Robot learning must produce policies that generalize to new combinations of constraints, teammates, and environments. To achieve this, we must structurally factor the policy, which is a choice that dictates what generalizes, what requires retraining, and what remains entangled. Existing methods span a wide spectrum, from expecting structure to emerge from data scaling, to hand-designing it via hierarchies, skill libraries or learned specializations. In this paper, we study what we argue is the most fundamental factorization in robotics: separating the world from the task. We investigate the conditions under which this factorization is principled. World factors are properties of the embodied system and the environment; they exist independently of intent. Task factors are defined by the task's logic over what the world admits. We formalize this asymmetry through Bayesian model evidence: it aligns with the data-generating process, maintains high likelihood through an analytical world model, and reduces the Occam razor's penalty on task parameters. We instantiate this factorization by pairing AICON, a differentiable graph of recursive estimators and interconnections that is compositional, operates without task-specific data, and propagates cost gradients to actuators, with a compact, learned policy that modulates gradient paths. Gradients serve as the interface between the two factors: they carry world structure through the graph and task structure through costs, enabling low-dimensional learning while preserving structural generalization. We test the world/task factorization across three problems that encompass heterogeneous robots, environments, task logic and sensorimotor modalities. Our framework outperforms end-to-end baselines and analytical heuristics in all settings, generalizes zero-shot to out-of-distribution configurations, and transfers to real hardware without retraining.

Multi-view object association is an important computer vision problem that underlies many multi-camera perception tasks. While this task is naturally formulated as a constrained one-to-one matching problem, recent works heavily rely on pairwise ranking metrics like AP and FPR-95 for model evaluation. We highlight a fundamental mismatch between these metrics and the actual assignment objective. Theoretically, we show that AP and FPR-95 can be imperfect even when the assignment is already correct, and that Sinkhorn-based normalization can make them perfect. Conversely, optimal pairwise ranking can still lead to incorrect assignments. We validate this mismatch in practice by using our Sinkhorn-based normalization as a controlled post-processing stress test. We show that optimizing just a few post-processing parameters significantly boosts AP and FPR-95 without corresponding improvements in assignment-level metrics such as ACC and IPAA.

Can a visually plausible food mesh be trusted to estimate the volume of consumed food? \method investigates this question using selected paired before- and after-consumption states from the MetaFood CVPR 2026 Continuous 3D Reconstruction While Eating Challenge. The submitted workflow follows a curated reconstruction protocol: SAM~3 segments the food and plate regions; Hunyuan3D/SAM~3D generates a dimensionless food mesh; the plate diameter provides the metric scale; the plate geometry is removed in Blender; and the remaining mesh is hole-filled, made watertight, and integrated to estimate volume. MoGe-2 is used only as an auxiliary cue for initial dish-diameter estimation when direct plate measurement is uncertain; it is not the primary scale source for the reported challenge result. \method ranks first, with an average Chamfer distance of 8.31 across 34 meshes using rigid ICP without scale correction. On 17 before- and after-pairs, it achieves 33.87\% state-level volume MAPE and zero monotonicity violations, while consumed-volume MAPE remains 53.74\%. The results show that surface reconstruction, metric scale, controlled mesh cleanup, watertight volume integration, and physical depletion consistency should be evaluated separately for dietary assessment. Source code and evaluation scripts will be available at \href{https://github.com/GCVCG/PerBite-CVPR-MetaFood-2026}{github.com/GCVCG/PerBite-CVPR-MetaFood-2026}.

This paper investigates the entropy dynamics of Chain-of-Thought (CoT) and uncovers a consistent two-phase structure: an Uncertainty Region of exploration transitioning sharply to a Confidence Region of convergence. We demonstrate that the Confidence Region possesses two critical properties: 1) High Reliability -- answers in the confidence region become highly accurate and stable, and 2) High Redundancy -- models generate unnecessary tokens long after reaching the correct answer. These properties unlock more efficient and reliable inference strategies: 1) Early Exit leverages reliability and redundancy to terminate computation safely when returns diminish, and 2)Test-Time Scaling uses the Confidence Region signal to prioritize converged trajectories. To operationalize these insights, we formulate Confidence Region detection as a sequential change-point detection problem, being the first to apply classical change-point methods to monitor CoT reasoning. Using the Cumulative Sum (CUSUM) algorithm, a statistically optimal change-point detector, we develop a training-free framework for real-time inference control. Experiments show our approach establishes a superior Pareto-frontier for early exit. CUSUM achieves 63.06% accuracy with 11.1% token reduction, outperforming DEER and Dynasor by 3.28% and 4.36% in accuracy respectively. For test-time scaling, CUSUM-weighted voting consistently outperforms self-consistency.

Algorithm Selection (AS) aims to automatically identify the most suitable optimization algorithm for a given problem instance by leveraging measurable problem characteristics and historical performance data. In this study, we investigate the generalization ability of AS models across both synthetic and real-world optimization landscapes. We consider two widely used academic benchmark suites (BBOB and CEC) and two real-world problem sets (robotics trajectory optimization tasks and unmanned aerial vehicle path-planning problems). Through a systematic cross-benchmark evaluation, we analyze how AS models transfer between domains, identify where generalization succeeds or breaks down, and highlight the challenges that arise when applying AS in realistic, domain-specific contexts. Our findings provide insights into the robustness of current AS approaches and inform the development of more reliable, broadly applicable AS systems for real-world optimization.

Large Reasoning Models (LRMs) rely on long reasoning traces, making inference expensive. While low-bit quantization reduces per-token decoding cost, we show that aggressive 2-bit inference can fail to deliver end-to-end speedup because instability in the generation process inflates total token count. Instead of merely lowering answer accuracy, 2-bit quantization often produces much longer traces with repetitive loops, budget exhaustion, delayed commitment, and unclosed reasoning segments. We analyze full reasoning traces of Qwen3 reasoning models across mathematical and commonsense benchmarks and show that accuracy degradation is tightly linked to these process-level failures. To address them, we introduce two lightweight controls: FP16 planning, which gives the 2-bit model a short high-precision outline, and loop rescue, which detects repetitive traces and either commits to an earlier answer or falls back to FP16. On MATH-500, loop rescue improves Qwen3-8B accuracy from 17.2% to 74.2%, while planning plus loop rescue improves Qwen3-32B from 65.0% to 87.2%. Overall, our results show that extreme low-bit reasoning becomes practical when its failures are treated as controllable generation pathologies: with lightweight detection and selective FP16 support, 2-bit inference can recover accuracy while preserving real end-to-end speed. Our code is available at: https://github.com/brain-lab-research/quantized-reasoning.

PlanarBench tests whether LLMs can draw planar graphs as ASCII art given only an edge list -- a spatial reasoning task that resists memorization because edge order, edge orientation, and node labels are all permutable. We evaluate 91 models on the 199 simplest non-isomorphic connected planar graphs (2 - 7 vertices). Edge count is the dominant difficulty predictor ($r = -0.85$) -- a finding not reported in prior LLM graph benchmarks, which use only node count as the difficulty axis.

In Automated Essay Scoring (AES), benchmarking practices have fostered minimalist evaluation practices, in contrast with the broader-view recommendations of evaluation frameworks, such as the argument-based validation framework (ABV), which argued in favor of a multidimensional assessment of systems, especially in the context of high-stakes language tests. In this paper, we introduce an enhanced and more practical version of the ABV framework, incorporating fairness analysis, correlations with linguistic features, prediction error evaluation, and model agreement compared with human raters. Applying this framework to French AES, we compare 8 model architectures on a corpus of 27k exam essays (2 raters each) and a generalization corpus of 961 essays (at least nine raters each). Our analyses illustrate the benefits of applying the ABV framework to better understand the capabilities and pitfalls of AES models, while also advancing the state-of-the-art for French AES.

Pre-training has become a fundamental paradigm in modern machine learning, with one of its key empirical benefits being reduced downstream sample complexity as the scale of pre-training data increases. However, existing theoretical frameworks for pre-training do not fully explain this phenomenon. In this paper, we introduce complexity minimization, a novel meta-representation learning framework designed to enable theoretical analysis of this scaling behavior, which learns representations by evaluating the downstream model complexity best suited to each domain and minimizing the worst-case such complexity across source domains. Our end-to-end theoretical analysis, spanning pre-training through downstream regression, shows that this framework provably captures this scaling behavior; in particular, we show that the error rate of few-shot adaptation improves as the amount of meta-training data grows. Empirically, we demonstrate that incorporating complexity regularization into existing meta-learning methods consistently improves downstream sample efficiency.

Blind image quality assessment (BIQA) aims to predict perceived image quality without access to a reference image. Classical natural scene statistics (NSS) descriptors and modern vision-language model (VLM) embeddings address this problem from fundamentally different perspectives, yet whether combining them yields complementary benefits and how to weight their contributions per input image remains unexplored. We propose a distortion-aware fusion framework that integrates a 138-dimensional NSS descriptor with two complementary VLM embeddings, SigLIP and CLIP-H, through a multiplicative gating mechanism that learns per-input stream weights conditioned on image content. Unlike static concatenation fusion, the proposed gating network suppresses or amplifies each stream's contribution based on the input, producing weights that correlate positively (Spearman rank correlation rho=0.33) with the per-distortion NSS contribution measured by independent ablation on KADID-10k. The framework requires no end-to-end fine-tuning of the VLM backbones and is trained with a hybrid loss combining mean squared error, Pearson linear correlation, and pairwise ranking objectives. We evaluate on three standard benchmarks: KonIQ-10k (SROCC=0.9142, PLCC=0.9279), KADID-10k (SROCC=0.9715, PLCC=0.9733, surpassing recent state-of-the-art methods), and LIVE Challenge in-the-Wild (SROCC=0.8527, PLCC=0.8802 with cross-dataset pretraining and fine-tuning). A per-distortion analysis on KADID-10k reveals that NSS features contribute most on noise and color-shift distortions where pixel statistics are directly affected, and least on perceptual distortions such as color saturation changes. The learned gate values validate these findings, confirming that the model autonomously discovers distortion-stream affinity patterns consistent with the manual per-distortion study.

General agentic intelligence hinges on the ability to interact with diverse real-world tools to complete complex tasks, a capability fundamentally tied to the quality of interaction data. To bypass the prohibitive costs of human annotation, prevailing paradigms depend entirely on Large Language Models (LLMs) to scale the synthesis of agentic environments and tasks. However, such unconstrained generation often degenerates into biased random sampling of LLMs' internal priors, failing to capture the diversity and difficulty of real-world domains or construct high-fidelity, long-horizon tasks. In this work, we introduce Grounded Agentic Interaction Synthesis (GAIS), a framework that automates the scalable construction of diverse environments and complex tasks via a two-phase grounding mechanism. Specifically, we construct protocol-anchored environments derived from real-world Model Context Protocol (MCP) servers to ensure functional diversity and difficulty. Subsequently, we employ structure-guided planning to navigate these environments, actively enforcing logical dependencies and adversarial policies to generate complex tasks. Experiments on BFCL, $τ^2$-Bench, and ACEBench demonstrate that GAIS-synthesized data significantly outperforms state-of-the-art baselines, enabling base models to match or even surpass their official instruction-tuned counterparts. Furthermore, GAIS exhibits superior data efficiency and scalability, achieving exceptional capabilities with significantly less data while maintaining continuous growth where baselines stagnate. Our code and dataset are publicly available at https://github.com/Eric8932/GAIS.

Diffusion models have shown remarkable success in video generation. However, whether such models are truly aware of the 3D structure underlying visual observations, rather than simply reproducing plausible 2D projections, remains an open question. In this work, we investigate this question through human motion control, a task that requires precise modelling of 3D human geometry, motion, camera viewpoint, and scene context. Unlike prior methods that rely on rendered 2D motion guidance videos, we propose a render-free framework that conditions video generation directly on compressed 3D human mesh tokens. This representation preserves full 3D geometric information while enabling a unified token-based generation pipeline that processes video tokens jointly with motion tokens in a DiT-based architecture. This design requires the model to reason jointly about appearance, 3D structure, and camera viewpoint during video generation. Experimental results demonstrate strong performance on human motion control benchmarks, while reducing artifacts induced by view-dependent 2D guidance and trajectory-pose mismatches during editing. These findings suggest that video diffusion models, when equipped with mesh tokenization, can better capture complex 3D human structures and their interactions with the surrounding environment.

Modern deep learning models for forecasting groups of time series rely on increasingly longer observation windows. However, the benefit of increasing the window size is often simply attributed to capturing long-range dependencies, and broader discussion on how global forecasting models leverage input observations has been limited. In this paper, we show that forecasting groups of time series involves two objectives: (i) generative process identification (GPI), i.e., inferring the specific process generating the input sequence, and (ii) conditional forecasting (CF), i.e., predicting future values given input observations. From this perspective, optimal predictions can be interpreted as an average over plausible data-generating processes, weighted by their likelihood given the input window. This suggests another explanation for the benefits of long context windows: they reduce the uncertainty about which specific process is generating the input time series during operation. We prove that even for processes with memory length $P$, an input window size strictly larger than $P$ is necessary to achieve the minimum attainable error. Finally, we show how decoupling GPI and CF can improve computational scalability without compromising accuracy. Experiments on synthetic and real-world data validate our insights and their relevance for designing forecasting architectures.

We introduce CARTE 1 (Culturally Anchored Regional-Territorial Evaluation), a multiplechoice benchmark for evaluating the ability of large language models (LLMs) to perform fine-grained reasoning over geographically grounded and regionally differentiated knowledge within France. While prior benchmarks focus on national-level cultural understanding, they largely overlook intra-country variation and the need to distinguish between closely related regional contexts. CARTE addresses this gap by introducing 2,431 questions spanning the 13 metropolitan regions of France and covering 14 thematic domains, including culture, language, demographics, economy, environment, and mobility. We further introduce CARTE-LV, a subset targeting Linguistic Variation across French regions, enabling focused evaluation of language-related differences. We evaluate 27 LLMs ranging from 1B to 12B parameters under few-shot settings. Our experiments reveal performance disparities across regions and model scales, suggesting systematic gaps in pretraining coverage and limited robustness to intra-national variation.

Abundant procedural knowledge on the Web holds great potential for helping agents solve long-horizon tasks. However, such knowledge is often multimodal, heterogeneous, noisy, and implicitly assumes human executors, making it difficult to use directly as the skills required by agents. To bridge the gap between human-oriented guides and agent-executable skills, we formalize this problem as guide-to-skill learning: converting in-the-wild guides into executable skills and continuously improving them from trajectories observable to the agent. To evaluate the capability of existing agents on this task, we introduce MMG2Skill-Bench, the first benchmark designed for this problem. We further propose MMG2Skill, a closed-loop framework that compiles guides into editable skills, conditions a fixed vision-language model (VLM) agent on these skills during execution, and revises the skills from trajectory-level root-cause feedback without using benchmark scores. Across GUI control, open-ended gameplay, and strategic card play with six VLM backbones, MMG2Skill consistently outperforms vanilla baseline agents in every model-domain setting, achieving macro-average gains of +12.8 to +25.3 percentage points across backbones. Ablation studies show that directly prompting agents with raw guides can degrade performance, while both structured skill construction and trajectory-driven revision are necessary for the observed improvements. On success-inferable tasks, analyzer-based early stopping further prevents late-stage performance regressions and saves 25%-53% of attempts when the success signal is properly calibrated.

Industrial anomaly detection has historically been a unimodal task. Recent multimodal vision-language models have produced systems that admit textual input alongside the image and are presented as enabling text-guided zero- and few-shot inspection. Yet these methods are evaluated with protocols inherited from unimodal benchmarks that hold the textual condition constant and therefore cannot measure whether language conditions the decision; whether reported gains reflect text guidance or strong pretrained visual features remains open. We introduce Text-Guided Anomaly Detection (TGAD), a structured benchmark that progressively increases the functional role of language across three scenarios: a controlled prompt-sensitivity setting on MVTec AD; a component-tagged extension of MVTec AD that requires the model to restrict its assessment to an instructed part; and the new Assembled Panel Dataset (APD), a realistic industrial setting that requires both defect-type and component-location knowledge. We evaluate one representative model per paradigm: generative large vision-language, training-free discriminative, and embedding-adaptive discriminative. In all three, the textual interface conditions the decision only superficially: prompt content is absorbed unless the object noun is removed (the generative model's I-AUROC drops from 97.4 to 82.6); component-level instructions do not constrain the decision once defects outside the instructed part are admitted as normal (from 90.3 to 66.3); and when both combine on APD, image-level discrimination collapses below the MVTec level, in one case below chance (71.2, 50.5, 31.5). These results suggest that standard benchmarks overstate the text-guided capabilities of current multimodal anomaly detection systems, and that a protocol of this kind is a prerequisite for models that can be reliably controlled through language for industrial deployment.

As Large Language Model (LLM) agents increasingly leverage the Model Context Protocol (MCP) to operate in complex environments, the expansion of their action spaces offers agents unsafe capabilities and underscores the risk of power-seeking. While broad action space and greater environment influence are essential for task fulfillment, they create a fragile risk surface where minor errors or hallucinations are magnified into catastrophic failures. In response, we propose SafeMCP, a {server-side} defense plugin that constrains tool acquisition via predictive reasoning regarding future safety risks. SafeMCP utilizes an internal world model for look-ahead reasoning to implement a two-tier defense: proactive tool filtering to constrain hazardous power expansion and immediate intervention as a fail-safe. To train SafeMCP, we introduce a three-stage pipeline comprising environmental dynamic grounding, safe policy initialization, and reinforcement learning (RL) with dual verifiable rewards. Experiments on PowerSeeking Bench, ToolEmu, and AgentHarm show that SafeMCP achieves a safe equilibrium, effectively mitigating risks while preserving agent utility.

We show that the Vehicle Routing Problem (VRP) can be reformulated as a Graph Edit Distance (GED) maximization problem. Under a simple edge-deletion cost model, minimizing total route cost is equivalent to maximizing the total weight of edges deleted from the complete instance graph. This formulation models VRP at the edge level, where solutions are defined by selected edges rather than route sequences, enabling structural analyses that are difficult in classical formulations: per-edge attribution of solution quality, decomposition of the optimality gap, characterization of solution sparsity, and identification of edges that are hard to reach by greedy construction. Theoretically, we establish a merge-decomposition theorem showing that Clarke-Wright savings equal per-merge GED increments, and an approximation-transfer theorem that turns GED approximation ratios into VRP cost bounds. Using this reformulation, we analyze 90 CVRP benchmark instances with known optimal solutions. We find that optimal routing graphs use only 5.5% of available edges, that approximately 3.0% of optimal edges are consistently not found by Clarke-Wright heuristics under repeated restarts, and that the cost gap decomposes into missed optimal edges and substituted non-optimal edges of comparable total weight. The edge-additive objective provides a natural per-edge supervision signal for future graph neural network approaches to edge prediction, suggesting a potential connection to graph neural network approaches that we leave for follow-up work.

Recent breakthroughs in instruction-based image editing have captured significant attention, as models are now capable of handling real-world editing demands with the practicality required by everyday users. However, editing models trained primarily for single-turn edits often break down in multi-turn editing--the natural interactive setting where a user iteratively refines an image based on the model's own previous outputs. This failure stems from the all-or-nothing requirement, where a single failed turn compromises the entire sequence, and error propagation, where exposure bias leads to compounding editing errors. To address these challenges, we introduce MT-EditFlow, a flow-matching reinforcement learning framework designed to optimize reward signals for sequential image editing. MT-EditFlow integrates a multi-turn perspective with a multi-reward formulation to provide a unified structure applicable to both GRPO and NFT-based reinforcement learning methods. We systematically analyze and optimize the reward signal by investigating effective scoring strategies for turn-level aggregation, VLM reasoning modes to trade off reward bias and variance, and advantage fusion levels to prevent reward hacking. Our findings reveal that broadcasting the aggregated advantage across the entire editing trajectory effectively bridges the gap between local planning and global multi-turn task success. Extensive experiments demonstrate that MT-EditFlow significantly improves performance across diverse base models. Notably, it boosts FLUX.1-Kontext-dev by 6.85 points in turn-3 overall performance, surpassing state-of-the-art open-source models such as Qwen-Image-Edit. By maintaining high marginal success rates and reducing exposure bias, MT-EditFlow provides a foundation for more reliable and natural human-AI collaboration in visual content creation.

Schema-constrained information extraction from diverse educational and labor-market corpora remains an open challenge in natural language processing because existing pipelines rely primarily on lexical-surface methods that cannot recover implicit competencies, lack grounding in shared taxonomies, and provide no formal measures of extraction reliability or document-level completeness. To address these limitations, this paper proposes a four-stage NLP framework that combines (i) schema-constrained prompting of a two-model frontier-LLM ensemble against a JSON Schema-enforced seven-slot competency formalism, (ii) Sentence-BERT (SBERT) alignment of the extracted records against an eleven-domain ESCO v1.2.1 controlled vocabulary, (iii) a two-tier adjudication protocol that resolves inter-model disagreements, and (iv) a verification mechanism that combines per-slot Cohen's kappa, schema conformance, and document-level completeness audits. The framework is instantiated for a critical application in higher-education quality assurance, namely curriculum-labor market alignment for the ABET-accredited BSc Computer Science program at the United Arab Emirates University. The pipeline extracts 400 competency records from the 85-course 2025-2026 study plan and aligns them, under a five-scope analysis ranging from the computing core to a probability-weighted student trajectory, with 30 job postings (483 requirement clauses) at an SBERT cosine threshold of 0.50. The extractor achieves Cohen's kappa of 0.79 on the skill slot, with 100% schema conformance and 100% document-level completeness. The alignment surfaces interpretable supply-demand gaps of 25.0% in general and transversal skills, 13.8% in algorithms and computational theory, and 12.2% in software engineering and project management, with a near-zero 1.8% gap in artificial intelligence and data science despite 38.6% supply coverage.

Vehicle re-identification focuses on retrieving images of the same vehicle from a gallery given a query image. Upon closer inspection of commonly used datasets, we observe that vehicles with few visual differences-e.g., the same make, model, and color-appear in both the training and test sets. As a result, methods that effectively memorize the training data tend to perform well on these test sets but struggle to generalize to other datasets. In this paper, we address this issue by proposing a novel evaluation approach that more effectively measures generalization capability to unseen vehicle types. To further study generalization performance, we also propose splitting the evaluation based on view, allowing us to differentiate the effect of viewpoint robustness from that of same-view re-identification. Our findings reveal that most state-of-the-art methods struggle with unseen vehicle types, and that their robustness to viewpoint changes and attention to detail are limited to vehicle types seen during training.

We consider LLM-based algorithm development through a case study on contractionorder optimisation for tensor networks with OpenEvolve. We pay particular attention to the choice of the LLM as well as design choices such as evaluation metric and test instances. Our results highlight both the promise of verifier-guided evolutionary coding agents for algorithm development/improvement and the continuing importance of evaluation, validation, and interpretation -- and corresponding challenges -- by the human scientist.

Open-set test-time adaptation (TTA) updates models on new data in the presence of input shifts and unknown output classes. While recent methods have made progress on improving in-distribution (InD) accuracy for known classes, their ability to accurately detect out-of-distribution (OOD) unknown classes remains underexplored. We benchmark robust and open-set TTA methods (SAR, OSTTA, UniEnt, and SoTTA) on the standard corruption benchmarks of CIFAR-10-C at the small scale and ImageNet-C at the large scale. For CIFAR-10-C, we use OOD data from SVHN and CIFAR-100 in their respective corrupted forms of SVHN-C and CIFAR-100-C. For ImageNet-C, we use OOD data from ImageNet-O and Textures in their respective corrupted forms of ImageNet-O-C and Textures-C. ImageNet-O is nearer to ImageNet, as unknown but related object classes (like ''garlic bread'' vs. ''hot dog'' for food, or ''highway'' vs. ''dam'' for infrastructure), while Textures is farther from ImageNet, as non-object patterns (like ''cracked'' mud, ''porous'' sponge, ''veined'' leaves). We evaluate the accuracy and confidence of TTA methods for InD vs. OOD recognition on CIFAR-10-C and ImageNet-C. We verify the accuracy of each method's own OOD detection technique on CIFAR-10-C. We also evaluate on ImageNet-C and report both accuracy and standard OOD detection metrics. We further examine more realistic settings, in which the proportions and rates of OOD data can vary. To explore the trade-off between InD recognition and OOD rejection, we propose a new baseline that replaces softmax/multi-class output with sigmoid/multi-label output. Our analysis shows for the first time that current open-set TTA methods struggle to balance InD and OOD accuracy and that they only imperfectly filter OOD data for their own adaptation updates.

Reliable autonomous UAV swarms in Search and Rescue (SAR) missions require fault-tolerant coordination capable of sustaining operations despite agent degradation. This paper introduces the Intelligent Replanning Drone Swarm (IRDS), a distributed coordination architecture designed for resource-constrained environments. The proposed framework employs a Reverse-Auction market mechanism where agents bid to service search sectors based on a distance-weighted cost function, coupled with a geometric consensus protocol for target verification. We evaluate the approach through physics-based simulations (N=8 agents, 8x8 grid) subjected to stochastic fault injection. Results indicate that the swarm autonomously reallocates tasks from failed agents with low latency relative to the total mission duration, maintaining a mission success rate of 93% under 25% workforce degradation. The proposed framework demonstrates a robust, empirically tested method for self-healing aerial robotic coordination.

While prompt engineering is instrumental in maximizing the capabilities of Large Language Models (LLMs) during inference, the role of prompts during training remains critically underexplored. Prevailing fine-tuning paradigms typically treat training prompts as mere surface forms, assuming that semantically equivalent instructions yield identical learning outcomes. However, we reveal that this equivalence is deceptive: while paraphrased prompts often lead to comparable in-task performance, they induce drastically different cross-task impacts regarding catastrophic forgetting and generalization. Crucially, these impacts are positively correlated across tasks, indicating the existence of superior prompts that consistently yield better performance. Furthermore, we discover that these superior prompts can be robustly identified by task loss prior to learning. Leveraging these insights, we introduce State-Adaptive Prompt Optimization (SAPO), a lightweight yet effective training strategy that shifts task formulation from a static input to a dynamic, state-adaptive variable. Comprehensive experiments on diverse benchmarks confirm its effectiveness, which significantly mitigates forgetting while improving generalization, achieving substantial performance gains over state-of-the-art methods. These results provide insights into how training prompts shape learning dynamics and offer a practical recipe for robust fine-tuning. Our code is available at https://github.com/Eric8932/SAPO.

This paper presents a comparative study of parameter-efficient fine-tuning (PEFT) methods, including LoRA and QLoRA, applied to the task of adapting large language models to the Bashkir language, a low-resource agglutinative language of the Turkic family. Experimental evaluation is conducted on a Bashkir text corpus of 71k documents (46.9M tokens) using models of various architectures: DistilGPT2, GPT-2 (base, medium), Phi-2, Qwen2.5-7B, DeepSeek-7B, and Mistral-7B. To improve the reliability of results, each configuration was trained with three different random seeds. The lowest perplexity on the test set was obtained for GPT-2 medium with full fine-tuning (3.34). Meanwhile, QLoRA applied to Mistral-7B (3.79) and Phi-2 (3.81) achieved comparable quality with over 40 times fewer trainable parameters. However, we also observed cases of significant quality degradation when using PEFT for certain architectures (e.g., DeepSeek-7B with rank 8, perplexity = 129.55), indicating that the outcome depends critically on the choice of the base model and its tokenizer. Additionally, a qualitative analysis of generated texts based on Bashkir prompts revealed that models with the best perplexity do not necessarily produce the most coherent outputs: QLoRA-tuned models generated monolingual Bashkir continuations, whereas the fully fine-tuned model with the lowest perplexity frequently switched to English. The results suggest that QLoRA on 7B-scale models offers an effective compromise between quality and computational cost for Bashkir. To ensure reproducibility, open data, code, and trained adapters will be released upon acceptance.

Acoustic side-channel attacks (ASCA) on keyboards pose a significant security risk, as keystrokes can be inferred from typing acoustics, revealing sensitive information. Prior ASCA studies are limited by small-scale datasets with restricted diversity in users, keyboards, and environments, constraining analysis across devices, microphones, and noise conditions. We introduce HEAR, a dataset designed to study ASCA along three axes: keyboard generalization, noise adaptation, and user bias. HEAR contains recordings from 53 participants using 37 laptop keyboards, collected in three realistic settings: (1) external microphone capture, (2) device microphone capture without network noise, and (3) VoIP-based streaming capture. This enables controlled evaluation across users, keyboards, and environments. On HEAR, we establish an ASCA benchmark spanning conventional features and pre-trained representations from raw audio and spectrograms in unimodal and multimodal settings. We propose DECKER, a domain-invariant keystroke inference framework with four stages: (1) Keyboard Signature Normalization to reduce device coloration, (2) domain-adversarial disentanglement to suppress keyboard identity, (3) supervised cross-keyboard contrastive alignment to enforce key consistency, and (4) Acoustic Style Randomization to synthesize unseen keyboard responses. We further explore sentence-level inference using an LLM-based post-processing layer to refine keystroke sequences via linguistic context. Results on HEAR show DECKER improves keystroke identification over strong baselines, particularly in cross-keyboard and cross-user settings, with further gains from language-model rectification. These findings highlight that ASCA remains effective across diverse users, devices, and noisy environments, underscoring its practical security risk.

As artificial intelligence (AI), including machine learning (ML) models and foundation models (FMs), are increasingly deployed in high-stakes domains, ensuring their trustworthiness has become a central challenge. However, the core trustworthy AI objectives, such as fairness, robustness, privacy, and explainability, are hard to achieve simultaneously, especially while preserving utility. This position paper argues that causality is necessary to understand and balance trade-offs in performance and multiple objectives of trustworthy AI. We ground our arguments in re-interpreting trustworthy AI trade-offs as incompatible invariance requirements under different changes to the data-generating process. We then illustrate this argument through case-study analyses from the literature and a stylized synthetic-data simulation, showing that causality provides a unifying framework for understanding how trade-offs in trustworthy AI arise and how they can be softened or resolved through selective invariance. This perspective applies to both classical ML models and large-scale FMs. Finally, we outline open challenges and opportunities for using causality to build both trustworthy and high-performing AI.

This paper presents the first comprehensive comparative analysis of modern machine learning architectures for transliteration between Tajik (Cyrillic script) and Persian (Arabic script). A key contribution is the creation and validation of a unique parallel corpus aggregated from multiple heterogeneous sources, including crowdsourced projects, lexicographic pairs, parallel texts of "Shahnameh", diplomatic articles, texts of "Masnavi-i Ma'navi", official terminology lists, and transliterated correspondences. The initial dataset comprised 328,253 sentence pairs; a representative subset of 40,000 pairs was formed using stratified random sampling. The experiment compared six classes of models: rule-based baseline, LSTM with attention, character-level Transformer, G2P Transformer (trained from scratch), pre-trained multilingual models (mBART, mT5 with LoRA), and byte-level ByT5. Results demonstrate the overwhelming superiority of ByT5 (chrF++ 87.4 for Tajik to Farsi, 80.1 for reverse). The G2P Transformer significantly outperformed mBART (72.3 vs. 62.2 chrF++) despite limited data. Models using subword tokenization (mT5) failed completely (chrF++ less than 18.5). The findings demonstrate that for accurate transliteration of the Tajik-Farsi pair, architectures operating at the byte or character level are unequivocally more effective than traditional multilingual Seq2Seq models relying on subword tokenization.

Human evaluation remains the primary standard for assessing modern AI systems, yet annotator disagreement, bias, and variability make system rankings fragile under standard majority vote aggregation. Majority vote discards annotator reliability and item-level ambiguity, often yielding unstable comparisons across annotator subsets. We introduce STABLEVAL, a disagreement-aware evaluation framework that models latent item correctness and annotator-specific confusion patterns to produce posterior expected item credit and calibrated agent-level scores. Unlike label-denoising approaches such as Dawid-Skene, STABLEVAL is explicitly designed for stable and uncertainty-aware system evaluation rather than hard label recovery. We formalize ranking stability as a first-class evaluation objective and analyze how aggregation methods preserve or distort underlying annotator behavior. Across controlled synthetic experiments and multiple real-world human-annotated benchmarks, majority vote exhibits increasing score error and ranking instability under annotator heterogeneity and adversarial noise, while STABLEVAL yields more stable and statistically grounded system rankings. These results demonstrate that modeling disagreement is essential for robust and reproducible AI evaluation.