optimizing first for objective correctness, then for alignment to aggregated human preferences. This paradigm fails in human-facing applications where solving a problem correctly is insufficient if the response mismatches the user's needs. This challenge intensifies in just-in-time scenarios where no prior user interaction history exists due to cold-start conditions or privacy constraints. LLMs need to proactively identify what they don't know about the user, strategically elicit preference values through questioning, then adapt their reasoning processes and responses accordingly -- a complicated chain of cognitive processes which we term personalized reasoning. We introduce PrefDisco, an evaluation methodology that transforms static benchmarks into interactive personalization tasks using psychologically-grounded personas with sparse, context-dependent preferences, and define PrefAlign as a fine-grained rubric-based metric for measuring preference alignment. PrefDisco builds scenarios where identical questions require different reasoning chains depending on user context, as optimal explanation approaches vary by individual expertise and preferences while maintaining factual accuracy. Evaluation of 21 frontier models across 10 tasks reveals 29.0% of naive personalization attempts produce worse preference alignment than generic responses, yet generic responses also fail to serve individual user needs. These findings suggest personalized reasoning requires dedicated development rather than emerging naturally. PrefDisco provides a foundation for developing systems that can adapt to individual users in education, healthcare, and technical domains where personalization is critical.

We introduce a novel formulation for continuous space-time video super-resolution. Instead of decoupling the representation of a video sequence into separate spatial and temporal components and relying on brittle, explicit frame warping for motion compensation, we encode video as a continuous, spatio-temporally coherent 3D Video Fourier Field (VFF). That representation offers three key advantages: (1) it enables cheap, flexible sampling at arbitrary locations in space and time; (2) it is able to simultaneously capture fine spatial detail and smooth temporal dynamics; and (3) it offers the possibility to include an analytical, Gaussian point spread function in the sampling to ensure aliasing-free reconstruction at arbitrary scale. The coefficients of the proposed, Fourier-like sinusoidal basis are predicted with a neural encoder with a large spatio-temporal receptive field, conditioned on the low-resolution input video. Through extensive experiments, we show that our joint modeling substantially improves both spatial and temporal super-resolution and sets a new state of the art for multiple benchmarks: across a wide range of upscaling factors, it delivers sharper and temporally more consistent reconstructions than existing baselines, while being computationally more efficient. Project page: https://v3vsr.github.io.

Large Language Models (LLMs) today are powerful problem solvers across many domains, and they continue to get stronger as they scale in model size, training set size, and training set quality, as shown by extensive research and experimentation across the industry. Training a frontier model today requires on the order of tens to hundreds of yottaflops, which is a massive investment of time, compute, and energy. Improving pretraining efficiency is therefore essential to enable the next generation of even more capable LLMs. While 8-bit floating point (FP8) training is now widely adopted, transitioning to even narrower precision, such as 4-bit floating point (FP4), could unlock additional improvements in computational speed and resource utilization. However, quantization at this level poses challenges to training stability, convergence, and implementation, notably for large-scale models trained on long token horizons. In this study, we introduce a novel approach for stable and accurate training of large language models (LLMs) using the NVFP4 format. Our method integrates Random Hadamard transforms (RHT) to bound block-level outliers, employs a two-dimensional quantization scheme for consistent representations across both the forward and backward passes, utilizes stochastic rounding for unbiased gradient estimation, and incorporates selective high-precision layers. We validate our approach by training a 12-billion-parameter model on 10 trillion tokens -- the longest publicly documented training run in 4-bit precision to date. Our results show that the model trained with our NVFP4-based pretraining technique achieves training loss and downstream task accuracies comparable to an FP8 baseline. These findings highlight that NVFP4, when combined with our training approach, represents a major step forward in narrow-precision LLM training algorithms.

Autoregressive (AR) models are promising for image generation, yet continuous-token AR variants often trail latent diffusion and masked-generation models. The core issue is heterogeneous variance in VAE latents, which is amplified during AR decoding, especially under classifier-free guidance (CFG), and can cause variance collapse. We propose SphereAR to address this issue. Its core design is to constrain all AR inputs and outputs -- including after CFG -- to lie on a fixed-radius hypersphere (constant $\ell_2$ norm), leveraging hyperspherical VAEs. Our theoretical analysis shows that hyperspherical constraint removes the scale component (the primary cause of variance collapse), thereby stabilizing AR decoding. Empirically, on ImageNet generation, SphereAR-H (943M) sets a new state of the art for AR models, achieving FID 1.34. Even at smaller scales, SphereAR-L (479M) reaches FID 1.54 and SphereAR-B (208M) reaches 1.92, matching or surpassing much larger baselines such as MAR-H (943M, 1.55) and VAR-d30 (2B, 1.92). To our knowledge, this is the first time a pure next-token AR image generator with raster order surpasses diffusion and masked-generation models at comparable parameter scales.

Evaluating language models fairly is increasingly difficult as static benchmarks risk contamination by training data, obscuring whether models truly reason or recall. We introduce BeyondBench, an evaluation framework using algorithmic problem generation to create mathematically grounded problems on the fly, ensuring each test remains uncontaminated. Our framework covers 44 algorithmic tasks with 117 variations across three difficulty levels: the Easy Suite (29 tasks) for arithmetic and statistics, the Medium Suite (5 tasks, 49 variations) for sequence patterns and reasoning, and the Hard Suite (10 tasks, 68 variations) for NP-complete and constraint satisfaction problems. Each task draws from a space exceeding 10^15 unique instances, with deterministically verified solutions. We evaluated 101 language models (85 open-source, 16 closed-source), spanning 0.5B to 141B parameters and multiple quantization schemes, using three-fold evaluation for robustness. Results reveal consistent reasoning deficiencies, with performance degrading sharply as complexity increases. In Hard Suite evaluations, Gemini-2.5-pro, Llama-3.3-70B, and Qwen2.5-72B achieved accuracies of 56.21%, 27.16%, and 33.37% respectively. Performance drops significantly without tool usage, with GPT-5, GPT-5-mini, and GPT-5-nano showing declines of 16.81%, 15.86%, and 43.95% in overall accuracy. Contamination resistance rests on three guarantees: (i) the problem space vastly exceeds any static dataset, (ii) every instance has a deterministically verifiable solution, and (iii) isomorphic transformations yield semantically equivalent but syntactically novel problems. BeyondBench redefines reasoning evaluation via genuine algorithmic problem-solving. Our leaderboard is at https://ctrl-gaurav.github.io/BeyondBench/, Python package at https://pypi.org/project/beyondbench/, and codebase at https://github.com/ctrl-gaurav/BeyondBench.

Diffusion-based planners have shown strong potential for autonomous driving by capturing multi-modal driving behaviors. A key challenge is how to effectively guide these models for safe and reactive planning in closed-loop settings, where the ego vehicle's actions influence future states. Recent work leverages typical expert driving behaviors (i.e., anchors) to guide diffusion planners but relies on a truncated diffusion schedule that introduces an asymmetry between the forward and denoising processes, diverging from the core principles of diffusion models. To address this, we introduce BridgeDrive, a novel anchor-guided diffusion bridge policy for closed-loop trajectory planning. Our approach formulates planning as a diffusion bridge that directly transforms coarse anchor trajectories into refined, context-aware plans, ensuring theoretical consistency between the forward and reverse processes. BridgeDrive is compatible with efficient ODE solvers, enabling real-time deployment. We achieve state-of-the-art performance on the Bench2Drive closed-loop evaluation benchmark, improving the success rate by 7.72% and 2.45% over prior arts with PDM-Lite and LEAD datasets, respectively. Project page: https://github.com/shuliu-ethz/BridgeDrive.

Reinforcement learning (RL) and sim-to-real transfer have advanced rigid-object manipulation. However, policies remain brittle for articulated mechanisms due to contact-rich dynamics that require both stable grasping and simultaneous free in-hand articulation. Furthermore, articulated objects and robot hands exhibit under-modeled joint phenomena such as friction, stiction, and backlash in real life that can increase the sim-to-real gap, and robot hands still fall short of idealized tactile sensing, both in terms of coverage, sensitivity, and specificity. In this paper, we present an original approach to learning dexterous in-hand manipulation of articulated tools that has reduced articulation and kinematic redundancy relative to the human hand. Our approach augments a simulation-trained base policy with a sensor-driven refinement learned from hardware demonstrations. This refinement conditions on proprioception and target articulation states while fusing whole-hand tactile and force-torque feedback with the policy's action intent through cross-attention. The resulting controller adapts online to instance-specific articulation properties, stabilizes contact interactions, and regulates internal forces under perturbations. We validate our method across diverse real-world tools, including scissors, pliers, minimally invasive surgical instruments, and staplers, demonstrating robust sim-to-real transfer, improved disturbance resilience, and generalization across structurally related articulated tools without precise physical modeling.

Automatic drum transcription (ADT) is traditionally formulated as a discriminative task to predict drum events from audio spectrograms. In this work, we redefine ADT as a conditional generative task and introduce Noise-to-Notes (N2N), a framework leveraging diffusion modeling to transform audio-conditioned Gaussian noise into drum events with associated velocities. This generative diffusion approach offers distinct advantages, including a flexible speed-accuracy trade-off and strong inpainting capabilities. However, the generation of binary onset and continuous velocity values presents a challenge for diffusion models, and to overcome this, we introduce an Annealed Pseudo-Huber loss to facilitate effective joint optimization. Finally, to augment low-level spectrogram features, we propose incorporating features extracted from music foundation models (MFMs), which capture high-level semantic information and enhance robustness to out-of-domain drum audio. Experimental results demonstrate that including MFM features significantly improves robustness and N2N establishes a new state-of-the-art performance across multiple ADT benchmarks.

Policy gradient methods usually rely on entropy regularization to prevent premature convergence. However, maximizing entropy indiscriminately pushes the policy towards a uniform distribution, often overriding the reward signal if not optimally tuned. We propose replacing the standard entropy term with a self-regulating complexity term, defined as the product of Shannon entropy and disequilibrium, where the latter quantifies the distance from the uniform distribution. Unlike pure entropy, which favors maximal disorder, this complexity measure is zero for both fully deterministic and perfectly uniform distributions, i.e., it is strictly positive for systems that exhibit a meaningful interplay between order and randomness. These properties ensure the policy maintains beneficial stochasticity while reducing regularization pressure when the policy is highly uncertain, allowing learning to focus on reward optimization. We introduce Complexity-Regularized Proximal Policy Optimization (CR-PPO), a modification of PPO that leverages this dynamic. We empirically demonstrate that CR-PPO is significantly more robust to hyperparameter selection than entropy-regularized PPO, achieving consistent performance across orders of magnitude of regularization coefficients and remaining harmless when regularization is unnecessary, thereby reducing the need for expensive hyperparameter tuning.

LLMs serve as the backbone in SpeechLLMs, yet their behavior on spontaneous conversational input remains poorly understood. Conversational speech contains pervasive disfluencies -- interjections, edits, and parentheticals -- that are rare in the written corpora used for pre-training. Because gold disfluency removal is a deletion-only task, it serves as a controlled probe to determine whether a model performs faithful structural repair or biased reinterpretation. Using the DRES evaluation framework, we evaluate proprietary and open-source LLMs across architectures and scales. We show that model performance clusters into stable precision-recall regimes reflecting distinct editing policies. Notably, reasoning models systematically over-delete fluent content, revealing a bias toward semantic abstraction over structural fidelity. While fine-tuning achieves SOTA results, it harms generalization. Our findings demonstrate that robustness to speech is shaped by specific training objectives.

Autonomous exploration in structured and complex indoor environments remains a challenging task, as existing methods often struggle to appropriately model unobserved space and plan globally efficient paths. To address these limitations, we propose GUIDE, a novel exploration framework that synergistically combines global graph inference with diffusion-based decision-making. We introduce a region-evaluation global graph representation that integrates both observed environmental data and predictions of unexplored areas, enhanced by a region-level evaluation mechanism to prioritize reliable structural inferences while discounting uncertain predictions. Building upon this enriched representation, a diffusion policy network generates stable, foresighted action sequences with significantly reduced denoising steps. Extensive simulations and real-world deployments demonstrate that GUIDE consistently outperforms state-of-the-art methods, achieving up to 18.3% faster coverage completion and a 34.9% reduction in redundant movements.

Learning-based methods excel at robot motion generation but remain limited in contact-rich physical interaction. Impedance control provides stable and safe contact behavior but requires task-specific tuning of stiffness and damping parameters. We present Diffusion-Based Impedance Learning, a framework that bridges these paradigms by combining generative modeling with energy-consistent impedance control. A Transformer-based Diffusion Model, conditioned via cross-attention on measured external wrenches, reconstructs simulated Zero-Force Trajectories (sZFTs) that represent contact-consistent equilibrium behavior. A SLERP-based quaternion noise scheduler preserves geometric consistency for rotations on the unit sphere. The reconstructed sZFT is used by an energy-based estimator to adapt impedance online through directional stiffness and damping modulation. Trained on parkour and robot-assisted therapy demonstrations collected via Apple Vision Pro teleoperation, the model achieves sub-millimeter positional and sub-degree rotational accuracy using only tens of thousands of samples. Deployed in realtime torque control on a KUKA LBR iiwa, the approach enables smooth obstacle traversal and generalizes to unseen tasks, achieving 100% success in multi-geometry peg-in-hole insertion.

Speech Language Models (SpeechLMs) model tokenized speech to capture both semantic and acoustic information. When neural audio codecs based on Residual Vector Quantization (RVQ) are used as audio tokenizers, they produce multiple discrete tokens per time step, yielding inherently multi-level representations. To process these multi-level tokens together, prior work typically adopts hierarchical architectures to capture this structure. In contrast, recent progress in NLP has progressively reduced architectural inductive biases, moving toward simpler and more scalable single-Transformer architectures. In this work, we propose Llama-Mimi, which flattens multi-level RVQ tokens produced by the Mimi neural audio codec into a single sequence and models them autoregressively with a Transformer decoder. We show that Llama-Mimi outperforms a CSM-based hierarchical model on most tasks and achieves the best performance on acoustic consistency. Our models, code, and speech samples are publicly available.

In Social Robot Navigation (SRN), the availability of meaningful metrics is crucial for evaluating trajectories from human-robot interactions. In the SRN context, such interactions often relate to resolving conflicts between two or more agents. Correspondingly, the shares to which agents contribute to the resolution of such conflicts are important. This paper builds on recent work, which proposed a Responsibility metric capturing such shares. We extend this framework in two directions: First, we model the conflict buildup phase by introducing a time normalization. Second, we propose the related Engagement metric, which captures how the agents' actions intensify a conflict. In a comprehensive series of simulated scenarios with dyadic, group and crowd interactions, we show that the metrics carry meaningful information about the cooperative resolution of conflicts in interactions. They can be used to assess behavior quality and foresightedness. We extensively discuss applicability, design choices and limitations of the proposed metrics.

Evaluating tabular generators remains a challenging problem, as the unique causal structural prior of heterogeneous tabular data does not lend itself to intuitive human inspection. Recent work has introduced structural fidelity as a tabular-specific evaluation dimension to assess whether synthetic data complies with the causal structures of real data. However, existing benchmarks often neglect the interplay between structural fidelity and conventional evaluation dimensions, thus failing to provide a holistic understanding of model performance. Moreover, they are typically limited to toy datasets, as quantifying existing structural fidelity metrics requires access to ground-truth causal structures, which are rarely available for real-world datasets. In this paper, we propose a novel evaluation framework that jointly considers structural fidelity and conventional evaluation dimensions. We introduce a new evaluation metric, $\textbf{global utility}$, which enables the assessment of structural fidelity even in the absence of ground-truth causal structures. In addition, we present $\textbf{TabStruct}$, a comprehensive evaluation benchmark offering large-scale quantitative analysis on 13 tabular generators from nine distinct categories, across 29 datasets. Our results demonstrate that global utility provides a task-independent, domain-agnostic lens for tabular generator performance. We release the TabStruct benchmark suite, including all datasets, evaluation pipelines, and raw results. Code is available at https://github.com/SilenceX12138/TabStruct.

Forecasting product demand in retail supply chains presents a complex challenge due to noisy, heterogeneous features and rapidly shifting consumer behavior. While traditional gradient boosting decision trees (GBDT) offer strong predictive performance on structured data, they often lack adaptive mechanisms to identify and emphasize the most relevant features under changing conditions. In this work, we propose AttnBoost, an interpretable learning framework that integrates feature-level attention into the boosting process to enhance both predictive accuracy and explainability. Specifically, the model dynamically adjusts feature importance during each boosting round via a lightweight attention mechanism, allowing it to focus on high-impact variables such as promotions, pricing, and seasonal trends. We evaluate AttnBoost on a large-scale retail sales dataset and demonstrate that it outperforms standard machine learning and deep tabular models, while also providing actionable insights for supply chain managers. An ablation study confirms the utility of the attention module in mitigating overfitting and improving interpretability. Our results suggest that attention-guided boosting represents a promising direction for interpretable and scalable AI in real-world forecasting applications.

Language use offers valuable insight into affective disorders such as bipolar disorder (BD), yet past research has been cross-sectional and limited in scale. Here, we demonstrate that social media records can be leveraged to study longitudinal language change associated with BD on a large scale. Using a novel method to infer diagnosis timelines from user self-reports, we compared users self-identifying with BD, depression, or no mental health condition. The onset of BD diagnosis corresponded with widespread linguistic shifts reflecting mood disturbance, psychiatric comorbidity, substance abuse, hospitalization, medical comorbidities, interpersonal concerns, unusual thought content, and altered linguistic coherence. In the years following the diagnosis, discussions of mood symptoms were found to fluctuate periodically with a dominant 12-month cycle consistent with seasonal mood variation. These findings suggest that social media language captures linguistic and behavioral changes associated with BD and might serve as a valuable complement to traditional psychiatric cohort research.

Code-switching (CS) presents a significant challenge for general Auto-Speech Recognition (ASR) systems. Existing methods often fail to capture the sub tle phonological shifts inherent in CS scenarios. The challenge is particu larly difficult for language pairs like Vietnamese and English, where both distinct phonological features and the ambiguity arising from similar sound recognition are present. In this paper, we propose a novel architecture for Vietnamese-English CS ASR, a Two-Stage Phoneme-Centric model (TSPC). TSPC adopts a phoneme-centric approach based on an extended Vietnamese phoneme set as an intermediate representation for mixed-lingual modeling, while remaining efficient under low computational-resource constraints. Ex perimental results demonstrate that TSPC consistently outperforms exist ing baselines, including PhoWhisper-base, in Vietnamese-English CS ASR, achieving a significantly lower word error rate of 19.06% with reduced train ing resources. Furthermore, the phonetic-based two-stage architecture en ables phoneme adaptation and language conversion to enhance ASR perfor mance in complex CS Vietnamese-English ASR scenarios.

Traversing narrow gates presents a significant challenge and has become a standard benchmark for evaluating agile and precise quadrotor flight. Traditional modularized autonomous flight stacks require extensive design and parameter tuning, while end-to-end reinforcement learning (RL) methods often suffer from low sample efficiency, limited interpretability, and degraded disturbance rejection under unseen perturbations. In this work, we present a novel hybrid framework that adaptively fine-tunes model predictive control (MPC) parameters online using outputs from a neural network (NN) trained offline. The NN jointly predicts a reference pose and cost function weights, conditioned on the coordinates of the gate corners and the current drone state. To achieve efficient training, we derive analytical policy gradients not only for the MPC module but also for an optimization-based gate traversal detection module. Hardware experiments demonstrate agile and accurate gate traversal with peak accelerations of $30\ \mathrm{m/s^2}$, as well as recovery within $0.85\ \mathrm{s}$ following body-rate disturbances exceeding $1146\ \mathrm{deg/s}$.

The growing complexity of urban mobility and the demand for efficient, sustainable, and adaptive solutions have positioned Intelligent Transportation Systems (ITS) at the forefront of modern infrastructure innovation. At the core of ITS lies the challenge of autonomous decision-making across dynamic, large scale, and uncertain environments where multiple agents traffic signals, autonomous vehicles, or fleet units must coordinate effectively. Multi Agent Reinforcement Learning (MARL) offers a promising paradigm for addressing these challenges by enabling distributed agents to jointly learn optimal strategies that balance individual objectives with system wide efficiency. This paper presents a comprehensive survey of MARL applications in ITS. We introduce a structured taxonomy that categorizes MARL approaches according to coordination models and learning algorithms, spanning value based, policy based, actor critic, and communication enhanced frameworks. Applications are reviewed across key ITS domains, including traffic signal control, connected and autonomous vehicle coordination, logistics optimization, and mobility on demand systems. Furthermore, we highlight widely used simulation platforms such as SUMO, CARLA, and CityFlow that support MARL experimentation, along with emerging benchmarks. The survey also identifies core challenges, including scalability, non stationarity, credit assignment, communication constraints, and the sim to real transfer gap, which continue to hinder real world deployment.

This work presents a comprehensive evaluation of how quantization affects model bias, with particular attention to its impact on individual demographic subgroups. We focus on weight and activation quantization strategies and examine their effects across a broad range of bias types, including stereotypes, fairness, toxicity, and sentiment. We employ both probability- and generated text-based metrics across 13 benchmarks and evaluate models that differ in architecture family and reasoning ability. Our findings show that quantization has a nuanced impact on bias: while it can reduce model toxicity and does not significantly impact sentiment, it tends to slightly increase stereotypes and unfairness in generative tasks, especially under aggressive compression. These trends are generally consistent across demographic categories and subgroups, and model types, although their magnitude depends on the specific setting. Overall, our results highlight the importance of carefully balancing efficiency and ethical considerations when applying quantization in practice.

This paper tackles the critical challenge of optimizing multi-modality trackers by effectively adapting pre-trained models for RGB data. Existing fine-tuning paradigms oscillate between excessive flexibility and over-restriction, both leading to suboptimal plasticity-stability trade-offs. To mitigate this dilemma, we propose a novel significance-regularized fine-tuning framework, which delicately refines the learning process by incorporating intrinsic parameter significance. Through a comprehensive investigation of the transition from pre-trained to multi-modality contexts, we identify that parameters crucial to preserving foundational patterns and managing cross-domain shifts are the primary drivers of this issue. Specifically, we first probe the tangent space of pre-trained weights to measure and orient prior significance, dedicated to preserving generalization. Subsequently, we characterize transfer significance during the fine-tuning phase, emphasizing adaptability and stability. By incorporating these parameter significance terms as unified regularization, our method markedly enhances transferability across modalities. Extensive experiments showcase the superior performance of our method, surpassing current state-of-the-art techniques across various multi-modal tracking benchmarks. The source code and models are publicly available at https://github.com/zhiwen-xdu/SRTrack.

Controllable human voice generation, particularly for expressive domains like singing, remains a significant challenge. This paper introduces Vevo2, a unified framework for controllable speech and singing voice generation. To tackle issues like the scarcity of annotated singing data and to enable flexible controllability, Vevo2 introduces two audio tokenizers: (1) a unified music-notation-free prosody tokenizer that captures prosody and melody from speech, singing, and even instrumental sounds, and (2) a unified content-style tokenizer that encodes linguistic content, prosody, and style for both speech and singing, while enabling timbre disentanglement. Vevo2 consists of an auto-regressive (AR) content-style modeling stage, which aims to enable controllability over text, prosody, and style, as well as a flow-matching acoustic modeling stage that allows for timbre control. Particularly, during the speech-singing joint training of the AR model, we propose both explicit and implicit prosody learning strategies to bridge speech and singing voice. Moreover, to further enhance the Vevo2's ability to follow text and prosody, we design a multi-objective post-training task that integrates both intelligibility and prosody similarity alignment. Experimental results show that the unified modeling in Vevo2 brings mutual benefits to both speech and singing voice generation. Additionally, Vevo2's effectiveness across a wide range of synthesis, conversion, and editing tasks for both speech and singing further demonstrates its strong generalization ability and versatility. Audio samples are are available at https://versasinger.github.io/.

Reliable evaluation of machine learning models for neonatal seizure detection is critical for clinical adoption. Current practices often rely on inconsistent and biased metrics, hindering model comparability and interpretability. Expert-level claims about AI performance are frequently made without rigorous validation, raising concerns about their reliability. This study aims to systematically evaluate common performance metrics and propose best practices tailored to the specific challenges of neonatal seizure detection. Using real and synthetic seizure annotations, we assessed standard performance metrics, consensus strategies, and human-expert level equivalence tests under varying class imbalance, inter-rater agreement, and number of raters. Matthews and Pearson's correlation coefficients outperformed the area under the curve in reflecting performance under class imbalance. Consensus types are sensitive to the number of raters and agreement level among them. Among human-expert level equivalence tests, the multi-rater Turing test using Fleiss k best captured expert-level AI performance. We recommend reporting: (1) at least one balanced metric, (2) Sensitivity, specificity, PPV and NPV, (3) Multi-rater Turing test results using Fleiss k, and (4) All the above on held-out validation set. This proposed framework provides an important prerequisite to clinical validation by enabling a thorough and honest appraisal of AI methods for neonatal seizure detection.

Foundation models have shown strong performance in multi-object segmentation with visual prompts, yet histopathology images remain challenging due to high cellular density, heterogeneity, and the gap between pixel-level supervision and clinical segmentation intent (e.g., selectively segmenting nuclei of a specific type). In practice, such intents are expressed through diverse and noisy prompts, causing prompt-intent misalignment and inconsistent predictions. We introduce SAMPO (Segmentation Anything Model with Preference Optimization), a preference-aligned fine-tuning framework that explicitly aligns pathology foundation models with clinical segmentation intent. SAMPO is the first to adapt Direct Preference Optimization (DPO) to pure vision foundation models, enabling accurate segmentation from minimal and imperfect prompts. The framework features three key components: (1) online prompt-centric preference mining to synthesize preference pairs across prompt qualities; (2) multi-mask preference learning to leverage output ambiguity for fine-grained ranking supervision; and (3) a hybrid loss combining preference optimization with pixel-level supervision for stable training. Trained on two datasets covering four tasks and evaluated on corresponding test sets and 12 external validation datasets, SAMPO consistently improves segmentation accuracy, robustness to prompt variations, and clinical intent adherence in dense histopathology images.

Although EDM aims to unify the design space of diffusion models, its reliance on fixed Gaussian noise prevents it from explaining emerging flow-based methods that diffuse arbitrary noise. Moreover, our study reveals that EDM's forcible injection of Gaussian noise has adverse effects on image restoration task, as it corrupts the degraded images, overextends the restoration distance, and increases the task's complexity. To interpret diverse methods for handling distinct noise patterns within a unified theoretical framework and to minimize the restoration distance, we propose EDA, which Elucidates the Design space of Arbitrary-noise diffusion models. Theoretically, EDA expands noise pattern flexibility while preserving EDM's modularity, with rigorous proof that increased noise complexity introduces no additional computational overhead during restoration. EDA is validated on three representative medical image denoising and natural image restoration tasks: MRI bias field correction (global smooth noise), CT metal artifact removal (global sharp noise) and natural image shadow removal (local boundary-aware noise). With only 5 sampling steps, competitive results against specialized methods across medical and natural tasks demonstrate EDA's strong generalization capability for image restoration. Code is available at: https://github.com/PerceptionComputingLab/EDA.

We consider the maximum causal entropy inverse reinforcement learning (IRL) problem for infinite-horizon stationary mean-field games (MFG), in which we model the unknown reward function within a reproducing kernel Hilbert space (RKHS). This allows the inference of rich and potentially nonlinear reward structures directly from expert demonstrations, in contrast to most existing approaches for MFGs that typically restrict the reward to a linear combination of a fixed finite set of basis functions and rely on finite-horizon formulations. We introduce a Lagrangian relaxation that enables us to reformulate the problem as an unconstrained log-likelihood maximization and obtain a solution via a gradient ascent algorithm. To establish the theoretical consistency of the algorithm, we prove the smoothness of the log-likelihood objective through the Fréchet differentiability of the related soft Bellman operators with respect to the parameters in the RKHS. To illustrate the practical advantages of the RKHS formulation, we validate our framework on a mean-field traffic routing game exhibiting state-dependent preference reversal, where the kernel-based method reduces policy recovery error by over an order of magnitude compared to a linear reward baseline with a comparable parameter count. Furthermore, we extend the framework to the finite-horizon non-stationary setting. We demonstrate that the log-likelihood reformulation is structurally unavailable in this regime and instead develop an alternative gradient descent algorithm on the convex dual via Danskin's theorem, establishing smoothness and convergence guarantees.

This paper examines the $L_p$ and $W^1_p$ norm approximation errors of ReLU neural networks for Korobov functions. In terms of network width and depth, we derive nearly optimal super-approximation error bounds of order $2m$ in the $L_p$ norm and order $2m-2$ in the $W^1_p$ norm, for target functions with $L_p$ mixed derivative of order $m$ in each direction. The analysis leverages sparse grid finite elements and the bit extraction technique. Our results improve upon classical lowest order $L_\infty$ and $H^1$ norm error bounds and demonstrate that the expressivity of neural networks is largely unaffected by the curse of dimensionality.