Papers

Panels of inexpensive LLM judges increasingly make accept-or-escalate decisions. In factuality settings, accepting a claim because several reference-free judges agree can create a hidden risk: agreement may reflect shared false-negative blind spots rather than independent evidence. We introduce JuryProbe, an empirical consensus-risk diagnostic for reference-free factuality judge panels, paired with a calibration-based routing policy. JuryProbe estimates consensus risk from a labeled calibration probe using false-negative-only (FN-only) judge correlation and false-consensus lift; when flagged high-risk, reference-free majority accepts are routed to the same judges with trusted references. On audited FEVER corruptions, reference-free panels show correlated false negatives (FN-only correlations 0.402 and 0.368; lifts 3.13x and 18.13x), while unanimous false consensus drops to zero under a trusted-reference best-case diagnostic on both minimal-pair and non-minimal-pair evidence. In flagged settings, the routed policy is by construction equivalent to grounding every reference-free majority accept (verified in 34/34 splits): improvement comes from accept-conditioned grounding, while the diagnostic determines whether to activate it. A fixed, pre-specified rule flags 8-10 of 10 splits across synthetic, benchmark-authored, and scientific families and 0 of 10 on a negative control, where standing down avoids 28% of reference acquisitions at a 0.004 increase in false accepts. False-accept reduction persists under weak BM25 retrieval at substantial coverage cost, while stale stand-down labels require periodic recalibration. JuryProbe provides no formal risk guarantee and does not establish reliable stand-down on natural panels; its supported contribution is an empirical diagnostic of high-risk panel error dependence.

Federated learning (FL) is vulnerable to multi-level attacks. However, existing methods address them separately, leaving FL exposed to data leakage, unauthorized reuse, and malicious gradient manipulation. In this work, we propose an FL framework that couples keyed context-provenance watermarking with verifiable lattice-based secure aggregation of Real-World Anchored Watermarking and Lattice-Based Zero-Knowledge Secure Aggregation. At the data layer, we propose a Kerckhoffs-compliant scheme that utilizes Physical Anchor Metadata (PAM) to ensure data provenance. PAM is defined as a context-provenance token derived from trusted infrastructure data (time, location, and server ID) and then subjected to a keyed HMAC-SHA-256 transformation to produce a watermark payload that cannot be generated without the client's secret key. We further design FMGAN, a GAN-based robust image watermarking framework that embeds this transformed payload using a feature fusion module and a Mamba-guided linear attention mechanism. At the computation layer, we adopt a lattice-based zero-knowledge secure aggregation (LZKSA) protocol that verifies key correctness, L2 norm bounds, and cosine similarity constraints over committed gradients without revealing private updates. The RLWE-based design guarantees post-quantum security. Extensive experiments validate the complementary protection of the two layers under composite attack scenarios. To our knowledge, no prior verification workflow has jointly evaluated both layers in a hybrid, end-to-end trustworthy FL framework.

We introduce FlavorBench: a benchmark for Compiling Dense Deterministic Answer Maps from a Versioned Culinary Embeddings Model. We test 27 frontier large language model endpoints on 534 substitution, pairing and constraining tasks for tasks that request a 3-ingredient portfolio from 8 candidates and score all 56 resulting portfolios. We conducted multiplicity-controlled paired tests on 101 of 351 model contrasts for this task-set. The largest point estimate on this task-set was achieved by Grok 4.6 at 65.1. The same rankings for this task-set were also achieved on several independently-compiled panels (using a variety of familiar metrics, task filters, etc.) and 3 public Epicure checkpoints. We present a 3-seed post-training study where LoRA SFT of a Qwen3-0.6B checkpoint on 270 optimal answers for Epicure to score on this task-set resulted in a 13.3 point gain on 84 anchor-disjoint maps (compared to format and label-matched control; 95% CI: 6.52, 20.29; p = 0.000170).

Convolutional Neural Networks (CNNs) face a dual challenge: vulnerability to adversarial attacks and prohibitive training cost. Adversarial training is effective but expensive, a burden that grows as learning shifts to the energy-constrained edge. This paper addresses both through GPU undervolting during training. Reducing supply voltage introduces stochastic perturbations that act as implicit regularization, improving robustness while lowering power. We characterize undervolting-induced faults at the bit level, then train LeNet, VGG-6, and MobileNetV3 on MNIST and CIFAR-10 under two training regimes, standard and adversarial, each at nominal and undervolted voltage, and evaluate all models against adversarial attacks. In both regimes, the undervolted model consistently achieves higher adversarial accuracy than its nominal-voltage counterpart, showing that hardware-induced faults strengthen even adversarial training. Because dynamic power scales quadratically with supply voltage, these robustness gains arrive with substantial energy savings. GPU undervolting is therefore a readily deployable hardware-level defense requiring no algorithmic change, and opens a promising direction in which robustness and energy efficiency move together.

AI agents extend large language models from single prompt-response interactions to long-running, goaldirected workflows that issue many model calls, invoke tools, and interact with external environments. These workflows enable tasks such as software repair, data analysis, and experiment management, but their repeated model invocations can incur substantial carbon emissions. This paper characterizes the carbon emissions of OpenClaw agent workloads using WildClawBench, and shows that emissions depend on token consumption, context cache reuse, and the carbon intensity of the grid. Our characterization identifies deadline flexibility as an opportunity for carbon-aware execution: agent tasks can wait for lower-carbon-intensity periods or shift to lower-carbon grids. However, doing so requires handling uncertain execution time for temporal shifting and cached context recomputation during spatial shifting. We present AgentDecarbonizer, a carbon optimizer for AI agents that runs alongside OpenClaw. Given a task prompt and user-specified deadline, AgentDecarbonizer conservatively estimates task duration and selects deadline-feasible execution schedules, while accounting for cache recomputation overhead during spatial shifting. Evaluated on WildClawBench workloads with 60 agent tasks across four grids, AgentDecarbonizer reduces carbon emissions by up to 57.9 % compared with a carbon-agnostic baseline and by up to 37.5 % compared with a baseline that selects the carbon-optimal grid at task start time.

Mixed-type data containing both numerical and categorical variables arise in many scientific and real-world applications. Existing representation learning and generative modeling approaches typically focus either on reconstruction accuracy or unconditional data generation, but often fail to recover the full conditional distribution of the data while preserving interpretable structural relationships between heterogeneous variable types. In this work, we introduce Conditional-Independence-Regularized Distributional Autoencoders, a framework for learning low-dimensional representations of mixed-type data through conditional distribution matching and structural regularization. Our method combines an energy-score-based objective for numerical variables, a likelihood-based objective for categorical variables, and an auxiliary conditional independence regularization term encouraging the learned representation to capture the dependence between numerical and categorical components. We provide theoretical analysis showing that the optimal representation balances unexplained numerical variability, conditional entropy of categorical variables, and residual conditional dependence. Empirically, the proposed method achieves strong performance on both synthetic and real-world datasets, substantially improving categorical distribution recovery, achieving competitive overall conditional distribution recovery, and preserving mixed-type dependence structure. The code has been made available at GitHub.

Diffusion models (DMs) have emerged as powerful generative priors for MRI reconstruction with promising results. Yet DM-based methods require extensive iterative refinement, limiting their practical deployment. Consistency models (CMs) provide a compelling alternative, aiming to map out the diffusion trajectory in a single pass, enabling faster generation. In this work, we propose CM-RED, a novel MRI reconstruction method that integrates a pretrained CM into the regularization by denoising (RED) scheme. Our method builds on accelerated proximal gradient RED (RED-APG), and further incorporates controlled noise injection during the update steps to enhance generative diversity and accelerate convergence. Extensive experiments on the fastMRI knee and brain datasets demonstrate that CM-RED achieves high-quality reconstructions across multiple anatomies, contrast weights, acceleration factors, and undersampling patterns, using only 4 network function evaluations (NFEs). The proposed method consistently outperforms existing DM- and CM-based approaches in both quantitative metrics and visual fidelity, and exhibits strong robustness to hyperparameter variations, highlighting CM-RED as an efficient and effective generative framework for accelerated MRI reconstruction. The source code and pretrained models are publicly available at https://github.com/MerveGulle/CM-RED.

Domain shift across clinical centers using different imaging hardware or acquisition protocols remains a fundamental barrier to deploying deep learning models for prostate cancer (PCa) detection. Existing test-time adaptation (TTA) methods address distribution shift through entropy minimization or augmentation-based self-supervision, correcting for statistical differences in image appearance but ignoring the anatomical structure of the target domain. We propose ANT, a segmentation-guided TTA framework that adapts a pretrained cancer detection encoder to the target domain by solving an auxiliary prostate segmentation task at test time, supervised by pseudo-masks from a frozen pretrained segmentation network. By aligning encoder representations to prostate anatomy in the target domain, ANT corrects domain-specific feature drift while preserving cancer-discriminative structure. The model was trained on 693 patients imaged with an earlier-generation micro-ultrasound scanner in a multi-center clinical trial, and evaluated on 118 patients acquired with a newer-generation system across two centers in another clinical trial. Under a leave-one-center-out protocol with identical evaluation conditions across all methods, ANT improves mean AUC by 2.9% and 3.6% at the biopsy-core and patient levels, respectively, over no adaptation, outperforming TTA baselines. Code is available at: https://github.com/ObedDzik/ant.git.

The critical failure modes in deployed large language models (LLMs) are cross-dimensional: a model can score 99.3 in safety alignment while refusing one in three benign queries, or improve across every capability metric while losing 21 points in privacy. Existing evaluation frameworks that assess safety, security, and privacy independently cannot detect these patterns. We introduce aiXamine, a unified black-box platform that evaluates LLM trustworthiness across safety, security, and privacy as interdependent properties. aiXamine orchestrates 46 tests across nine services through an automated red-teaming pipeline, producing hierarchical risk profiles, from prompt-level diagnostics to cross-service trade-off analytics, that enable reproducible comparison of proprietary and open-weight systems under identical conditions. Applying aiXamine to over 120 LLMs through more than 5,000 test runs, we conduct the largest joint safety, security, and privacy study to date and uncover three cross-dimensional phenomena invisible to single-axis evaluation. First, safety enforcement incurs a quantifiable safety tax: stronger alignment systematically increases over-refusal, forcing providers to choose between protection and utility. Second, privacy is near-orthogonal to other trustworthiness dimensions and not captured by standard alignment. Third, we identify and formally characterize distillation-induced robustness collapse: off-policy distillation without on-policy correction causes entropy collapse, catastrophically destroying robustness (56.9$\to$2.6) on the same base architecture. These findings, compounded by diminishing returns from scale and category-dependent safety behaviors, demonstrate that trustworthiness is inherently multi-dimensional: progress along one axis does not guarantee, and can actively undermine, progress along others, yet current alignment methods treat it as a single objective.

Disaster damage is spatial: buildings rarely fail in isolation. Yet using spatial context for damage classification remains surprisingly underexplored, and many pipelines still rely primarily on per-building appearance cues even when the dominant uncertainty is spatially structured. Complicating matters, the right neighbourhood is not the same across events. Floods, hurricanes, and wildfires can exhibit very different clustering behaviour, making spatial reasoning valuable but easy to misuse - naive context aggregation can improve visual coherence while oversmoothing boundaries or propagating structured errors. We study this tension on xBD (the dataset used in the xView2 challenge) in a controlled post-localization, classification-only setup: each building is represented by a pre/post combined (PPC) patch cropped from the provided polygons, and spatial context is modelled with GPS-derived building graphs. Our approach keeps local evidence "close" by preserving strong spatial relationships in disaster damage patterns, while bringing only the right neighbours "closer" through a disaster-type-conditioned graph model that injects a learnable multi-scale spatial kernel prior into attention, allowing the effective neighbourhood scale to adapt across disaster types rather than being learned as a single global smoothing rule. To discourage coherence-by-smoothing, we add a residual de-correlation loss that penalizes positive Moran's~I in prediction residuals. We evaluate the method under event and dataset shift with a leave-one-event-out (LOEO) protocol on xBD and cross-dataset transfer from xBD to Ida-BD. The model improves macro-F1 and substantially reduces residual spatial autocorrelation under zero-shot event shift, indicating better use of spatial context rather than naive smoothing and enabling more reliable transfer to unseen events within known disaster types.

NVIDIA provides a SASS disassembler but no public SASS assembler for recent data-center GPUs, limiting controlled machine-code rewriting. We present F2Asm, which learns exact 128-bit SASS encoders from paired disassembly and original CUBIN instruction words. To our knowledge, F2Asm is the first system to learn SASS instruction encoders as vector-valued affine maps over $\mathbb{F}_2$ and the first open-source NVIDIA SASS assembler to support Rubin SM107. F2Asm uses Gaussian elimination over $\mathbb{F}_2$ to incrementally build a compact basis, detect inconsistencies, and reject inputs outside the learned span. F2Asm separates target-specific control bits, relocation rules, and CUBIN metadata from its learning algorithm. We train encoders for Hopper SM90/SM90a, Blackwell SM100, and Rubin SM107 using 3,225 CUBINs from pinned NVIDIA and third-party production libraries, CUDA 13.3 packages, and CUDA 13.4 Developer Preview archives. In round-trip tests, F2Asm reassembles each CUBIN's disassembled SASS, and all compared executable text sections match the originals exactly. Joint training with F2Asm yields one shared encoder for five Blackwell SM targets and another for three Rubin SM targets, providing strong evidence of a common SASS encoding scheme for instructions shared within each family. Continual training extends the Rubin encoder to all 17,159 previously unsupported cuTile and GROMACS queries with 1,504 additional basis rows, matching the derived lower bound.

Joint-embedding predictive architectures are selected almost universally by linear probing and effective rank. We report a case where both read healthily while the representation carries zero usable instance information. We repair it, and a second failure appears: the repaired metric saturates on a target carrying no structural information. Our corpus is a scientific-reasoning graph over 57,903 articles, each a subgraph. A Graph-JEPA predicts one masked aspect from a subgraph's remaining aspects, attaining linear-probe accuracy 0.871 and effective rank 18-47, yet retrieval recovers 0.00 of 14.4 bits (MRR 1.9e-4 vs chance 1.99e-4, p=0.98). Three upper bounds on the same pool and code recover nearly everything (+14.28, +14.34, +14.22 bits), ruling out corpus, masking, pool, and metric as causes. We trace this to variance allocation - frozen inputs place 86.05% of variance on subgraph identity and 0.40% on aspect identity, while trained latents place 0.39% and 99.61%. This is a property of the objective's optimum: the degenerate solution is a global minimum of the coupled predictor/EMA-target objective, present already at init. A repaired configuration reaches 14.377 of 14.379 bits, above the 13.865-bit oracle; reverting the loss to regression drops it to 0.307 bits, confirming it. Yet the repair licenses nothing about reasoning: the target is reducible, since intra-subgraph edges are a deterministic function of node census. The oracle reaches 96.4% of the ceiling, and our largest effect is the learning-rate schedule, not architecture. Bits and a reasoning probe show no relation across ten cells. A data-derived target fails a quality gate - 25.96% of nodes are duplicate placeholders, and the rest is more generic than supporting evidence. Rank, probes, and metrics can all saturate on an unsupportive evaluation. We release a harness with a reducibility audit and target gate.

Deploying high-accuracy neural networks on resource-constrained edge devices remains challenging, as existing approaches treat training, compression, and hardware synthesis as separate stages, leaving a gap between software-trained models and efficient end-to-end deployment with limited support for interpretability. We propose Bern2Edge, an end-to-end framework that uses knowledge distillation to convert a pretrained teacher feed-forward network into hardware-efficient representations via Bernstein polynomial activations. This representation enables two deployment paths: (i) a high-fidelity LUT-based realization that preserves model fidelity under compression, and (ii) a symbolic rule-based representation derived from Bernstein activation geometry, enabling interpretable inference with explicit input-space constraints. The resulting BNNs achieve up to 2.12 percentage-point (pp) accuracy improvement over ReLU under identical compression constraints. At the system level, Bern2Edge achieves up to 99.8% latency reduction and 95.2% BRAM reduction relative to a W8A8 quantized teacher on an AMD Xilinx KV260 FPGA, while maintaining accuracy within 0.5 pp, and further deploys on a low-power Spartan-7 XC7S15 FPGA. The rule-based path reduces DSP usage by up to 89.0% at a cost of 1.5 pp in total accuracy.

AI tools increasingly support tasks across the scientific research cycle, from experiment design and manuscript preparation to peer review. At the same time, the continuing growth in conference submissions has increased the burden on meta-reviewers, who must synthesize reviewer feedback, author rebuttals, and manuscript revisions. To address this concern, this paper introduces Metag, a dataset to accelerate the development of meta-reviewing agents, specifically to identify changes made to scientific articles during the review-rebuttal process. Each instance contains a reviewer concern, the author's proposed resolution, and the manuscript diffs implementing the stated change. Metag is collected by obtaining manuscript versions from before the review deadline and after acceptance, computing differences between the two documents, and asking human annotators to align these differences with action items from OpenReview discussions. The resulting dataset consists of 349 high-quality action items tied to paper differences and will enable building methods to empower meta reviewers to quickly identify whether authors have addressed reviewer statements and where in the paper those changes have been made, resulting in additional transparency and traceability throughout peer review. The dataset is publicly available at https://github.com/microsoft/Metag-dataset.

We develop analytical and particle-based methods for uncertainty propagation in random neural network models, where both the inputs and network parameters are allowed to be random. Building on the piecewise-linear structure of the Leaky ReLU activation function, we derive a local approximation of the neural network output with respect to perturbations in both its inputs and parameters. This approximation is exact for perturbations that preserve the network activation pattern, and it allows us to compute analytical expressions for the probability density function and characteristic function of the network output, together with closed-form approximations for its mean and covariance. We extend this uncertainty propagation framework to autonomous dynamical systems whose one-step evolution map is represented by a random neural network. Repeated application of this map defines an autoregressive model, for which we derive recursive equations to propagate uncertainty in both the state and network parameters over time. These equations explicitly account for the state-parameter cross-covariance that develops under successive iterations of the network. Numerical experiments on the Lorenz-63 system and the Kuramoto-Sivashinsky equation demonstrate accurate uncertainty propagation through the predictability horizon and the applicability of the proposed framework to high-dimensional dynamical systems.

Optimal learners are tailored to exploit the i.i.d.\ data assumption underlying the classic PAC model. What if an i.i.d.\ training sample were corrupted with correctly labeled examples drawn from an otherwise unrelated, even adversarial source? This model of learning with monotone adversarial corruptions was recently introduced by Larsen et al. (2026), who demonstrated that all known optimal binary learners suffer increased error rates in this setting, from $O(d / n)$ in the PAC model to $\Omega (d \log(n / d) / n)$ under monotone corruption. Mehrotra (2026) proved this logarithmic factor to be necessary for binary classification, but left open the consequences of corruption for more general learning settings, such as multiclass classification and partial binary concept classes. As our primary result, we demonstrate that monotone adversaries are frighteningly more powerful in each of these settings. We exhibit a learnable multiclass problem, of DS dimension only 2, that becomes altogether unlearnable under a monotone adversary, and show an analogous result for partial binary concept classes. These results are achieved by an adaptive adversary permitted to view the original i.i.d.\ training set $S$ and to insert $b < \infty$ corrupted datapoints into $S$. In the multiclass example, the adversary need only insert a linear number $b = |S| = n$ of datapoints. We complement these impossibility results by proving that every class remains learnable when the number of adaptive additions is $o(n)$, which our previous multiclass lower bound proves to be tight. We further observe that the classic multiclass error rate of $O(d_{\mathrm{DS}} / n)$ remains achievable against adaptive adversaries restricted to a known constant budget $b = O(1)$, against semi-adaptive adversaries viewing only a $p$-fraction of $S$ for $p \in (0, 1)$, and against oblivious adversaries that cannot view $S$.

Feature selection is a highly relevant task in a data-driven knowledge discovery project. Several techniques have been developed aiming at finding the features that influence most an outcome to predict, including mutual information and, in recent years, the data-based sensitivity analysis. The present research focus on analyzing the advantages and disadvantages of each of these two techniques, by applying both to a bank telemarketing case. Thereafter, a logistic regression model is built on the tuned set of features identified by each of the two techniques as the most influencing set of features on the success of a telemarketing contact, in a total of 13 features for mutual information and 9 features for the data-based sensitivity analysis. The latter performs better for lower values of false positives while the former is slightly better for a higher false positive ratio. Thus, mutual information becomes a better choice if bank managers intend to reduce slightly the cost of contacts without risking losing a high number of successes. Such results show that mutual information, although not recent, is still a valid method for feature selection. On the other side, the data-based sensitivity analysis selection achieved good prediction results with less features.

Kernel density estimation converts finite samples into probability densities, but its performance depends critically on bandwidth selection. Classical selectors prescribe the sample-to-bandwidth rule analytically or asymptotically, or solve a new optimization for each sample. An amortized framework is proposed that instead learns this mapping across a distribution of density-estimation tasks by optimizing the logarithmic score. A truncated-and-renormalized bounded-support formulation enables stable learning across heterogeneous tasks, while affine standardization allows a selector trained on a single reference interval to transfer across bounded intervals. Experiments under Gaussian sampling, a multi-family benchmark, and randomized Gaussian-mixture training show that the amortized selector consistently and substantially outperforms Silverman's rule, the Sheather–Jones selector, and least-squares cross-validation, with especially large gains in small and heterogeneous samples. Finite Gaussian mixtures provide a generic training mechanism supported by their $L^1$ approximation property. Selectors trained in this way generalize strongly across different density structures, allowing the same trained selector to be applied directly to finite samples from unknown densities without specifying or fitting a distributional family. This combination of broad applicability and strong empirical performance makes the framework attractive for a wide range of applications in which finite samples or ensembles must be converted into continuous probability densities.

Subliminal trait transfer allows a student model to acquire behavioral dispositions from teacher-generated data in which the trait is not semantically expressed. Recent work explains how such signals enter gradients, but not how they survive source removal or acquire different signs under later training. We treat parameters and optimizer moments as a single trainer state and derive an exact transport-valuation identity separating observer-independent propagation of the source perturbation from the value assigned by a future continuation and behavioral readout. State surgery identifies the first moment as a causal carrier. Transplanting it alone leaves parameters, hidden states, and outputs unchanged at the cut, yet source-free updates generate growing parameter and hidden-state differences; transplanting parameters with the first moment recovers the terminal behavioral response. Sending the same source-induced difference through matched futures produces negative, near-zero, and positive Qwen effects (-0.658, +0.008, and +0.658 seed means). This ordering recurs in all 12 Llama-3.2-1B seeds after eight updates, while state-difference norms remain nearly equal across routes. Both contrasts grow in every paired seed when the continuation extends to sixteen updates. A full-horizon costate predicts all 42 Qwen route-mean signs and all 21 resolved Llama ordinary-route signs. Observer-independent transport also replicates across Qwen, SmolLM2, and Llama, while the complete-state recurrence predicts physical, hidden, and fixed-head responses in non-LoRA MNIST systems, including CNNs trained with AdamW and momentum SGD. Together, these results identify a two-stage mechanism for subliminal trait transfer: optimizer state transports the source perturbation, and later training determines its behavioral value.

Selecting the optimal neural-operator prediction during deployment is challenging when high-fidelity reference solutions are unavailable. We demonstrate that under a squared Hilbert-space loss, ranking a finite model library depends strictly on the low-dimensional span of candidate differences, allowing us to score all models simultaneously using a single anchor-based linearized response of the governing equation. This shared physical diagnostic accurately recovered over 99.6% of pairwise preferences and 99.0% of optimal checkpoints across diverse Fourier and convolutional operator libraries for fluid, reaction-diffusion, and wave dynamics. Furthermore, the corrected physical proxy frequently outperformed the best individual candidates, and we establish computable sufficient conditions that rigorously certify exact decisions for strongly monotone discretizations. By exploiting the local dynamical response rather than raw defect magnitude, this framework enables the reliable and highly efficient deployment of scientific surrogates without requiring ground-truth data.

Classification of edible oils in processed foods is important for food quality, fraud prevention, and regulatory compliance. This study develops a Mutually Exclusive, Collectively Exhaustive framework integrating spectral organization, interpretable classification, Physics-Informed Artificial Intelligence (PI-AI), and Frugal AI-based feature reduction. Five edible oils were analyzed in pure form and within a fried-potato-chip matrix using t-SNE, K-means clustering, Decision Trees, and Non-Negative Least Squares (NNLS)-based spectral decomposition. Unsupervised analyses showed stronger class organization and separability in pure oils, while food-matrix effects caused substantial spectral overlap. Decision Trees achieved 100% classification accuracy for pure oils using only four Raman variables from 1866 spectral features. These variables represented only 0.21% of the available spectral information while retaining perfect test-set performance. Two variables associated with lipid unsaturation (about 1650 cm-1) and hydrocarbon-chain organization (about 1127 cm-1) remained important after NNLS matrix correction. Their combined contribution increased from 50% in pure oils to about 62% and 89% in paper-subtracted and paper-plus-potato-subtracted datasets, respectively. NNLS-based PI-AI improved food-matrix classification by separating oil signatures from paper and potato contributions. Optimized post-pruned models achieved nearly 80% test accuracy using only five and four Raman variables, respectively. The four-feature representation reduced the data footprint by 99.44% without loss of accuracy. These findings demonstrate that Raman-based oil identification can use compact, physically meaningful, and interpretable spectral representations, supporting Frugal AI, Edge AI, portable sensing, and embedded food-quality monitoring.

Neural PDE operators are increasingly trained on reusable solver archives, yet validation often relies on clean prediction error and parameter-agnostic plausibility checks. We introduce cross-parameter relinking, a data-poisoning primitive that makes a triggered input select a valid solution from the same PDE family under an incorrect physical parameter. We term this a wrong-physics backdoor: the output remains physically plausible but is wrong for the intended parameter. The attack exploits tensor-to-parameter provenance failures in multi-parameter archives by stamping the surrogate input and relinking its supervision to a cached alternate-parameter solution for the same latent sample. Across 476 attack campaigns, we evaluate Burgers, advection-diffusion, two-dimensional Navier-Stokes, and an elliptic Poisson case. Fourier Neural Operators and DeepONet provide the primary evidence, with Transformer, GRU, and LSTM models as support. FNO reaches a backdoor success rate of 1.0000 on both advection-diffusion and two-dimensional Navier-Stokes while retaining low clean relative L2 error. Clean-label, label-only, and shuffled controls show that high attack success alone is insufficient: successful attacks must move predictions toward the intended alternate-physics target while preserving bounded clean error. These results expose a structural validation gap: smoothness or generic solver-like behavior is insufficient unless the provenance of the intended physical parameter is also verified.

Accounting for information flow on the path space of trajectories of a nonnegative martingale yields exact variational identities for it, even at arbitrary random times. This recovers the widely used classical concentration inequalities, from Ville to PAC-Bayes, and measures what each one discards. The tail a bound controls is itself a relative entropy, resolved by the chain rule into per-step conditional divergences. The discarded slack has an exact form in each of three geometries: a Gibbs tilt for the Azuma-Hoeffding and PAC-Bayes bounds, the crossing itself for Ville's and for pooled tests, and a dominating certificate for the $L^p$ maximal bound. That certificate's optional-stopping deficit resolves per step into Bregman divergences of the running maximum. On a path-time space, the same identity gains one factor that prices anticipation: an arbitrary random time carries an e-process “peeking penalty.” The partition function can be read as a coalescent–a prefix-sharing probability of independent copies–and geometric mixtures of test martingales gain a pooling benefit for multi-model safe testing.

Deep neural networks are often overconfident, assigning high confidence even to incorrect predictions. Consequently, users lack a reliable signal for deciding when a prediction can be trusted. Post-hoc confidence estimation addresses this by training a lightweight auxiliary head over a frozen classifier. Existing targets, however, suffer from inherent ambiguity: they assign overlapping confidence values to correct and incorrect predictions, while errors near the decision boundary receive confidence scores indistinguishable from correct predictions. In this work, we propose $TCP_\alpha$, a novel confidence target that resolves these limitations by introducing a margin-controlled penalty for misclassified samples. We prove that $TCP_\alpha$ guarantees complete separation between the target values of correct and incorrect predictions, with a separation margin that is independent of the number of classes and increases monotonically with the penalty parameter. Since accurate classifiers naturally produce very few errors, learning these targets results in a severely imbalanced regression problem. We therefore present a systematic study of training strategies for learning under this imbalance and identify an effective training configuration through extensive ablation studies. We evaluate the proposed approach on rāga identification, investigate its robustness under domain shift, and further validate it on frame-wise ornamentation detection without modifying the selected configuration. Across all settings, $TCP_\alpha$ consistently outperforms existing confidence targets for failure prediction. Rejecting only the least-confident 8% of predictions improves the base model's macro-F1 from 0.89 to 0.98, while fine-tuning the confidence head with only 5% labeled samples from a new corpus effectively restores performance under domain shift.

Despite their growing importance for contact-free radio frequency (RF) based healthcare monitoring, different radio technologies such as frequency-modulated continuous wave (FMCW) radar, impulse radio ultra-wideband (IR-UWB), and Wi-Fi sensing are rarely compared under identical deployment conditions, as existing studies typically differ in hardware, datasets, and evaluation methodologies. In addition, the performance of ceiling-mounted radars, despite their practical deployment and cost advantages in healthcare environments, remain underexplored. Therefore, this paper presents a controlled comparison and analysis of ceiling-mounted FMCW, IR-UWB, and Wi-Fi sensing using synchronized recordings from 20 participants across six room layouts. All technologies are evaluated with the same convolutional neural network (CNN) on both a fine-grained 10-class human activity recognition (HAR) task and a coarse 4-class sleep monitoring task. IR-UWB achieves the highest cross-subject activity recognition performance (89.0% macro F1), while FMCW generalizes best to unseen room layouts (83.8% macro F1). For sleep monitoring, all technologies exceed 92% macro F1 in unseen environments. The results reveal a fundamental trade-off between recognition performance and environmental robustness, which can be explained through differences in range resolution, antenna diversity, Doppler resolution, and spatial information retention. These findings provide practical guidelines for the design of healthcare-oriented RF sensing systems.

Recursive self-improvement (RSI) asks whether an AI system can improve the process that produces AI systems, so that the next system inherits the improvement. That process is the training algorithm: a better objective or update rule improves the compute\mbox{-}capability exchange rate for every subsequent run, including the one that produces the next agent. Whether RSI is feasible therefore turns on whether an agent can design training algorithms. No benchmark isolates that ability: existing suites are won by collecting data or by tuning hyperparameters, and none tells a change to how a run is executed apart from a change to how the model learns. We present AI4AI\mbox{-}Bench, 10 frozen research repositories spanning 10 training algorithm families. In each task, an agent has 4 hours on one B300 to rewrite the training algorithm; its code is then rerun from scratch for up to 12 hours and scored by a fixed evaluator hidden from the agent, against the repository's original algorithm under the same procedure. Because the 10 metrics are incommensurable, every task is mapped onto one scale on which $0$ is an uninformative model, $0.1$ is the algorithm the repository ships, and $1.0$ is the task optimum. Across 29 configurations of 6 systems on all 10 tasks the mean score is $0.166$, and the best system reaches $0.250$: even the strongest closes under a fifth of the distance between the algorithm that was already there and the optimum. The submissions show where that distance went: most never change how the model learns at all, and the minority that do average $0.226$ against $0.126$ for the rest. More reasoning effort mostly buys the willingness to go there, taking that minority from $8\%$ of submissions to $64\%$ and the mean score from $0.094$ to $0.196$. We release the task suite, the evaluators and every scored submission, so that the measurement can be repeated as these systems change.

Predictive models over structured electronic health records (EHRs) remain central to machine learning for healthcare, but few have jointly emphasized quantitative laboratory information and interpretability with respect to input medical events. We present BERT-LER, a BERT-style model for coded EHR timelines pretrained and fine-tuned from a de-identified EHR dataset of 75 million patients, that encodes laboratory test results as discrete tokens while retaining graded information through percentile-based binning, paired with Integrated Gradients for token-level attributions grounded in the input EHR sequence. We benchmark our approach on the public EHRShot benchmark suite and on an asthma severity progression study based on real-world data. This addresses a methodological gap in EHR foundation-style modeling by unifying laboratory value representation and explainability in a single framework, while assessing whether both predictive performance and explanations generalize beyond standard clinical prediction tasks. Across EHRShot and asthma tasks, BERT-LER achieves predictive performance that is competitive with, and on laboratory-related tasks often exceeds, publicly available benchmark models, and provides attributions that align with clinically known risk factors. Our architecture and explainability approach can be applied to many therapeutic areas and prediction tasks using language models trained on structured EHRs.

Sparse pursuit after dictionary learning can yield a precise atom support even when its physical interpretation is not justified by the calibration data, especially for highly coherent dictionaries where alternative calibration-compatible dictionaries may assign different physical meanings to the same selected support. We develop resolution-aware physical-support inference that jointly accounts for uncertainty in the learned dictionary and in the representation of a deployment signal. Our cross-dictionary confidence correspondence retains calibration-compatible dictionaries and deployment-compatible sparse representations, then projects the surviving explanations onto physical-support space. For local coherent-atom classes with separation scale s, once the deployment data resolve the coherent-block explanation and its atom support, the minimax physical resolution from N calibration signals satisfies $\delta_{\mathrm{opt}}(N,s)\asymp\min\{s,\frac{1}{\sqrt{N}s^2}\}$, with relative resolution governed by the orientation-information scale $Ns^6$. Deployment replication improves physical localization only when orientation changes cannot be absorbed by adjusting the active coefficients. For computation, we introduce active endpoint bracketing (AEB), an adaptive finite-bank procedure that evaluates only candidates that can still affect the physical report and otherwise safely coarsens or abstains. Finite-bank experiments, including a four-region synthetic application, show that a point-valued plug-in selector can be physically overprecise, whereas AEB avoids unsupported refinement with fewer candidate evaluations.

The causal dynamics of sleep-disordered breathing are complex and vary across patient populations, hindering the development of targeted interventions. We learn dynamic causal graphs of sleep-disordered breathing from Home Sleep Apnea Test (HSAT) recordings, revealing systematic differences in causal structure across sex and age subcohorts. We do so using the PCMCI+ algorithm on windowed fractional variables derived from 105 HSAT recordings, exploiting domain knowledge via edge blacklisting and employing bootstrap aggregation to address small subcohort sizes. The learned graphs show that temporal self-dependencies and the apnea-desaturation relationship persist across all cohorts, while other relationships vary substantially.

Semantic caches reuse an LLM response when the incoming query embedding lies near a cached query, but proposed eviction policies have rarely been compared under one protocol. Using CLEVER, we evaluate FIFO, LRU, LFU, ARC, GDSF, a single-pass streaming adaptation of SISO, and a semantic-redundancy policy across three ordered, deduplicated query corpora, three cache capacities, and two encoders. No evaluated policy improves on LFU by more than 0.041 percentage points in any of the eighteen settings. Replacement is not irrelevant: FIFO and streaming SISO trail LFU by as much as 8.67 and 8.55 points, respectively, at tight capacity. We explain the missing upside with a conditional packing result. Under exact lookup and insert-on-miss, a newly inserted entry cannot have a resident neighbor within the hit radius, so a geometry-aware eviction rule receives little new redundancy signal. A separate audit exposes a larger problem with the evaluated operating point. At MiniLM's median nearest-neighbor threshold, only 2.1-3.9% of sampled LMSYS and QQP hits are judged answer-substitutable, reducing raw hit rates of 51-60% to quality-adjusted rates of 1.1-2.2%. The cross-encoder study further shows that thresholds do not transfer between embedding models. LFU is the strongest simple default in this protocol; deployment decisions should first establish answer validity and then test sub-point policy differences with exact search.