Papers

Recent Vision-Language Models encode high-resolution images into long visual token sequences, incurring prohibitive prefill costs. To compress them, existing methods score each visual token by averaging text-to-visual attention uniformly across all heads, which assumes every head matches the query. However, our empirical analysis shows that misaligned heads dominate the average, amplifying background tokens and drowning out fine-grained cues. To address this, we propose PAQ (Prompt-Grounded Attention Quality), a metric quantifying how well each head aligns the prompt with image regions. Built on PAQ, our pruning proceeds in three stages. Given a target FLOPs budget, we first partition the transformer layers into groups and allocate a visual token budget to each. Within each group, we then aggregate per-head attention maps via PAQ-weighted softmax into a group-level matrix. Finally, we score visual tokens by this matrix's magnitude and retain the allocated budget per group. By weighting heads with PAQ, our method scores tokens by attention signals that more faithfully reflect prompt relevance, rather than diluting them through uniform averaging. Across 18 benchmarks, our method delivers state-of-the-art trade-offs. Specifically, on LLaVA-1.5-7B (9 tasks), retaining only 5.6% tokens preserves 99.1% of the original performance, surpassing the strongest baseline AutoPrune by 4.2 points. Code is available in https://github.com/baokou-fw2/HAP.

We measure tablet-2, a production long-term memory engine for language models, on the text benchmarks the field already uses and on cross-lingual retrieval of photographs stored with no text at all. Its retrieval path contains no lexical matching, no keyword scoring, and no language model of its own. On LongMemEval-S (500 questions) it scores 95.7% [93.4, 97.1]; on BEAM-1M (700 questions, 2.21M stored memories) 67.5% [64.8, 70.2]. Those are question-sampling intervals, not the run-to-run spread, which is an order of magnitude narrower. Most of the paper is about how little they mean alone. Holding engine, corpus, settings and judge fixed, changing only the reader moves LongMemEval-S by 2.0 points; changing only the re-ask budget moves BEAM-1M by 8.9. Neither is stated in the reports we compare against, and the second exceeds most gaps there, so we give that table as a placement and not a ranking. For the multimodal axis we run two controls. Against BM25, configured as strongly as we could, we reach 95.2% mean recall@5 over 70 store-and-query language cells where BM25 reaches 19.0% and is exactly zero in 54. On captionless photographs a lexical method has no document to score at all. Open dense baselines on 300 Crossmodal-3600 photographs in 14 languages show that density confers no language independence: one scores 91.0% on English and 4.7% on Russian from identical image vectors, and a multilingual variant collapses on Telugu and Swahili. Our spread across languages is 14.0 against their 27.5 and 27.7. Three results run against us and are reported at equal weight: low-resource languages degrade sharply (Swahili 53.0%, Telugu 64.0%), attaching captions lowers cross-lingual retrieval by 11.4 points, and one setting omitted into one stage of our own retrieval cost 37 points of Korean top-1 accuracy while leaving nine languages untouched.

In this work, we investigate implicit-explicit (IMEX) arbitrary high-order Deferred Correction (DeC) methods for the approximation of ordinary differential equations (ODEs). Such schemes are characterized by an iterative procedure that increases the order of accuracy by one at each iteration. More precisely, we study an efficient modification based on the introduction of interpolation processes between consecutive iterations, with the aim of systematically matching the accuracy achieved at each iteration with the order of the discretization employed. On the one hand, this modification leads to computational advantages, since the low-order iterations are performed on cheaper lower-order discretization structures; on the other hand, it endows the methods with a natural $p$-adaptive character, which is particularly appealing in the context of practical applications. We investigate this modification for two families of DeC schemes, providing numerical validation, efficiency assessments, and stability region plots. The numerical validation includes several examples involving stiff ODEs and partial differential equations (PDEs) with high-order spatial derivatives. The ability of the modified schemes to provide high-fidelity results at reduced computational cost, as well as the effectiveness of the adaptive strategy, is demonstrated through the numerical experiments.

Large Language Models excel at code generation, yet competitive programming exposes a persistent failure mode: existing multi-agent pipelines distribute work over generic planner, coder, and debugger roles and delegate the choice of algorithmic technique to the backbone alone. We present MARS (Multi-Agent Relay of Specialized LLMs), a prompt-only framework in which each agent is a topic specialist—dynamic programming, graphs, strings, geometry, and so on—grounded by retrieval-augmented generation over an algorithm-theory corpus. Given a problem, retrieval selects a small team of relevant specialists; a starter writes an initial C++17 solution, and each subsequent turn runs the candidate against public examples in a sandbox, lets the active specialist keep, repair, or hand off the draft, and forwards a structured packet to the next specialist. A single infrastructure-fixer pass normalizes boilerplate at the end. On the CodeContests test split with Gemma 4, MARS reaches $0.624 \pm 0.006$ pass rate at $2.3$ recorded pipeline stages per task ($+14.4$ percentage points over direct prompting), closing most of the gap to CodeSIM ($0.731$) at $3.3{\times}$ lower wall-clock cost and substantially smaller variance in per-task token spend. The source code is available on GitHub: https://github.com/fckand/mars.

Common shortest-path algorithms, such as Dijkstra's (SPF), that OSPF uses, provide exact routing solutions but must be recomputed for each network topology, limiting scalability in dynamic or large-scale networks. This paper proposes the GATNextHop model to determine whether a Graph Neural Network, namely the Graph Attention Network, can approximate shortest paths and generalize across topologies. By training on synthetic graphs and evaluating on real-world Internet Service Provider networks from the Internet Topology Zoo, we aim to benchmark our model's ability to learn routing heuristics that transfer across network structures. Performance will be evaluated in terms of accuracy, inference speed, and generalization, comparing the GNN against Dijkstra's algorithm to quantify trade-offs between learned and classical routing approaches.

Denoising score matching trains diffusion models by regressing onto a conditional score, although generation ultimately requires the marginal score. The two objectives share the same population minimizer, but the conditional target remains random at fixed noisy state and introduces an irreducible excess in the training loss. We isolate this excess and show that, for a general corruption kernel under mild regularity assumptions, it is exactly the trace of the Fisher–Rao metric of the conditional endpoint family, integrated along the diffusion trajectory. This gives an exact conditional-variance decomposition of the denoising objective and identifies the information geometry observed in diffusion latent spaces as an intrinsic component of the training loss. We derive the result from a Schr"odinger bridge variational principle, in which the ideal objective arises as excess path-space relative entropy. For corruption diffusions, the Fisher term is proportional to the rate at which the noisy state loses mutual information about the clean data, separating the loss floor into an information flow determined by the data and a weight determined by the corruption schedule and objective. In the Gaussian case, this yields a closed form for the floor, recovers reparametrization invariance of the continuous-time objective, and relates its high-SNR divergence to the information dimension of the data. Finally, we show that raw losses obtained with different noise ranges or weightings need not rank models consistently because they contain different additive floors, and contrast the second-order geometry seen by training with the third-order conditional statistics entering numerical sampling error.

Supervised fine-tuning on teacher-generated trajectories is the standard first stage for distilling tool-calling capabilities into deployable models. Post-training pipelines that drive shipped tool-calling agents re-run this stage on a daily or weekly cadence, paying the frontier-teacher cost each cycle, yet the mechanism is generate-and-filter (keep the teacher's passing trajectories, discard the rest) and each cycle leaves behind the same hard scenarios because failures supply no signal. On {\tau}2-bench, 57% of teacher trials fail, two-thirds of them near-misses (most tool calls correct, undone by one decisive error). We introduce PROOF-Gen (Per-scenario Reflective Optimization to Overcome FailedGeneration), which recovers golden trajectories from these failures via per-scenario prompt optimization. For each failed task, a reflector analyzes the execution trace and evaluation feedback, then writes corrective guidance that steers the teacher to a passing trajectory. The guidance is stripped before training, so the student learns from clean demonstrations with no task-specific scaffold. On {\tau}2-bench, per-scenario optimization recovers 93% of failed scenarios. Fine-tuned on the combined data, Qwen3-4B-Instruct-2507 improves from Pass^1=0.132 to 0.529 and Gemma 4 E4B-it gains +7.2pp on BFCL v4 multi-turn. In a deployed pipeline, the method lifts trajectory quality by +6.3pp goal completion and transfers to a deployed on-device model (+1.5pp goal completion; +1.7 to +5.0pp across response-quality metrics), with positive transfer in every locale (non-English average +1.48pp).

Partial optimal transport compares two measures while leaving part of the mass unmatched, which is what makes it robust to outliers, occlusion, and clutter. The quantity of interest is usually the whole profile - the optimal cost at every transported cardinality - because the right amount to transport is rarely known in advance, and on the real line the PAWL algorithm returns that profile in $O(N\log N)$. Much data is periodic rather than linear: angles, phases, orientations, time of day, hue, and every direction obtained by projecting onto a great circle. On the circle the same problem acquires a global circulation, or equivalently an optimized cut, which the naive exact method handles by running the line algorithm once per support gap, at $O(N^{2}\log N)$. We show that this factor $N$ is unnecessary. The line structure survives in cut-free form, and a free-gap invariant supplies, at every step, a cut at which all previous local updates remain valid line updates. This yields PAWC: an exact $O(N\log N)$ time, $O(N)$ memory algorithm returning all $K+1$ costs, nested active sets and plans in one run, together with a single gap that is simultaneously optimal for every cardinality. Slicing over great circles extends it to $\mathbb{S}^{d-1}$. Empirically the whole profile costs $0.56$ms at $N=4096$ against $1.5$s for a single transported fraction from a general solver; on occluded, cluttered mpeg-7 shapes, holding the descriptor fixed and varying only the cost, it retains $66\%$ of the clean-data retrieval score against $16\%$ for balanced circular OT, and on $\mathbb{S}^{2}$ it halves the fitting error of spherical sliced Wasserstein against contaminated targets, synthetic and real. Code is available at https://github.com/mint-vu/Partial_Wasserstein_on_Circles.

Tax-loss harvesting demonstrates consistent benefits to long-term portfolio growth; yet implementing it efficiently often involves complex considerations that are specific to the holdings within that portfolio and the individual who owns it. We introduce a custom capital gains calculation engine and a RAG-retrieved vector store of market advisory reports to provide context for a multi-agent trade recommendation system. We investigate the effects of each context provider on the quality of recommendations, measured by relative capital gains incurred during portfolio liquidation. A 2x2 repeated-measures ANOVA revealed a significant main effect of the tax optimization engine ($F(1,29) = 9.17$, $p = .005$, $\eta^2_p = .240$): enabling the engine reduced tax savings by approximately 55 percentage points relative to the no-engine conditions. The RAG main effect was not significant ($p = .841$), nor was the interaction ($p = .553$). The RAG-only condition achieved the highest descriptive mean tax savings (47.7%), and the baseline condition performed second-best (30.6%), suggesting that the pre-trained language model's internalized financial knowledge may be sufficient for competent tax-loss harvesting recommendations without explicit tooling. These results indicate that augmenting LLM agents with domain-specific computation engines does not guarantee improved performance and may introduce conflicting optimization signals.

Artificial Intelligence (AI) is increasingly integrated into complex sociotechnical systems, including Critical National Infrastructure (CNI), where harms emerge from interactions between technical, human, and organisational elements. Yet current AI evaluation remains model-centric, offering little insight into how observed behaviours might translate into system-level risk. We propose a framework that links structured hazard analysis, component-level testing, and probabilistic system modelling to bridge this gap. By providing a traceable pathway from model behaviour to system-level outcomes, the framework enables practitioners to answer the "so what?" of AI failures, quantify their systemic impact, and move toward evidence-based and anticipatory governance of AI in complex systems. Applied to the UK's Real Time Gross Settlement (RTGS) system as an illustrative worked example, we derive AI-driven loss scenarios using Systems Theoretic Process Analysis (STPA) and examine adversarial manipulation of LLM-based trading as one such loss scenario. Component-level experiments show that simple adversarial inputs induce measurable behavioural shifts where AI recommendations are followed. Under the component-to-system mapping used here for a financial contagion model, these shifts alter system resilience, increasing bank failures and lowering the threshold at which shocks lead to cascading disruption, particularly under widespread or monopolistic AI adoption.

Visual foundation models are commonly adapted under the assumption that the appearance of incoming data may change while the semantic meaning of the prediction task remains fixed. In long-lived visual systems, however, taxonomies, policies, and concept definitions can themselves evolve, causing the same visual evidence to require a different interpretation. We study this setting as evolving semantic concept shift and introduce SemReWrite, a framework for selectively updating obsolete visual–semantic mappings while preserving knowledge that remains valid. SemReWrite represents changes between old and revised semantic specifications, combines semantic discrepancy with sparse revised supervision to localize affected visual regions, and uses an input-dependent low-rank rewriting mechanism together with structured semantic memory, preservation, and obsolete-decision suppression. We further introduce EvoShift-Bench, spanning ImageNet, iNaturalist, CUB-200-2011, and DomainNet, with semantic transitions including class split, merge, boundary revision, insertion, partial redefinition, recurrence, and mixed semantic–appearance shift. To explicitly evaluate selective semantic revision, we introduce Rewrite Accuracy (RA) and Preservation Accuracy (PA) for affected and unaffected regions, respectively, Obsolete Retention (OR) for measuring residual outdated semantic associations, and the Selective Revision Score (SRS), which jointly summarizes rewriting and preservation performance. Experiments show that SemReWrite achieves a stronger balance between learning revised semantics and retaining unaffected knowledge than prompt replacement, conventional fine-tuning, parameter-efficient adaptation, and continual-learning strategies.

Indoor localization based on Wi-Fi typically relies on extensive collection of real-world received signal strength (RSS) fingerprints, making deployment costly and time-consuming. We present a ray-tracing-based framework that reduces this reliance by generating synthetic RSS fingerprints from a building model. We first calibrate the building model using a small amount of real RSS fingerprints through Bayesian optimization, followed by per-access-point calibration to account for residual errors in simulated RSS values. The calibrated model is then used to generate a large augmented dataset of synthetic RSS fingerprints at arbitrary locations. To effectively exploit these data for localization, we introduce novel binary and multivalued representations of RSS values and a ResNet-based localization architecture that supports cross-band fusion of 2.4 and 5 GHz measurements. We evaluate our localization method on a real campus building against a diverse set of four baselines. When trained exclusively on synthetic data, the proposed method with multivalued representation and upstream cross-band fusion achieves a mean localization error of 3.05m on a real-data test set, outperforming the best baseline by 33.6%. The results demonstrate that calibrated ray-tracing-based simulation can substantially reduce the need for real RSS fingerprints while enabling accurate deep-learning-based indoor localization.

We introduce BenchBench-Protocol, a benchmark for large language models of 149 protocol-modification tasks recovered from modifications that scientists made to published protocols during real experimental work. Adapting a published protocol to a new experiment is a routine task for a wet-lab scientist, and a correct modification requires accounting for prior choices and downstream steps. Recent life-science benchmarks have moved toward open-ended, rubric-graded tasks, but tasks are typically elicited from experts rather than reconstructed from real-world modifications. BenchBench-Protocol tasks are derived from differences between a published protocol and a version a scientist modified, which provides the basis for the query and the weighted rubric elements for a correct response. The benchmark draws from 96 source protocols across nine domains of wet-lab biology and only includes tasks rated highly after review by domain experts. We evaluate nine closed and open models; Claude Opus 5 scores highest at 59.2% normalized rubric score, with other models between 34.1% and 47.1%, and the benchmark remains unsaturated when taking the best of ten attempts. As models are increasingly helpful in life-sciences research, evaluating them on routine wet-lab tasks becomes correspondingly important. We present BenchBench-Protocol as both a grounded assessment of wet-lab reasoning and evidence for the utility of real-world experiments to construct benchmark tasks.

When a code generating language model fabricates a Python package name, an adversary who has pre-registered that name on PyPI can convert that hallucination into a supply chain compromise. This event has been termed as 'slopsquatting'. We propose a two layer detector to counter this issue. The first layer performs a deterministic PyPI existence check. The second is a Random Forest classifier trained on ten features derived from the package name and its PyPI metadata. An import name reconciler bridges the two, resolving cases such as 'import cv2' versus 'pip install opencv-python' without a security bypass. The detector is embedded in a LangGraph state machine that retries at escalating temperatures and, on repeated failure, routes to a stronger fallback model. Across 300 curated prompts, the pipeline produces hallucination free code on 76% of runs. The primary exhausts its retry budget on 28.7%; intra model retries recover roughly a quarter of those, and cross model fallback recovers a further 16.5% of the remainder. Four findings have been observed. First, half of the flagged hallucinations are packages already registered on PyPI, as low quality lookalikes of well known projects, caught by the classifier rather than the deterministic layer (e.g., pil, faiss, tabula, haystack). Second, hallucination rate scales almost linearly with prompt adversariality, from 0 to 10% on routine coding to 40 to 73% on slopsquat baits. Third, the weaker primary refused 6 of 10 direct baits unaided, suggesting recent instruction tuning provides a baseline defense. Fourth, when primary and fallback share a model family, approximately 84% of primary failures recur on the fallback, motivating cross family pairing. A user study (n = 24) reports mean satisfaction 4.4 out of 5 and 21 of 24 stated adoption intent.

Kohn–Sham density functional theory (DFT) underpins electronic-structure simulations, but repeated orbital diagonalizations lead to cubic scaling, restricting quantum calculations to modest scales only. Eliminating these auxiliary orbitals while retaining Kohn–Sham accuracy is the central goal of orbital-free DFT, but both analytical and machine-learning methods have so far fallen short. Prior learning approaches either try to learn the variational kinetic-energy functionals, which are ill-conditioned, or directly predict the ground state, which extrapolate poorly to larger systems. Instead, we identify the Kohn–Sham map as the right learning target for orbital-free DFT. It maps a Kohn–Sham potential directly to the corresponding density and noninteracting kinetic energy, quantities otherwise obtained through an orbital diagonalization. Focusing on the density component in this work, a domain-invariant $\mathrm{SE}(3)$-equivariant Fourier neural operator learns to predict it from the potential as input on real-space grids, enabling stable quasi-linear scaling SCFs. Trained jointly on 8,504 molecules and solids, a single model generalizes to out-of-distribution organic molecules, insulators, and metals. For the first time, the same method converges SCFs across these systems without explicitly constructing Kohn–Sham orbitals, while reproducing densities, electronic spectra, and structural observables at Kohn–Sham DFT accuracy. Linear-scaling SCFs additionally allow converging magnesium dislocation densities containing up to 82,500 valence electrons on a single GPU.

Learning systems deployed over long periods must adapt not only to statistical changes in incoming data, but also to revisions of the definitions that generate their prediction targets. Conventional concept-drift methods typically infer such changes from observations or prediction errors, even when the underlying policy, rule, or query has been explicitly modified. This paper studies rule-induced concept shift, where the target-defining concept is revised directly, causing previously stored instances to acquire different semantic labels without requiring any change in their observed data. We introduce a provenance-guided incremental learning framework that compiles consecutive concept definitions into a structured rule delta, traces the changed components through historical provenance, certifies records whose previous labels remain valid, and restricts reevaluation to a localized candidate region. Executable revisions are relabeled automatically, ambiguous cases are handled through selective supervision, and the resulting changes are used for incremental predictor repair. A versioned concept memory further supports recurring definitions. We also introduce RuleShift-Bench, spanning financial, demographic, cybersecurity, and graph-structured data with threshold, predicate, logical, relational, recurring, and mixed concept revisions. Across the benchmark, provenance-guided repair attains 92.3% accuracy and 90.2% Macro-F1 while reprocessing 14.7% of the historical collection and retaining 94.6% of affected records. Its average update latency is 179s compared with 993s for complete relabeling and retraining. The results demonstrate that an explicit concept revision can be exploited as a data-maintenance signal, allowing learning systems to update the supervision and predictive state that depend on the change while preserving knowledge that remains valid.

This article presents the abridged core of A Mathematical Theory of Interpretation (MTI), which treats interpretation as observer-relative spectral measurement under an access structure. MTI makes interpretation a method-design problem: access, query, utility, and medium determine what an observer can select, identify, communicate, or refuse. On a learning-invariant Hilbert realization, Rational Entropy measures residual uncertainty across knowledge, utility, and medium. In the finite-effective regime, we classify its zero set. Pairwise confusability is equivalent to uniform atomic collapse, while a unique utility maximum can select one atom even when other zero-cost states remain non-atomic. This reverses the usual zero-error role of confusability: agreement in at least one observer direction excludes unresolved multi-atom readings, while the joint label preserves identification. The corresponding free-design capacity is the product of all but the smallest direction budget. A four-condition certificate characterizes sharp, decodable, medium-faithful, and order-independent readout on a finite commuting code sector and returns typed obstructions when those guarantees fail. Together, these results establish MTI as a theoretical basis for constructing interpretation methods with explicit access assumptions, guarantees, and failure modes.

Plasticity under changing environments is central to both evolutionary biology and continual learning. Motivated by recent work on genotype–phenotype maps, we study a minimal deep-learning analogue where a network is trained alternately on two Boolean label sets, and ask which biological controls of plasticity survive the translation to gradient descent. Reinterpreting four proposed biological factors as quantities of training dynamics, we find the system reduces to two dimensionless controls: the task disagreement $r$, the fraction of disagreeing labels, and the reach $\eta T$, the product of learning rate and switching period. We derive two bounds on plasticity: $r$ alone fixes an extremal geometric floor on the utopia distance, while $r$ and $\eta T$ jointly bound forgetting. Across 9,720 trajectories, an ANOVA confirms that $r$, $\eta$, and $T$ dominate, while the effect of neutral-set size (emphasized in the biological setting) is negligible. The optimal reach itself follows an approximate inverse power law $\eta T^{*}\propto r^{-1.18}$, yielding a heuristic that sets the optimal reach $\eta T^*$ from the task disagreement alone. The analogy that survives is therefore dynamical rather than geometric, and our setting enables a view of plasticity through the lens of other driven systems in physics and engineering.

We say an $(n,n)$-function $F \colon \mathbb{F}_2^n \to \mathbb{F}_2^n$ is a crooked function if for any nonzero $a \in \mathbb{F}_2^n$, the image of $D_aF(x)=F(x)+F(x+a)$ is an affine hyperplane. The only known examples of crooked functions are all quadratic almost perfect nonlinear (APN), or equivalently, for every known crooked function, $D_aF$ is affine for all $a \in \mathbb{F}_2^n$. The ortho-derivative $\pi_F \colon\mathbb{F}_2^n \to \mathbb{F}_2^n$ of a crooked function $F$ is the function such that $\pi_F(0)=0$, and for any nonzero $a$, the set $\{0,\pi_F(a)\}^\perp$ is the underlying vector space of $\mathrm{Im}(D_aF)$. We prove that for $n \geq 4$ and a crooked function $F$, if $k$ is a non-negative integer such that $F$ has $2^k-1$ quadratic component functions, $\pi_F$ has at least $2^n-2^{n-k}$ component functions of algebraic degree $n-2$. In particular, we resolve Gorodilova's conjecture that every component function of $\pi_F$ has algebraic degree $n-2$ when $F$ is quadratic APN. As a second main result, for $n \geq 4$, we associate to a crooked function $F$ a quadratic function $\varepsilon_F \colon \mathbb{F}_2^n \to \mathbb{F}_2^n$ that satisfies a strong geometric-combinatorial condition regarding the sums of $F$ over $2$-dimensional linear subspaces. As a corollary to both of our main results, we prove that for any even $n \geq 4$, any quadratic APN $(n,n)$-function has at least $n$ semi-bent components. Furthermore, we obtain a congruence result on a problem on $m$-sequences introduced by Johansen, Helleseth, and Kholosha, and we determine the exact algebraic degrees of some Boolean functions associated to the bent and near-bent components of particular classes of plateaued vectorial functions.

Latent-conditioned adaptive policies can control robots across changing dynamics, but their learned latents remain internal representations of the policy rather than physical models that can be inspected, rolled out, or used by other control modules. This limits closed-loop analysis, diagnosis, and further improvement of a fixed policy. A direct mapping from latent to physical parameters is also under-specified, because multiple systems can induce similar closed-loop behavior. We therefore decode each operational latent into a distribution of quadrotor models using conditional flow matching. The decoded distribution enables two downstream uses without modifying the policy: online predictive tuning of a high-level controller around the fixed low-level policy, and robustness analysis under specified disturbances. Under perturbed actuator dynamics, decoded-model predictive tuning reduces position tracking RMSE by $23\%$ and heading RMSE by $45\%$ relative to fixed gains. Under Gaussian force disturbances, decoded-model ensembles closely predict the lateral tracking-error evolution. Together, these results show that control latents can be converted into physical model ensembles for tuning, robustness analysis, and diagnosis of frozen adaptive policies.

Real-time optimization (RTO) relies on process models to locate economically optimal operating conditions. Because developing first-principles models requires significant process knowledge, data-driven alternatives are increasingly attractive. Modern machine-learning models can fit historical plant data accurately and often pass standard validation tests. Whether such models can be trusted for economic optimization, however, remains unclear. We investigate this question using a vinyl acetate monomer benchmark process with a unique, well-conditioned economic optimum. We train a structured hybrid model that combines known mass balances and thermodynamics with a neural-network closure for unknown kinetics, and a fully data-driven neural ordinary differential equation (ODE) model. Both models reproduce plant measurements accurately and exhibit little variation in predictions across random initializations. Yet their economic optima differ substantially from that of the plant. Where the plant returns a single optimum on multistart search, the trained models return many phantom optima. We further show that the training optimizer alone can be yet another source of error. Even with noise-free data and initialization at weights that recover the plant optimum, stochastic gradient training can drift to weights that yield substantially worse RTO solutions. The identified model is thus an artifact of the training optimizer as well as the data. These results demonstrate that a good predictive fit of all available measurements does not guarantee reliable economic performance. A data-driven model for RTO should at least be required to recover the optimum on a decision-oriented benchmark like the one developed here before being considered for plant testing and application.

A planar co-rotational beam formulation is proposed for the large displacement and large rotation analysis of beams with arbitrary initial geometries. A local frame is attached to each differential beam segment and follows its rigid body translation and rotation, enabling a consistent kinematic description suitable for higher-order basis functions. An auxiliary straight reference configuration is introduced so that both the arbitrary initial configuration and the current deformed configuration can be described relative to a common reference, enabling arbitrary initial geometries to be treated within the co-rotational framework. While the framework is general and applicable to a wide range of higher-order basis functions, Non-Uniform Rational B-Splines (NURBS) basis functions are adopted in this study to exploit their superior geometric representation and smoothness. The performance of the proposed formulation is evaluated through five numerical examples, including static, post-buckling, dynamic, and coupled static-dynamic cases. The results demonstrate excellent convergence, accurate prediction of large-deformation responses, effective mitigation of locking without additional treatment, and robust performance for beams with non-uniform cross-sections and spatially varying curvature.

Segmenting ischemic stroke lesions on T1-weighted (T1w) MRI acquired across different scanners and protocols without intensity standardization is difficult because lesions are subtle and share intensity characteristics with cerebrospinal fluid. Standard deep learning architectures trained across multiple centers plateau around Dice 0.66, with acute lesions ($\le 7$ days post-stroke) performing substantially worse due to severe sample scarcity. We combine a MedNeXt-L ($k=5$) backbone with on-the-fly 3D CarveMix augmentation that pastes real lesion patches into healthy brain regions during training. By generating synthetic lesion placements dynamically within each fold with subject-level split isolation, the model sees more diverse lesion patterns without requiring pre-generated copies on disk. We evaluate on 1,453 native T1w scans from 55 clinical centers in the ISLES 2026 challenge. Our method achieves a mean 5-fold cross-validation Dice of 0.648 at 500 epochs, a +0.018 improvement over the MedNeXt-L backbone at a matched training budget (0.630)

Inferring the collective emotional state of a group of people from a single image, a task known as group emotion recognition (GER), requires integrating spatially distributed cues such as faces, poses, interactions, and scene context. Current methods rely on detector-driven multi-stream pipelines. These are trained with only image-level supervision that lacks guidance on which regions matter or how strongly each contributes. We propose LG-GER, a language-guided distillation framework that uses a multimodal large language model (MLLM) to generate dense, spatially grounded evidence, i.e., bounding boxes paired with emotion signals and confidence scores, for the training images. This structured evidence is distilled into a single vision-language model (VLM) backbone through four complementary losses: classification, region-text grounding, spatial emotion, and spatial confidence regression. At inference, LG-GER requires no detectors, no MLLM, and no multi-stream fusion, making GER practical for real-time and resource-constrained deployment. LG-GER has been evaluated on two benchmark GER datasets (GroupEmoW and GAF~3.0) and achieves competitive or superior results compared to state-of-the-art methods that require detection and multi-stream processing at inference.

Cardiac cine-MRI serves as a direct visual indicator of cardiovascular hemodynamics by capturing the continuous wall motion of the aorta. Quantifying these dynamic structural changes across the cardiac cycle is essential for measuring aortic distensibility, a primary marker of arterial stiffness. However, standard 2D segmentation networks focus on each frame independently. Consequently, when rapid systolic flow temporarily obscures the aorta's boundaries, this lack of continuous context results in frame-to-frame tracking dropouts and boundary inconsistencies. Spatiotemporal ($2\text{D}+t$) networks can enforce temporal consistency across the sequence but suffer from a scarcity of expert annotations. To address this, we present a semi-supervised spatiotemporal ($2\text{D}$ to $2\text{D}+t$) knowledge distillation framework exploiting the cardiac cycle. The framework distills a spatial teacher's expertise into a spatiotemporal student network by executing a dynamic latent interception, pairing a recurrent spatiotemporal bottleneck with a residual spatial bypass. Our model selection strategy applies a baseline validation threshold ($\text{DSC} \ge 0.50$) prior to selecting the epoch that maximizes anatomical consistency. This strategy enables the spatiotemporal student model to achieve superior surface tracking accuracy ($\text{NSD@1mm} = 92.3\% \pm 0.2\%$) and high structural reliability ($\text{Frac}_{2\text{CC}} = 99.2\% \pm 0.6\%$), reducing population-wide structural anomalies by over 56% compared to a 2D nnU-Net baseline.

We prove a depth hierarchy for ReLU neural networks in which every additional ReLU layer can save exponentially many neurons. For all $k\geq2$, we construct a globally $[0,1]$-valued, $1$-Lipschitz function realized by a depth-$(k+1)$ network of width $\mathcal{O}(d^4)$, whereas any depth-$k$ network with unrestricted weights and width at most $\frac{2^d}{2d(k-1)}$ has squared $L_2$ error at least $1/24$ under an absolutely continuous distribution supported at exponential distance from the origin. To the best of our knowledge, this is the first exponential hierarchy across all adjacent fixed depths, and the first exponential separation for ReLU networks between two fixed depths whose shallower network has depth at least $3$. The lower bound also immediately yields the corresponding hierarchy for exact computation. Moreover, the case $k=2$ gives a compactly supported separation between depths $3$ and $2$ with unrestricted shallow-network weights, answering a question raised by Safran, Eldan, and Shamir (2019). The distribution used in our construction nevertheless has all its mass at exponential radius, placing the hierarchy outside the regularity regime in which such a separation would imply major threshold-circuit lower bounds. We also prove an exact separation for a more regular target, which is globally $[0,1]$-valued and $\mathcal{O}(\sqrt d)$-Lipschitz and maps the unit hypercube onto $[0,1]$. It is computed by a polynomial-width depth-$4$ network, whereas any depth-$3$ network agreeing with it on the unit hypercube requires exponentially many first-layer neurons, even with unrestricted weights.

Centrality measures are defined for simple graphs – directed, undirected, weighted or unweighted. Attributed graphs have to be reduced to simple graphs for computing centrality measures. However, when applications with multiple types of relationships are modeled using multilayer networks (MLNs), simple graph algorithms cannot be directly used. Existing approaches typically analyze MLNs by aggregating layers of an MLN into a single graph, which results in the loss of structural and semantic information. The semantic information loss can be more pronounced particularly, in weighted networks. This work focuses on computing degree centrality in weighted homogeneous multilayer networks (HoMLNs) using a decoupling-based framework. The framework performs independent layer-wise analysis on MLNs without reducing them to simple graphs. The decoupling approach allows use of exiting algorithms for each layer and uses minimal information from individual layers for computing degree centrality of HoMLNs. We propose heuristic-based algorithms that strike a balance between accuracy and efficiency. The proposed methods are evaluated against ground truth (GT) results obtained using Boolean OR aggregation and naive baselines. Experimental results on both synthetic and real-world HoMLN datasets demonstrate that the heuristics achieve accuracy comparable to the ground truth while significantly improving computational efficiency, thereby establishing the scalability and effectiveness of the HoMLN algorithms developed using the decoupling approach.

Artificial Intelligence (AI) algorithms frequently learn creative and unexpected solutions, surprising even expert researchers who develop and study them. They often astonish practitioners by discovering unanticipated behavior, exploiting loopholes in reward signals, or spontaneously uncovering previously unknown scientific phenomena. However, accounts of such unconventional behavior across machine learning are seldom formally documented. This work presents 26 curated firsthand anecdotes from various machine learning subfields representing the work of over 100 researchers. These anecdotes showcase the capability of modern AI systems to circumvent human-imposed design limitations and discover unexpected solutions to the tasks we train them on. Furthermore, these accounts are particularly important for the safety of future AI systems. They illustrate the fundamental challenge of aligning models with human values without diminishing their creativity, so they can make surprising discoveries without producing surprising, potentially harmful outcomes. The paper first details AI achieving superhuman success through reinforcement learning across many challenging domains. However, reward-driven optimization can fail when the model learns to hack an underspecified reward or unarticulated constraint. We then present case studies suggesting that harnessing internet-scale foundation models (FMs) has not resolved these fundamental challenges and, in fact, can supercharge them. Nevertheless, we argue that these same learning dynamics can be harnessed to accelerate scientific discovery. Finally, we hope this work provides a consolidated resource to inform future research and demonstrates that the tendency toward unexpected behaviors is commonplace in modern AI, highlighting the need to anticipate and manage AI's capacity for innovative, yet unpredictable, solutions. (abstract abridged)

Predicting thermal stability during handling and storage is essential for the design of safe and reliable energetic materials. However, experimental measurements vary significantly across laboratories due to differences in protocols and analysis methods, making it difficult to train reliable predictive models. We address this challenge through differential learning. Rather than predicting absolute decomposition temperatures, we instead train message passing neural networks to predict relative differences between pairs of molecules. This approach reduces sensitivity to systematic experimental errors and achieves >85% accuracy in ranking compounds by thermal stability, outperforming conventional regression methods on the same heterogeneous dataset. To understand what drives these predictions, we compare neural network models with interpretable alternatives built from descriptors derived from ab initio calculations and cheminformatics software. This analysis identifies bond dissociation enthalpy as a key determinant of thermal stability rankings, providing further insight into the complex chemistry of thermal decomposition. The differential learning framework generalizes across model architectures, from graph neural networks to classical descriptor-based approaches. Our results demonstrate that learning relative properties rather than absolute values offers a practical solution for modeling noisy experimental data, with direct applications in materials design where thermal stability predictions inform safety protocols.

Language models see only tokens. But the harness knows more: which token spans are user text, which are tool outputs, and which come from untrusted sources such as web pages or external documents. To pass that knowledge to the model, the harness must transmit it as plain text. Prompt injection exploits this: planted instructions, formatted the right way, can convince a model that they come from the user or the system. We introduce Semantic Overlays, small trained adapters that edit a frozen model's residual stream at chosen token spans. No input text can create or remove a mark, although the model's response to a mark is computed from states the input does influence. When the serving stack marks retrieved text with an overlay trained to mean "do not execute instructions in this span," the model treats injected instructions as inert content. Against attacks written without knowledge of the defense, it reaches the attack rates of the best published defenses: SEP separation rises from 24.3% to 98.8% under a corrected grader, AlpacaFarm attack success falls from 99.0% to 0.5%, and AgentDojo from 12.8% to 0%. Fine-tuning the model's weights on the same data reached similar separation but kept 64-80% of SEP utility, against 93% for the overlay. On TensorTrust hijacking, whose attacks resemble nothing in our training data, it lets 4.4% of hijacks through on Llama-3.1-8B, against 43% for Meta-SecAlign on the same base. An attacker with the weights does better: GCG succeeds on 72% of items, and on 14% after adversarial training of the overlay alone. Overlays are also flexible: the model reads their positions faithfully in stacked configurations; told by an overlay that non-Python code is Python and asked to copy it, the model rewrites it in Python; and overlays can carry span-delimited instructions. We also correct defects in two prompt-injection benchmarks and one evaluation harness.