Driver gaze is commonly modeled as a spatial heatmap, but heatmaps alone are difficult for humans to interpret because they do not explain which road object or region is being monitored or why an attention shift may matter. This study examines whether minimal human-grounded supervision can steer a vision--language model toward interpretable descriptions of driver attention shifts. Using selected high-change gaze moments from the Berkeley DeepDrive-Attention dataset, we compare zero-shot, one-shot, and LoRA fine-tuned VLM conditions against human-refined reference descriptions and expert ratings. Results show that fine-tuning with 80 expert-refined attention examples improves ROUGE-L, METEOR, Entity Alignment F1, and Human Alignment Score relative to unsteered VLM outputs. The findings suggest that language-based descriptions can complement gaze heatmaps by making driver attention more accessible for human-factors analysis, driver-monitoring review, and situation-awareness support.

Retrieval-Augmented Generation (RAG) enhances the factual accuracy of large language models (LLMs) by conditioning outputs on external knowledge sources. However, when retrieval involves private or sensitive data, RAG systems are susceptible to extraction attacks that can leak confidential information through generated responses. We propose Privacy-Aware Decoding (PAD), a lightweight, inference-time defense that adaptively injects calibrated Gaussian noise into token logits during generation. PAD integrates confidence-based screening to selectively protect high-risk tokens, efficient sensitivity estimation to minimize unnecessary noise, and context-aware noise calibration to balance privacy with generation quality. A \renyi Differential Privacy (RDP) accountant rigorously tracks cumulative privacy loss, enabling explicit per-response $(\varepsilon, δ)$-DP guarantees for sensitive outputs. Unlike prior approaches requiring retraining or corpus-level filtering, PAD is model-agnostic and operates entirely at decoding time with minimal computational overhead. Experiments on three real-world datasets demonstrate that PAD substantially reduces private information leakage while preserving response utility, outperforming existing retrieval- and post-processing-based defenses. Our work takes an important step toward mitigating privacy risks in RAG via decoding strategies, paving the way for universal and scalable privacy solutions in sensitive domains. Our code is available: https://github.com/wang2226/PAD.

Socially interactive humanoid robots must engage with humans through their bodies, adapting in real time to a partner's movement, intent, and abilities. This requires models that understand not just how bodies move, but what movement means in a shared social context. Yet evaluation frameworks for interactive motion generation do not measure whether generated follower motion is legible within a shared movement vocabulary, nor whether it is appropriate to the partner's proficiency level. This gap has two causes: existing frameworks rely on kinematic metrics such as FID and beat alignment that cannot measure either property, and existing datasets lack the move annotations and proficiency variation needed. Salsa is well-suited as an evaluation domain: improvised, dyadic, and governed by a move vocabulary and judging criteria covering timing, musicality, technique, difficulty, partnering, and originality. We present CoMPAS3D, a motion capture dataset of improvised partner salsa paired with an evaluation framework covering kinematic quality, two objective metrics (move legibility and proficiency appropriateness), and six competition-based subjective dimensions. The dataset includes 3 hours of improvisation by 18 dancers spanning beginner, intermediate, and professional levels, with over 2,800 expert-annotated segments covering move types, errors, and stylistic elements. We define three benchmarks: move classification (analogous to transcription), proficiency estimation (fluency assessment), and follower generation (dialogue response). Fine-tuned vision-language models perform strongly on objective metrics applied to ground-truth motion sequences. Applied to Duolando and InterGen, the metrics reveal failures that kinematic metrics miss. Human evaluations confirm the gap between generated and ground-truth motion. CoMPAS3D, annotations, benchmark code, and baseline results are publicly available.

Accurate predictions of spatio-temporal systems are crucial for tasks such as system management, control, and crisis prevention. However, the inherent time variance of many spatio-temporal systems poses challenges to achieving accurate predictions whenever stationarity is not granted. In order to address non-stationarity, we propose a Distribution and Relation Adaptive Network (DRAN) capable of dynamically adapting to relation and distribution changes over time. While temporal normalization and de-normalization are frequently used techniques to adapt to distribution shifts, this operation is not suitable for the spatio-temporal context as temporal normalization scales the time series of nodes and possibly disrupts the spatial relations among nodes. In order to address this problem, a Spatial Factor Learner (SFL) module is developed that enables the normalization and de-normalization process. To adapt to dynamic changes in spatial relationships among sensors, we propose a Dynamic-Static Fusion Learner (DSFL) module that effectively integrates features learned from both dynamic and static relations through an adaptive fusion ratio mechanism. Furthermore, we introduce a Stochastic Learner to capture the noisy components of spatio-temporal representations. Our approach outperforms state-of-the-art methods on weather prediction and traffic flow forecasting tasks.Experimental results show that our SFL efficiently preserves spatial relationships across various temporal normalization operations. Visualizations of the learned dynamic and static relations demonstrate that DSFL can capture both local and distant relationships between nodes.

Autonomous unmanned aerial vehicles (UAVs) increasingly rely on reinforcement learning (RL) for navigation. However, global navigation satellite system (GNSS) spoofing attacks can induce out-of-distribution observation shifts that corrupt value estimation and degrade mission performance. Existing robust RL approaches typically improve resilience against specific attack models but often fail to generalize to attacks not encountered during training. To address this limitation, we propose a curriculum-guided adaptation framework that progressively exposes a robust policy to gradient-based adversarial observation perturbations of increasing intensity while aligning temporal-difference (TD) error distributions across curriculum stages. Rather than adapting to a particular attack model, the proposed approach preserves TD-error consistency to promote transferability across attack conditions. We further derive a TD-space generalization certificate showing that if the TD-error distribution induced by a test-time attack remains sufficiently close to that of the final curriculum stage, the resulting performance degradation is bounded. The framework is evaluated in a UAV deconfliction environment with dynamic 3D obstacles under previously unseen fixed and dynamic GNSS spoofing attacks. Under fixed spoofing conditions, the curriculum-adapted policy achieved near-perfect mission success rates, compared with 20-56% for standard and robust RL baselines. Under dynamic obstacle-luring spoofing attacks, it achieved the highest episodic rewards while reducing mission completion steps by up to 45% across increasing aerial traffic densities.

Sign Language Recognition (SLR) involves the automatic identification and classification of sign gestures from images or video, converting them into text or speech to improve accessibility for the hearing-impaired community. In Bangladesh, Bangla Sign Language (BdSL) serves as the primary mode of communication for many individuals with hearing impairments. This study fine-tunes state-of-the-art video transformer architectures -- VideoMAE, ViViT, and TimeSformer -- on BdSLW60 (arXiv:2402.08635), a small-scale BdSL dataset with 60 frequent signs. We standardized the videos to 30 FPS, resulting in 9,307 user trial clips. To evaluate scalability and robustness, the models were also fine-tuned on BdSLW401 (arXiv:2503.02360), a large-scale dataset with 401 sign classes. Additionally, we benchmark performance against public datasets, including LSA64 and WLASL. Data augmentation techniques such as random cropping, horizontal flipping, and short-side scaling were applied to improve model robustness. To ensure balanced evaluation across folds during model selection, we employed 10-fold stratified cross-validation on the training set, while signer-independent evaluation was carried out using held-out test data from unseen users U4 and U8. Results show that video transformer models significantly outperform traditional machine learning and deep learning approaches. Performance is influenced by factors such as dataset size, video quality, frame distribution, frame rate, and model architecture. Among the models, the VideoMAE variant (MCG-NJU/videomae-base-finetuned-kinetics) achieved the highest accuracies of 95.5% on the frame rate corrected BdSLW60 dataset and 81.04% on the front-facing signs of BdSLW401 -- demonstrating strong potential for scalable and accurate BdSL recognition.

Graph Neural Networks (GNNs) have become essential tools for analyzing graph-structured data in domains such as drug discovery and financial analysis, leading to a growing demand for model transparency. Recent advances in explainable GNNs have addressed this need by revealing important subgraphs that influence predictions, but these explanation mechanisms may inadvertently expose these models to security risks. This paper investigates how such explanations potentially leak critical decision logic that can be exploited for model stealing. We propose {\method}, a novel stealing framework that integrates explanation alignment for capturing decision logic with guided data augmentation for efficient training under limited queries, enabling effective replication of both the predictive behavior and underlying reasoning patterns of target models. Experiments on molecular graph datasets demonstrate that our approach shows advantages over conventional methods in model stealing. This work highlights important security considerations for the deployment of explainable GNNs in sensitive domains and suggests the need for protective measures against explanation-based attacks. Our code is available at https://github.com/beanmah/EGSteal.

Due to the proficiency of self-attention mechanisms (SAMs) in capturing dependencies in sequence modeling, several existing dynamic graph neural networks (DGNNs) utilize Transformer architectures with various encoding designs to capture sequential evolutions of dynamic graphs. However, the effectiveness and efficiency of these Transformer-based DGNNs vary significantly, highlighting the importance of properly defining the SAM on dynamic graphs and comprehensively encoding temporal and interactive dynamics without extra complex modules. In this work, we propose TIDFormer, a dynamic graph TransFormer that fully exploits Temporal and Interactive Dynamics in an efficient manner. We clarify and verify the interpretability of our proposed SAM, addressing the open problem of its uninterpretable definitions on dynamic graphs in previous works. To model the temporal and interactive dynamics, respectively, we utilize the calendar-based time partitioning information and extract informative interaction embeddings for both bipartite and non-bipartite graphs using merely the sampled first-order neighbors. In addition, we jointly model temporal and interactive features by capturing potential changes in historical interaction patterns through a simple decomposition. We conduct extensive experiments on several dynamic graph datasets to verify the effectiveness and efficiency of TIDFormer. The experimental results demonstrate that TIDFormer excels, outperforming state-of-the-art models across most datasets and experimental settings. Furthermore, TIDFormer exhibits significant efficiency advantages compared to previous Transformer-based methods.

Fusing heterogeneous information remains a persistent challenge in modern data analysis. While significant progress has been made, existing approaches often fail to account for the inherent heterogeneity of object patterns across different semantic spaces. To address this limitation, we propose the Cooperation of Experts (CoE) framework, which encodes multi-typed information into unified heterogeneous multiplex networks. By overcoming modality and connection differences, CoE provides a powerful and flexible model for capturing the intricate structures of real-world complex data. In our framework, dedicated encoders act as domain-specific experts, each specializing in learning distinct relational patterns in specific semantic spaces. To enhance robustness and extract complementary knowledge, these experts collaborate through a novel large margin mechanism supported by a tailored optimization strategy. Rigorous theoretical analyses guarantee the framework's feasibility and stability, while extensive experiments across diverse benchmarks demonstrate its superior performance and broad applicability. Our code is available at https://github.com/strangeAlan/CoE.

In deep learning, functional similarity evaluation quantifies the extent to which independently trained models learn similar input--output relationships. In model stitching, functional similarity is framed as representation forward compatibility, i.e., whether the representations of two models can be aligned to solve a given task. Recent studies, however, highlight a critical limitation: models relying on different information cues can still produce compatible representations, making them appear misleadingly similar (Smith et al., 2025). We attribute this failure to standard model stitching being inherently blind to the invariance properties of the stitched models. To address this limitation, we introduce the forward--backward compatibility requirement under which we formulate the invariance-aware model stitching. Through analyzing key stitching configurations, we study the interplay between forward and backward compatibility, showing that invariance-aware model stitching provides a more principled approach to functional similarity evaluation while revealing functional discrepancies previously obscured.

3D Gaussian Splatting (3DGS) has shown a powerful capability for novel view synthesis due to its detailed expressive ability and highly efficient rendering speed. Unfortunately, creating relightable 3D assets and reconstructing faithful geometry with 3DGS is still problematic, particularly for reflective objects, as its discontinuous representation raises difficulties in constraining geometries. Volumetric signed distance field (SDF) methods provide robust geometry reconstruction, while the expensive ray marching hinders its real-time application and slows the training. Besides, these methods struggle to capture sharp geometric details. To this end, we propose to guide 3DGS and SDF bidirectionally in a complementary manner, including an SDF-aided Gaussian splatting for efficient optimization of the relighting model and a GS-guided SDF enhancement for high-quality geometry reconstruction. At the core of our SDF-aided Gaussian splatting is the mutual supervision of the depth and normal between blended Gaussians and SDF, which avoids the expensive volume rendering of SDF. Thanks to this mutual supervision, the learned blended Gaussians are well-constrained with a minimal time cost. As the Gaussians are rendered in a deferred shading mode, the alpha-blended Gaussians are smooth, while individual Gaussians may still be outliers, yielding floater artifacts. Therefore, we introduce an SDF-aware pruning strategy to remove Gaussian outliers located distant from the surface defined by SDF, avoiding floater issue. This way, our GS framework provides reasonable normal and achieves realistic relighting, while the mesh from depth is still problematic. Therefore, we design a GS-guided SDF refinement, which utilizes the blended normal from Gaussians to finetune SDF. With this enhancement, our method can further provide high-quality meshes for reflective objects at the cost of 17% extra training time.

This paper presents the Robust Recurrent Deep Network (R2DN), a scalable parameterization of robust recurrent neural networks for machine learning and data-driven control. We construct R2DNs as the feedback interconnection of a linear time-invariant system and a 1-Lipschitz deep feedforward network, and directly parameterize the weights so that our models are stable (contracting) and robust to small input perturbations (Lipschitz) by design. Our parameterization uses a structure similar to the previously-proposed recurrent equilibrium network (REN), but without the requirement to iteratively solve an equilibrium layer at each time-step. This speeds up both model inference and backpropagation on GPUs, and makes it computationally feasible to scale up the network size, batch size, and input sequence length in comparison to RENs. We compare R2DNs to RENs on three representative problems in nonlinear system identification, observer design, and learning-based feedback control. We find that training and inference are both up to an order of magnitude faster with similar test set performance, and that they scale more favorably with respect to model expressivity.

In the past decade, considerable research effort has been devoted to securing machine learning (ML) models that operate in adversarial settings. Yet, progress has been slow even for simple "toy" problems (e.g., robustness to small adversarial perturbations) and is often hindered by non-rigorous evaluations. Today, adversarial ML research has shifted towards studying larger, general-purpose language models. In this position paper, we argue that the situation is now even worse: in the era of LLMs, the field of adversarial ML studies problems that are (1) less clearly defined, (2) harder to solve, and (3) even more challenging to evaluate. As a result, we caution that yet another decade of work on adversarial ML may be failing to produce meaningful progress.

This paper presents a differentiable optimization framework for the design of constrained Linear Differential Microphone Arrays (LDMAs). The proposed method leverages a non-uniform delay-and-sum beamformer as a light-weight base system model, proving its ability to achieve the optimal beampattern of LDMAs by jointly optimizing microphone positions and filter weights. The formulation enables the optimized design of a filter with a distortion-free constraint in the desired sound direction, while also imposing constraints on microphone positioning to ensure consistent performance. Through evaluation on multiple metrics, including Mean Squared Error (MSE), Directivity Index (DI), White Noise Gain (WNG), and computation time, and comparison with state-of-the-art methods, this approach demonstrates a flexible, directive, robust, and hardware-efficient design.

Vision-Language Models (VLMs) bring powerful understanding and reasoning capabilities to multimodal tasks. Meanwhile, the great need for capable aritificial intelligence on mobile devices also arises, such as the AI assistant software. Some efforts try to migrate VLMs to edge devices to expand their application scope. Simplifying the model structure is a common method, but as the model shrinks, the trade-off between performance and size becomes more and more difficult. Knowledge distillation (KD) can help models improve comprehensive capabilities without increasing size or data volume. However, most of the existing large model distillation techniques only consider applications on single-modal LLMs, or only use teachers to create new data environments for students. None of these methods take into account the distillation of the most important cross-modal alignment knowledge in VLMs. We propose a method called Align-KD to guide the student model to learn the cross-modal matching that occurs at the shallow layer. The teacher also helps student learn the projection of vision token into text embedding space based on the focus of text. Under the guidance of Align-KD, the 1.7B MobileVLM V2 model can learn rich knowledge from the 7B teacher model with light design of training loss, and achieve an average score improvement of 2.0 across 6 benchmarks under two training subsets respectively. Code is available at: https://github.com/fqhank/Align-KD.

Physics-informed neural networks (PINNs) have emerged as a powerful deep learning approach for solving partial differential equations (PDEs) in the physical sciences, yet their behavior remains largely opaque and is typically understood through failure mode analyses rather than explicit interpretability. To address this issue, we introduce PINNfluence, a training data attribution framework for interpreting PINNs based on influence functions. By extending influence functions to composite physics-informed training objectives, we enable fine-grained attribution between predictions, loss components, and training data points. Through benchmark experiments across various PDEs, we demonstrate that influence patterns provide granular diagnostics that distinguish structural characteristics across well-trained and poorly-trained PINNs. PINNfluence thus opens a new avenue for understanding and improving the reliability of PINNs through the lens of their data.

While vision-language models like CLIP have shown remarkable success in open-vocabulary tasks, their application is currently confined to image-level tasks, and they still struggle with dense predictions. Recent works often attribute such deficiency in dense predictions to the self-attention layers in the final block, and have achieved commendable results by modifying the original query-key attention to self-correlation attention, (e.g., query-query and key-key attention). However, these methods overlook the cross-correlation attention (query-key) properties, which capture the rich spatial correspondence. In this paper, we reveal that the cross-correlation of the self-attention in CLIP's non-final layers also exhibits localization properties. Therefore, we propose the Residual Cross-correlation Self-attention (RCS) module, which leverages the cross-correlation self-attention from intermediate layers to remold the attention in the final block. The RCS module effectively reorganizes spatial information, unleashing the localization potential within CLIP for dense vision-language inference. Furthermore, to enhance the focus on regions of the same categories and local consistency, we propose the Semantic Feedback Refinement (SFR) module, which utilizes semantic segmentation maps to further adjust the attention scores. By integrating these two strategies, our method, termed ResCLIP, can be easily incorporated into existing approaches as a plug-and-play module, significantly boosting their performance in dense vision-language inference. Extensive experiments across multiple standard benchmarks demonstrate that our method surpasses state-of-the-art training-free methods, validating the effectiveness of the proposed approach. Code is available at https://github.com/yvhangyang/ResCLIP.

Optimizing costly black-box functions within a constrained evaluation budget presents significant challenges in many real-world applications. Surrogate Optimization (SO) is a common resolution, yet its proprietary nature introduced by the complexity of surrogate models and the sampling core (e.g., acquisition functions) often leads to a lack of explainability and transparency. While existing literature has primarily concentrated on enhancing convergence to global optima, the practical interpretation of newly proposed strategies remains underexplored, especially in batch evaluation settings. In this paper, we propose \emph{Inclusive} Explainability Metrics for Surrogate Optimization (IEMSO), a comprehensive set of model-agnostic metrics designed to enhance the transparency, trustworthiness, and explainability of the SO approaches. Through these metrics, we provide both intermediate and post-hoc explanations to practitioners before and after performing expensive evaluations to gain trust. We consider four primary categories of metrics, each targeting a specific aspect of the SO process: Sampling Core Metrics, Batch Properties Metrics, Optimization Process Metrics, and Feature Importance. Our experimental evaluations demonstrate the significant potential of the proposed metrics across different benchmarks.

Large language models (LLMs) are prone to hallucinate unintended text due to false or outdated knowledge. Since retraining LLMs is resource intensive, there has been a growing interest in model editing. Despite the emergence of benchmarks and approaches, existing unidirectional editing and evaluation paradigms have failed to explore the reversal curse. In this paper, we study bidirectional language model editing, aiming to provide a rigorous evaluation to assess if edited LLMs can recall the editing knowledge bidirectionally. A metric of reverse generalization is introduced and a benchmark dubbed Bidirectional Assessment for Knowledge Editing (BAKE) is constructed to evaluate if post-edited models can recall the edited knowledge in the reverse direction of editing. We conduct extensive experiments using a variety of editing methods and LLMs. The results show that while most editing methods are able to accurately recall editing facts along the modification direction, they exhibit substantial systematic deficiencies when evaluating in the reverse direction. To further investigate the underlying causes of reversal curse and to explore potential strategies for mitigation, a detailed analysis is conducted from three perspectives. Our findings reveal that although In-Context Learning (ICL) can mitigate the reversal curse to a certain extent, it lacks continuity, is limited by the input length, and may introduce hallucinations. Therefore, combining the advantages of ICL and other editing methods is a promising direction for developing new editing paradigms.

Learning models whose predictions are invariant under multiple environments is a promising approach for out-of-distribution generalization. Such models are trained to extract features $X_{\text{inv}}$ where the conditional distribution $Y \mid X_{\text{inv}}$ of the label given the extracted features does not change across environments. Invariant models are also supposed to generalize to shifts in the marginal distribution $p(X_{\text{inv}})$ of the extracted features $X_{\text{inv}}$, a type of shift we call an $\textit{invariant covariate shift}$. However, we show that proposed methods for learning invariant models underperform under invariant covariate shift, either failing to learn invariant models$\unicode{x2014}$even for data generated from simple and well-studied linear-Gaussian models$\unicode{x2014}$or having poor finite-sample performance. To alleviate these problems, we propose $\textit{weighted risk invariance}$ (WRI). Our framework is based on imposing invariance of the loss across environments subject to appropriate reweightings of the training examples. We show that WRI provably learns invariant models, i.e. discards spurious correlations, in linear-Gaussian settings. We propose a practical algorithm to implement WRI by learning the density $p(X_{\text{inv}})$ and the model parameters simultaneously, and we demonstrate empirically that WRI outperforms previous invariant learning methods under invariant covariate shift.

Statistical information is ubiquitous but drawing valid conclusions from it is prohibitively hard. We explain how knowledge graph embeddings can be used to approximate probabilistic inference efficiently using the example of Statistical EL (SEL), a statistical extension of the lightweight Description Logic EL. We provide proofs for runtime and soundness guarantees, and empirically evaluate the runtime and approximation quality of our approach.

Diffusion models have demonstrated impressive image generation capabilities. Personalized approaches, such as textual inversion and Dreambooth, enhance model individualization using specific images. These methods enable generating images of specific objects based on diverse textual contexts. Our proposed approach aims to retain the model's original knowledge during new information integration, resulting in superior outcomes while necessitating less training time compared to Dreambooth and textual inversion.

Training with noisy class labels impairs neural networks' generalization performance. In this context, mixup is a popular regularization technique to improve training robustness by making memorizing false class labels more difficult. However, mixup neglects that multiple annotators, e.g., crowdworkers, typically provide class labels. Therefore, we propose an extension of mixup, which handles multiple class labels per instance while considering which class label originates from which annotator. Integrated into our multi-annotator classification framework annot-mix, it performs superiorly to eleven (mostly state-of-the-art) approaches in an evaluation study with eleven datasets comprising noisy class labels from either human or simulated annotators. Our code is publicly available through our GitHub repository at https://github.com/ies-research/multi-annotator-machine-learning/tree/annot-mix

Fully-observable non-deterministic (FOND) planning is at the core of artificial intelligence planning with uncertainty. It models uncertainty through actions with non-deterministic effects. In this work, we present a collection of techniques that establish explicit best-first policy-space search as a method competitive with the state of the art for solving FOND planning tasks. We study how to define equivalence relations between policies, allowing part of the search space to be pruned. We show it is possible to use group theory techniques to effectively compute canonical symmetries between states. We also present two contributions that go beyond just policy-space search: we present a procedure that infers in polynomial time a solution policy function given just the specification of its domain set, and an integer-programming formulation procedure that, given a solution policy defined over complete states, yields a set of resource-efficient models that are capable of finding a partial-state policy that represents it unambiguously with the fewest partial states possible.

The choice of \emph{which} tokens to mask is a central, under-examined design decision in masked language modeling (MLM). Standard pretraining masks tokens uniformly at random, but several studies show that more informative masking targets can improve downstream performance. We study masking as a \emph{task-adaptive} component of the fine-tuning pipeline and introduce \textbf{Typhoon}, a masking strategy that uses the gradient of the task loss with respect to one-hot token inputs to estimate, online, how much each token type contributes to the objective. Typhoon maintains an exponential moving average of per-token-type saliency and calibrates these scores into a masking distribution whose expected masking rate matches a target budget, under a token-independence approximation. We formalize the method and evaluate it against random masking and whole-word masking on two GLUE tasks, MRPC and CoLA, across three BERT-family backbones (TinyBERT, DistilBERT, and BERT-base) and five random seeds per configuration ($90$ training runs in total). Our main finding is that, once seed variance is accounted for, no masking strategy is reliably better than the others on these tasks: on MRPC the gap between Typhoon and the best baseline stays within $0.004$ $F_1$, across all twelve Typhoon comparisons no paired test reaches significance, and every $95\%$ confidence interval contains zero. Typhoon's apparent advantage in single-run experiments does not survive this more careful evaluation. We read this as a cautionary, reproducibility-focused result -- gradient-based task-adaptive masking is competitive but not clearly better than resource-free random masking at this scale -- and we describe a clean modern reimplementation to support follow-up work.

In robotics systems, vast amounts of visual data are easily captured at high resolution using low-cost, low-power hardware. Yet, limited bandwidth and on-device compute resources prevent full utilization when transmitted via conventional codecs like JPEG/MPEG. Newer codecs, like AV1/AVIF, improve the rate-distortion trade-off, but demand far more resources for encoding, impractical without custom ASICs. Recent asymmetric autoencoders deliver high quality under extreme power and bandwidth constraints, but add prohibitive decoding cost and use bespoke formats that ignore decades of infrastructure built around standards like JPEG. To address these limitations, we introduce a compression framework for cloud robotics based on a Sensor Embedded Autoencoder paired with a One-Time Transcode for Efficient Reconstruction (SEAOTTER). Because the sensor, cloud, and consumer stages face very different power and bandwidth budgets, SEAOTTER combines the compactness of a learned latent with the broad usability of a standard JPEG file. Since naive transcoding degrades performance, we propose a learnable JPEG color and quantization transform that enables increased accuracy for global, dense, and vision-language-based perception. Using SEAOTTER, we train both general-purpose and task-aware transcoding pipelines for a pre-trained, frozen encoder. At a compression ratio of 200:1 and compared to AVIF, we observe 7 times faster encoding, 3.5 times faster decoding, and +8% ImageNet top-1 accuracy, while retaining compatibility with JPEG infrastructure. Our code is available at https://github.com/UT-SysML/seaotter .

Recently, diffusion models operating on VAE latents or mel-spectrograms have become the dominant paradigm for zero-shot TTS. Although these compressed representations improve generation efficiency, they inevitably suffer from information loss and non-end-to-end training. Theoretically, directly modeling raw waveforms circumvents these issues; however, this direction remains underexplored and is often deemed difficult due to the extremely long sequence length of audio signals. To overcome this, we propose WavTTS, the first raw waveform generative TTS model that substantially narrows the gap with latent-space generative models. Built upon the flow matching with Diffusion Transformer (DiT), WavTTS directly models speech waveforms via a simple patchification strategy, while integrating multi-scale mel-spectrogram supervision to provide perceptual guidance during training. Furthermore, we investigate the impact of prediction targets and noise scheduling in waveform diffusion, and develop an effective schedule design to improve generation quality. Evaluations on open-source benchmarks demonstrate that WavTTS closely approaches the performance of current state-of-the-art latent generative zero-shot TTS models, while substantially outperforming previous end-to-end speech generation models. Our findings demonstrate the feasibility of scaling diffusion-based TTS directly in the waveform space, opening a new direction for end-to-end speech generation.

The rapid advancement of instruction-guided audio generation has highlighted the critical need for robust alignment evaluation. Current automated evaluation methods heavily rely on holistic scoring from general-purpose large language models, which struggle to decouple complex instructions, lack interpretability, and fail to capture fine-grained attribute mismatches. To address this, we introduce a novel dynamic rubric-based evaluation paradigm that adaptively decomposes complex audio captions into a variable number of independent, verifiable binary rubric items. To rigorously benchmark this capability, we propose the AnyAudio-Judge Bench, a comprehensive, bilingual benchmark comprising 7,920 meticulously curated samples across four diverse audio domains (speech, sound, music, and mixed), featuring deliberately constructed hard negatives. Furthermore, we construct a large-scale corpus of 105K samples with explicit Chain-of-Thought (CoT) rationales to train our dedicated evaluator, the AnyAudio-Judge model. By employing a training pipeline that combines Supervised Fine-Tuning (SFT) and Group Relative Policy Optimization (GRPO), our model successfully aligns its reasoning paths with the rubric-based scoring mechanism. Extensive experiments demonstrate that AnyAudio-Judge not only significantly enhances zero-shot alignment detection compared to state-of-the-art baselines, but also provides precise and interpretable reward signals that substantially improve instruction alignment in downstream reinforcement learning for audio generation.

In this study, we conduct a comprehensive comparative analysis of generative and discriminative deep learning-based speech enhancement methods, specifically in noise reduction tasks. Our investigation focuses on evaluating their effectiveness under high and low signal-to-noise ratio conditions, considering both matched and mismatched training scenarios. We further investigate the impact of training data volume, model convergence speed, and interpret the performance differences in terms of objective results for the considered training paradigms. Additionally, we compare the complexity-performance trade-off and the practical viability of these approaches. To further strengthen the evaluation, we study the hallucination characteristics of generative approaches in terms of word error rate and phoneme similarity. The insights derived from this study provide empirical evidence to assist researchers and practitioners in understanding whether the perceptual gains of different approaches justify their computational cost in practical applications.

We introduce D^3S Consensus, a physics-based, closed-form algorithm that unifies depth-from-defocus (DfD) and stereo to achieve highly accurate depth estimation throughout an extended working range beyond the depth-of-field (DoF) of cameras. Given a pair of dual-defocus stereo images, the method estimates an overdetermined set of depth using a novel DfD theory, Dual Differential Defocus (D^3), and (S)tereo in a coupled fashion. It then picks the most confident depth prediction from the set by enforcing consensus between these physically independent cues to reject unreliable estimates. Analysis shows that D^3S achieves a comparable working range under the same error tolerance with 10x smaller baseline than previous triangulation-based depth estimation systems. This enables compact passive binocular rangefinders with substantially smaller form factors than conventional stereo and DfD designs. We demonstrate the first D^3S prototype with only 4 mm baseline and 12 mm EFL. It generates up to 900 x 1800-pixel depth maps with 1-cm mean absolute error over 0.3-1.64 m from a snapshot acquisition. This has surpassed the reported accuracy of certain commercially available stereo cameras with much larger form factors.