arXiv:2606.02577v1 Announce Type: new
Abstract: Scaling robot learning requires large-scale, diverse demonstrations, yet real-world data collection via teleoperation remains prohibitively expensive and time-consuming. While video diffusion models offer a promising avenue for data scaling, existing generative approaches are often limited to superficial visual augmentation, or suffer from embodiment hallucinations that yield physically infeasible motions. We present a generalizable embodiment-centric world model that achieves scalable data generation by synthesizing photorealistic demonstrations with novel objects, in novel scenes, and from novel viewpoints. Our approach anchors generation to rendered robot motion while conditioning on explicit scene and object priors, effectively decoupling trajectory execution from environment synthesis. This formulation has the potential to unlock two powerful data scaling capabilities: (1) retrieval and rebirth, which repurposes existing trajectories into entirely new contexts without new motion data; and (2) prop-free teleoperation, where operators manipulate empty air and the model hallucinates the target objects and scene afterwards, eliminating reset time. We demonstrate with real-world experiments that our generated data consistently improves downstream policy performance and significantly reduces real-world data requirements across diverse manipulation tasks.
Science Journals
arXiv:2507.02792v5 Announce Type: replace
Abstract: Text-to-image (T2I) diffusion models have shown remarkable success in generating high-quality images from text prompts. Recent efforts extend these models to incorporate conditional images (e.g., canny edge) for fine-grained spatial control. Among them, feature injection methods have emerged as a training-free alternative to traditional fine-tuning-based approaches. However, they often suffer from structural misalignment, condition leakage, and visual artifacts, especially when the condition image diverges significantly from natural RGB distributions. Through an analysis of existing methods, we identify a key limitation: the sampling schedule of condition features, previously unexplored, fails to account for the evolving interplay between structure preservation and domain alignment throughout diffusion steps. Inspired by this observation, we propose a flexible training-free framework that decouples the sampling schedule of condition features from the denoising process, and systematically investigate the spectrum of feature injection schedules to achieve a better balance between structural alignment and appearance quality. We further enhance the sampling process by introducing a restart refinement schedule, and improve the visual quality with an appearance-rich prompting strategy. Together, these designs enable training-free controllable generation that is both structure-rich and appearance-rich. Extensive experiments demonstrate that our method achieves state-of-the-art performance under complex and diverse conditions. Owing to its generality, our framework naturally supports compositional conditional generation and generalizes across architectures in a plug-and-play manner, from UNet-based diffusion models to modern DiT backbones such as FLUX.
arXiv:2506.22881v3 Announce Type: replace
Abstract: Density ratio estimation is a core concept in statistical machine learning because it provides a unified mechanism for tasks such as importance weighting, divergence estimation, and likelihood-free inference, but its potential in vision and language models has not been fully explored. Modern vision-language encoders such as CLIP and SigLIP are trained with contrastive objectives that implicitly optimize log density ratios between joint and marginal image-text distributions, which implicitly learn similarity scores proportional to log density ratios. However, prior work has largely focused on their embedding utility, and the density-ratio structure induced by contrastive learning has not been systematically examined or exploited in multimodal applications. To address this gap, we reinterpret CLIP-style models as pretrained and general-purpose density ratio estimators and show that this perspective enables new algorithmic capabilities. We present a unified explanation of how contrastive objectives estimate density ratios and propose two practical applications: Importance Weight Learning and KL divergence estimation. Our Importance Weight Learning method requires only a single additional prompt and improves F1 scores by up to 7 points. We further show that CLIP-based density ratios support estimation of KL divergences that quantify how conditioning on an image or text alters the distribution of the other modality. Through qualitative examples and an N-gram analysis of captions, we find that these divergences capture semantic diversity and mode structure in multimodal data. Leveraging this property, we introduce a simple KL-guided data curation method that achieves performance competitive with LAION2B filtering.
arXiv:2606.02071v1 Announce Type: cross
Abstract: In this article, we consider the words with cyclic indices. For given $s$, we consider the pair $(\iota,\kappa)$ of indices such that the word of length $s$ from $\iota$ is equal to the word of length $s$ from $\kappa$. We give a characterization of such pairs for a cyclic Fibonacci word, and give the number of them.
arXiv:2406.10359v3 Announce Type: replace
Abstract: This paper addresses the problem of identifying a nonlinear state-space model, along with an adequate model order, from a given input-output training dataset. To this end, a novel framework, termed state-space neural network with ordered variance (SSNNO), is proposed. In SSNNO, the state variables are ordered according to their variances computed using the training data. This ordering is achieved by introducing a variance-regularization term into the loss function used for SSNNO training and it facilitates a distinction between significant states, which exhibit high variance from the other residual states with near-zero variance. The number of significant states is indicative of a suitable model order. The variance-regularization mechanism is designed to minimize the number of significant state variables, thereby promoting a minimal order of the identified state-space model without significantly compromising its prediction accuracy. A systematic procedure is then introduced to obtain a reduced-order state-space model from the trained SSNNO, yielding a reduced-order SSNNO (R-SSNNO). The existence of an SSNNO with variance-ordered states, based solely on input-output data, as well as an upper bound on its output prediction error, are formally established. A practical and robust method is proposed for ensuring variance-ordered states in an SSNNO, even when the network is trained using local optimization algorithms. The effectiveness of the proposed method for identification of nonlinear state space models is demonstrated through simulation studies on a nonlinear continuous stirred-tank reactor process. The identified model is further used for state estimation and prediction in a model predictive control implementation.
arXiv:2603.25484v3 Announce Type: replace
Abstract: Migrating stateful microservices in Kubernetes requires careful state management because in-memory state is lost when a container restarts. For StatefulSet-managed workloads, the problem is amplified by identity constraints that prohibit two pods with the same ordinal from running simultaneously, forcing a sequential stop-restore cycle with unavoidable downtime. This paper presents SHADOW (Seamless Handoff And Zero-Downtime Orchestrated Workload Migration), a Kubernetes-native framework that implements the Message-based Stateful Microservice Migration (MS2M) approach as a Kubernetes Operator. SHADOW introduces the ShadowPod strategy, where a shadow pod is created from a CRIU checkpoint image on the target node while the source pod continues serving traffic, allowing concurrent operation during message replay. For StatefulSet workloads, an identity swap procedure with the ExchangeFence mechanism re-checkpoints the shadow pod, creates a StatefulSet-owned replacement, and drains both message queues to guarantee zero message loss during the handoff. An evaluation on a bare-metal Kubernetes cluster with 280 migration runs across four configurations and seven message rates shows that, compared to the sequential baseline on the same StatefulSet workload, the ShadowPod strategy reduces the restore phase by up to 92%, eliminates service downtime, and reduces total migration time by up to 77%, with zero message loss across all 280 runs.
arXiv:2604.10688v2 Announce Type: replace
Abstract: On-policy reinforcement learning has become the dominant paradigm for reasoning alignment in large language models, yet its sparse, outcome-level rewards make token-level credit assignment notoriously difficult. On-Policy Distillation (OPD) alleviates this by introducing dense, token-level KL supervision from a teacher model, but typically applies this supervision uniformly across all rollouts, ignoring fundamental differences in signal quality. We propose Signal-Calibrated On-Policy Distillation Enhancement (SCOPE), a dual-path adaptive training framework that routes on-policy rollouts by correctness into two complementary supervision paths. For incorrect trajectories, SCOPE performs teacher-perplexity-weighted KL distillation to prioritize instances where the teacher demonstrates genuine corrective capability, while down-weighting unreliable guidance. For correct trajectories, it applies student-perplexity-weighted MLE to concentrate reinforcement on low-confidence samples at the capability boundary rather than over-reinforcing already mastered ones. Both paths employ a group-level normalization to adaptively calibrate weight distributions, accounting for the intrinsic difficulty variance across prompts. Extensive experiments on six reasoning benchmarks show that SCOPE achieves an average relative improvement of 11.42% in Avg@32 and 7.30% in Pass@32 over competitive baselines, demonstrating its consistent effectiveness.
arXiv:2606.02344v1 Announce Type: new
Abstract: Oblivious Transfer (OT) is a fundamental cryptographic primitive enabling privacy-preserving computation and constitutes a core building block for secure multi-party computation while supporting a wide range of security-sensitive applications: private information retrieval, zero-knowledge proofs, and password-authenticated key exchange, to cite a few. While recent advances in OT extension have significantly reduced amortised costs, their reliance on batches of random base OTs and substantial pre-computation phases limits their practicality in scenarios where the number of transfers is modest or where communication latency and client-side computation are critical constraints. In such settings, efficient base OT protocols remain both relevant and necessary. In this work, we introduce $I$-$(OT)^2$, a novel base 1-out-of-2 OT protocol grounded in the quadratic residuosity problem, specifically designed to minimise receiver-side computation and interaction. Our construction is particularly appealing on client--server architectures in which the receiver operates on low-power hardware, such as Internet of Things (IoT) devices. Through a lightweight offline pre-computation phase, $I$-$(OT)^2$ shifts the on-transfer computational burden almost entirely to the Sender, while reducing online communication to only six messages and four digests exchanged. We provide a detailed description of the protocol, accompanied by a formal proof of its security. Moreover, to demonstrate the viability of $I$-$(OT)^2$, we also present an open-source proof-of-concept implementation (in C language) evaluated on real IoT hardware. Results are staggering: for 128-bit security using a 3072-bit RSA modulus, the receiver incurs an average online cost per OT as low as 2.80 {\mu}s on desktop platforms and 39.90 {\mu}s on IoT devices, more than 10$\times$ faster than the well known SimplestOT.
arXiv:2603.26028v2 Announce Type: replace
Abstract: Medical Visual Question Answering (MedVQA) models often exhibit limited generalization due to reliance on dataset-specific correlations, such as recurring anatomical patterns or question-type regularities, rather than genuine diagnostic evidence. Existing causal approaches are typically implemented as static adjustments or post-hoc corrections. To address this issue, we propose a Learnable Causal Trimming (LCT) framework that integrates causal pruning into end-to-end optimization. We introduce a Dynamic Anatomical Feature Bank (DAFB), updated via a momentum mechanism, to capture global prototypes of frequent anatomical and linguistic patterns, serving as an approximation of dataset-level regularities. We further design a differentiable trimming module that estimates the dependency between instance-level representations and the global feature bank. Features highly correlated with global prototypes are softly suppressed, while instance-specific evidence is emphasized. This learnable mechanism encourages the model to prioritize causal signals over spurious correlations adaptively. Experiments on VQA-RAD, SLAKE, SLAKE-CP and PathVQA demonstrate that LCT consistently improves robustness and generalization over existing debiasing strategies.
arXiv:2603.26779v2 Announce Type: replace
Abstract: Large Language Models (LLMs) have demonstrated impressive reasoning capabilities, yet they struggle with spatial tasks that require mental simulation, such as mental rotation. This paper investigates whether equipping an LLM with an external ``Imagery Module'' -- a tool capable of rendering and rotating 3D models -- can bridge this gap, functioning as a ``cognitive prosthetic.'' We conducted experiments using a dual-module architecture in which a reasoning module (an MLLM) interacts with an imagery module on 3D model rotation tasks. Performance was lower than expected, with accuracy reaching at most 62.5%. Further investigation suggests that even when the burden of maintaining and manipulating a holistic 3D state is outsourced, the system still fails. This reveals that current frontier models lack the foundational visual-spatial primitives required to interface with imagery. Specifically, they lack: (1) the low-level sensitivity to extract spatial signals such as (a) depth, (b) motion, and (c) short-horizon dynamic prediction; and (2) the capacity to reason contemplatively over images, dynamically shifting visual focus and balancing imagery with symbolic and associative information.
arXiv:2605.02016v3 Announce Type: replace
Abstract: Therapy and life-coaching apps have been rapidly growing in number, flavors, and popularity. However, their users routinely share highly sensitive and personal information, such as traumas, fantasies, desires, relationship difficulties, and other mental health concerns. This prompts the need for an empirical analysis of privacy practices in this ecosystem, and particularly the alignment between these apps' privacy policies and their actual behavior. In this paper, we present a comprehensive analysis of 25 popular Android mental health and life-coaching apps, combining static analysis, dynamic network capture, and LLM-assisted privacy policy extraction validated against manual annotation.
Our findings highlight serious concerns and substantial transparency gaps. First, every app embeds at least one tracker SDK that its privacy policy does not name, and 68% of apps fail to disclose at least half of the trackers detected in their APKs; Talkie alone embeds 20 while naming none. Second, we identify 16 permission-policy contradictions across 13 apps, i.e., a dangerous permission is declared in the manifest but omitted from the policy, including 6 apps that request camera or microphone access without disclosing photo, video, or audio collection. Third, 48% of apps disclose third-party AI processing (e.g., via OpenAI, Anthropic, Groq), with one app sending journal entries to all three simultaneously, while 7 apps use only generic language that leaves recipients unidentified. Taken together, our findings demonstrate that current disclosure practices fall short of the transparency required for meaningful informed consent. We argue for a significantly updated regulatory framework governing therapy apps in the spirit of the professional and ethical standards that bind licensed human therapists.
arXiv:2603.27223v2 Announce Type: replace
Abstract: We present EuraGovExam, a multilingual and multimodal benchmark sourced from real-world civil service examinations across five representative Eurasian regions: South Korea, Japan, Taiwan, India, and the European Union. Designed to reflect the authentic complexity of public-sector assessments, the dataset contains over 8,000 high-resolution scanned multiple-choice questions covering 17 diverse academic and administrative domains. Unlike existing benchmarks, EuraGovExam embeds all question content--including problem statements, answer choices, and visual elements--within a single image, providing only a minimal standardized instruction for answer formatting. This design demands that models perform layout-aware, cross-lingual reasoning directly from visual input. All items are drawn from real exam documents, preserving rich visual structures such as tables, multilingual typography, and form-like layouts. Evaluation results show that even state-of-the-art vision-language models (VLMs) achieve only 86% accuracy, underscoring the benchmark's difficulty and its power to diagnose the limitations of current models. By emphasizing cultural realism, visual complexity, and linguistic diversity, EuraGovExam establishes a new standard for evaluating VLMs in high-stakes, multilingual, image-grounded settings. It also supports practical applications in e-governance, public-sector document analysis, and equitable exam preparation.
arXiv:2603.27224v4 Announce Type: replace
Abstract: Memory leaks remain prevalent in real-world C/C++ software. Static analyzers such as CodeQL provide scalable program analysis but frequently miss such bugs because they cannot recognize project-specific custom memory-management functions and lack path-sensitive control-flow modeling. We present MemHint, a neuro-symbolic pipeline that addresses both limitations by combining LLMs' semantic understanding of code with Z3-based symbolic reasoning. MemHint parses the target codebase and applies an LLM to classify each function as a memory allocator, deallocator, or neither, producing function summaries that record which argument or return value carries memory ownership, extending the analyzer's built-in knowledge beyond standard primitives such as malloc and free. A Z3-based validation step checks each summary against the function's control-flow graph, discarding those whose claimed memory operation is unreachable on any feasible path. The validated summaries are injected into CodeQL and Infer via their respective extension mechanisms. Z3 path feasibility filtering then eliminates warnings on infeasible paths, and a final LLM-based validation step confirms whether each remaining warning is a genuine bug. On eight real-world C/C++ projects totaling over 3.6M lines of code, MemHint detects 54 unique memory leaks (53 confirmed/fixed) at approximately $1.7 per detected bug, compared to 19 by vanilla CodeQL and 3 by vanilla Infer.
arXiv:2404.11326v5 Announce Type: replace
Abstract: In recent years, change detection and anomaly detection models based on CNN and transformer have achieved remarkable success across various datasets based on paired data. However, most such methods exhibit limited crossdataset generalization due to domain-specific designs and typically rely on large amounts of paired labeled data. In this paper, based on visual-language pre-training model, we introduce a Single-temporal multimodal Contrastive Learning (SCL) foundation models for change detection without training on the target dataset. To further improve the model's ability to learn context of textual and visual information, we propose a Dynamic Text-vision Context Optimization (DTCO) module for prompt learning. Meanwhile, to address the data dependency issue of existing methods, we introduce a controllable generation and Single-temporal trAINing strategy (SAIN). This allows us to train the model using a large number of existing single-temporal images without the need for paired label. Extensive experiments on various realworld change detection datasets demonstrate the superior performance and generalization of SCL, outperforming state-of-the-art methods under the evaluated settings. Code is available at https://github.com/Kane-Du/scl-cd.git.
arXiv:2604.11063v2 Announce Type: replace
Abstract: The high-precision solution of the radial Dirac equation is fundamental to relativistic quantum chemistry, essential for reliable pseudopotential generation and all-electron electronic structure methods. Capturing both the non-polynomial singularities at the origin and the state-dependent asymptotic decay on semi-infinite domains presents a significant computational challenge. In this work, we propose the Adaptive Log-Laguerre Spectral Method (ALLSM), a novel coupled spectral-element solver that seamlessly integrates three advanced mathematical methodologies into a unified framework. Specifically, Generalized Log-Orthogonal Functions (GLOFs) are deployed in the near-core region to intrinsically approximate complex $r^s$ singular behaviors without requiring prior knowledge of the exact analytical exponent $s$. Concurrently, an adaptive Laguerre spectral method is employed to dynamically capture diverse exponential tails on $[0, \infty)$, avoiding artificial domain truncation. To structurally guarantee spectral purity across this bipartite basis, the framework rigorously incorporates the Inverse Dirac Operator Method (IDOM), effectively eliminating variational collapse and spurious states. Validated across diverse physical regimes, including Coulomb, finite-nucleus, and screened potentials, the proposed solver restores exponential convergence and consistently achieves relative accuracies of $10^{-10}$. This work provides a robust, pollution-free computational kernel for atomic structure calculations, establishing a highly reliable numerical standard for complex molecular simulations.
arXiv:2307.06647v4 Announce Type: replace
Abstract: We propose DeepIPCv2, an end-to-end autonomous driving framework that integrates LiDAR-based environmental perception with command-specific control learning. Unlike prior camera-reliant models, DeepIPCv2 employs point cloud segmentation and multi-view projection to construct robust scene representations. These features are fused and decoded through a combination of gated recurrent units, command-specific multi-layer perceptrons, and PID controllers to estimate both waypoints and navigational control commands. This design enhances maneuverability and addresses action imbalance in driving datasets. To validate the model, we constructed a dataset covering diverse illumination conditions and conducted ablation studies and comparative tests against recent methods, including TransFuser. Results demonstrate that DeepIPCv2 achieves the lowest total metric error and the fewest driving interventions, highlighting both its robustness to illumination changes and its improved control accuracy. By releasing the codes at https://github.com/oskarnatan/DeepIPCv2 later, we aim to support reproducibility and future advancements in end-to-end autonomous driving research.
arXiv:2303.04345v2 Announce Type: replace
Abstract: Federated learning (FL) is a promising framework that models distributed machine learning while protecting the privacy of clients. However, FL suffers performance degradation from heterogeneous and limited data. To alleviate the degradation, we present a novel personalized Bayesian FL approach named pFedBayes. By using the trained global distribution from the server as the prior distribution of each client, each client adjusts its own distribution by minimizing the sum of the reconstruction error over its personalized data and the KL divergence with the downloaded global distribution. Then, we propose a sparse personalized Bayesian FL approach named sFedBayes to enhance the inference efficiency. To overcome the extreme heterogeneity in non-i.i.d. data, we propose a clustered Bayesian FL model named cFedbayes by learning different prior distributions for different clients. Theoretical analysis gives the generalization error bound of three approaches and shows that the generalization error rates of the proposed approaches achieve minimax optimality up to a logarithmic factor. Moreover, cFedBayes achieves a cluster-level generalization error bound, rather than a single uniform bound in pFedBayes. Numerous experiments demonstrate that the proposed approaches have better performance than other advanced personalized methods on private models in the presence of heterogeneous and limited data.
arXiv:2507.22714v3 Announce Type: replace
Abstract: The interaction between acoustic waves and turbulent grazing flow over an acoustic liner is investigated using Lattice-Boltzmann Very-Large-Eddy simulations. A single-degree-of-freedom liner with 11 streamwise-aligned cavities is studied in a grazing flow impedance tube. The conditions replicate reference experiments from the Federal University of Santa Catarina. The influence of grazing flow (with a centerline Mach of 0.32), acoustic wave amplitude, frequency, and propagation direction relative to the mean flow is analysed. Impedance is computed using both the in-situ and the mode-matching methods. The in-situ method reveals strong spatial variations; however, averaged values throughout the sample show minimal differences between upstream and downstream propagating waves, in contrast to the mode-matching method. Flow analyses reveal that the orifices displace the flow away from the face sheet, with this effect amplified by acoustic waves and dependent on the wave propagation direction. Consequently, the boundary layer displacement thickness (${\delta}$*) increases along the streamwise direction compared to a smooth wall and exhibits localised humps downstream of each orifice. The growth of ${\delta}$* alters the flow dynamics within the orifices by weakening the shear layer at downstream positions. This influences the acoustic-induced mass flow rate through the orifices, suggesting that acoustic energy is dissipated differently along the liner. The role of near-wall flow features highlights the need to consider a spatially evolving turbulent flow when studying the acoustic-flow interaction and measuring impedance. The spatial development of the turbulent flow may also partly explain the upstream-downstream impedance differences, as current eduction methods do not account for it.
arXiv:2506.22271v3 Announce Type: replace
Abstract: Neural networks often map low-dimensional embeddings to high-dimensional output spaces. Usually, the output layer is linear, which can create a "rank bottleneck" that limits the functions a model can represent. Such bottlenecks are ubiquitous in link prediction models, such as knowledge graph embeddings (KGEs), as the output space of entities can be orders of magnitude larger than the embedding dimension. We investigate how rank bottlenecks limit model expressivity for fitting the training data. While previous work focused on sufficient bounds on the embedding dimension required for specific KGEs, we show necessary bounds for all KGEs with a linear output layer, which grow with graph size and connectivity. We also consider a non-linear output layer using mixtures to break the bottleneck without significant parameter overhead. Empirically, we show that models using this non-linear layer improve in ranking performance and probabilistic fit for large and dense datasets at a low parameter cost, as predicted by our theory. Our work reveals how linear output layers limit KGEs and motivates non-linear alternatives for scaling to large and dense graphs.
arXiv:2606.02553v1 Announce Type: new
Abstract: Autoregressive (AR) video diffusion enables variable-length synthesis, but long-horizon generation often suffers from accumulated errors and identity drift. For efficiency, existing methods commonly adopt sliding-window attention during generation. This creates an irreversible generation trajectory: once the active window accumulates appearance errors, subsequent generations can only condition on this degraded trajectory and drift further away. We address this limitation by formulating long video generation as a retrieval-augmented generation (RAG) problem. Rather than relying solely on the recent window, we treat previously generated latents as a dynamic, searchable history. We propose LongLive-RAG, a general retrieval framework for AR video generation. At each new block, LongLive-RAG uses a query embedding to retrieve relevant historical latents. This lightweight retrieval step adds only a small overhead relative to generation and lets the generator condition on non-local context instead of only the recent window. To make retrieval more discriminative, we introduce the Window Temporal Delta Loss that suppresses redundant local similarity and encourages embeddings to capture meaningful temporal changes. Together, these components help reduce error accumulation caused by sliding-window attention. Experiments across multiple AR backbones and generation lengths show improved long-video quality and the best average VBench-Long rank. To our knowledge, among open-ended AR long video generation methods, LongLive-RAG is the first to formulate self-generated latent history as content-addressable retrieval memory. Code is available at https://github.com/qixinhu11/LongLive-RAG.
arXiv:2603.27446v2 Announce Type: replace
Abstract: This paper investigates the nonlinear dynamics and phase transitions in power packet network connected with routers, conceptualized as macroscopic information-ratchets. In the emerging paradigm of cyber-physical energy systems, the interplay between stochastic energy fluctuations and the thermodynamic cost of control information defines fundamental operational limits. We first formulate the dynamics of a single router using a Langevin framework, incorporating an exponential cost function for information acquisition. Our analysis reveals a discontinuous (first-order) phase transition, where the system adopts a strategic abandon of regulation as noise intensity exceeds a critical threshold $D_c$. This transition represents a fundamental information-barrier inherent to autonomous energy management. Here, we extend this model to network configurations, where multiple routers are linked through diffusive coupling, sharing energy between them. We demonstrate that the network topology and coupling strength significantly extend the bifurcation points, with collective resilient behaviors against local fluctuations. These results provide a rigorous mathematical basis for the design of future complex communication-energy network, suggesting that the stability of proposed systems is governed by the synergistic balance between physical energy flow and the thermodynamics of information exchange. It will serve to design future complex communication-energy networks, including internal energy management for autonomous robots.
arXiv:2604.11424v2 Announce Type: replace
Abstract: Speech Language Models (SLMs) exhibit strong semantic understanding, yet often fail to translate this capacity into expressive acoustic realization, producing speech with flattened prosody and misaligned emotion. We identify this mismatch as the semantic understanding-acoustic realization gap. Existing approaches typically rely on externally specified proxies, such as emotion labels or style prompts, which require annotations and struggle to capture dynamically evolving expressive intent throughout dialogue. To overcome these limitations, we propose SASLM (Self-Aware Speech Language Model), a proxy-free framework that bridges what the model thinks and how it speaks through self-aware intent-realization alignment: (1) Intent-Aware Bridging self-distills expressive intent from the model's own evolving semantic generation states via a Variational Information Bottleneck (VIB), thereby guiding expressive speech realization without external expressive supervision; while (2) Realization-Aware Alignment reflectively aligns generated acoustics with intended expression through self-reward optimization, progressively improving intent-realization consistency during speech generation. Despite using only 3B parameters and 800 hours of expressive speech data, SASLM achieves state-of-the-art performance on EchoMind among open-source systems, surpassing models over 10 times larger and approaching commercial systems.
arXiv:2604.13517v4 Announce Type: replace
Abstract: Temporal credit assignment in reinforcement learning is often approached by introducing value estimates at multiple discount factors. A natural next step is to let the actor dynamically route among these temporal heads, using either differentiable attention or heuristic uncertainty weights. This paper argues that such routing can create a numerical shortcut rather than a reliable temporal abstraction. We study this issue in a controlled PPO setting on LunarLander-v2, using the environment as a visual sandbox for diagnosing failure modes. First, we formalize Surrogate Objective Hacking: a differentiable softmax router exposed to the PPO surrogate receives a direct gradient toward advantage heads that are numerically favorable for the current update, even when this routing change does not correspond to improved physical control. Because unnormalized advantages at different discount factors have different effective scales, this creates a scale-discrepancy vulnerability. Second, we identify the Paradox of Temporal Uncertainty in gradient-free error-based routing: short-horizon heads can receive the largest routing share because their prediction targets are easier, even when they are less aligned with delayed task success. As a structural response, we study Target Decoupling: the critic may retain multi-timescale auxiliary heads, but the actor is updated only with the long-horizon advantage. Target Decoupling is not presented as a broad performance booster; in this run set it removes the exploitable actor-side routing pathway and improves the observed worst-seed return. Code is available at https://github.com/ben-dlwlrma/Representation-Over-Routing.
arXiv:2606.02548v1 Announce Type: new
Abstract: Word Error Rate (WER) is the dominant metric for automatic speech recognition (ASR), but it can overestimate errors when references and hypotheses encode the same words in different scripts. This issue is common in multilingual settings where ASR models may emit romanized text. We propose Script-Normalized WER (SN-WER), a training-free, evaluation-only scoring method that transliterates both reference and hypothesis text into a language-specific canonical script before computing WER. We evaluate SN-WER on 5 Indic languages, 2 datasets, and 3 ASR models. On curated FLEURS data, SN-WER reduces inflated model gaps by up to 12%, while on noisier Common Voice data the reductions are smaller or inconsistent, indicating genuine recognition weaknesses rather than only script mismatch. Controlled stress tests show a 67% attenuation of artificial romanization-induced WER inflation, while lexical-substitution controls show near-identical sensitivity to semantic errors, with Delta SN-WER / Delta WER approximately 1.09. SN-WER is robust to transliterator choice, normalization changes, and shows low token-collision rates below 0.1% in the evaluated Indic setting. We argue that SN-WER should be reported alongside WER and CER as a companion metric for script-insensitive ASR evaluation, especially when transcripts feed downstream search, indexing, or multilingual LLM pipelines.
arXiv:2511.02086v2 Announce Type: replace
Abstract: Purpose: In this paper, we develop and clinically evaluate a depth-only, markerless augmented reality (AR) registration pipeline on a head-mounted display, and assess accuracy across small or low-curvature anatomies in real-life operative settings. Methods: On HoloLens 2, we align Articulated HAnd Tracking (AHAT) depth to Computed Tomography (CT)-derived skin meshes via (i) depth-bias correction, (ii) brief human-in-the-loop initialization, (iii) global and local registration. We validated the surface-tracing error metric by comparing "skin-to-bone" relative distances to CT ground truth on leg and foot models, using an AR-tracked tool. We then performed seven intraoperative target trials (feet x2, ear x3, leg x2) during the initial stage of fibula free-flap harvest and mandibular reconstruction surgery, and collected 500+ data per trial. Results: Preclinical validation showed tight agreement between AR-traced and CT distances (leg: median |Delta d| 0.78 mm, RMSE 0.97 mm; feet: 0.80 mm, 1.20 mm). Clinically, per-point error had a median of 3.9 mm. Median errors by anatomy were 3.2 mm (feet), 4.3 mm (ear), and 5.3 mm (lower leg), with 5 mm coverage 92-95%, 84-90%, and 72-86%, respectively. Feet vs. lower leg differed significantly (Delta median ~1.1 mm; p < 0.001). Conclusion: A depth-only, markerless AR pipeline on HMDs achieved ~3-4 mm median error across feet, ear, and lower leg in live surgical settings without fiducials, approaching typical clinical error thresholds for moderate-risk tasks. Human-guided initialization plus global-to-local registration enabled accurate alignment on small or low-curvature targets, improving the clinical readiness of markerless AR guidance.