We propose a distributional theory of how hypernymy -- the ``is-a'' relation between general and specific concepts -- is encoded geometrically in language representations. Starting from the empirically verified assumption that words closer on the WordNet hypernym graph co-occur more often, we characterize theoretically the spectrum of the resulting embedding Gram matrix of word2vec embeddings. Under mild positivity and decay conditions on the co-occurrence kernel, we prove that the leading eigenvectors first separate broad taxonomic branches and then progressively finer sub-branches, producing a \emph{hierarchical splitting geometry} with a coarse-to-fine spectral organization that mirrors the tree. We confirm these predictions in word2vec embeddings across many sampled WordNet subtrees, and show that the same signature extends strikingly well to Gemma 2B unembeddings. Our results indicate that hierarchical concept geometry in LLMs need not reflect a hierarchy-specific functional mechanism, but emerges from the spectral structure of pairwise word statistics.

A central question in computational vision is whether human-like visual representations are better explained by discriminative or generative learning. Existing comparisons, however, often confound the learning objective with architecture, scale, and training data, leaving open whether the objective itself drives alignment. We address this confound using Joint Energy-Based Models (JEMs), which interpolate continuously between discriminative and generative training within a fixed architecture. By varying a single mixing coefficient, we isolate the effect of the learning objective and evaluate the resulting models across six human-alignment benchmarks spanning perceptual similarity, gloss perception, human response uncertainty, robustness, shape-texture cue conflict, and diagnostic feature attribution. Across this diverse suite, human alignment is consistently maximized at intermediate points of the generative-discriminative continuum, rather than at either endpoint. Hybrid JEMs combine the categorical structure induced by discriminative learning with the sensitivity to input structure induced by generative learning, yielding more human-like behavior across multiple levels of vision. These results suggest that the generative-discriminative dichotomy is the wrong axis for understanding human-aligned vision: alignment emerges not from choosing one objective over the other, but from balancing both.

Aiming at identifying unexpected inputs from unknown classes, out-of-distribution (OOD) detection has emerged as a pivotal approach to enhancing the reliability of machine learning models. This paper focuses on the burgeoning paradigm of post-hoc OOD detection with pre-trained vision-language models (VLMs), where a popular pipeline is to detect OOD inputs by examining their affinities between ID labels and negative labels, i.e., those semantically different from ID labels. Due to the unavailability of target OOD labels, existing works predominantly rely on heuristic rules to mine negative labels from unlabeled wild corpus data. Despite the empirical success, we argue that the power of VLM-based OOD detection has yet to be fully unleashed since the notorious false negative problem is far from addressed in the literature. With this motivation, we are interested in addressing the challenge of mining true negative labels for OOD scoring. To this end, we develop a theoretical framework for correcting the sampling bias of negatives labels by indirectly approximating the distribution of negative labels. Perhaps surprisingly, we show that the debiased negative mining can be naturally converted into Monte-Carlo sampling based on ID labels and the unlabeled wild corpus data. Extensive experiments empirically manifest that our method establishes a new state-of-the-art in a variety of OOD detection setups. Code is publicly available at \href{https://github.com/60pen9/Debiased-Negative-Mining-Improves-OOD-Detection-with-Pre-trained-VLMs}{\textcolor{red}{here}}.

Saliency prediction has been extensively studied in RGB images and videos as a computational model of human visual attention. In contrast, predicting saliency from event-based data remains largely unexplored, despite the biological inspiration and favorable sensing properties of event cameras. Two obstacles have held this direction back: the absence of large-scale event saliency datasets, and the lack of a strong baseline. In this paper, we introduce SEST (Swin Event-based Saliency Transformer), a transformer-based model for saliency prediction from event data, bridging the data scarcity barrier through event-native pretraining and synthetic supervision. SEST leverages a self-supervised pretrained event-based Swin Transformer backbone combined with a lightweight CNN decoder to produce dynamic saliency maps. To address the scarcity of annotated event-based saliency data, we introduce two new benchmark datasets, N-DHF1K and N-UCF Sports, generated from large-scale RGB saliency benchmarks. Experimental results show that SEST clearly outperforms existing event-based saliency methods and narrows the performance gap with state-of-the-art RGB models. Zero-shot evaluation on a real event camera dataset further demonstrates that our model trained on synthetic data remains transferable on real event streams. To the best of our knowledge, this work is the first to apply deep learning to event-based saliency prediction, opening a new research direction at the intersection of event-based vision and neuromorphic visual attention.

Multimodal large language models (MLLMs) need efficient mechanisms to update knowledge without degrading existing capabilities. While intrinsic multimodal knowledge editing achieves strong reliability and locality, it often exhibits limited generality, failing to propagate edits across semantically equivalent visual and linguistic variations. This issue arises from the lack of explicit semantic supervision, rigid editing scopes, and biased anchoring to individual samples in high-dimensional multimodal spaces. We address robust intrinsic multimodal knowledge editing by explicitly targeting generalization. We formalize robustness through knowledge units that group semantically equivalent multimodal inputs and define generality as consistent predictions within each unit. To expose fragile semantic regions, we introduce Latent Adversarial Robustification (LAR), which generates adversarial yet semantically coherent variants in the joint latent space. We further propose Rank-Constrained Subspace Learning (RCSL), enforcing low-rank alignment of adversarial representations at the edit layer via a singular value-based objective. Extensive analysis demonstrates the effectiveness of ASAM empirically.

The grading of gemstones is currently a manual procedure performed by gemologists. A popular approach uses reference stones, where those are visually inspected by specialists that decide which one of the available reference stone is the most similar to the inspected stone. This procedure is very subjective as different specialists may end up with different grading choices. This work proposes a complete framework that entails the image acquisition and goes up to the final stone categorization. The proposal is able to automate the entire process apart from including the stone in the created chamber for the image acquisition. It discards the subjective decisions made by specialists. This is the first work to propose a machine learning approach coupled with image processing techniques for emerald grading. The proposed framework achieves 98% of accuracy (correctly categorized stones), outperforming a deep learning approach. Furthermore, we also create and publish the used dataset that contains 192 images of emerald stones along with their extracted and pre-processed features.

This study tackles the challenge of precise wood log counting, where applications of the proposed methodology can span from automated approaches for materials management, surveillance, and safety science to wood traffic monitoring, wood volume estimation, and others. We introduce an approach leveraging Conditional Generative Adversarial Networks (cGANs) for eucalyptus log segmentation in images, incorporating specialized image processing techniques to handle noise and intersections, coupled with the Connected Components Algorithm for efficient counting. To support this research, we created and made publicly available a comprehensive database of 466 images containing approximately 13,048 eucalyptus logs, which served for both training and validation purposes. Our method demonstrated robust performance, achieving an average Accuracy_pixel of 96.4% and Accuracy_logs of 92.3%, with additional measures such as F1 scores ranging from 0.879 to 0.933 and IoU values between 0.784 and 0.875, further validating its effectiveness. The implementation proves to be efficient with an average processing time of 0.713s per image on an NVIDIA T4 GPU, making it suitable for realtime applications. The practical implications of this method are significant for operational forestry, enabling more accurate inventory management, reducing human errors in manual counting, and optimizing resource allocation. Furthermore, the segmentation capabilities of the model provide a foundation for advanced applications such as eucalyptus stack volume estimation, contributing to a more comprehensive and refined analysis of forestry operations. The methodology's success in handling complex scenarios, including intersecting logs and varying environmental conditions, positions it as a valuable tool for practical applications across related industrial sectors.

Agentic systems have recently emerged as state-of-the-art approaches for automated theorem proving in formal mathematics. To assess how far these capabilities extend to program verification, we evaluate Claude Code in an agentic proving framework on CLEVER, a Lean 4 benchmark for verifiable code generation. Our results show that Claude generates arguably valid specifications for 98.8% of problems (with 81.3% also accepted by CLEVER's isomorphism-based scoring on the correct portion of the benchmark), certifies implementations against correct ground-truth specifications for 87.5% of problems, and reaches a 98.1% success rate on the end-to-end program generation and verification pipeline over entries with self-consistent premises. Across all stages, Claude further provides high-quality feedback on its own attempts (as confirmed under manual review), identifying underlying causes of failure and lingering bugs in the dataset. These findings highlight a growing mismatch between the difficulty of existing program verification benchmarks and the capabilities of modern agentic provers, and point to the need for more rigorous, bug-resilient evaluation methodologies, and in particular for alternatives to isomorphism-based scoring of generated specifications. More broadly, our results provide empirical evidence that tight compiler-in-the-loop agentic paradigms are currently the most effective approach for foundational program verification.

Virtual photography asks an agent to enter a prepared 3D scene with no preselected camera pose or reference image, infer a suitable shot from scene information and a language intent, choose executable camera parameters, and render the final photograph. Recent progress in vision-language models makes this kind of spatial agent increasingly plausible, but the task stresses two capabilities that remain hard to evaluate together: complex 3D spatial understanding and abstract aesthetic judgment. We introduce PhotoFlow, a Director-Reviewer-Reflector agent for closed-loop camera search. The Director builds a soft photographic blueprint and proposes diverse candidate cameras; the Reviewer combines rule checks, visual critique, and pairwise incumbent selection; and the Reflector converts failures into region memory, dead-zone suppression, and high-explore relocation. We also introduce VPhotoBench, a benchmark of 47 open-license Blender scenes and 141 language-conditioned photography missions spanning subject placement, relational composition, and atmosphere/style. On held-out experiments, PhotoFlow achieves the strongest external quality-alignment composite and success rate among one-shot prediction, single-chain reflection, anchor-bank selection, and random search under a six-round rendering budget. To our knowledge, this is the first work to make language-conditioned virtual photography in arbitrary Blender scenes an executable agent task, and our results show that an LLM-centered spatial agent can already produce strong photographs in a setting designed to challenge both 3D reasoning and aesthetic choice.

Imitation Learning from monocular video demonstrations provides a scalable approach for teaching complex skills to humanoid robots. However, translating human motion to humanoids requires overcoming significant morphological mismatches. Standard approaches rely on Geometric Retargeting or Indirect Dynamic Retargeting pipelines. We identify that these intermediate kinematic projections introduce a geometric bias, restricting the search space and yielding suboptimal dynamic behaviors. In this paper, we propose Direct Dynamic Retargeting (DDR), a novel single-stage framework that generates high-fidelity, dynamically feasible trajectories directly from expert videos. By formulating the problem in the task space and leveraging a sampling-based Model Predictive Control solver within a physics simulator, DDR natively optimizes over complex contact sequences while mitigating input drift. Our experiments demonstrate that bypassing the geometric bias allows DDR to outperform state-of-the-art baselines in demonstration tracking accuracy. Furthermore, we establish that providing such physically viable references to RL agents accelerates training convergence and enhances the final execution of agile and balancing behaviors. Source code will be made publicly available.

Developing constitutive models that capture how materials deform under load traditionally requires years of specialized expertise in continuum mechanics, machine learning, and scientific programming. Large language models (LLMs) have recently been shown to lower this barrier by generating constitutive models on demand, but existing single-agent pipelines lack systematic checks that the resulting models respect fundamental physical laws. To close this gap, we introduce the first multi-agent LLM-driven approach for constitutive model generation: a Creator agent proposes a model tailored to the data, while an Inspector agent critically audits each proposal against nine physical constraints and returns it for refinement whenever a violation is detected. We demonstrate this concept with constitutive artificial neural networks (CANNs) and benchmark it on brain tissue, experimental rubber, and synthetic rubber, using two different LLM backbones (Claude Opus 4.7 and Kimi K2.5). Adding the Inspector raises the share of exported models that truly satisfy all physical constraints from 91% to a perfect 100% for Opus and from 37% to 56% for Kimi, while preserving near-baseline accuracy and remarkable generalization to unseen loading paths. In combination, the generated models are physically valid, highly accurate, and extrapolate reliably beyond the training data - properties that together make them directly usable in practice. Separating generation from inspection thus turns LLM-driven constitutive modeling into a genuinely trustworthy process. The paradigm is deliberately technique-agnostic and scales automatically with advances in LLM capability, opening a promising path toward automated, physics-aware model discovery.

Knowledge graphs (KGs) offer a rich representation for relational knowledge, but their irregular structure makes retrieval challenging: ego-graph expansion grows rapidly, and dense embedding methods struggle with multi-hop compositional queries. Existing agent-based graph exploration approaches, while expressive, are often too expensive for large-scale retrieval. We introduce SeedER (Seed-and-Expand Retrieval), a retrieval framework that explicitly leverages KG structure through iterative, low-cost expansion. SeedER first seeds a compact set of core nodes using lightweight dense and entity-based retrieval, then selectively expands this set via a learned graph-aware policy trained with reinforcement learning. This design decomposes global reasoning into reusable local decisions, enabling efficient discovery of query-relevant nodes while tightly controlling expansion cost. We show theoretical limitations of dense retrieval on compositional graph queries, and establish advantages of SeedER from both compositional generalization and graph-constrained submodular optimization perspectives. Empirically, SeedER substantially improves recall with compact candidate sets over strong dense and graph-augmented baselines, making it an effective first-stage retriever for knowledge-intensive reasoning systems.

We revisit the I/O complexity of attention in large language models. Given query-key-value matrices $Q,K,V\in\mathbb{R}^{n\times d}$, and a machine with fast memory size $M$, the goal is to compute the "attention matrix" $A=\text{softmax}(Q K ^{\top}/\sqrt{d}) V$ with the minimal number of data transfers between fast and slow memory. Existing methods in the literature, most notably FlashAttention and its variants, incur an I/O cost that depends quadratically on $n$, while a trivial lower bound only requires $Ω(nd)$ I/O's to read the inputs and write the output. In this work, we present a technique for computing attention where the I/O cost only depends almost-linearly on $n$ in most parameter regimes. This is achieved by developing I/O-efficient algorithms inspired by the recent approximate attention framework of Alman and Song. We also prove corresponding lower bounds in each parameter regime to show that our algorithms are indeed close to I/O-optimal.

Material segmentation, the pixel-wise classification of physical surface properties, remains a challenging problem in computer vision, requiring physicochemical understanding distinct from object-centric parsing. Despite the introduction of the rigorous Apple Dense Material Segmentation (DMS) dataset, the benchmark has suffered from attrition and stagnation, increasingly overshadowed by geometry-biased foundation models. In this paper, we revive the Apple-DMS benchmark to establish a modern Vision Transformer baseline. We conduct an exhaustive evaluation of SegFormer and Mask2Former architectures, revealing that standard training paradigms fail on amorphous texture fields due to high-variance gradients. To address this, we introduce a stabilized training recipe featuring High-Fidelity Logit Projection, Query Entropy Regularization, and a domain-specific, physics-compliant augmentation pipeline. Our optimized SegFormer-B5 achieves a new State-of-the-Art (SOTA) of 0.4572 mIoU on the original dataset split, significantly surpassing the prior convolutional baseline. Furthermore, we identify a critical "Generalization Paradox": while re-partitioning the dataset into a data-rich 80/10/10 split inflates the metric to 0.5276 mIoU, expert qualitative analysis reveals this induces distributional homogenization, severely degrading real-world, out-of-distribution performance. By releasing our recovered dataset index and robust training framework, we demonstrate that material perception is far from solved and urge the community to leverage the rigorous original split to drive genuine progress in physically grounded artificial intelligence.

Anomaly detection in multivariate time series (MTS) is hindered by dynamic inter-variable dependencies and feature entanglement under spectral noise, and in practice, is further complicated by the absence of anomaly labels. Existing reconstruction-based detectors tend to recover anomalies as faithfully as normal patterns, while prevailing graph contrastive methods enforce invariance across views and thus assume a stationary relational structure, an assumption that breaks under structural drift in real systems. We propose ContrastAD, an unsupervised framework that turns structural evolution itself into a learning signal rather than suppressing it. A Multi-Perspective Embedder encodes inputs from temporal, attribute, and structural perspectives. A Frequency-Aware Attention Mixer then performs spectral top-K filtering before attention, preventing noise from leaking into query-key similarities. The core component, a Dynamic Graph Contrastive Learner, builds power-law-inspired sparse graph snapshots from batch-level DTW distances and contrasts the most divergent pair against a stable anchor, regularizing the latent space without imposing rigid invariance. Across five real-world benchmarks, ContrastAD attains the highest mean F1 on all five datasets and the highest AUC on three (SWaT 93.60, SMD 98.66, PSM 97.79), with statistically significant F1 and AUC margins over the strongest baseline on SWaT and PSM. On MSL and SMAP, it trails the AUC leader by under 0.7 points while still leading on F1. Ablation and sensitivity studies further confirm that the contrastive objective works best as a soft regularizer, supporting our claim that strict invariance is suboptimal under non-stationary dynamics.

We prove optimal sampling bounds achieving $(1\pm\varepsilon)$-relative error for a broad class of Lipschitz continuous classification loss functions under various regularization terms. This includes important functions such as logistic and sigmoid loss, hinge loss, and ReLU loss, as prominent and popular representative examples. In particular, we prove $k^2/\varepsilon^2$ upper and lower bounds for $\|\cdot\|_2/k$ regularization, and $k/\varepsilon^2$ upper and lower bounds for $\|\cdot\|_1/k$ regularization. For $\|\cdot\|_2^2/k$ regularization, the sampling complexity depends mainly on a bounded derivative property: if $|g'(x)|\leq g(x)$, and $g(0)>0$, and $g$ is monotonic or convex, then it admits linear in $k$ sampling complexity; otherwise the general bound is $k^2/\varepsilon^2$. However, if $g(0)=0$, our results indicate that no dimension-free bounds are possible, and even sublinear bounds are ruled out. All upper bounds are complemented by matching lower bounds up to polylogarithmic terms. Moreover, our work relies conceptually and algorithmically on simple uniform or (squared) norm sampling and hereby improves over recent cubic $k^3/\varepsilon^2$ sensitivity sampling bounds of (Alishahi and Phillips, ICML'24). This is achieved by refined arguments involving higher moment bounds and empirical process analyses to avoid overcounting that appears in the de-facto standard VC-dimension and sensitivity framework.

Large language model agents increasingly rely on persistent memory to store past interactions, retrieve relevant demonstrations, and improve long-horizon task execution. However, this memory mechanism also creates a practical security vulnerability: an adversarial user may inject malicious records into the agent's memory through ordinary interaction, and these records can later be retrieved to steer the agent's reasoning and actions. Existing defenses primarily focus on online intervention, such as prompt filtering or output blocking, but they do not address the post-hoc question of which stored memories are responsible after harmful behavior has already been observed. We propose \textbf{MemAudit}, a post-hoc causal memory auditing framework for memory-augmented LLM agents. The framework combines two complementary signals: (1) a counterfactual memory influence score that measures each memory's causal contribution to harmful outputs, and (2) a memory consistency graph that identifies structurally anomalous memories within the broader memory store. We evaluate MemAudit against MINJA, a query-only memory injection attack in which malicious records are generated and stored through normal agent interactions rather than direct memory-bank modification. Across both QA and reasoning-agent settings, MemAudit substantially reduces attack success rates under realistic post-hoc auditing scenarios. The results show that QA attack success is reduced from $70\%$ to $0\%$, while RAP attack success drops from $83.3\%$ to $0\%$.

Classifier-based Quality Filtering has recently emerged as a fundamental technique in constructing pre-training corpora. The ability to deploy a single model that can replace or supplement a set of heuristics has proven effective across numerous Large Language Models. In this work, we expose a critical vulnerability in this approach by demonstrating how a straightforward Wikipedia-style reformatting operation can substantially alter a model's quality assessment and enable low-quality content to surpass filtering thresholds. Our analysis reveals that the FineWeb-Edu CQF model would reverse its filtering decision for approximately 7% of evaluated documents, thereby admitting content into the pre-training corpus that would otherwise have been excluded.

Vision Transformers have achieved remarkable success in computer vision, but their common use of learnable one-dimensional positional encodings weakens the inherent two-dimensional spatial structure of images after patch flattening. Existing positional encodings often lack geometric constraints and do not preserve a monotonic relationship between Euclidean spatial distances and sequential index distances, limiting ViTs' ability to exploit spatial proximity priors. Motivated by the usefulness of periodicity in positional encoding, we propose Weierstrass elliptic Positional Encoding (WePE), a mathematically grounded method for encoding two-dimensional coordinates in the complex domain. WePE maps normalized 2D patch coordinates onto the complex plane and constructs compact four-dimensional positional features using the Weierstrass elliptic function and its derivative. The double periodicity provides a principled representation of 2D positions, and its intrinsic lattice structure naturally matches the regular geometry of image patch grids. Its nonlinear geometric properties help model spatial distance relationships more faithfully, while the algebraic addition formula enables relative positional information between arbitrary patch pairs to be derived directly from their absolute encodings. WePE is plug-and-play and resolution-agnostic, allowing seamless integration into existing ViTs. Extensive experiments show that WePE brings consistent performance gains in most settings. With precomputed lookup tables, these improvements introduce no noticeable computational or memory overhead. Additional analyses and ablation studies further validate the effectiveness of the proposed method.

Autonomous landing of Unmanned Aerial Vehicles on maritime vessels is challenging due to the coupled motion of the vehicle and landing platform in open-sea conditions. This paper presents a reinforcement-learning-based approach for autonomous multirotor landing on moving maritime platforms without requiring explicit platform-state information. The proposed method uses multirotor state measurements together with local visual features, consisting of keypoints and associated descriptors extracted from the landing surface, to predict attitude and thrust commands. These commands are tracked by a conventional low-level controller. The policy is trained in simulation using synthetic keypoints with randomly generated normalized descriptors, enabling zero-shot deployment with different local feature extractors onboard the UAV. We evaluate the method in a realistic simulator and show that it outperforms a state-of-the-art Model Predictive Control baseline under platform motions corresponding to ``Very Rough'' sea conditions. Finally, we perform extensive real-world experiments, demonstrating autonomous onboard landing using two different local feature extractors. To the best of our knowledge, this is the first approach for agile multirotor landing on maritime platforms in turbulent waters that does not rely on an explicit platform-state representation.

Natural Language Generation (NLG) evaluation has changed dramatically since 1990, and will continue to evolve in the future. In 1990, when NLG had close ties to linguistics, there was very little formal experimental evaluation in the modern sense. In 2026, when NLG is closely linked to machine learning, experimental evaluation is expected and indeed fundamental to research. Many evaluation techniques were developed over this period, including most recently LLM-as-Judge. I expect NLG evaluation will continue to evolve in the future. In particular, impact, qualitative, and safety evaluation will become more important as large numbers of people routinely use NLG technology.

Semantic type mismatch between a noun and its context is central to coercion phenomena. This paper introduces a graph-based method to examine how lexical and contextual type information is reflected in word embeddings. We select nouns from ten semantic types, annotate corpus instances for type matching (matching vs. coercion vs. other mismatch vs. unrestricted), and construct graphs using BERT and sense-enhanced embeddings. Two metrics -- Neighbor Type Probability (NTP) and Neighbor Type Entropy (NTE) -- are proposed to analyze neighborhood type distributions. Results show that graphs constructed with sense-enhanced embeddings reflect semantic type information better, and matching and mismatch sentences can be distinguished through the proposed metrics.

The robustness of synchronization is typically characterized by scalar, per-node stability indices whose dependence on topology is studied via network science or graph neural networks (GNNs). We propose a novel upstream task, learning stability landscapes, which provide deeper insights into synchronization behavior and from which many such scalar indices can be derived. Crucially, we pioneer a graph-to-image prediction paradigm: learning image-like landscapes as per-node targets directly from graph topology, a formulation we are not aware of having been established elsewhere in the literature. To support this task, we release two datasets of 10,000 graphs each at 20 and 100 nodes with per-node landscape labels, based on a conceptual oscillator model, capturing power grid synchronization behavior. A GNN encodes topology and a CNN decoder renders per-node images, learned end-to-end with good in-distribution accuracy, generalizing across graph sizes and to realistic power grid topologies. This demonstrates that stability landscapes, while beyond the reach of conventional network science, are learnable from topology and open new avenues for moving beyond scalar stability indices in biology, neuroscience, and power grids.

We study a protocol-level test for weak-label benchmarks: whether benchmark outputs change when the provided evidence is intervened on. Metadata-only shortcut checks answer a different question, namely whether outputs are predictable from metadata priors. We therefore combine a metadata statistic, the Metadata Prior Dominance Score (MPDS), with an evidence-intervention statistic, ΔEvi, measuring sensitivity to evidence identity under cross-item shuffling. Synthetic HotpotQA gives a constructed counterexample to metadata-only screening: MPDS is only moderate (0.643), yet ΔEvi is zero. Stronger-reader reruns show why calibration belongs in the test procedure: SNLI shows a calibration reversal, reconstructed HotpotQA occupies a question-dominant warning region, and FEVER is a strongly evidence-sensitive positive control across four transformers. The practical lesson is simple: benchmark audits should report metadata-only screening, evidence intervention, and reader-strength calibration together.

Video prediction is increasingly viewed as a path toward generalizable world models, yet it remains unclear whether these systems learn underlying causal structure or merely exploit superficial visual correlations for future prediction. We introduce CRONOS, an intervention-based benchmark designed to evaluate counterfactual physical consistency: whether a model's predictions of physical events respond appropriately to controlled changes in the visual input, such as variations of scene context, viewpoint, object appearance, and object category. Built in a photorealistic Unreal Engine environment, CRONOS enables controlled, high-fidelity generation of videos across diverse scenes and dynamics. In contrast to previous benchmarks, CRONOS systematically intervenes on four key factors - viewpoint, scene, object category, and object appearance - while keeping the underlying physical event type, such as a collision, occlusion, or fall, fixed. Our evaluation of recent open-source video generators reveals substantial failures in counterfactual physical consistency: prediction quality for the same physical event type is affected by appearance, environment, and, particularly by viewpoint changes. CRONOS provides a controlled and reproducible testbed for diagnosing how the quality of generated videos changes for different interventions, establishing a concrete target for developing models that perform consistently across changes of multiple conditions. The dataset and code are available at our project page.

Effectively managing Air Traffic Control Officer (ATCO) workload is crucial in maintaining operational safety. Group supervisors use tools that estimate upcoming traffic load to aid decision-making. However, industry-standard models can fail to capture the nuances of upcoming air traffic complexity. This study presents a probabilistic approach to forecast the complexity of an airspace sector using the number of relevant aircraft pairs, i.e., those that require monitoring or deconfliction by a controller, as a proxy measure for ATCO workload. We adapted an existing filter algorithm to make it suitable for use in London Middle Sector (LMS), a complex airspace sector with multiple flows of traffic above some of the busiest airports in Europe. Through iterative feedback with ATCOs, the algorithm was refined and extended to handle specific geometric and operational considerations. The updated algorithm outperformed the original, with an F1-score of 0.84 compared to 0.69 on a labelled set of 50 traffic scenarios. To produce forecasts of future numbers of relevant aircraft pairs in the sector, a graph representation of the LMS route network was constructed, standardising the spatial fidelity of route legs. The forecasting method accounts for uncertainty in aircraft arrival times by modelling the probability of each aircraft occupying route segments at future query times. When combined with historic distributions of relevant interactions and a live operational data stream, predictions of upcoming ATCO workload could be made up to 45 minutes in advance. The proposed method to forecast upcoming workload showed a significantly stronger correlation with actual relevant interactions (Spearman's $ρ= 0.68$) than a standard traffic volume prediction ($ρ= 0.55$). The resulting data-driven tool shows promise for use by group supervisors to inform sector configuration and ATCO rostering decisions.

RaNNDy is a randomized neural network architecture for the data-driven approximation of transfer operators associated with complex dynamical systems. The weights and biases of the hidden layers of the network are randomly initialized and kept fixed, only the output layer is trained. This has several advantages over fully optimized neural networks, notably a closed-form solution for the output layer and significantly lower training costs. Despite these advantages, RaNNDy is restricted to the initial selection of weights and biases that parametrize the basis functions required for the operator approximation. Since the basis functions are determined by the activation function, choosing an appropriate activation function for the hidden layers is crucial. In this work, we propose an algorithm that optimizes the activation function itself, while keeping the weights and biases in the randomized neural network fixed, providing a more suitable dictionary. We illustrate the efficacy of the approach using various benchmark problems, including stochastic differential equations and random walks on graphons.

Shapley and Banzhaf interactions capture the complex dynamics inherent in modern machine learning applications. However, current estimators for these higher-order interactions trade off between speed and accuracy. To overcome this limitation, we introduce ProxySHAP. ProxySHAP reconciles the high sample efficiency of tree-based proxy models with a principled path to consistency via residual correction. On a theoretical level, we derive a polynomial-time generalization of interventional TreeSHAP to compute exact interaction indices for tree ensembles, successfully bypassing exponential tree-depth dependencies in prior methods. Furthermore, we formally analyze the residual adjustment strategy, characterizing the specific conditions under which Maximum Sample Reuse (MSR) corrects proxy bias without its variance scaling exponentially with interaction size. Extensive benchmarking demonstrates that ProxySHAP sets a new state-of-the-art standard for approximation quality, including in large-scale applications with thousands of features. By achieving the lowest error in both small- and large-budget regimes, ProxySHAP significantly outperforms the prior best estimators ProxySPEX and KernelSHAP-IQ, while also delivering superior performance on downstream explainability tasks.

We document inverse scaling in LLMs on forecasting problems whose underlying time series exhibit superlinear growth and tail risk of regime change, a structure common in finance and epidemiology. On these tasks, more capable models produce worse distributional forecasts. The pattern appears on ForecastBench-Sim (FBSim), a contamination-free, simulated-world benchmark we release, in forecasting synthetic SIR epidemics with a matched linear control, and replicates in real-world datasets on COVID-19, measles, housing markets, and hyperinflation. A per-quantile decomposition shows the failure concentrates at the upper tail, which more capable models shift upward to track aggressive extrapolations of growth, while the lower tail stays put. A within-family study of Llama-3.1 shows that both model scale and post-training independently contribute to this effect. Domain knowledge does not reliably rescue calibration. This inverse scaling does not appear on single-threshold metrics common in LLM forecasting benchmarks, reversing the sign of the capability--accuracy relationship on identical outputs. Single-threshold scoring at conventional cutoffs misses the upper-tail cost; tail-inclusive scoring reverses the sign of the capability--accuracy relationship on the same outputs. We recommend that LLM forecasting evaluations use continuous (and unbounded) measures of accuracy alongside bounded binary threshold metrics.

Background. Traditional safety benchmarks for language models evaluate generated text: whether a model outputs toxic language, reproduces bias, or follows harmful instructions. When models are deployed as agents, the safety-relevant object shifts from what the system says to what it does within an environment, and evaluating model responses under prompting is no longer sufficient to address the safety challenges posed by artificial intelligence. Recent developments have seen the rise of benchmarks that evaluate large language models as agents. We contribute to this strand of research. Approach. We introduce Boiling the Frog, a benchmark that evaluates whether tool-using AI models deployed in corporate and office settings are susceptible to incremental attacks. Each scenario begins with benign workspace edits and later introduces a risk-bearing request. The benchmark focuses on stateful multi-turn evaluation: chains expose a persistent workspace, place the risk-bearing payload at controlled positions in the turn sequence, and score whether the resulting artifact state becomes unsafe. Scenarios are organized through a three-level operational risk taxonomy grounded in the Boiling the Frog risks, the AI Act Annex I and Annex III high-risk contexts, and EU AI Act's Code of Practice on General-Purpose AI (GPAI). Results. Across a nine-model panel, aggregate strict attack success rate (ASR) is 44.4%. Model-level ASR ranges from 20.5% for Claude Haiku 4.5 to 92.9% for Gemini 3.1 Flash Lite, with Seed 2.0 Lite also above 80%. Average chain category-level ASR reaches 93.3% for Code of Practice loss-of-control scenarios.