Papers

We study deterministic bounded-noise state reconstruction for nonlinear systems when the measurement schedule drives the sampled-output map through a rank-deficient configuration. Local reconstruction is governed by the leading pairwise separation order and its corresponding higher-order modulus, which together determine the leading minimax noise-to-state law. For corank-one cubic singular schedules, including overdetermined sampled maps, we derive the exact physical $\ell_\infty$ cubic modulus through quotient geometry in $\mathbb{R}^q/\operatorname{Im}DF$ and show how the required finite jet is computed from trajectory sensitivities. We then prove a uniform signed timing crossover: a transversal timing perturbation has transition scale $\varepsilon=O(\delta^{2/3})$, linking the regular, cubic-singular, and fold regimes. A damped pendulum provides closed-form crossover coefficients and exact nonlinear minimax validation, while a four-state downward cart–pole verifies the rectangular theory. The results show that measurement timing can change not only reconstruction conditioning but the exponent governing bounded-noise state uncertainty.

Many bug reports lack the necessary information required to reproduce issues. In early studies, large language models have been employed to generate follow-up questions to enhance incomplete reports. However, they ask a fixed number of questions regardless of whether the report is incomplete. We propose BRQGen, an iterative framework that identifies missing or ambiguous information in bug reports and generates targeted follow-up questions, only if necessary. Reports with no identified gaps are classified as sufficient. When the author responds, BRQGen evaluates whether the answers provide the requested information and generates further questions for any exposed gaps. We curated a novel dataset of 429 incomplete bug reports from Home Assistant and Brave Browser, each containing at least one maintainer follow-up question. Each question is mapped to the author's answer and assigned a label indicating whether the requested information could have been provided directly through the issue template. Earlier maintainer requests were associated with author responses, whereas delayed requests were associated with no response. Among responding authors, requests within one hour received median replies after 0.52 hours in Home Assistant and 1.45 hours in Brave, versus 9.76 and 13.49 hours after four hours. BRQGen covered 35.4% (Qwen 3.8 Flash) to 51.8% (DeepSeek V4.1 Flash) of the 1,073 ground-truth follow-up questions. DeepSeek V4.1 Flash managed to ask 45.7% of all detailed questions and 63.9% of all generic questions. Gemini 3 Flash Preview reached the highest macro-F1 for completeness assessment of 49.3%, but with a post-hoc analysis by getting only the gaps with high severities, Qwen 3.8 Flash reaches 73.6% macro-F1 score. Completeness assessment improved when at least one identified gap had high severity, and nearly all ground-truth questions were captured within the first two iterations.

Forensic analysis of low-resolution Word thumbnails presents a significant challenge because these artifacts may constitute the only remaining visual evidence of deleted documents. Standard Optical Character Recognition (OCR) tools perform poorly on such severely degraded images, creating a critical bottleneck in forensic evidence recovery. Although deep learning architectures, such as U-Net, can improve visual image quality, models optimized solely using pixel-based loss functions tend to favor visual smoothness over the structural fidelity required for accurate text recognition. To address this limitation, we propose a text-aware restoration framework designed for forensic triage. The proposed approach employs a U-Net architecture guided by a hybrid optimization strategy that integrates pixel-level fidelity with machine-readable text constraints. Specifically, an OCR-based Normalized Levenshtein Distance (NLD) metric is used during model selection to prioritize the recovery of an accurate character topology. Experimental results demonstrate that the proposed method significantly reduces the NLD, enabling reliable recovery of machine-readable text from severely degraded thumbnails.

We present Gen Through the Lens, a camera-controllable image-to-video generation framework that synthesizes videos from a given reference image while enabling realistic and physically consistent camera operations. Unlike previous models that treat camera motion as abstract conditioning or rely on a single coordinate system, our approach explicitly decomposes and fuses control signals across three coordinate spaces: the camera coordinate (modeled by intrinsic parameters), the world coordinate (represented by spatial geometry hints), and the object coordinate (described by 3D bounding boxes of target subjects). Through cross-coordinate conditioning fusion, these heterogeneous cues are unified to guide the generative process, enabling precise control of camera motion, focus, and framing while preserving temporal and spatial coherence. Extensive experiments demonstrate that our method produces controllable and visually consistent results across diverse camera operations, outperforming existing approaches in both realism and editability.

Large language model serving is increasingly limited not only by model size and GPU speed, but also by scheduling during autoregressive decoding, where continuous batching interleaves requests at token granularity as sequences arrive and finish. We study whether lightweight scheduling policies can improve latency-throughput tradeoffs without changing model architectures, kernels, or deploying learned schedulers. We implement a Python/PyTorch continuous-batching testbench with real Hugging Face execution and evaluate eight policies across 1,152 GPU-model-scheduler-workload runs using ShareGPT-derived traffic, six Qwen, Gemma, and Mistral models, and Modal L40S, A100, and H100 GPUs. Scheduler choice changes throughput, TTFT, completion latency, and tail behavior under fixed model, GPU, workload, batch size, and prefill budget. Aging shortest-prompt-first is the strongest practical baseline in many L40S and mixed-workload settings, while oracle remaining-length scheduling shows that job-size knowledge alone is insufficient because batching and hardware effects also shape performance. The custom dataset has been uploaded to Hugging Face.

Qwen3-8B reviewing Llama-3.1-8B in math adds 33.2 points, yet Qwen3-8B alone is 6.4 points more accurate than the result. Reviewing another model's answer is usually judged by its gain over the generator. We show that this confuses verification, which adds accuracy beyond either model alone, with capability transfer, where a stronger reviewer supplies answers it could have given on its own. Measured against the stronger model alone, across 48 pairs of 12 open-weight models (3B to 72B) in math, knowledge, code and reasoning, review beats it in only 9 of 96 review directions, and does worse the larger the capability gap. The reason is anchoring: the reviewer mostly follows the generator, discarding its own correct answers. Review costs 2.6 to 2.8 times as much as the stronger model alone, while a selector using the same two calls is 2.7 to 4.2 points more accurate than review and cheaper. Studies of verification should report the stronger model alone, not only the gain over the generator.

In sequential learning, a group of agents sequentially receives a noisy private signal and observes the predictions of prior agents to estimate a hidden ground truth. While extensively studied for binary signals, we formalize and analyze sequential learning when the hidden ground truth is an ordering of $n$ items, and both private signals and inferences involve pairwise comparisons. A key focus in sequential learning is herding, where agents disregard their private signals to follow the crowd. We show that, while local herding in our model may lead agents to make incorrect pairwise inferences, the global hidden order can still be approximated. Specifically, by aggregating all pairwise comparison outputs, the highest-ranked items can be retrieved. We investigate top-$k$ retrieval across three scenarios: (1) passive aggregation, where information is gathered from the social learning process without intervention; (2) strategic selection, where agents choose their pairwise comparisons dynamically; and (3) aggregation with stubborn nodes, where a very small set of stubborn agents, who base their inferences solely on their private signals, join the mechanism. With query complexity as the number of agents needed for a high-probability estimation, we propose top-$k$ retrieval algorithms with query complexities sublinear in $O(n^2)$, the total number of pairs.

Large language model (LLM) code-repair systems fix faulty programs by repeatedly generating and testing revisions. Each revision is costly, making the choice of which repair to pursue important for efficiency. Existing search policies rely on test pass rates and search statistics, ignoring the details of test failures and proposed edits. We introduce CodeTriage, a learned policy that prioritizes edits before their revised programs are generated and tested. Its 0.6B-parameter model reads the test output, failure diagnosis, and proposed edit, then predicts a distribution over the number of repair steps needed to reach a fix through that edit. These predictions prioritize all pending edits across programs before generating and testing the revised code. On held-out repair tasks, CodeTriage finds fixes with 21% fewer program revisions and 12% fewer output tokens than the strongest baseline on each metric. Without retraining, it achieves the highest average success across search budgets for each of 11 unseen repair LLMs. Ablations show that both the failure context and edit text, as well as comparisons across programs, improve search. We also release a corpus of 12.2 million program versions from four LLMs, organized into repair trees with proposed edits and test results.

We study eigenpair estimation for the tensorized, density-weighted Laplace–Beltrami operator from $n$ independent data points sampled from a multi-manifold mixture distribution where manifolds may intersect. Let $d$ denote the largest intrinsic dimension of the manifold components and $s\geq2$ the smoothness of the sampling densities. For simple, nontrivial eigenpairs, we establish the minimax rate $n^{-s/(d+2s)}$ for approximating the eigenvalue and eigenfunction of the tensorized Laplace–Beltrami operator, and show that a continuum plug-in estimator based on kernel density estimation attains the same rate when the supporting manifolds are known. We further analyze a manifold-agnostic graph Laplacian estimator and prove that its eigenpairs converge to those of the tensorized Laplace–Beltrami operator at rate $n^{-1/(2.5d-1)}$, up to logarithmic factors, improving upon previously known spectral convergence rates for multi-manifold learning. Numerical experiments on synthetic multi-manifold data corroborate the rates guaranteed by our theorems. These results characterize the information-theoretic difficulty of multi-manifold spectral estimation and quantify the additional cost of learning without oracle component information.

Predictive video representations do not by themselves specify how a multimodal system should handle missing sensors. We transfer frozen V-JEPA2 features to a set of heterogeneous sensors comprised by RGB video, depth video, and radar time series. A shared fusion model uses modality queries, fixed sensor slots, and explicit availability to classify breathing activity and pose from all seven non-empty sensor subsets. The contribution is a controlled evaluation of subset supervision, pooling capacity, and representation transfer on 39 OMuSense-23 subjects with five subject-group folds and three seeds. With matched downstream parameter counts, the final model reaches recording-level activity macro-F1 of $75.23%\pm0.60$ averaged across seven subsets. Supervising all subsets in each batch improves the matched query-pooling control by 1.923 percentage points (conditional 95% subject-bootstrap interval [0.647, 3.209]), while query pooling shows no clear advantage over matched mean pooling under balanced supervision. Radar probes support transfer beyond pooled wavelet pixels, 64-token inference approximately preserves mean activity performance, and the tested Consensus-JEPA loss adds no measured benefit against its exact control.

Reinforcement learning with verifiable rewards spends much of its rollout budget on states that yield no learning signal. We recast this as a question about the sampling distribution itself: under a fixed rollout budget, which distribution over states minimizes the variance of the policy-gradient estimator? The answer admits a closed-form Neyman allocation, $\mu_{\mathrm{var}}^\pi(s) \propto d^\pi(s)\,\sigma_\pi(s)$, which connects naturally to GRPO's group-relative advantage and induces an emergent reverse curriculum as the policy improves, with no externally imposed schedule. Building on this principle, we propose VSR (Variance-driven State Resampling), a plug-in modification to GRPO using a prefix replay buffer and a zero-cost variance proxy. On mathematical reasoning benchmarks and long-horizon agentic tasks (Sokoban, BabyAI), VSR outperforms GRPO and GiGPO on most benchmarks with fewer generated tokens per update.

Workflow scheduling in user device-centric edge-cloud computing has emerged as a challenging problem due to the need to jointly optimize execution performance, energy consumption, and monetary cost. Modern workflows consist of heterogeneous tasks, where compute-intensive components are well-suited for Infrastructure-as-a-Service (IaaS) resources, while fine-grained, event-driven tasks are more efficiently executed using the Function-as-a-Service (FaaS) model. We formulate this scheduling problem for application workflows over a hybrid User Device (UD)-IaaS-FaaS computing continuum as a multi-objective Mixed-Integer Linear Program (MILP). The formulation models the workflow execution mechanism by incorporating cumulative capacity constraints for IaaS resources and elastic resource provisioning for FaaS. To overcome the intractability of MILP, we propose a learning-based framework that combines graph neural networks, surrogate performance modeling, and a multi-sample REINFORCE approach. The proposed approach is workflow-aware, leveraging task pairing to improve data locality and structured task placement decisions. Agent training is organized into three stages: small-scale learning of UD-IaaS trade-offs, intermediate locality aware co-location, and large-scale handling of parallelism and aggregation bottlenecks in workflows. The proposed framework is validated through real-world experiments using the Pegasus Montage workflow on a custom UD-IaaS-FaaS testbed comprising a user device, multiple IaaS nodes, and serverless FaaS instances. Extensive experimental results demonstrate the improved trade-offs across makespan, energy consumption, and cost. Compared with the genetic algorithm and greedy scheduling approaches, the proposed approach reduces makespan by 10-17.5%, energy consumption by 3-9%, and cost by 4-10%, while remaining within 3.5-5.6% of the MILP optimum.

Computational fluid dynamics (CFD) is computationally expensive for iterative building design. Deep learning surrogates offer faster predictions, but generalization across varying room dimensions remains challenging. This study compares two geometry-conditioned neural network surrogate models for rapid indoor airflow and temperature field prediction under varying room sizes. During the architectural design stage, this method is relevant for real-time assessment of whether conditioned supply air will effectively provide fresh and cool air throughout rooms with varying heights and lengths. The canvas-based neural network (CaNN) explicitly encodes room size by preserving different occupied regions within a fixed canvas, whereas the coordinate-based neural network (CoNN) uses normalized coordinate channels. Both employ U-Net autoencoders to separately predict airflow and temperature fields. Evaluation uses simplified rooms with varying heights and lengths as a controlled benchmark. The results show that CaNN achieves higher agreement with CFD results than CoNN, demonstrating its potential to support performance-driven building design.

False retry permission (FRP) occurs when an agent retries a timed-out tool call without support from the visible contract. RetryBench separates this error from realized recovery. Successful protected recovery does not imply reliable rejection of unsupported retries. In a 720-response key control, valid keys usually enabled safe progress, but offered changed keys elicited unsupported retries in Sol (1/80), Terra (55/80), and Luna (60/80). A later 240-response control removed the candidate and required models to supply a key. Luna still made 8/80 unsupported dispatches; Sol and Terra made none in 80 responses each. Accepting a displayed bad key therefore cannot explain all observed errors, although changed output requirements and conditions preclude a pure display-effect comparison. Removing the weak-contract label increased retries in Sol and Terra's coded-action conditions in a six-endpoint study. A separate 1,536-response canonical-action block also found a positive Terra contrast (115/128 versus 124/128 retries), concentrated in two task/order conditions; the other five contrasts remained unresolved after multiplicity correction. An explicit-contract cue control elicited deferral in all 762 returned responses, leaving recovery unresolved. These results support reporting permission and recovery together across declared interface variants. The evidence covers four purposive tasks, fixed GPT endpoints, and one recovery decision; it identifies neither internal mechanisms nor episode-level safety.

Text wrapped in a trigger can be made behaviorally invisible to a language model: omitted from summaries and quotations, ignored in question answering, and left unflagged by downstream filtering. We study this selective blindness as a dual-use information-flow primitive, a supply-chain backdoor, and a white-box auditing problem. We first implement it exactly as a serving-time span veto: blocking information flow out of a marked span makes every outside activation bit-identical under same-length substitution of the span's content, a noninterference property we verify across ten models, seven mixer families, and 0.5B-120B parameters. The veto serves as a deployable quarantine, an exact reference for calibrating leakage, and a teacher. Distilled into weights, it yields a backdoor: checkpoints trained on six model families from 4B to 120B ignore marked context at little benchmark cost, and their outputs show no measurable semantic trace of the suppressed content above the veto reference on our held-out suite; an in-domain variant suppresses planted vulnerabilities during AI code review. Detection is therefore a white-box problem. Simpler backdoors remain internally transparent – linear probes recover suppressed content at 0.57-0.79 accuracy versus 0.025 chance – but hardening the attack until probes reach chance collapses attention from outside the span into it below the range of base models. Across the attacks we evaluate, probe visibility trades off against routing visibility; we do not establish that this tradeoff is fundamental.

A stored fact may stop being valid when a user changes a preference or an event replaces an earlier one. We study whether a language model can learn these changes in its parameters and answer correctly from only a question and the current time. Engram Injection stores facts in LoRA adapters and supplies the clock through FiLM conditioning on a frozen Qwen2.5-3B-Instruct model. We compare it with additive clock conditioning, prompting with the full dated history, and retrieving four dated records. We train six matched adapters across three seeds after fixing a sampler that could admit test-age examples into training. The training audit reports no such examples among 127,760 evaluated training inputs. On 62 pairs of questions asked before and after an update, at an elapsed age of seven days excluded from training, both adapter methods answer zero pairs correctly under the primary blinded assistant review with order code 1 held fixed. Full-history prompting answers 59/62 pairs correctly, and retrieval answers 62/62. Separate experiments show that adapters can store facts and that removable memory units can support local deletion. Combining fact selection with time-based version selection remains unsuccessful. These results separate learning a fact from learning when to use it.

This paper proposes a fully autonomous execution model for multi-GPU applications that completely excludes the involvement of the CPU beyond the initial kernel launch. In a typical multi-GPU application, the host serves as the orchestrator of execution by directly launching kernels, issuing communication calls, and acting as a synchronizer for devices. We argue that this orchestration, or control flow path, causes undue overhead and can be delegated entirely to devices to improve performance in applications that require communication among peers. For the proposed CPU-free execution model, we leverage existing techniques such as persistent kernels, thread block specialization, device-side barriers, and device-initiated communication routines to write fully autonomous multi-GPU code and achieve significantly reduced communication overheads. We demonstrate our proposed model on two broadly used iterative solvers, 2D/3D Jacobi stencil and Conjugate Gradient(CG). Compared to the CPU-controlled baselines, the CPU-free model can improve 3D stencil communication latency by 58.8% and provide a 1.63x speedup for CG on 8 NVIDIA A100 GPUs. The project code is available at https://github.com/ParCoreLab/CPU-Free-model.

We study the approximate channel simulation problem in the normal approximation regime. Our starting point is the fact established by Cao et al.~\cite{Cao_channel_simulation}, that the minimal codebook size $M_\epsilon^*(W)$ for a fixed error tolerance $\epsilon>0$ is closely related to the $\epsilon$-smooth channel max-information, $I_{\max}^\epsilon(W)$. Using this, we refine earlier large-blocklength approximations to $M_\epsilon^*(W^{\otimes n})$ for discrete memoryless channels. Specifically, we provide tight third-order expansions for Gallager-symmetric channels and demonstrate that the third-order term is channel-dependent. As in corresponding channel coding expansions, we identify a substantial rate gap between singular and nonsingular channels. Notably, the direction of this gap is reversed relative to channel coding: The third-order term is $-\frac{1}{2}\log n$ for singular channels but $O(\log\log n)$ for nonsingular channels.

We prove that reachability for constraint automata over finite binary words is undecidable. Eight registers suffice, with conjunctive guards using only equality, proper prefix, proper suffix, and infix incomparability, without word constants or concatenation. Finite-run nonemptiness, Buechi nonemptiness, and bounded Buechi nonemptiness are undecidable as consequences. For Boolean guards over equality, prefix, suffix, and prefix and suffix incomparability, Buechi nonemptiness is PSPACE-complete with arbitrarily many registers and word constants. The upper bound also allows positive infix atoms and yields PSPACE-completeness of prefix–suffix LTL satisfiability.

Currently, there is no common methodology or set of practices for evaluating memory-system performance in integrated simulation infrastructures. As a result, studies adopt different evaluation approaches, which can lead to conflicting conclusions even for the same simulation infrastructure. We propose a validation methodology that evaluates memory-system performance from three complementary views: the memory simulator, the CPU–memory interface and the simulated application. We use these views to detect, understand and correct memory-simulation inaccuracies arising from simulator integration and simplified modeling of microarchitectural features. The methodology reveals that the application's perception of memory performance can become effectively decoupled from the detailed memory simulation. Most importantly, we identify and correct a fundamental memory-model mismatch between delayed-response cycle-accurate memory simulators and ZSim, a representative parallel CPU simulator. We demonstrate the broader applicability of these enhancements by applying them to ZSim integrations with Ramulator 2, DRAMsim3 and DRAMSys 4, covering all major public cycle-accurate memory simulators. After the corrections and enhancements, all integrations reproduce the main bandwidth–latency trends and key characteristics of the actual system. Across the evaluated simulators, STREAM bandwidth differs from actual-hardware measurements by less than 5% in most experiments, while pointer-chase latency remains 15–18% lower than on actual hardware. All simulator enhancements are released as open source.

Open Radio Access Networks (O-RAN) empower operators with choice and customization through their disaggregated and interoperable architecture. However, ensuring interoperability in O-RAN presents challenges due to complex testing environments and evolving specifications. This paper addresses these challenges by proposing the Generative AI and Graph Neural Network-based Open RAN Testing and Automation (GNG-OTA) framework. GNG-OTA utilizes state-of-the-art AI techniques including generative AI, natural language processing, and graph neural networks to enable accurate, reliable, efficient, and automatic O-RAN interoperability and performance testing.

We study stationary discrete-time mean-field games with finite state and action spaces under the average-reward criterion, with an entropy-regularized objective and population-dependent dynamics and rewards. Under a minorization condition, the soft Q-operator is a span-seminorm contraction for the full transition kernel, so relative value iteration does not require the minorizing measure. When the Lipschitz constant of the kernel is below the minorization mass and the temperature exceeds an explicit threshold, the equilibrium fixed-point operator is a contraction: an equilibrium exists, its population distribution is unique, and a fixed-point iteration converges to it geometrically. We then learn the equilibrium model-free from a simulator, combining soft Q-learning at a fixed population distribution with an empirical transition kernel. We prove a finite-sample bound for soft Q-learning in the span seminorm, with temperature-independent constants, and derive convergence rates for the learned population distribution and policy. A congestion-aware routing game illustrates the results.

AI-text detectors are already applied to Kazakh writing, yet no study reports how often publicly available detectors flag Kazakh text whose human authorship is dated before ChatGPT. We present AINA-KK: 3,811 human passages from encyclopedic, news, web and educational pages that provably existed before ChatGPT, and 5,597 machine and gray-zone passages from 9 generators, including the Kazakh-built KazLLM and Sherkala, with script and length stress tests. Five English-trained detectors from the Hugging Face Hub flag up to 81.4% of human passages and 79.6% of published student essays; recalibrated, they catch at most 1.3% of machine text. Calibrated zero-shot detectors catch at most 50.3% of machine text from in-distribution generators. Fine-tuned encoders rank text almost perfectly, but mDeBERTa at its default threshold flags 24.4% of 50-word human excerpts, partly because machine outputs are shorter than human text; calibrated, it catches only 47.1% of KazLLM and 62.7% of Sherkala passages. Without script handling, machine-transliterated Latin-2021 text defeats every detector. Fixes that need no new data (calibration, script canonicalization, script-augmented and length-matched training, and an OR with a zero-shot scorer) form AINA-KK-Det. On test it flags 0.9% of human passages (95% CI 0.4-2.7; 1.5% on development data) and catches 85.2% of machine text from known generators, 92.9% from two unseen Gemini-family Gemma-4 models and 61.3% from the Kazakh-built ones; its configuration was chosen with test results in view. We release the data, code and detector at https://github.com/ozxmn/aina-kk.

We study the behavior of the $k$-th-order nonhomomorphicity of Boolean functions under restriction to affine subspaces. For odd $k\ge3$ and $0\le t<n$, we introduce the $t$-restriction $k$-th-order nonhomomorphicity $P_{k,t}(f)$, defined as the minimum of $P_k(f_{\mid A})$ over affine spaces of codimension at most $t$. For a direct sum of $m$ monomials of degrees at least two, we determine $P_{k,t}$ exactly. When $t<m$, a minimum is obtained by setting one variable equal to zero in each of the $t$ monomials of lowest degree. When $m\le t<n$, the minimum is zero. The proof combines multiplicativity of even Walsh moments with a contraction argument for affine restrictions. We establish a general formula for the Walsh transform of a Boolean function restricted to an affine subspace of arbitrary codimension, yielding an exact spectral characterization of the $t$-restriction normalized nonlinearity $\rnnl_t(f)$. As odd $k$ tends to infinity, $(1-2P_{k,t}(f))^{1/(k+1)}$ converges to $1-2\rnnl_t(f)$. We further prove that every non-affine function has a hyperplane restriction with strictly smaller odd-order nonhomomorphicity. Finally, we determine $\rnnl_1(T_{d,n})$ exactly for all threshold functions with $n\ge3$. This proves the two threshold-function conjectures left open in~\cite{SergeAnaPante2025}, including the previously unresolved odd-majority case, where a minimizing hyperplane is $x_1+x_2=0$.

Collision-resilient airframes allow quadrotors to survive impacts, but do not ensure autonomous post-collision recovery. Innovation gating in stock onboard estimators often rejects valid external velocity measurements after high-speed impacts, compromising recovery. We present REFLEX, a collision-recovery flight mode that runs entirely on the flight controller, requiring neither contact sensors nor a companion computer. Its proprioceptive recurrent policy stabilizes the vehicle without position, velocity, heading, or altitude estimates. The translational estimate is reinitialized while REFLEX is active, and a trust gate checks estimator consistency before returning control to the stock autopilot. The policy is trained in simulation using proximal policy optimization, with episodes initialized from post-impact states measured in 47 real collisions. In hardware experiments, REFLEX recovered 19 of 22 wall impacts at 5.03 $\pm$ 0.09 m/s, compared with 1 of 14 for the matched stock autopilot. These included 15 consecutive recoveries in a single flight. A separate exploratory flight demonstrated two additional recoveries at 6.02 and 6.05 m/s within the same flight, exceeding the previously reported velocity for autonomous wall-collision recovery by 64%.

Over the past decade, Metaverse technology (AR/VR) has seen widespread adoption in numerous areas, including gaming, social networking, and online collaboration. Although recent studies have exposed vulnerabilities in VR systems, they often apply methods (e.g., pre-installed malicious apps or scripts) from conventional mobile platforms directly, demonstrating impractical attacks that fail to investigate unique privacy leakages specific to VR platforms. In this paper, we present an orthogonal attempt to analyze the attack surface of the power side channel on AR/VR's network connection by capturing the through-wall wireless power in the network transmission medium (e.g., Wi-Fi signals) non-intrusively in a comparatively long distance (e.g., 5.0m-8.0m) and further reconstructing multi-level privacy-aware user behaviors. We propose MetaSpy to validate the feasibility and practicality of our disclosed wireless power-based traffic analysis attack. It first captures through-wall wireless power transmitted from the Wi-Fi router of an adjacent room, identifies the device type, then infers (i) user interactions such as browsing websites, running AR/VR apps, and performing various in-app activities, (ii) unseen VR-specific activities in VR games (e.g., Population: ONE), and (iii) avatar-based user re-identification and virtual keystroke inference corresponding to head movements. We evaluated the effectiveness of MetaSpy in four commercial-off-the-shelf (COTS) VR headsets, and the empirical results demonstrate its competitive accuracy in unveiling multi-level user privacy within virtual scenes in through-wall settings at a longer distance (1.8x-80x) than most of the recent studies, and it shows resilience to a series of external interference and impact factors. In addition, we also propose and evaluate effective countermeasures to protect data confidentiality in AR/VR wireless transmission.

Large-scale personalization is shifting from embedding-table retrieval to LLM-based recommenders that generate the Semantic ID (SID) of the item to recommend, where each SID is a short sequence of discrete hierarchical content codes. A single instruction-tuned model can then unify retrieval, content understanding, and language, and recommend brand-new titles from their content alone using SID; then, how should the model generate the SID? Traditionally, models rely on a single language modeling head that generates tokens autoregressively, typically using constrained beam search. In contrast, an emerging body of work explores multi-token prediction (MTP), where dedicated heads predict all tokens in parallel, substantially reducing the decoding latency associated with beam search. We report an industry-scale, apples-to-apples comparison of these two generation paradigms: single-head autoregressive decoding vs. MTP. We introduce a new auxiliary ranking loss, defined over the joint SID score, to train the MTP objective. We evaluate the SID training objectives on both quality and training/inference cost, inside a multi-task LLM that jointly does personalized retrieval, content question-answering (QA), and instruction following over a large streaming-media catalog. Our experiments show single-head beam search consistently beats MTP on most product-critical metrics: ranking and content QA, and trains at roughly 2.6x lower dollar cost on Qwen3-8B scale; MTP wins only on raw inference throughput. Around this finding, we detail an efficient training and inference stack and report end-to-end gains over the previous non-SID baseline equipped with an auxiliary ranking head: +77% item cold-start ranking, +22% content-QA accuracy, +2.8x faster training, with language metric regression traced to SID vocabulary expansion. We address the language regression by reinforcement learning post training.

Elastic buckling remains a critical concern in the design of deep spherical shell caps, given that experiments reveal a significant reduction of critical loads from those predicted by linearized eigen-buckling analysis. This discrepancy has been attributed to initial geometric imperfections and is often reconciled through a dimensionless parameter, namely the knockdown factor (KDF). However, conventional KDFs (NASA SP-8007; Eurocode 3) are highly conservative, as they represent the lower bound of an extensive body of experimental data. This study aims to narrow the conservatism in the KDFs for spherical shell caps through machine learning (ML) regression models trained on experimental data on spherical shell buckling. A hybrid model is proposed by combining Gaussian Process Regression (GPR) with Extreme Gradient Boosting (XGBoost), and its improved performance is demonstrated over conventional ML algorithms, i.e., the Artificial Neural Network (ANN), Support Vector Regressor (SVR), Random Forest Regressor (RFR), and Histogram-based Gradient Boosting (HGBoost). Each of these models uses six input features and a single output, i.e., the KDF. The most critical feature affecting the KDF is ascertained through a sensitivity analysis. The study concludes with a reliability-based design framework that relies on the predictive ML model to obtain safety-consistent, yet less conservative, KDFs.

Belief propagation (BP) is attractive for quantum low-density parity-check (QLDPC) decoding because of its low complexity and low latency, but short cycles and degeneracy can limit its error-correction performance. Reinforcement-learned sequential scheduling (RL-S) substantially improves this performance by adapting the variable node (VN) update order, although Q-table storage and heap-based selection introduce additional overhead. We propose bucket RL-S (B-RL-S), which quantizes learned Q-values into integer priorities and schedules VN updates using dynamically maintained first-in, first-out buckets. Without retraining or changing the BP message updates, B-RL-S reduces policy storage and refreshes only priorities affected by local residual syndrome changes. A fixed number of priority levels enables constant-time VN selection and queue relocation. For the $[[288,12,18]]$ bivariate bicycle code over the depolarizing channel at $p=0.05$, B-RL-S with eight priority levels achieves approximately one fifth the block error rate (BLER) of flooding BP followed by order-10 ordered-statistics decoding. The respective caps are $1000$ sequential iterations for B-RL-S and $1000$ flooding iterations for the BP stage of the reference decoder. These results highlight strong error correction with inexpensive scheduling and no additional post-processing stage.

DiffusionNFT (NFT) fine-tunes diffusion models toward task-specific rewards, but optimizing these rewards does not guarantee fidelity to the full prompt. Although classifier-free guidance (CFG) strengthens prompt alignment, NFT is formulated as a CFG-free method, and applying CFG to the fine-tuned velocity field can reduce the optimized reward. This motivates optimizing rewards over the guided output distribution while seeking to retain its prompt alignment. However, applying standard NFT directly to guided samples can cause reward degradation or training collapse. We trace this failure to a mismatch between the reference velocity used in the NFT loss and the score of the guided sample distribution, which biases the improvement direction. We propose GuidanceNFT, which estimates the matching reference velocity through online flow matching on guided samples and uses it to correct both regression terms. With an exact reference, the corrected objective recovers NFT's intended update. In our experiments, the learned correction sustains reward improvement under both explicit and built-in guidance. On SD3.5-M, GuidanceNFT achieves higher optimized rewards than NFT with and without CFG across GenEval, text-rendering, and multi-reward objectives, together with better overall prompt alignment as measured by off-domain evaluation metrics. Experiments on models with built-in guidance, including Model Guidance and ELF, further demonstrate improved ImageNet generation, a better perplexity-entropy tradeoff on OpenWebText, and higher accuracy on GSM8K.