Science & Mathematics

Autonomous industrial inspection requires strictly deterministic real-time inference under tight memory and computational constraints. However, existing Video Instance Segmentation (VIS) models suffer from prohibitive scaling: dense temporal memory grows with spatial resolution and sequence length, while fixed-capacity decoders fail to exploit the inherent sparsity of industrial scenes, creating a practical memory wall on resource-constrained edge platforms. We present DetM-Tracker, a streamlined VIS framework for autonomous industrial inspection, where target objects often exhibit highly similar appearance under constrained "micro" fields of view. DetM-Tracker eliminates architectural redundancy through three synergistic components: (1) Compact Scalar Memory, which condenses high-dimensional instance features into four-scalar descriptors, reducing temporal memory complexity from $O(T \cdot N \cdot R)$ to $O(T \cdot \hat{N})$. This mechanism effectively compresses frame-level representations into a bank of active instances, where $R$ denotes resolution, $N$ and $\hat{N}$ represent global and window-specific instance counts, respectively and $T$ the temporal window size; and (2) Dynamic Instance Channels, which adaptively allocate query capacity according to the actual number of visible objects; and (3) Entropy-Gated Skip Connection, which concentrates computational budget on high-uncertainty spatial regions to refine object boundaries in a single forward pass without auxiliary refinement heads. Our approach effectively bridges the gap between modern VIS models and the practical demands of industrial edge deployment, reducing inference latency by approximately 20$\times$ and VRAM consumption by 6$\times$ compared to state-of-the-art dense-feature baselines.

Satellite precipitation products face a persistent trade-off: real-time estimates are timely but less accurate, whereas the most reliable products are often released months after observation. This delay keeps high-quality precipitation information outside the decision window for flood warning and disaster response. Here we report Huayu-Global, a cross-constellation deep-learning system that combines three geostationary sensor streams from five satellites through a common retrieval framework and generates coherent quasi-global precipitation fields every thirty minutes. Across more than a million independent station records from two periods outside the training years, Huayu-Global improves event detection and precipitation-amount estimation relative to the real-time products evaluated and reaches delayed-product-level skill in key metrics. Huayu-Global therefore turns operational geostationary observations into a timely, high-skill precipitation layer for flood forecasting, water management and disaster response. Its cross-constellation design also provides a scalable foundation for incorporating future geostationary missions into hydrological and weather-prediction systems.

Personalized shopping benchmarks ask which item fits the user; we ask how far an agent may commit on that item, and pose learning the individual's auto-purchase line as a task. The boundary is evidence-conditioned and three-valued: whether committing is appropriate depends on the sufficiency of product information (warrant), reversibility, price against the user's norm, and urgency, not only on fit; and the natural output is don't buy, confirm first, or buy automatically. Across sixteen instruction-tuned models, autonomous commitment is misaligned with a human authorization reference in which fifteen annotators, role-playing the delegating principal from a given profile, choose the furthest action they would allow without confirmation: models commit even when fit is low and when their own evidence flag reports the warrant missing. This is overreach. The line is individual: these raters disagree, yet the disagreement is structured and partly predictable from a person's own other labels. A shallow learner on a few dozen of a person's own labels predicts their held-out auto-purchase line at a median of 0.79 by 30-40 labels, outperforming every evaluated prompt$\times$model configuration, including a per-annotator oracle over 80 configurations; the learned boundary does not transfer across domains. We release the counterfactual benchmark, human authorization reference, and annotation protocol after acceptance.

Decision models return normalized probability distributions over candidate choices, which downstream applications can use to select among choices. Although their providers emphasize the reliability of these probabilities, calibration on tasks with a single correct choice does not establish their reliability when multiple choices are correct. We evaluate three decision models (Jev, Liquid d1, and GPT-6 Luna Decisions) on three diverse datasets (MetaTool, HotpotQA, and ECtHR) to study probability allocation and thresholding. Even when models rank every correct choice above every distractor, in 28% to 50% of those lists, at least one correct choice receives a probability below the uniform probability of 1/n, where n is the number of choices. We also present each model with five identical copies of a correct choice and observe probability disparities, with mean gaps ranging from 17 to 79 percentage points. We further show how these allocation disparities affect the recovery of multiple correct choices under probability thresholding. Decision-model probabilities can rank correct choices, but a probability threshold does not identify the set of correct choices.

Personalized reward models require user-specific preference comparisons, but collecting such feedback is costly and provides only limited supervision for each user. At the same time, large pools of responses can be generated without user identities or preference labels, yet existing personalized reward models leave this data unused. Directly learning from these responses is difficult because pairwise objectives constrain relative scores within a prompt, whereas pseudo-labeling individual responses requires scores that are comparable across prompts for the same user. We propose \textbf{PerSPL} (Personalized Semi-supervised Preference Learning), a backbone-agnostic framework that uses pair-mean score centering to stabilize this cross-prompt score coordinate and iteratively turns unlabeled responses into user-specific training signals. Against seven representative baselines across four benchmarks---PersonalLLM, HelpSteer2, Reddit TLDR, and Chatbot Arena---PerSPL achieves the highest reported mean accuracy among the evaluated methods for both seen and unseen users. The two latter benchmarks provide real-user preference comparisons in summarization and open-domain chatbot interaction, respectively. These results indicate that unlabeled responses can provide additional supervision for personalized reward modeling when per-user preference labels are scarce.

Quality prediction in multistage manufacturing is challenging when inter-stage lags vary over time, as measurements taken at the same timestamp may reflect different process states. Existing approaches often preserve production order but fuse timestamp-aligned features or model delayed dependencies implicitly, without identifying the upstream history relevant to each downstream observation. We propose Delay-Guided Attention for Multi-stage Forecasting (DAMF), which explicitly estimates inter-stage lag distributions and uses them to guide cross-stage information transfer for final quality prediction. For each adjacent stage pair, a lag predictor compares the current downstream representation with candidate lagged upstream representations. An occurrence-magnitude factorization separately estimates the probability of a positive lag and its conditional magnitude, yielding a sample- and stage-pair-specific distribution over discrete lag states without requiring plant-specified delays at inference. DelayAlignment incorporates this soft distribution as a temporal prior into differentiable cross-stage attention, while gated residual fusion regulates information propagation across production stages. For interpretation, segment-wise Viterbi decoding converts pointwise lag distributions into temporally coherent lag paths, whereas forecasting retains the soft distributions for alignment. On the real-world LiquidSugar dataset, DAMF improves final quality prediction over the compared baselines. Evaluations on two controlled-delay benchmarks show more accurate and stable lag recovery, and ablation studies support the contributions of lag prediction, delay-guided attention, and structured temporal decoding.

Piano transcription models have achieved very high performance in recent state-of-the-art systems, but their performance tends to degrade in ensemble settings where the piano is played alongside other instruments. While some approaches attempt to address this issue by incorporating multi-instrumental datasets, the problem remains challenging. In particular, when other instruments produce high-energy sustained tones while relatively low-energy piano notes occur simultaneously, the magnitude components of their frequency representations overlap, making it difficult to accurately distinguish piano notes. To address this issue, we propose an improved piano transcription model that jointly leverages magnitude and phase information, enabling more robust piano transcription performance in ensemble settings. We evaluate the proposed model on two piano ensemble datasets with precise MIDI note annotations, which we obtained by revisiting and extending existing datasets. Notably, we constructed a new piano concerto dataset, PCD-K, by playing a piano part on a Disklavier and recording the audio and MIDI pairs with pre-recorded orchestral tracks. Experimental results demonstrate that phase information is highly effective in improving performance, outperforming magnitude-based approaches in ensemble settings.

Retrieval-augmented generation can expose unwanted concepts through retrieved evidence even when the generator is moderated. We study query-side concept suppression with a fixed encoder and document index. Compositional Concept Unlearning (C$^2$U) routes a profile-derived direction through an augmented atom bank, adding only a bounded edited-minus-original delta to the raw query. Across five datasets, four encoders, and three seeds, the validation-selected configuration reduces pooled binary target Hit@5 by 4.25 percentage points with non-positive control degradation and target-exposure changes. Under the evaluated query transforms, it suppresses target density without increasing budgeted control-query exposure. Separately, ArXiv four-concept composition improves over an averaged-direction comparator after independent validation gates and first-use test evaluation.

We present a plaintext-recovery attack on both the DBDH-based construction and the CBDH-based variant of Datta, Komargodski, and Waters (Journal of Cryptology, 2023). The attack succeeds with probability one in the claimed static security model using one corrupt authority and two users, each unauthorized even when granted the corrupt authority's attribute. The flaw is an invalid extension of the LSSS matrix-normalization lemma originating with Rouselakis and Waters (FC,~2015): preserving complete-share distributions does not preserve coordinate-wise ciphertext masks, which a corrupt authority can expose to remove identity binding from other users' shares. For both variants, replacing these masks with independent full linear sharings of zero preserves correctness and restores ciphertext-distribution invariance. We prove static security of both repaired variants in the random oracle model under their respective BDH assumptions, using the original hardcore-bit convention for the CBDH variant. Both proofs allow adaptive identity-hash queries. The modification retains the original matrix-width and row-labeling restrictions, keys, decryption, and ciphertext size, requiring only additional encryption randomness and finite-field arithmetic. We also attack both multi-authority attribute-based inner-product functional encryption constructions of Datta and Pal (PKC, 2023) and establish correctness and matrix-substitution invariance of the same repair. We prove static security of the repaired small-universe construction under asymmetric DBDH. For the large-universe construction, we identify a missing source-group encoding in the original reduction and give a complete static security proof under an augmented $L$-DBDH assumption, whose generic bilinear group hardness we establish.

Current LLM safety benchmarks predominantly rely on binary safe or unsafe labels, lacking explanatory depth and suffering from low consistency across automated judges. To address this, we introduce DiagSafe, a diagnostic evaluation framework that moves beyond binary classification by employing prompt-level rubrics. Grounded in globally applicable AI safety policies, each rubric provides explicit criteria along two dimensions: safety, which targets concrete harms with specific exemptions to suppress false positives, and usefulness, which independently evaluates instruction fulfillment. To ensure a robust and realistic evaluation, we construct a bilingual (English-Chinese) benchmark grounded in a comprehensive 84-label safety taxonomy, consisting of two subsets: DiagBase, which covers single- and multi-turn prompts with direct harmful queries and template-based jailbreaks, and DiagAttack, which embeds harmful intents into complex instruction-following tasks to form challenging instruction attacks. Extensive experiments demonstrate that DiagSafe significantly enhances the reliability of LLM-as-a-judge evaluation, achieving 95.34% agreement with human annotations -- substantially higher than the 69%-87% achieved by the baselines -- and high inter-judge consistency (Cohen's $κ$ = 0.864). Furthermore, applying our framework reveals that leading LLMs differ markedly in their unsafe rates, and each model also exhibits unique vulnerabilities along different dimensions. In addition, we observe that models are particularly susceptible in Chinese and multi-turn contexts. Ultimately, DiagSafe provides a reliable and explainable tool for auditing and optimizing LLM alignment.

Barycenter models, based on stochastic and deterministic optimal transport bridges between probability measures, have proven to be useful for supervised and unsupervised learning for (probability) measure-valued data. In this paper, we utilize the directional asymmetry and dual drift-diffusion control arising in the recently proposed Schrödinger-Bass Bridge (SBB) between two measures, and propose \textit{SBB-encoder}, \textit{SBB-decoder}, \textit{SBB-symmetric}, and \textit{SBB-Markovian} barycenter models. For each of these novel models, our first theoretical result establishes the existence of the SBB-barycenter under mild assumptions. Our second theoretical result characterizes the decoder SBB-barycenter of Gaussians where we show that for Gaussian input measures, the decoder formulation admits a Gaussian minimizer, reducing barycenter computation to a finite-dimensional optimization and yielding a compact algorithm. For general input measures, we derive a hybrid Wasserstein gradient particle descent scheme for all four formulations using the SBB dual representation. Finally, we utilize these algorithms for solving distributional-inverse/interpolation problems arising in the context of spatial images and single-cell RNA sequencing data, where we show that the SBB-decoder and Markovian barycenter allow us to interpolate between successive time points. Results show that our method outperforms existing barycenter models for measure interpolation and inference.

Recent generative medication recommendation systems ultimately produce sets of drugs, but drug--drug interaction (DDI) constraints depend on the set as a whole. A natural post-hoc approach is to first predict medications and then locally repair interacting pairs by removing one drug at a time. We ask whether this local repair step itself introduces avoidable error. Using fixed medication scores from FLAME's frozen classifier on MIMIC-III, we compare increasingly strong local repair strategies with an exact global decoder while holding the scores, output size, and DDI constraint fixed. The global decoder jointly selects the highest-scoring feasible medication set using mixed-integer linear programming. Across validation and test splits, global decoding consistently achieves higher Jaccard similarity than greedy repair, degree-aware repair, and degree-aware repair strengthened with exhaustive 1-swap local search. Its advantage is small under loose DDI constraints and becomes larger as the constraint tightens; on the test split it reaches +0.0064 Jaccard over the strongest local baseline at the strict DDI operating point. The finding also holds across a fine accuracy--DDI sweep at matched realized DDI exposure and under three score representations. At the strict setting, global decoding retains more true-positive medications while removing more false positives. These results show that medication recommendation is not only a representation-learning problem: even with a frozen clinical model, how its scores are converted into a constrained medication set can materially affect the final prediction.

Streaming video understanding (SVU) requires video-LLMs to process incoming video continuously while answering user queries with low latency. Recent methods reduce query-time cost by prefilling frames into KV caches during idle time and reusing the cached KVs at query time. To bound per-frame computation over unbounded video streams, these methods restrict the context available during prefill. This creates a fundamental context mismatch: the restricted context may omit information needed to properly contextualize video tokens, yet, still contain information irrelevant to the eventual query, which is unknown during prefill. We propose FrameBlend, a simple yet effective plug-and-play method to address this mismatch through selective KV recomputation, without requiring any additional training. At query time, FrameBlend re-prefills selected video tokens using query-relevant context and replaces their cached KVs with recomputed ones. Crucially, this recomputation enables shorter attention windows during idle-time prefill, substantially reducing prefill cost while improving accuracy. Experiments across three SVU frameworks and five streaming and offline benchmarks demonstrate accuracy gains of up to 3.9 pp. On RVS-Ego, FrameBlend reduces per-frame prefill FLOPs by over 20% and total processing time by 14-59% for retrieval-based frameworks, including the cost of query-time recomputation. These results demonstrate that a simple change in how contextualization is distributed between idle time and query time can improve the accuracy-efficiency trade-off for SVU.

A quantum computer could recover the key of any Ethereum account whose public key appears on chain. We prove that no sidechain or bridge that leaves Ethereum's rules unchanged can protect such accounts: their funds must move into a contract whose release condition relies only on quantum-safe primitives. We design \emph{PQ-Vault}, a custody system for unmodified Ethereum combining this vault, a quantum-safe proof-of-stake sidechain whose checkpoints carry compact certificates, and an optimistic dispute protocol. We give the first security proof of compact certificates against quantum attackers and a matching attack: with classically chosen parameters, the guarantee falls from 128 to about 60 bits, and a simple rule restores it. We show that a common seed-update rule in hash-chain proof of stake lets an attacker steer chosen slots and profit from splitting its stake, and give an election whose steering gain and inclusion delay are provably bounded against quantum attackers. Since compressing certificates with a pairing-based SNARK breaks quantum security, we use fraud proofs that need a single signature check. For 65,536 validators, a certificate is 0.8 MB, 260 times smaller than all signatures, and verifies in about 20 ms; the vault accepts a checkpoint for about 420,000 gas.

Visual question answering from low-altitude UAV imagery must deliver timely and accurate judgments to support disaster response under limited edge computing resources. Routine questions may be answered through direct decisions, whereas more challenging questions may benefit from step-by-step reasoning. However, existing research primarily focuses on improving models' visual understanding and generative reasoning capabilities, with limited exploration of efficient direct decision-making and its adaptive coordination with generative reasoning. We develop Disaster-S1 based on a vision-language model adapted to the disaster domain. Trained for option discrimination, it outputs a probability distribution over answer options in a single multimodal forward pass for fast decision-making. We further propose DisasterRoute, a dual-system routing framework that combines prediction confidence with disaster evidence to decide whether to accept the initial answer or invoke the disaster reasoning model DisasterVL for review. By allocating inference computation on demand, the framework retains the efficiency of direct decision-making and improves overall question-answering accuracy through selective review. Experiments on DisasterBench show that Disaster-S1 achieves superior accuracy with low latency, while DisasterRoute further improves accuracy through selective review at substantially lower latency than full reasoning.