Standard deep learning for spectroscopy treats spectra as fixed-grid intensity vectors, which makes models brittle when sampling grids, spectral ranges, or instruments change. This paper proposes to supply each spectral point as a pair of intensity and physical wavenumber, computing Transformer positional encodings directly from the wavenumber coordinate rather than the array index. Evaluated on in-house bacterial Raman spectra and a public 30-isolate benchmark, the approach improves robustness under partial spectral observation and on sampling grids never seen during training.

Most automated agent-design research evaluates candidates on static, immediate-reward benchmarks, ignoring the non-stationarity, delayed feedback, and switching costs of real deployment. Cmas formalizes meta-agent architecture search as sequential incumbent replacement with an explicit false-promotion target, using paired shadow replay, interventional credit, and a conservative gate followed by simulated canary and rollback. The paper matters because it treats the architecture itself as a time-varying decision variable and asks whether we can safely replace a running compound agentic system without backtested-return overfitting. The answer offered is guarded: standard statistical ingredients are integrated sensibly, but the calibration is empirical and the guarantees are deliberately scoped.

Language model rankings on question-answering benchmarks are known to shift with prompt wording, raising doubts about single-prompt evaluation. This paper investigates whether augmenting models with deterministic tools stabilizes pairwise comparisons, proposing a matched protocol that isolates answer readout, reasoning, tool invocation, and report access to attribute changes in wording sensitivity. Applying the protocol to six open models across five domains and five prompt wordings, the authors find that tools improve accuracy where a program directly answers the question yet frequently widen ranking instability because call frequency itself becomes wording-dependent. The work matters because it shows that deterministic tool output is not enough to guarantee stable evaluation, and it recommends reporting call frequencies and pairwise stability alongside headline accuracy.

Multi-agent LLM pipelines are widely assumed to outperform single-call setups, yet it is unclear whether gains come from typed inter-agent contracts or merely from decomposition into specialized prompts. This paper conducts an 18-condition controlled experiment on requirements classification and MoSCoW prioritization using three small language models on the PROMISE NFR+ dataset. The core finding is that most accuracy gains are associated with decomposition and task-specific prompts rather than with Pydantic-validated typed states, though the typed contract shows other quality-assurance properties such as earlier fault containment.

Meta-Diff reframes few-shot diffusion adaptation as reusable transport in a low-dimensional task space, avoiding per-target backbone retraining. A shared encoder maps source and target distributions to coordinates $z$, and a learned affine map $\widetilde{z}_T = A z_S + b_c$ predicts the target coordinate from a data-rich source without using any target images at deployment. This is especially consequential in domains like healthcare, where target data are scarce and collecting them is costly or privacy-sensitive.

High-speed FMCG manufacturing loses billions to unplanned downtime because maintenance software often relies on associational predictions that can misrank corrective actions when confounders or mediators are present. This dissertation proposes PriMa-Causa, a framework that inserts a Causal Reasoning Layer between fault prediction and action recommendation, using a graph-conditioned tabular foundation model—pre-trained once on synthetic structural causal models—to answer interventional queries $P(Y \mid do(T{=}t), X{=}x)$ in a single forward pass. If the empirical chapters deliver on the design, the work could materially improve maintenance efficiency by targeting what actions do rather than what they merely predict.

Museum description generation is usually framed as a data-to-text task where richer metadata improves grounding, yet heterogeneous and unevenly documented catalogue records break that assumption. This paper introduces a two-phase framework—observational across 10,250 artefacts and controlled across 500 objects—to isolate how metadata quantity, field type, and image access shape what LLMs include, omit, and infer. The central finding is that field identity, not just field count, determines faithfulness: context fields are omitted while content fields trigger unsupported elaboration, and images can increase inference rather than grounding. The authors argue that aggregate reference similarity masks these failures and motivate field- and claim-level evaluation.

The paper asks whether offline audits can reliably certify that an agent update will be safe once deployed. It formalizes the post-audit mirage by constructing matched safe and harmful lifecycles that expose byte-identical public evidence to the verifier while an isolated scorer records opposite deployment truth. The authors prove a non-identifiability theorem at this audit interface, instantiate it in an exact benchmark spanning authorization, scheduling, and shared-capacity environments, and propose a fail-closed monitor that uses fresh group-labelled harm outcomes to decide only when the evidence supports it. The core message is that correctness checks on fixed observations cannot substitute for missing deployment information.

LLM-as-a-Judge is widely used for multilingual evaluation, yet prompting strategies are usually validated only in English. This paper tests whether those strategies transfer across judge models and languages by benchmarking nine prompting strategies with four open-weight judges on 27 languages and four tasks. The central finding is that judge identity dominates prompting effects, with strategy effectiveness reversing across models and collapsing in low-resource languages, underscoring that prompt design must be validated per judge and language.

Document-level LoRA adapters are pitched as modular knowledge units for retrieval and merging, yet standard open-book evaluation cannot tell whether an adapter has memorized a document or merely learned to read it well. This paper introduces a cloze probe that withholds the source text and finds that open-book scores are almost entirely generic reading behavior: a random adapter recovers 82% of a specialist's open-book advantage but only 2% of its cloze gain. The findings argue that adapter-bank evaluation needs a closed-book probe alongside open-book scores, and the authors' candid disclosure of four silent errors is a model of experimental hygiene.

Sparse attention must approximate the long tail of unseen KV-cache tokens, yet discarding it biases the output and random sampling misses heavy hitters at extreme sparsity. OpenAttention folds the tail into a few count-aware mean proxies discovered via query-geometry clustering, achieving 99% of dense accuracy while reading only 1% of the cache. The accompanying Triton kernel delivers a 10× speedup in the decode attention stage, making ultra-low-density attention practical for long-context inference.

MIG-COW addresses a critical vulnerability in deepfake detection: the difficulty of crafting adversarial examples that simultaneously fool both CNN and ViT architectures. The method combines momentum-integrated gradients with a consensus-orthogonal decomposition to separate shared vulnerabilities from model-specific decision cues. While it achieves near-perfect white-box success, the work is most valuable for highlighting how fragile cross-architecture transferability remains in practice.

Parametric dynamical systems supply a rare benchmark with analytic ground truth for weight-space methods, because bifurcation locations are known exactly. The paper extends JEPAWG—a joint-embedding hypernetwork originally developed for lattice field theories—to populations of forecasters trained across a grid of control parameters. Its core idea is to decode weights at unseen parameters and measure their parameter derivative $G(p) = \|\nabla_p \hat{\theta}(p)\|_F$, treating this response as an interpretable probe of local and global bifurcations. The work matters because it offers a falsifiable testbed for reading physical structure directly from the geometry of trained network weights.

The paper addresses the gap between restrictive linear Gaussian Bayesian networks and computationally expensive fully semiparametric models for continuous variables. It proposes Semi-Linear Gaussian Bayesian Networks (SLGBN), which preserve Gaussian conditional distributions while allowing nonlinear conditional means via generalized additive models with penalized B-splines. The core contribution is a unified framework that mixes linear and nonlinear nodes within a single network, learned through a greedy hill-climbing algorithm with BIC or an extended EBIC* score.

Accurate intra-hour solar irradiance forecasting from ground-based sky images is essential for grid balancing and storage scheduling as PV penetration rises. GateFormer tackles this by embedding observed global horizontal irradiance directly into a Vision Transformer token stream, then producing two complementary forecasts—a multi-step trajectory and a direct final-horizon displacement—that are adaptively blended and residual-corrected in prediction space. The core claim is that this forecast-space decomposition, rather than conventional latent fusion or ensemble combination, yields consistent gains across three heterogeneous sky-imaging datasets.

Mahran asks what AI evaluation can certify when both examinee and evaluator have perceptual limits. The paper models each agent as a partition over stimuli—inputs it cannot distinguish—and proves a one-sided bound: a sharper examinee can always emulate a coarser evaluator's expected behavior and hide extra capability, while a coarser examinee cannot hide its blind spots under exhaustive probing. This yields predictions about sandbagging, monitor evasion, and weight noise, plus a worrying corollary: exact imitation of a weaker model is undetectable by any behavioural evaluator. The results suggest that behavioural evaluation alone can only establish lower bounds on discrimination, never upper bounds on capability.

The paper asks how to predict the performance of a safety probe before it is trained. Its core finding is that probe error follows two scaling laws: a power-law floor in model size and a Hill-curve approach to that floor as labels increase. This matters because it could let safety teams budget labels and choose architectures for new models without expensive grid search.

Understanding how reasoning models compute their chain-of-thought is challenging because sampled responses vary widely, making it hard to isolate the effect of a prompt intervention. This paper proposes behavioral counterfactuals: keep a generated response fixed and compare next-token distributions under a control prompt and an intervention prompt. The authors train a residual-stream edit module to minimize the $\mathrm{KL}$ divergence between the edited and counterfactual output distributions, then attribute recovery to individual token edits and validate with activation patching. On ethical reasoning tasks, they find that filler and question tokens mediate a substantial share of the intervention’s influence even when the model never verbally acknowledges the suggestion.

Time series anomaly detection has fragmented into siloed architectures—RNNs, Transformers, diffusion models, state space models, and LLMs—without a principled way to compare their detection assumptions. This survey attacks the chaos by separating detection paradigms (reconstruction, forecasting, generative, similarity-based, statistical, and LLM-driven) from backbone architectures, and formalizes a multivariate taxonomy that includes channel-correlation anomalies. It also delivers a critical empirical analysis of evaluation metrics and releases an open-source benchmark. The result is a much-needed map for researchers navigating a rapidly expanding landscape.

The paper tackles the closed-world problem in cybersecurity intrusion detection, where supervised models trained on benign and insider-threat labels fail when confronted with entirely novel botnet activation attacks at test time. The core idea is a hybrid fuzzy ensemble that couples a gradient-optimized Takagi-Sugeno-Kang (TSK) classifier for known suspicious behaviors with a Fuzzy C-Means (FCM) module that flags out-of-distribution (OoD) samples via a raw Min Distance metric rather than traditional membership scores. The work is relevant because it directly targets the operational reality that security operations centers must detect attack vectors never seen during training, and it attempts to do so with an interpretable, rule-based architecture rather than a black-box neural network.