Papers

Deepfake detectors remain vulnerable to transfer-based black-box attacks, in which adversarial examples are generated on a source surrogate model and transferred to a target model, unknown to the attacker. Yet how source–target compatibility shapes attack success remains poorly understood. Prior studies evaluate limited detector pools and rarely disentangle architectural from training factors. We conduct a controlled evaluation of adversarial transferability across 60 detectors spanning six backbones, two pretraining regimes, and five training-data configurations, using two attack procedures: AutoAttack (AA) and the Carlini–Wagner attack with Expectation over Transformation (CW–EOT). Matched comparisons reveal significantly higher transfer when source and target share an exact backbone, architecture family, pretraining regime, or training data. This compatibility structure is attack-dependent: exact backbone compatibility has the largest effect under AA, whereas shared pretraining and training data have the largest effects under CW–EOT. When transfer is averaged across non-target sources, mean attack success rate (ASR) is $7.21\%$ under AA and $19.52\%$ under CW–EOT. By contrast, a multi-source oracle combining both attacks attains a \(64.48\%\) mean ASR after excluding exact backbone and training-data matches, showing that source averaging can substantially understate target vulnerability. We release 240,000 adversarially perturbed images, complete pairwise transfer results, detector configurations, and evaluation code. These findings establish source–target compatibility and source-model selection as central dimensions of credible transfer-based black-box robustness evaluation.

Few-shot medical image segmentation (FSMIS) seeks to delineate unseen structures from a small support set, but its standard formulation fixes task-defining evidence before inference. This assumption is fragile under acquisition shift, atypical pathology, ambiguous boundaries, and poor image quality. Adding clinician interaction and rapid adaptation is not sufficient: the binding constraint is deciding when asking or changing is warranted. We therefore reframe FSMIS as a three-layer sequential decision problem. First, decidable self-assessment separates errors that a bounded intervention can repair from those that no admissible intervention can reach. We formalize this distinction through a correctable set defined by the update operator and remaining interaction budget. Second, selective interaction allocates a distinct expert-attention budget by response-conditioned net expected value of information, yielding explicit accept, query, and defer actions. Third, bounded adaptation emphasizes reversibility and independent safety reassessment rather than speed. A complementary cross-case memory stores reproducible correction priors over failure modes instead of disease-specific mask priors. This structure links sparse support representation, cross-domain robustness, multi-level risk estimation, clinician feedback, and governed experience transfer. We state six hypotheses with an explicit dependency order and propose a minimal pilot that can falsify the foundational self-assessment claim before a clinician study. The central claim is not that interaction resolves domain shift, but that scarce expert attention should be used only when a bounded intervention is expected to reach a clinically better outcome.

Numerically solving McKean-Vlasov stochastic differential equations is computationally challenging due to the compounding costs of discretizing in time and in the distribution of the solution. Multilevel ideas have been proposed to provide speed-ups. In this work, we study the multilevel Monte Carlo method proposed by Ricketson (2015) for equations whose drift and diffusion terms depend on the law of the solution $X_t$ through the expectation $E[R(X_t)]$. The scheme follows the single-ensemble paradigm, where particles interact across levels at each timestep. While cross-level feedback makes this scheme attractive in practice, the correlations it introduces have so far confined its cost-error analysis to a model problem with linear drift, deterministic diffusion, and $R$ the identity. We use additional coarse particles to enforce geometrically decaying coupling errors towards the coarser levels. This allows us to prove our main contribution, an $L^p$-error of $O(\epsilon)$ at cost $O(\epsilon^{-2-\delta})$ for any $p\ge2$ and $\delta>0$ (with a constant that grows as $\delta\to0$), assuming only global Lipschitz bounds on the drift, the diffusion, and $R$. An exploratory experiment is consistent with the derived rates and suggests that in practice the constant does not grow significantly for small $\delta$. Our methodology and proof strategy may also be useful for other single-ensemble multilevel schemes, such as multilevel ensemble Kalman filters.

Terminology-aware translation asks for more than a correct translation: the output must use the exact terms a glossary prescribes. The standard recipe, fine-tuning on glossary-annotated translation pairs, hides an inefficiency: for most examples the glossary prescribes exactly what the model would have produced anyway, so they teach nothing about following a glossary. We therefore keep only the examples where the model's own translation contradicts the glossary. In a controlled study at fixed data volume, this selection alone raises term accuracy from 78.7% to 89.9%. The filtered data, built by a two-way synthetic pipeline on open models, is part of the instruction-tuning mixture of our public release SalamandraTA-7b-instruct v3.0, which, used exactly as released and wrapped in a document-level inference pipeline, forms the BSC submission to the WMT26 Terminology Shared Task Track 1. At the official WMT26 evaluation, our system achieves 94.2% term success at 74.6 chrF++, with only two of the twenty-two submissions outperforming it on both metrics. On last year's benchmark, it also surpasses our GRPO-based system, despite being trained solely with ordinary supervised fine-tuning.

The Arctic regions remain heavily dependent on fossil fuels for energy generation. At the same time, the Arctic is warming at a rate considerably faster than the global average, increasing the need for low-carbon and climate-resilient energy systems. This review documents the current energy systems and assesses state-of-the-art energy solutions applicable to Arctic and cold-climate regions, as well as the future direction of these emerging energy systems, and analyses the role of energy storage, heating requirements, and advanced energy solutions. A total of 88 research articles were systematically reviewed. The reviewed studies indicate the potential for wind and solar to reduce dependence on fossil-fuel based energy generation. For energy storage, the reviewed solutions include hydrogen for long-term storage, battery-based system for short-term storage and regulation, and thermal storage using boreholes to meet heating demand. Where a fully renewable system cannot provide the required reliability, diesel can serve as back up generation. Some barriers to widespread adoption are technological, such as the need for specialized planning tools and equipment resilient to harsh weather, and social and institutional, such as the need for governmental support, subsidies, and appropriate legal frameworks.

Nonadiabatic dynamics through a conical intersection requires both static and dynamic correlation and, at every geometry, an excited-state gradient and interstate nonadiabatic coupling (NACME). Analytic multireference derivatives at this level have been confined to datacenter-scale hardware. We report device-resident \mbox{DMRG-QD-NEVPT2} with analytic gradients and NACMEs within the $8$\,GB of a consumer GPU: a density-matrix-renormalization-group (DMRG) active-space reference supplies the static correlation, and quasi-degenerate SC-NEVPT2 dresses it through a multi-state effective Hamiltonian that remains valid through a conical intersection, where single-state perturbation theory develops a vanishing denominator and fails. Every gradient and NACME is one reverse-mode transpose of a single contraction graph: the DMRG sweep enters as a gauge-free differentiable node, and all two-electron work is AO-direct through Cholesky factors. The DMRG reference is exact within the active space ($\sim\!10^{-15}$\,Ha). The corrected adiabats close smoothly to a real $S_0/S_1$ conical intersection with a non-vanishing, $1/\Delta E$-divergent NACME—where linear-response TDDFT returns zero identically—and the dressing restores the differential dynamic correlation that ionic $\pi\pi^*$ states require. DMRG-NEVPT2 and quasi-degenerate NEVPT2 are each established; what is new is the combined stack—multi-state and through a conical intersection, with analytic gradients and couplings—realized device-resident on commodity hardware.

A natural way to cut reasoning-model inference cost is to repeatedly probe a single partial trajectory for its current answer and stop once probes agree – self-consensus. We ask whether any such rule is both safe and token-saving, and whether one can be selected once and reused. A preregistered sweep of 3,520 consensus rules, replayed on frozen trajectories from two models and three benchmarks, clears none of three acceptance gates fixed in advance; the frontier reproduces on a held-out split and on two unseen models – while a boundary-confidence control (DEER) swept through the same pipeline clears all three. The reason lies in the signal: agreement establishes that the current answer persists under a fixed probing procedure, not that the reasoning has terminated – a consensus-termination gap. Stopping on it commits non-terminal answers. At a rule still saving 32% of the tokens, one stop in nine fires on an answer the trajectory itself later abandons, and most of those stops cut off a correction it would otherwise have made. Widening the agreement window does not remove them: the share levels off near 7%, and by then the saving has fallen to 8%. Probe re-wording and a hand-labelled error taxonomy show the agreed answer is often a placeholder the model had not settled on. Used on its own as the stop signal, agreement fails not because it is insufficiently strict, but because it repeatedly measures the wrong object.

Repository-level code generation requires generated code to be compatible not only with the target repository but also with its dependency environment. Existing retrieval-based methods mainly retrieve context from the local repository, leaving external API usage dependent on the model's pretrained knowledge, which can be insufficient for unseen or version-specific APIs. Moreover, current retrieval strategies largely focus on one-hop evidence and overlook the structural relationships among code components. We propose CrossCoder, a cross-repository code generation framework that explicitly incorporates external libraries into the retrieval context through a unified knowledge graph over repository and library entities. CrossCoder identifies important nodes via planning and semantic retrieval, then selectively expands neighboring nodes to retrieve richer multi-hop contextual evidence for generation. To further evaluate dependency-version compatibility, we introduce VersionExec, an execution-based benchmark derived from BigCodeBench that evaluates generation under different dependency versions. Experimental results on RepoExec, DevEval, and VersionExec demonstrate that CrossCoder consistently improves both functional correctness (up to 6.3% on pass@1) and robustness to dependency-version changes.

Consistent submodular maximization aims to maintain a high-utility subset of size at most $k$ as new elements arrive one by one, while changing only a limited number of elements per update. Previously, [DFL+25](STOC '25) established an information-theoretic bound of $2/3$ and gave the first polynomial-time randomized algorithm with a $0.51$-approximation guarantee. However, a fundamental question remained open: Is it possible to design an efficient algorithm for consistent submodular maximization that matches the classical offline $(1-1/e)$-approximation? In this paper, we answer this question in the negative by proving that the sharp approximation threshold achievable with polynomially many value queries and worst-case constant recourse is \[\beta:=2-\sqrt{2}\approx0.5858<1-1/e.\] Specifically, we develop an improved poly-time randomized algorithm, which can attain a $(\beta-\varepsilon)$-approximation in expectation at each fixed time with $O(\varepsilon^{-2})$ worst-case recourse. The cornerstone of this algorithm is a novel anchored sampling scheme, which uses only value queries on subsets of the elements seen so far to construct a solution that remains competitive under unknown future arrivals. Conversely, we establish a matching lower bound by showing that any fixed improvement over $\beta$ requires either exponentially many value queries or $\Omega(k)$ changes per update. Furthermore, we characterize the impact of curvature by establishing a sharp threshold of $1-(\sqrt{2}-1)\vartheta$ for consistent submodular maximization, where $\vartheta\in[0,1]$ denotes the total curvature. Surprisingly, for weighted coverage objectives, we devise a poly-time randomized algorithm that can match the optimal $(1-1/e-\varepsilon)$-approximation with $O(\varepsilon^{-1})$ worst-case recourse, thereby separating coverage maximization from general submodular maximization in the consistent settings.

Real-world video understanding requires integrating visual, audio, textual, and temporal evidence distributed across a video. Yet many pipelines use a fixed video context and single-pass inference, limiting adaptive evidence acquisition when observations are incomplete, ambiguous, or conflicting. We present VLX-VR, an agentic-aware video reasoning model trained within a video reasoning framework defined by a Think–Memory–Observation loop. At each step, VLX-VR determines the needed evidence, invokes read_memory or write_memory, incorporates the returned Observation, and decides whether to continue or produce the task output. We train VLX-VR with multimodal data, including videos and agent trajectories, using reinforcement learning to learn evidence acquisition, memory use, and termination. On MINERVA, VLX-VR achieves state-of-the-art performance among the models included in our comparison, with 78.79% accuracy. Under the original three duration groups, its accuracies are 76.70%, 78.73%, and 80.92%, with a cross-duration accuracy variance of 2.97~$\mathrm{pp}^2$. On correctly answered samples, 96.20% of VLX-VR's reasoning traces are consistent with the MINERVA reference reasoning traces and the evidence described by them, while approximately 75.80% of all evaluated samples satisfy both answer correctness and this evidence-grounded trace criterion. These results show strong performance and broadly stable behavior across durations, while counting, state changes, causal reasoning, and spatial perception remain challenging.

Understanding the historical allocation and distribution of research funding advances our knowledge of how scientific research is supported across fields, institutions, and regions. However, large-scale analyses are hindered by the lack of comprehensive funder name disambiguation solutions, as funder names often exhibit spelling variations, translations, abbreviations, and inconsistent levels of granularity. In this paper, we present a framework for developing multilingual, multi-functional funder name disambiguation models and demonstrate its application to research publications in biodiversity conservation. To construct a training dataset, we integrated the Research Organization Registry (ROR), which provides unique identifiers for research organizations, with two publication datasets: the Web of Science (WoS) and the Crossref Open Funder Registry (OFR). We used multi-task learning with Contrastive Loss and Multiple Negatives Ranking Loss to fine-tune three open-weight embedding models from the Sentence Transformer, Gemma, and Qwen3 families. The best-performing models achieved accuracy above 0.90 when matching WoS funder names to ROR identifiers, outperforming general-purpose LLMs, including GPT-5.2, Claude-Sonnet-4.6, and Gemini-2.5-Flash, by more than 0.1. For funder names not indexed in ROR, we constructed a similarity network among funder names and identified clusters within it. Finally, we analyzed the disambiguation results and highlighted challenges arising from limited knowledge of smaller funders and funders from non-English-speaking countries. This work provides a reusable framework for funder name disambiguation with potential applicability across different model architectures and datasets, featuring cost-effective training data creation and multi-task learning and disambiguation.

We certify an honest-decapsulation failure upper bound for ML-KEM-768 in an explicit random-function/centered-binomial (ROM/CBD) abstraction. Domain-separated public-matrix streams are modeled as independent uniform ring elements and secret/noise polynomials as independent CBD2 primitives; this is not an information-theoretic statement about the fixed SHAKE instantiation of FIPS 203. Recent formal assessments identify rigorous justification of ML-KEM's heuristic decapsulation-failure scale as an open problem; within the explicit ROM/CBD abstraction studied here, we obtain a dependency-preserving certified upper bound at that scale. The analysis preserves dependencies induced by the public matrix and by both ciphertext-compression terms. Its terminal chain has three components: an exact graph-coupled full-ideal reference for the joint c_u/c_v residual; a proper-ideal bivariate Fourier transport whose rare |T|>=3 branch is closed by an exhaustive three-factor anti-concentration replay; and exact bit-specific FIPS decoding events followed only by a 256-coordinate union bound. A formal partial-Fourier lemma makes the spectral-to-total-variation step explicit. The reduced rational certificate satisfies Pr[K' != K] <= P_* <= 2^-164.81, with -log2(P_*) = 164.810716201343121.... The 164.81 threshold is exact but numerically tight: the certified exponent exceeds it by only about 0.0007162 bit, and 164.82 is not certified. The result is an upper bound for an arbitrary message fixed independently of the public and secret randomness, under honest encryption and decapsulation. It is not an exact DFR, not a fixed-SHAKE equivalence theorem, not a new IND-CCA reduction, and not an adaptive delta-correctness result.

We study offline inference for the optimal value in reinforcement learning under finite state and action spaces. Two new nuisances are derived as fixed points of a self-induced Bellman equation, in which we approximate the maximum Bellman operator by its softmax correspondence. We propose a debiased estimator through the Neyman orthogonality and establish its asymptotic normality under diverging horizons even when the behavior policy changes with time, as long as the nuisances have the statistical rates that can be achieved by many machine learning methods. We provide a concrete estimating procedure for these nuisances and show they can lead to valid inference. Synthetic experiments validate the numerical performance of our inference method, and we implement it in real-life decision-making problems, including bike repositioning and AI agentic tool use.

High Performance Computing-Quantum Computing (HPCQC) platforms expose multiple Quantum Processing Units (QPUs) that may differ in size, topology, native gates, and noise characteristics. For current noisy devices, errors compound along the compiled circuits quickly, and minimizing them, that is, maximizing the circuits' execution fidelity, is essential for reliable results. Fidelity depends on the compilation to a specific target device: the same high-level circuit may produce different executables and, therefore, different expected fidelities across QPUs. We present a low-overhead fidelity-aware scheduling framework for multi-QPU systems based on a Graph Neural Network (GNN) that estimates, before compilation, the expected fidelity of each circuit on each available QPU. Then, a tunable scheduler uses these estimates to control the trade-off between execution fidelity and parallelism. Results show that this framework allows for approximating an exhaustive fidelity-based assignment, saving computational resources compared to a brute-force approach that compiles each circuit on every device.

Bounded SQL equivalence checkers search for a counter-example database of bounded size, and a search that comes back empty proves nothing. We supply missing theory: computable bounds $B$ such that agreement on all databases with at most $B$ tuples per relation implies equivalence. We work in the combined-semantics framework, which captures SQL's mix of duplicate-eliminating (DISTINCT) and duplicate-preserving computation over set-valued relations. For conjunctive queries we prove a bound linear in the query size for fixed multiset width: inequivalent queries already disagree on a database with at most $2^w |Q|$ tuples, where the width $w$ counts only the columns the queries actually read, independently of the total number of multiset variables. Declared keys shrink the bound to $2^{kw} |Q|$ for the smaller key-width $kw$, acyclic foreign keys leave it unchanged, and the result extends to several classes of queries with comparisons, for which equivalence had not previously been characterized. For these fragments, bounded search becomes a terminating, complete decision procedure.

This project investigates how software architecture practices in Free/Libre Open Source Software (FLOSS) are shaped by socio-technical and ethical factors, and how education can support more explicit, inclusive, and reflective architectural work. Motivated by FLOSS's role in digital sovereignty, it is observed that architectural decisions are often undocumented and scattered across issues, pull requests, and mailing lists. While prior research has studied architectural artifacts, erosion, and communication, the interplay between architectural work, governance arrangements, and ethical commitments in FLOSS remains underexplored. The research follows a three-phase design: (1) multi-method case studies of 3-4 domain-pairs of architecturally non-trivial FLOSS projects, (2) framework and intervention design with practitioners and educators, and (3) pilot evaluations in projects and courses. It will produce (i) cross-case empirical evidence on FLOSS architecture practices, (ii) a conceptual framework linking architecture practices to socio-technical conditions and ethical dimensions, and (iii) lightweight practices and teaching formats that render architectural work more explicit and inclusive.

Grapheme-to-phoneme conversion (G2P) refers to the task of converting a sequence of graphemes to a corresponding sequence of phonemes. While Filipino G2P is fairly straightforward due to its shallow orthography, the inclusion of prosodic features such as stress adds a layer of complexity that requires sentence-level context instead of single-word inputs. However, sentence-level data for Filipino typically do not include phoneme transcriptions, posing a challenge for training G2P models. As such, we investigate how to obtain sentence-level phoneme data for Filipino using available data and compare the resulting models with multilingual word-level G2P as well as measure how accurately they predict stress marker position for Filipino. We propose fine-tuning a ByT5-based model, pre-trained on multilingual word-level G2P data, on three sentence-level G2P datasets annotated with an LLM-assisted pipeline guided by data from Wiktionary. This approach produces models that perform well on the G2P task, achieving at best around 0.54% PER and 2.50% CER, a significant decrease compared to base model PER at around 19.74%, on a manually-corrected test set. The model is able to correctly classify most of the main stress classes in Filipino, but struggles particularly with malumi words. We show that a ByT5-based model performs well at sentence-level Filipino G2P and offers strong potential for Filipino homograph disambiguation.

Evaluating video captioning remains a critical challenge for Visual Large Language Models (VLLMs). Existing metrics primarily rely on matching generated text against ground-truth references. This paradigm suffers from the “one-to-many” nature of video description, where high-quality captions are often penalized for lexical mismatches or valid shifts in visual focus. Furthermore, such assessments are typically one-dimensional, failing to provide a fine-grained analysis of caption quality. To address this, we redefine caption quality through the lens of information fidelity: A caption must maximize the coverage of salient visual information while ensuring strict factuality. We introduce CapQuiz, a novel reference-free benchmark that assesses captions based on their utility in answering human-verified, fine-grained, multiple-choice questions derived from the video. CapQuiz features a hierarchical taxonomy of 10 question types (spanning Descriptive and Inferential categories) across 24 diverse video domains. Extensive experiments demonstrate that CapQuiz correlates significantly better with human judgments than existing metrics and offers interpretable insights into model performance.

Finger-specific motor intent is a clinically meaningful control signal for post-stroke neurorehabilitation, where residual muscle activity may remain measurable despite weak or incomplete movement. We study five-finger multilabel intent decoding from impaired-arm high-density surface electromyography (sEMG) in PhysioMio, a bilateral longitudinal dataset collected from stroke patients. A common processing protocol aligns movement labels, applies 20–450 Hz Butterworth filtering and Symlet-4 wavelet denoising, segments overlapping 200 ms windows, and extracts twelve time- and frequency-domain descriptors per channel. Direct LSTM, CNN, and GNN baselines reveal complementary behavior: the LSTM attains the highest subset accuracy (0.545), whereas the GNN attains the highest macro F1 (0.706) and macro AUPRC (0.776). Architecture search then identifies CNN-Large as the strongest single-split CNN, with 0.593 subset accuracy and 0.714 macro F1, while CNN-Micro provides a compact architecture for embedded inference. To match a four-sensor hardware design, we retrain CNN-Micro using channels associated with ECRB, ECRL, FDS, and FDP and exclude the ground electrode from model input. Across five seeds, cross-channel knowledge distillation improves the four-channel student over direct training, reaching $0.5219 \pm 0.0114$ subset accuracy, $0.7612 \pm 0.0038$ finger accuracy, and $0.6095 \pm 0.0058$ macro F1. The selected 123K-parameter model accepts nine windows of 48 features and has been exported to ONNX. These results establish a reproducible software path from post-stroke sEMG to compact five-finger intent prediction for subsequent hardware-in-the-loop evaluation.

The $\mathbb{Z}_2\mathbb{Z}_4\mathbb{Z}_8$-additive codes are subgroups of $\mathbb{Z}_2^{\alpha_1}\times\mathbb{Z}_4^{\alpha_2}\times\mathbb{Z}_8^{\alpha_3}$, and a $\mathbb{Z}_2\mathbb{Z}_4\mathbb{Z}_8$-linear Hadamard code is the Gray map image of such a code. A recursive construction of $\mathbb{Z}_2\mathbb{Z}_4\mathbb{Z}_8$-additive Hadamard codes $\mathcal H^{t_1,t_2,t_3}$, with all $\alpha_i\neq0$, $t_1\geq1$, $t_2\geq0$, and $t_3\geq1$, is known, as are the linearity, kernel dimension, and rank of the corresponding codes $H^{t_1,t_2,t_3}$ of length $2^t$, where $t+1=3t_1+2t_2+t_3$. Yet these invariants do not completely classify the family. Two infinite families of pairs of distinct types share the length, rank and kernel dimension, and for classified lengths $2^t$, $3\leq t\leq11$, such pairs were separated only by computer equivalence tests. In this paper, we introduce an equivalence invariant that resolves these cases. The kernel partitions the binary coordinates into blocks, two coordinates lying in the same block when every kernel word takes the same value in both; the kernel-block rank profile is the multiset of the dimensions of the linear span punctured on these blocks. Unlike rank and kernel dimension, which are global, this invariant records how much of the span survives on each block. We compute it for the whole family: there are $2^{t_1+t_2+t_3-1}$ blocks, all of size $2^{2t_1+t_2}$, and the profile takes at most two values $t_2+\binom{t_1+2}{2}$ and $t_2+2+\binom{t_1+1}{2}$, whose difference is $t_1-1$; it is constant precisely when $t_1=1$. Hence, the profile recovers $t_1$, and the length and kernel dimension recover $t_2$ and $t_3$. Two codes of the family with the same length are therefore equivalent if and only if their types coincide, and the number of pairwise nonequivalent such codes of length $2^t$ is $\lfloor(t^2+6)/12\rfloor$ for every $t\geq3$.

Training speaker diarization models on loose labels, such as speech segments with padded boundaries or filled pauses, often results in similarly loose model outputs. To obtain tighter boundaries, pseudo-labeling based on the averaged outputs of causal and anticausal models has been proposed. However, since the pseudo-labels are estimation-based, they can suffer from over-tightening, which increases missed detections that can propagate as unrecoverable errors to downstream tasks. This paper carefully analyzes the causes of over-tightening and proposes three approaches to address them: (i) removing pause filling rather than padding, (ii) introducing a burn-in phase to mitigate missed detections near the beginning of causal and anticausal predictions, and (iii) making pseudo-label-based co-training aware of the non-causal model used for final inference. Experimental results show that the proposed method reduces missed detections caused by over-tightening and improves both diarization accuracy and downstream multi-talker ASR performance.

Large Language Models (LLMs) have achieved remarkable progress across natural language processing (NLP) tasks, yet their capabilities degrade sharply for low-resource languages and dialectally diverse settings. Bangla, the world's sixth most spoken language, exemplifies this gap: existing resources overwhelmingly target Standard Bangla, leaving its regional dialects without the benchmarks needed to develop or evaluate dialect-aware systems. We address this gap with 5-Dialects-BN, the first multi-annotation Bangla dialect benchmark to align Romanized transliteration with dialectal text, Standard Bangla, English, and subjectivity labels across five regional varieties. The dataset comprises 6,000 manually annotated entries spanning five major dialects: Chittagong, Barisal, Noakhali, Sylhet, and Rangpur (Chittagong 1,900; Noakhali 1,500; Sylhet 1,200; Barisal 700; Rangpur 700), reflecting natural online availability. Each entry is enriched with five aligned annotations: the original dialectal text, a Romanized transliteration, an English translation, a Standard Bangla translation, and a subjectivity label (subjective vs. objective). Annotations were produced and cross-validated by native speakers and undergraduate linguistics students to ensure dialectal authenticity and semantic fidelity. The resulting resource supports a diverse suite of tasks, including dialect identification, dialect-to-standard normalization, machine translation, subjectivity classification, and parameter-efficient fine-tuning (e.g., LoRA) of multilingual LLMs. By providing a standardized, multi-annotation benchmark, 5-Dialects-BN enables principled evaluation of LLMs on dialectally diverse Bangla and lays a foundation for further research in low-resource, dialect-aware NLP.

Smartphones simultaneously transmit temporary identifiers over LTE, WiFi, and BLE. Existing privacy defenses analyze identifier randomization per protocol, implicitly assuming that these protections compose across protocols. We show that they do not: Even when each protocol leaks only temporary identifiers and the adversary is fully passive, unsynchronized identifier rotations allow cross-protocol stitching of device traces. We present CrossLink, an uncertainty-aware tracing algorithm that links identifiers across time, space, and protocols under noisy localization and mobility. We evaluate CrossLink using controlled lab experiments with commodity devices and large-scale mobility simulation. Under large-scale mobility simulation, CrossLink reconstructs full traces for 83% of users, versus 22% for the best single-protocol baseline, showing that location privacy must be analyzed jointly across protocols. We further show that CrossLink remains effective under partial coverage: strategically placed sniffers near LTE handover regions, mobile sniffers, and limited high-coverage subregions retain sufficient cross-protocol evidence to bridge observation gaps, achieving substantially higher linkability than random deployments.

We propose a variable and behavior-based node-specific proof-of-work (PoW) model for a directed acyclic graph (DAG)-based distributed ledger technology (DLT) network to mitigate decentralized network congestion control. Network congestion control for centralized communication systems is an established field of study, with detailed and continuous research being done on the subject. However, attention to congestion control in decentralized networks is relatively recent and underexplored, especially with DLT, such as blockchain and DAG-based networks. For the DLT networks, the network congestion is caused by factors such as transaction spamming, an increase in the user base, and the launch of new tokens. We focus on the congestion caused by the spamming of transactions within the blockchain and DAG-based DLT network. Based on the network throughput of transactions per second and consensus procedure, the DAG-based DLT needs to control network spamming more than the blockchain network. The PoW model within the DLT consensus framework is a limited deterrent against spamming. Our model provides equal opportunities for all stakeholders regardless of their computational resources. It prevents and penalizes any node that attempts to spam or dominate the network with more than the prescribed number of transactions. Since the system nodes compete to issue transactions with finite network resources, we display the system behavior through a non-cooperative game. Further, we show that our model enforces prescribed behavior amongst the nodes through the proof of the existence of Nash equilibrium in the game.

The methods for capturing the creative process come with associated tensions around memory recall, articulation, and communication during the act of making, as well as how to record these considerations. This paper has a twofold purpose: first, to offer an example of a mixed methodology, drawn from dance anthropology, sensory ethnography, and design, that applies embodied methods as an alternative for documenting creative making. Specifically, this incorporates the researcher-as-participant and the collation of fieldnotes, embodied knowledge/movement recall, with notation forms, and participant interviews. These are existing methods in dance anthropology; however, using them alongside exploratory prototyping and workshop approaches broadened this work into transdisciplinary practice. Second, it discusses the activation of somatosensory systems through wearable technology and the facilitation of heightened sensory awareness for the practitioner, leading to a subsequent ability to focus on creative decisions linked to reflection and metacognition.

Modern large language models (LLMs) can translate natural-language descriptions into operations research (OR) formulations. Post-training techniques including reinforcement learning and on-policy self-distillation have further improved this capability. However, three limitations remain in training LLMs for OR formulations. First, training commonly relies on synthetic formulations validated by human experts or stronger models, constraining scalable supervision. Second, credit assignment is either coarse or costly: outcome rewards score an entire trajectory without locating the responsible modeling decision, whereas process-level supervision requires an additional evaluator. Third, privileged self-distillation can induce style mismatch by using solver context unavailable at deployment. We find that a model can improve from solver-artifact feedback generated by its own rollouts, making self-distillation a practical, evaluator-free source of dense supervision. Therefore, we propose SOLID: Solver-Informed On-Policy LearnIng through Self-Distillation, a novel framework for self-improving OR language models without verified answers or external evaluators. SOLID executes candidate programs from multiple rollouts, clusters their objectives, and selects a majority-group artifact as a pseudo-reference. The model then performs updates using group-relative advantages and dense self-supervision signals. Across multiple OR benchmarks, SOLID improves solution accuracy for both general-purpose and OR-tuned models over outcome-only group-relative training. These results show that solver artifacts can support scalable self-improvement without trusted answers.

Cross-lingual sentence encoders enable scalable transfer across hundreds of languages, powering applications such as translation mining and zero-shot learning in low-resource settings. Although trained for sentence-level alignment, they are increasingly also applied to token-level tasks such as hallucination detection and sequence tagging, exposing a mismatch between training and usage. We propose SALT, a lightweight post-training method that improves token representations by injecting span-level supervision into existing sentence encoders. Across five multilingual token-level benchmarks, SALT achieves the best overall results on four of them, outperforming alternative fine-tuning strategies and competitive encoders. It also improves sentence-level performance on cross-lingual retrieval and classification tasks. These results demonstrate that span-level supervision is an effective signal for improving both token and sentence representations.

Real-world detectors must often interpret functional or ambiguous prompts, yet conventional models such as YOLO remain restricted to fixed class lists. Even open-vocabulary models like YOLO-World frequently misalign vague language with the intended objects. Building on our prior work Commonsense-Guided Open-World Object Detection Using LLMs and Visual-Semantic Matching, we address YOLO-World's limitations in grounding task-driven queries. We propose Vague2Detect, a hybrid pipeline in which a fine-tuned Sentence-BERT retrieves candidates from a structured household Knowledge Base (KB), and YOLO-World verifies their presence in the image. For prompts outside the KB, a large language model (GPT-3.5-turbo) generates candidate descriptions, dynamically expanding the KB to cover novel concepts. On a benchmark of household scenes using custom images and an Open Images V7 subset, YOLO-World alone achieves only 32% Vague Prompt Success Rate (VPSR), the ability to map ambiguous queries to correct detections. In contrast, Vague2Detect improves performance to 61% VPSR with high precision, and up to 85% when augmented with GPT fallback.

Speech-based Alzheimer's disease (AD) detection has recently benefited from multimodal foundation-model representations that integrate complementary acoustic and linguistic information. However, conventional probability averaging over these complementary classifiers is unreliable, because their posterior probabilities exhibit mismatched scales: identical values may reflect different confidence levels across models. We propose a robust rank aggregation framework that aggregates normalized prediction ranks instead of posterior probabilities. Each subject is scored by its percentile within a fixed training-cohort distribution of out-of-fold predictions; since rank ordering is invariant to monotonic transformations, this avoids probability-scale mismatch while preserving classifier confidence ordering. A confidence-gated Random Forest further corrects residual errors using clinically interpretable linguistic features, overriding the rank prediction only when the two disagree and the RF is highly confident, without additional deep model training or explicit posterior-probability calibration. On ADReSS2020 and ADReSSo2021, the method achieves accuracies of 95.83% and 90.14%, respectively, comparing favorably with previously reported results.

Recent controllable speech generation requires training data with fine-grained annotations of speaker traits, prosody, emotion, paralinguistic cues, acoustic scenes, and context. Existing workflows often rely on manual correction, paid hosted multimodal services, or fixed processing chains, which limits large-scale data processing through annotation cost, external-service dependence, or weak cross-stage recovery. We introduce SpeechAnnotator, a locally deployable, context-aware multi-agent framework built entirely from open-source models and tools. Supporting frontend modules first obtain speaker-aware segments and final segment transcripts, while prior evidence extractors attach heterogeneous segment-level cues. Three specialist agents then collaborate through shared state: the Planning Agent converts local audio evidence, speaker history, neighboring segments, and recording-level context into field-specific contracts; the Labeling Agent performs contract-guided multimodal prediction for directly observable attributes; and the Review Agent runs a bounded review loop that checks evidence support and cross-segment consistency, triggering relabeling only for unsupported or inconsistent fields. To address the fragmentation of existing evaluation resources across isolated tasks and narrow-domain test sets, we introduce SpeechAnnotator-Bench (SA-Bench), containing 8.87 hours of human-annotated audio across nine source formats, together with SpeechAnnotator-Eval (SA-Eval), which separates Timeline-Eval for speaker-aware timeline recovery, Closed-Eval for finite-set attributes, and Open-Eval for open-ended attributes. Experiments and ablations show that SpeechAnnotator provides a locally deployable alternative to commercial audio-capable systems, while the bounded review loop improves multidimensional annotation through evidence- and context-aware field-level recovery.