arXiv:2607.00274v1 Announce Type: new Abstract: Effective writing feedback is among the strongest drivers of student learning, yet producing it at scale is labor-intensive. LLMs offer a natural path to scaling writing support, but two gaps stand in the way: few public corpora capture how instructors actually deliver feedback in real classrooms, and no reliable method measures whether generated feedback aligns with what an instructor would write. We address both. SEFORA is a public corpus pairing instructor inline feedback with assignment prompts, rubrics, scores, and multi-draft revisions across various college writing genres, comprising 564 drafts and 8,240 instructor annotations. UniMatch is a reference-based evaluation framework for open-ended generation: it segments feedback into feedback units, scores their semantic correspondence under instructor-derived criteria, and aligns them via optimal matching to yield interpretable precision, recall, and F1. Across 74 experimental configurations spanning multiple LLMs, no setting exceeds 0.4 F1. UniMatch reveals that models struggle to identify the feedback instructors would prioritize, and performance degrades as models generate more.
Science Journals
arXiv:2607.00403v1 Announce Type: new Abstract: Large language models are increasingly used by participants on crowdsourcing platforms when responding to surveys, potentially undermining the validity of collected data. Our study aims to quantify the prevalence of this behavior and investigate methods to detect and prevent it. In a series of surveys (N = 250), we examined conditions such as platform choice, survey length, requests not to use AI, and disabling copy-paste functionality. We were able to identify distinct characteristics of LLM-assisted responses and found that their frequency varied widely, from under 10% on Prolific to over 80% on Mechanical Turk. Mitigation measures reduced LLM usage but did not necessarily improve data quality. No participants employed browser-use agents at the time of our survey, but we report on our own detection experiments. We recommend that researchers actively screen survey responses for LLM usage by recording and analyzing keystroke data and crafting instructions and questions aimed at AI.
arXiv:2607.00415v1 Announce Type: new Abstract: Authority bias poses a critical safety concern in language models: models systematically prioritize social cues from authority figures over factual consistency, swaying their answers based on source credibility rather than evidence. We mechanistically investigate this phenomenon using a controlled medical QA setting, where hints suggesting incorrect answers are attributed to personas of varying expertise. Across Llama-3.1-8B, Qwen3-8B, and Gemma-2-9B, we find that models respond in a graded manner proportional to perceived authority, a hierarchy that is never explicitly prompted but emerges from training. Logit lens analysis and linear/non-linear probing localize this effect to a critical late layer where correct answer representations are actively erased, an erasure that scales with authority level, resists mean vector intervention, and is only partially reversible through chain-of-thought reasoning. Our findings suggest that authority-induced sycophancy is not a surface-level output bias but mechanistic knowledge erasure, a precise, layer-localized overwriting of correct internal representations by high-status authority signals.
arXiv:2511.06160v2 Announce Type: replace Abstract: While recent safety guardrails effectively suppress overtly biased outputs, subtler forms of social bias emerge during complex logical reasoning tasks that evade current evaluation benchmarks. To fill this gap, we introduce a new evaluation framework, PRIME (Puzzle Reasoning for Implicit Biases in Model Evaluation), that uses logic grid puzzles to systematically probe the influence of social stereotypes on logical reasoning and decision making in LLMs. Our use of logic puzzles enables automatic generation and verification, as well as variability in complexity and biased settings. PRIME includes stereotypical, anti-stereotypical, and neutral puzzle variants generated from a shared puzzle structure, allowing for controlled and fine-grained comparisons. We evaluate multiple model families across puzzle sizes and test the effectiveness of prompt-based mitigation strategies. Focusing our experiments on gender stereotypes, our findings highlight that models consistently reason more accurately when solutions align with stereotypical associations. This demonstrates the significance of PRIME for diagnosing and quantifying social biases perpetuated in the deductive reasoning of LLMs, where fairness is critical.
arXiv:2511.06236v2 Announce Type: replace Abstract: In this paper, we study the Schr\"odinger equation with a Gaussian random potential (SE-GP) and develop an efficient numerical method to approximate the expectation of physical observables. The unboundedness of Gaussian random variables poses significant difficulties in both sampling and error analysis. Under time-splitting discretizations of SE-GP, we establish the regularity of the semi-discrete solution in the random space. Then, we introduce a non-standard weighted Sobolev space with properly chosen weight functions, and obtain a randomly shifted lattice-based quasi-Monte Carlo (QMC) quadrature rule for efficient sampling. This approach leads to a QMC time-splitting (QMC-TS) scheme for solving the SE-GP. We prove that the proposed QMC-TS method achieves a dimension-independent convergence rate that is almost linear with respect to the number of QMC samples. Numerical experiments illustrate the sharpness of the error estimate.
arXiv:2607.00502v1 Announce Type: new Abstract: While long-horizon mobile GUI agents typically rely on thought-action-observation loops, they struggle to separate persistent task states from transient screen observations. As execution histories grow, this entanglement imposes a severe context burden, causing agents to forget initial requirements, hallucinate progress, or repeatedly interact with stale interfaces. To address this, we introduce Task-State Representation (TSR), a training-free framework that explicitly decouples task state from sensory input. Acting as a lightweight external wrapper, TSR maintains three structured components: a global instruction summary, a dynamic progress tracker for subgoals, and a transition-aware action verifier. By continuously updating through pre- and post-action visual comparisons, TSR effectively guides the agent's reasoning without requiring architectural modifications. Experiments across four mobile GUI benchmarks validate TSR's effectiveness, yielding up to a 12 absolute point increase in success rate on complex cross-application and memory-intensive tasks.
arXiv:2607.00616v1 Announce Type: new Abstract: The fundamental limits of information flow in spatial networks are usually characterized under stationary spatial point processes, but this assumption cannot capture non-stationary regimes where the node intensity field evolves continuously in space and time. This paper develops Fluid-Spatiotemporal Stochastic Geometry (F-STSG), treating dynamic network topology as a hydrodynamic limit of the discrete node constellation. We formulate the identification of latent network dynamics as an inverse boundary value problem and, using the minimum kinetic energy principle from optimal transport, establish the existence and uniqueness of a scalar potential field governing the compressive evolution of network load. The resulting field-theoretic formulation couples continuous Lagrangian transport with discrete Eulerian interference geometry. Based on this model, we derive the information flux vector as a sufficient statistic for macroscopic advection and the material derivative as a kinematic predictor of topological divergence. We further characterize non-stationary network limits through energy-density scaling and source-channel interpretation, showing how coordination overhead, topology deformation, and control signaling requirements are linked to the kinematic entropy of the evolving network topology.
arXiv:2607.00687v1 Announce Type: new Abstract: Comparing transformer backbones for image segmentation is confounded: each is paired with a different decoder, recipe, and pretraining, so reported differences rarely reflect the backbone itself. We introduce the Lightweight Universal Mask Adapter (LUMA), a lightweight, backbone-agnostic mask-transformer head that treats any backbone as a black-box feature extractor, letting a set of queries read from its features through cheap cross-attention. LUMA matches the accuracy of EoMT, the state-of-the-art efficient ViT-segmenter, at lower cost, while attaching unchanged to isotropic, hierarchical, convolutional, and mixture-of-experts backbones alike. Holding this head fixed, we benchmark 20 backbones, 11 pretraining schemes and a range of resolutions on ADE20K and Cityscapes under one modern recipe. We find that ``efficient'' token mixers fail to deliver efficiency even at the high resolutions that motivate them, with plain ViT holding the throughput Pareto-front at every resolution. Additionally, the pretraining objective, not the architecture, the lever the field has tuned hardest, governs segmentation quality.
arXiv:2607.00796v1 Announce Type: new Abstract: Visual Reinforcement Learning (VRL) has achieved considerable success in solving control tasks. However, generalizing learned policies to new environments remains a major challenge, as agents often overfit to task-irrelevant features in the training environment. To solve this problem, we introduce the concept of decoupling observations into task-relevant and task-irrelevant representations. Building on this idea, we propose a self-supervised Task-Relevant Representation Decoupling (T2RD) algorithm for VRL. This algorithm consists of three components: task-relevant representation consistency, cross-reconstruction, and cross-dynamic prediction. The first two components achieve the decoupling of content and style features, but the resulting content representations are not necessarily task-relevant. To further refine task-relevant features from content representations, we design the third component that introduces dynamic prediction. T2RD achieves State-Of-The-Art (SOTA) generalization performance and sample efficiency in the DeepMind Control Suite and Robotic Manipulation tasks.
arXiv:2511.06701v3 Announce Type: replace Abstract: AI-Scientist systems risk manufacturing spurious discoveries through uncontrolled multiple testing. We present a functional architecture that enforces statistical rigor at two levels: a Haskell embedded domain-specific language (the Research monad) that makes it impossible to test a hypothesis without updating the error budget, and a declarative scaffold, backed by an OS-level sandbox, that makes validation data physically absent from the environment in which LLM-generated code runs. We ground the design in a machine-checked Lean~4 formalization of LORD++ online false-discovery-rate (FDR) control: we derive its error budget and prove both marginal and full FDR control, then close the gap to the implementation by verifying the budget's wealth invariant over IEEE~754 arithmetic in SPARK/Ada. To our knowledge this is the first verified chain from theorem to floating-point implementation for an online FDR procedure. In simulation, the architecture holds the false discovery rate near 1\% against a 5\% target, where a naive approach reaches 41\%. In end-to-end case studies, a valid test avoids the false discoveries a flawed one produces, yet still finds real effects when the data allow. An adversarial evaluation confirms that generated code cannot read the held-out data even when given its exact path.
arXiv:2511.07397v3 Announce Type: replace Abstract: Voice agents face a fundamental tension: the reasoning, retrieval, and tool use that make foundation models capable are iterative and slow, while conversational interaction demands responses on a millisecond timescale. Smaller, real-time models meet the latency bar but cannot match foundation models on complex tasks, leaving current voice agents to trade away either responsiveness or capability. We introduce conversational infill, where a small talker model both immediately generates contextually grounded responses to hide the latency of an external reasoner model and fluently integrates streamed reasoner knowledge into its responses during inference. We curate a 290,571-example synthetic dataset spanning six domains and demonstrate that this task is learnable across seven widely used small language models ranging from 135M to 1.7B parameters. Our system implementation, ConvFill, sustains millisecond-level time-to-first-response while closing the accuracy gap to within 6.3% of the corresponding frontier reasoner performance. In a live user study (n=18) with talker deployments running on an Apple M2 SoC, participants rank ConvFill on par with frontier models overall, prefer it for retrieval-heavy tasks, and rate it significantly more responsive. These results show that conversational infill unlocks a new point on the latency-capability Pareto frontier, offering a practical path toward voice agents that are both responsive and highly capable. Code, models, and datasets are available at https://github.com/vysri/conversational-infill.
arXiv:2511.12110v5 Announce Type: replace Abstract: Despite notable progress in text-guided medical image segmentation nowadays, these methods are limited to single-round dialogues and fail to support multi-round reasoning, which is important for medical education scenarios. In this work, we introduce Multi-Round Entity-Level Medical Reasoning Segmentation (MEMR-Seg), a new task that requires generating segmentation masks through multi-round queries with entity-level reasoning, helping learners progressively develop their understanding of medical knowledge. To support this task, we construct MR-MedSeg, a large-scale dataset of 177K multi-round medical segmentation dialogues, featuring entity-based reasoning across rounds. Furthermore, we propose MediRound, an effective baseline model designed for multi-round medical reasoning segmentation. To mitigate the inherent error propagation within the chain-like pipeline of multi-round segmentation, we introduce a lightweight yet effective Judgment & Correction Mechanism during model inference. Experimental results demonstrate that our method effectively addresses the MEMR-Seg task and outperforms conventional medical referring segmentation methods. The project is available at https://github.com/Edisonhimself/MediRound.
arXiv:2512.11173v4 Announce Type: replace Abstract: Achieving precise positioning of the mobile manipulator's base is essential for successful manipulation actions that follow. Most of the RGB-based navigation systems only guarantee coarse, meter-level accuracy, making them less suitable for the precise positioning phase of mobile manipulation. This gap prevents manipulation policies from operating within the distribution of their training demonstrations, resulting in frequent execution failures. We address this gap by introducing an object-centric imitation learning framework for last-meter navigation, enabling a quadruped mobile manipulator robot to achieve manipulation-ready positioning using only RGB observations from its onboard cameras. Our method conditions the navigation policy on three inputs: goal images, multi-view RGB observations from the onboard cameras, and a text prompt specifying the target object. A language-driven segmentation module and a spatial score-matrix decoder then supply explicit object grounding and relative pose reasoning. Using real-world data from a single object instance within a category, the system generalizes to unseen object instances across diverse environments with challenging lighting and background conditions. To comprehensively evaluate this, we introduce two metrics: an edge-alignment metric, which uses ground truth orientation, and an object-alignment metric, which evaluates how well the robot visually faces the target. Under these metrics, our policy achieves 74.58% success in edge-alignment and 89.42% success in object-alignment when positioning relative to unseen target objects. These results show that precise last-meter navigation can be achieved at a category-level without depth, LiDAR, or map priors, enabling a scalable pathway toward unified mobile manipulation. Project page: https://rpm-lab-umn.github.io/category-level-last-meter-nav/
arXiv:2512.18303v2 Announce Type: replace Abstract: Despite technical advancements, the human factor remains cybersecurity's most exploited vulnerability. Current research acknowledges this but remains fragmented, treating vulnerabilities as isolated, static traits. To address this, we introduce MORPHEUS, a holistic framework conceptualizing human-centric security as a dynamic, interconnected system. Grounded in the Cognition-Affect-Behavior (CAB) model and Attribution Theory, MORPHEUS consolidates 50 human factors influencing susceptibility to major cyberthreats (e.g., phishing, malware, password management, and misconfigurations). Beyond mere identification, the framework introduces a hierarchical Causal Pathway Architecture. Systematically mapping 302 empirical interactions (82.8% architecture-compliant), we reveal how cognitive, affective, and behavioral processes jointly shape security outcomes, distilling them into 12 recurring interaction mechanisms. MORPHEUS further links theory to practice through an inventory of 99 validated psychometric instruments for empirical assessment. We illustrate its applicability through in-depth operational scenarios for risk diagnosis and targeted interventions. Overall, MORPHEUS provides a comprehensive theoretical foundation for advancing human-centered cybersecurity.
arXiv:2601.11641v3 Announce Type: replace Abstract: While Diffusion Transformers (DiTs) have achieved notable progress in video generation, this long-sequence generation task remains constrained by the quadratic complexity inherent to self-attention mechanisms, creating significant barriers to practical deployment. Although sparse attention methods attempt to address this challenge, existing approaches either rely on oversimplified static patterns or require computationally expensive sampling operations to achieve dynamic sparsity, resulting in inaccurate pattern predictions and degraded generation quality. To overcome these limitations, we propose a \underline{\textbf{M}}ixture-\underline{\textbf{O}}f-\underline{\textbf{D}}istribution \textbf{DiT} (\textbf{MOD-DiT}), a novel sampling-free dynamic attention framework that accurately models evolving attention patterns through a two-stage process. First, MOD-DiT leverages prior information from early denoising steps and adopts a {distributed mixing approach} to model an efficient linear approximation model, which is then used to predict mask patterns for a specific denoising interval. Second, an online block masking strategy dynamically applies these predicted masks while maintaining historical sparsity information, eliminating the need for repetitive sampling operations. Extensive evaluations demonstrate consistent acceleration and quality improvements across multiple benchmarks and model architectures, validating MOD-DiT's effectiveness for efficient, high-quality video generation while overcoming the computational limitations of traditional sparse attention approaches.
arXiv:2607.01110v1 Announce Type: new Abstract: We study the expressive power of the simplest modal-temporal language, obtained by adding Prior's temporal operators \(G\) and \(H\) to the basic modal language with \(\Box\). This language is the standard bimodal combination of modal and tense logic; under its functional interpretation it is denoted \(L_{T\times W}\). To analyse its definability across five order types, we consider two semantic readings of the temporal operators: the standard reading (\(G,H\)), which includes the current instant, and the strict reading (\(G^{\ast},H^{\ast}\)), which always excludes it. We examine nine functional properties -- totality, non-totality, injectivity, surjectivity, monotonicity, strict monotonicity, antitonicity, strict antitonicity, and constancy -- over preorders, strict preorders, partial orders, linear orders, and strict linear orders. Our analysis reveals two different levels of expressive power. In the original multiflow setting, the language is quite weak and the two readings coincide. When we restrict the semantics to minimal functional frames (the \(O^{2}\) family), many properties become definable, and the choice of reading becomes crucial: the strict reading can define properties such as injectivity even in reflexive orders. The same definability patterns appear with indexed languages and with the Uniform Domain condition on the semantics of \(L_{T\times W}\). That three such different ways of controlling functional multiplicity lead to identical definability patterns indicates that the expressive limitations of the original framework come from the uncontrolled multiplicity of functions, not from any weakness of the operators. Even after controlling functional multiplicity, a set of properties remains undefinable in all non-linear orders, showing that the lack of connectivity is a fundamental obstacle.
arXiv:2601.14232v2 Announce Type: replace Abstract: Pixel-based reinforcement learning agents often fail under purely visual distribution shift even when latent dynamics and rewards are unchanged, but existing benchmarks entangle multiple sources of shift and hinder systematic analysis. We introduce KAGE-Env, a JAX-native 2D platformer that factorizes the observation process into independently controllable visual axes while keeping the underlying control problem fixed. By construction, varying a visual axis affects performance only through the induced state-conditional action distribution of a pixel policy, providing a clean abstraction for visual generalization. Building on this environment, we define KAGE-Bench, a benchmark of six known-axis suites comprising 34 train-evaluation configuration pairs that isolate individual visual shifts. Using a standard PPO-CNN baseline, we observe strong axis-dependent failures, with background and photometric shifts often collapsing success, while agent-appearance shifts are comparatively benign. Several shifts preserve forward motion while breaking task completion, showing that return alone can obscure generalization failures. Finally, the fully vectorized JAX implementation enables up to 33M environment steps per second on a single GPU, enabling fast and reproducible sweeps over visual factors. Code: https://avanturist322.github.io/KAGEBench/.
arXiv:2601.14660v2 Announce Type: replace Abstract: Agentic Large Language Models (LLMs) are models able to reason, plan, and execute tools over unstructured data. These abilities are enabling transformative applications in domains spanning from personal assistant, financial, and legal domains. While these systems can substantially improve productivity and service quality, effective agency typically requires access to sensitive personal or organizational information. However, this access introduces critical inference-time privacy risks, specifically regarding contextually appropriate information disclosure. While recent studies highlight the inability of agentic LLMs to consistently adhere to privacy norms, existing defenses often rely on auxiliary LLM-based monitors. However, these defenses are expensive and offer limited protection against attacks that are robust to semantic censorship. To contrast this background, this paper proposes a notion of privacy filters based on activation probing. We show that these filters are both computationally efficient and effective for both single-turn and multi-turn conversational settings. Furthermore, this work provides the first systematic investigation into probing model internals across a conversation trajectory, moving beyond static, single-prompt analysis to capture the evolving state of privacy-sensitive interactions.
arXiv:2601.17020v2 Announce Type: replace Abstract: With the rising popularity of interdisciplinary work and increasing institutional incentives in this direction, there is a growing need to understand how resulting publications incorporate ideas from multiple disciplines. Existing computational approaches, such as affiliation diversity, keywords, and citation patterns, do not account for how individual citations are used to advance the citing work. Although, in line with addressing this gap, prior studies have proposed taxonomies to classify citation purpose, these frameworks are not well-suited to interdisciplinary research and do not provide quantitative measures of citation engagement quality. To address these limitations, we propose a framework for the evaluation of citation engagement in interdisciplinary Natural Language Processing (NLP) publications. Our approach introduces a citation purpose taxonomy tailored to interdisciplinary work, supported by an annotation study. We demonstrate the utility of this framework through a thorough analysis of publications at the intersection of NLP and Computational Social Science.
arXiv:2601.17508v3 Announce Type: replace Abstract: We consider the permanent of a square matrix with non-negative entries. A tractable approximation is given by the so-called Bethe permanent that can be efficiently computed by running the sum-product algorithm on a suitable factor graph. While the ratio of the permanent of a matrix to its Bethe permanent is, in the worst case, upper and lower bounded by expressions that are exponentially far apart in the matrix size, in practice it is observed for many ensembles of matrices of interest that this ratio is strongly concentrated around some value that depends only on the matrix size. In this paper, for an ensemble of block-structured matrices where entries in a block take the same value, we numerically study the ratio of the permanent of a matrix to its Bethe permanent. It is observed that also for this ensemble the ratio is strongly concentrated around some value depending only on a few key parameters of the ensemble. We use graph-cover-based approaches to explain the reasons for this behavior and to quantify the observed value.
arXiv:2607.00844v1 Announce Type: cross Abstract: We consider the problem of designing input signals for an unknown linear time-invariant system in such a way that the resulting data, within a finite horizon, is suitable for identification with a desired accuracy. We consider both noise-free and noisy settings with $\ell_\infty$--bounded noise models. We will take into account general prior knowledge of the system parameters. Central in our study is the concept of universal inputs. An input is called universal for identification if, when applied to any system complying with the prior knowledge, it yields data suitable for accurate identification. We provide new methods for designing such universal inputs. Our results generalize the experiment design approach based on Willems et al.'s fundamental lemma that relies on persistently exciting inputs, and that is limited to prior knowledge on controllability. It turns out that for other types of prior knowledge, there exist universal inputs that outperform the persistently exciting ones, e.g., in terms of sample efficiency. Moreover, we investigate types of prior knowledge that enable experiment design for exact identification in the presence of noise.
arXiv:2607.00113v1 Announce Type: new Abstract: Background. Labeled data for security classification is scarce. Semi-supervised learning (SSL) propagates labels from a small labeled pool to larger unlabeled pools. Yet security applications often use SSL as a black box: default parameters, a fixed classifier, and no handling of pseudo-label-induced class imbalance. Aims. Recent work reports sizeable gains from optimizing SSL pipelines via joint search, AutoML, or per-component tuning. These gains are hard to attribute: they may reflect useful SSL-classifier interactions, or mostly from simply tuning the downstream classifier. We disentangle these effects for binary tabular security data with classical SSL and tree-based classifiers. Method. We build SemiScope as an analysis instrument, not a deployment recommendation. It uses Bayesian Optimization to jointly tune SSL settings, confidence filtering, oversampling, and the classifier. The key control, Tuned-Clf, fixes SSL to defaults but gets the same 100-trial classifier budget and validation-set threshold tuning as SemiScope. At 10% labels, we compare them with paired TOST using a +/-1.0 g-measure smallest effect of interest. Results. SemiScope beats every default SSL baseline on all five datasets, improving over the strongest by 0.7-12.7 points. Under the equal-budget control, Tuned-Clf is statistically equivalent to the full pipeline on 4 of 5 datasets; Phishing is inconclusive. Classifier HPO alone recovers a median 86% of SemiScope's gain over Default Self-Training (ST) + Random Forest (RF). Conclusions. The reusable contribution is the decomposition protocol. A simpler recipe suffices: use Self-Training, tune the classifier with Bayesian Optimization, and tune the decision threshold on validation data. It reaches within 1 g-measure of Supervised RF at 20-30% labels on four datasets and 40% on Drebin, at the same or lower label rate than Default ST + RF on every dataset.
arXiv:2607.00841v1 Announce Type: new Abstract: We present a real-time hardware implementation of a versatile detector emulator capable of reproducing realistic silicon photomultiplier signals. Our approach builds upon the open-source SimSiPM framework, originally developed to simulate the microscopic response of silicon photomultipliers, including photon detection efficiency, optical crosstalk, afterpulsing, and dark counts. SimSiPM provides idealized photon-level data with arbitrary temporal and amplitude resolution. In contrast, our emulator, built on a field-programmable gate array, translates this fine-grained simulation into physically realizable analog signals, maintaining real-time operation and finite hardware resolution. The system receives simulated photon events either via a 10-gigabit Ethernet stream or directly from the processing system of a system-on-chip, and performs on-chip temporal quantization, dividing time into bins equal to one clock cycle. All photon hits within a bin are accumulated, and their contribution is combined through a weighted temporal averaging scheme that preserves sub-bin precision. Signal shaping is executed entirely in hardware, using parallel one-pole recursive filters that synthesize the rise and the two decay components of the response. The resulting waveform is converted to analog through dual 16-bit digital-to-analog converters operating at 2.5 gigasamples per second. This architecture generates physically accurate detector signals in real time, rather than replaying precomputed waveforms. It also generalizes beyond silicon photomultipliers, providing a flexible framework for hardware-in-the-loop testing of front-end electronics. The proposed implementation demonstrates high throughput, low latency, and minimal processor overhead.
arXiv:2607.00978v1 Announce Type: new Abstract: Privacy-preserving perception is a critical requirement for deploying 3D scene understanding systems in real-world indoor environments, yet it remains underexplored in open-vocabulary 3D semantic segmentation. Existing methods typically rely on obtaining rich semantic cues from RGB images, which may expose privacy-sensitive visual information. Depth-only 3D geometry provides a privacy-preserving alternative, but the absence of appearance-based semantic cues makes open-vocabulary predictions highly uncertain and less reliable. Under this setting, we propose to convert uncertainty into a guidance signal to identify unreliable semantic responses and use semantic priors from foundation models to regularize their refinement. We present UTTO, an uncertainty-guided test-time optimization framework for depth-only open-vocabulary 3D semantic segmentation. Without additional training, experiments on ScanNet20, ScanNet40, and ScanNet200 demonstrate that UTTO consistently improves depth-only open-vocabulary 3D segmentation and outperforms representative baselines under privacy-preserving conditions.
arXiv:2601.20595v4 Announce Type: replace Abstract: Communication has become a first-order bottleneck in large-scale GPU workloads, and existing distributed compilers address it mainly by overlapping whole compute and communication kernels at the stream level. This coarse granularity incurs extra kernel launches, forces device-wide synchronizations at kernel boundaries, and leaves substantial slack when the slowest tile or kernel stretches the communication tail. We present Syncopate, a compiler and runtime that enables automatic fine-grained overlap inside a single fused kernel. Syncopate introduces a communication chunk abstraction that decouples communication granularity from kernel structure and backend mechanisms, allowing chunk-level plans to be ported from existing distributed compilers, written directly by users, or instantiated from reusable templates. Given a local Triton kernel and a chunk schedule, Syncopate performs transformations to align computation with chunk availability. Implemented as a source-to-source compiler on Triton, Syncopate delivers an average end-to-end speedup of 1.3$\times$ and up to 4.7$\times$ on multi-GPU workloads. Our code is open-sourced at https://github.com/tie-pilot-qxw/syncopate.