Science & Mathematics

Visual Place Recognition (VPR) is an important means for loop closure detection in visual Simultaneous Localization and Mapping (SLAM). While state-of-the-art VPR and SLAM approaches excel in urban and feature-rich natural environments, the vast and feature-poor environments of extraterrestrial surfaces like on the Moon or Mars pose severe challenges for existing approaches. This paper presents a novel benchmark, VPR-Space, to evaluate VPR capabilities on extraterrestrial surfaces. There are three contributions: (1) We collected sixteen VPR database-query datasets based on existing data from Space Analog Domains (SAD), terrestrial environments that resemble properties of the Moon or Mars. The collection includes a range of challenges regarding environment, viewpoint change, and illumination conditions. (2) To address the ill-posed nature of VPR, we propose a novel approach to generate the ground-truth information in VPR benchmarks. Instead of defining that a query and datbase image show the same place if and only if their camera poses are within fixed position and viewing direction thresholds, we propose to make the decision dependent on whether the match would be valuable in the downstream SLAM task. To avoid the overhead and restrictions of evaluating VPR in-the-loop within SLAM systems, we provide VPR ground-truth (GT) matrices based on the performance of an ensemble of state-of-the-art Relative Pose Estimators (RPE). Under this GT we require a VPR method to identify exactly those image pairs for which a subsequently applied RPE is expected to provide a valuable pose estimate in a SLAM context. (3) We evaluated nine state-of-the-art VPR approaches. The results show considerable room for improvement and indicate directions for future research. We provide a pre-release version of the dataset on github: $\texttt{https://github.com/rfechner/VPR-Space}$

We prove two universality theorems for neural networks consisting of a single neuron. First we prove a strong form of the Bovdi--Geche universality conjecture on neurons over finite fields, where the activation function is a certain modification of the discrete logarithm, using the Weil bound on multiplicative character sums over finite fields. Then we show a generalisation of Aizenberg's universality theorem for complex binary neurons with a weight vector independent of the target function, positively answering another question in the literature. The proof of the latter uses an equidistribution theorem for the dilates of certain maps into tori.

High harmonic generation in solids is typically described under the assumption of a stationary medium. This assumption is rarely questioned because the electronic properties of most media evolve on timescales much slower than the harmonic emission process itself. However, in strongly driven time-varying transparent conductive oxides the plasma frequency, carrier density, and effective mass can evolve on the same femtosecond timescale as harmonic generation. Under these conditions, harmonic generation and material evolution are no longer independent processes. Within this context, we consider a two-field pumping scheme in which a strong field generates a comb of high-order harmonics across a broad spectral window, while a weaker field simultaneously reshapes its own individual spectral lines. Distinct from geometry-dependent wave-mixing processes, the evolving electron plasma reshapes multiple collinear harmonic channels simultaneously. A fully self-consistent hydrodynamic-Maxwell framework reveals that even- and odd-order harmonics naturally emerge from free- and bound-electron dynamics without the typical need to invoke phenomenological nonlinearities. Our model captures the full spectral evolution on an ultra-broad bandwidth, while experiments performed on ITO confirm the distinct behavior of bulk-driven odd harmonics and interface-driven even harmonics under strong pumping. These results establish time-dependent plasma dynamics as a powerful new tool for broadband hyperspectral shaping of multiple harmonic channels beyond the framework of four-wave-mixing processes. Our findings are of particular relevance to ultrafast science, with potential applications in ultrafast pulse manipulation and the generation and control of tailored multi harmonic waveforms.

Neural architecture search (NAS) is challenging because of the vast, non-linear space of artificial neural network (ANN) topologies. Evolutionary methods can generate architectures, but their effectiveness is limited by difficulties in preserving functional organization during mutation and recombination. Naturally evolved biological networks may provide organizational principles that support adaptation and evolvability. Here, we introduce TRON (TRansfer Of Natural networks), a neuroevolution framework that uses experimentally characterized protein-protein interaction (PPI) networks as architectural priors. Instead of starting from random topologies, TRON reuses biological interaction patterns as structural building blocks. It combines classical evolutionary operators with neuroevolution-inspired mechanisms, including speciation, and introduces species quarantine to protect newly generated architectures during early maturation. We evaluated TRON against established evolutionary and neuroevolutionary methods on three classification benchmarks (CORA, CITESEER, and PUBMED) and three regression benchmarks (DIABETES, FRIEDMAN, and CALIFORNIA). TRON achieved the highest predictive performance among the evaluated methods on all classification datasets and most regression tasks. It showed substantially lower variability across independent runs, indicating greater robustness and convergence stability, at increased computational cost. These findings support the use of PPI networks as effective architectural priors and suggest that biologically derived interaction networks provide more favorable substrates for neuroevolution than random topologies. More broadly, TRON demonstrates the feasibility of transferring naturally evolved network architectures into artificial intelligence systems, offering a direction for improving evolvability in NAS through the reuse of biological organizational principles.

Satellite communication systems are increasingly relied upon for connectivity in remote areas and are expected to play a pivotal role in 6G networks, operating under high round-trip times and packet loss where forward error correction is essential. This work evaluates Reed-Solomon (RS) coding, Random Linear Network Coding (RLNC), and a combined RS+RLNC hybrid over a three-hop satellite channel whose loss statistics are derived from orbital geometry and link-budget calculations rather than assumed, comparing RS and RLNC at matched code rates of 0.800. The central finding is the \emph{bridging property} of RLNC: every RS code rate collapses above a crossover threshold $p^*$, while RLNC operates continuously through all failure regions with overhead growing as $1/(1-p)$. At rate~0.800, $p^* \approx 0.21$. Under the derived satcom channel with mean burst duration 166~minutes, RS$(255,204)$ is impractical despite an average loss rate of 1%, while RLNC requires 15.5~Mbit to recover all 1000 packets. An RS+RLNC hybrid reduces overhead by 19% relative to pure RLNC at BER$=10^{-3}$. To the authors' knowledge, this is the first analytical derivation of the rate-normalised crossover $p^*$ between RS and RLNC and the first experimental confirmation of the bridging property under a physically derived satellite channel model.

Continuous Integration and Delivery (CI/CD) are fundamental to modern software development, yet configuring pipelines such as GitHub Actions (GHA) workflows remains complex and error-prone. Recent work has explored Transformer-based models for GHA workflow recommendation, but existing approaches typically model workflow steps as a single homogeneous generation task. This formulation overlooks a key distinction: reusable Actions are structured, ecosystem-level components, whereas project-specific Scripts encode local build, test, and deployment semantics. Consequently, current models struggle to capture the different forms of knowledge required for accurate workflow completion. In this work, we present FLOWSTEP, a framework for GHA workflow step completion that explicitly distinguishes between Action and Script generation. FLOWSTEP combines large language models (LLMs) with contrastive learning to better capture semantic relationships among workflow contexts, Actions, and Scripts. Evaluated on established benchmarks, FLOWSTEP improves end-to-end workflow step prediction accuracy from 38.23% to 51.09%. Beyond prediction accuracy, we further study whether LLM-as-a-judge can provide a more semantic evaluation of generated workflow steps. By comparing LLM-based assessments with human annotations, we find that the best-performing model configuration reaches substantial agreement with human ground truth (Krippendorff's alpha = 0.701), suggesting that LLM-as-a-judge can serve as a scalable complement to manual evaluation.

Many high-throughput scientific experiments produce distributional outputs, requiring conditional generative models to interpret the data, and to predict outcomes at new inputs. Acquiring experimental data at new conditions (e.g., for a new antigen or a new cell type) is often expensive, motivating the need for automated experimental design methods. However, existing active learning frameworks are limited as they assume scalar or vector-valued outputs. We first show that the Wasserstein test risk of a conditional generative task decomposes into a mean term and a shape term, demonstrating why mean-based active learning is blind to distributional shape. To capture both, we propose the transport Neural Tangent Kernel (tNTK), a tractable kernel that measures how sensitive a generative model's predicted distribution is with respect to its parameters. The tNTK recovers the standard NTK in the deterministic limit and upper-bounds the posterior Fréchet variance of the prediction under small parameter perturbations. It admits a closed-form expression for affine Gaussian transport and tractable Monte Carlo estimators otherwise. Treating the tNTK as the covariance function of a Gaussian process, we greedily select inputs that minimize the total GP posterior variance across the pool, yielding a distribution-aware active learning strategy. We evaluate this strategy across four classes of conditional generative models, including conditional flow matching and conditional Monge gap, on synthetic distributions, intent-labeled sentences, and single-cell perturbation responses. In these experiments, the tNTK outperforms the baseline NTK computed from the conditional mean, demonstrating the value of distribution-aware active learning for generative-modeling tasks.

Adapting a pre-trained machine-learning interatomic potential for non-equilibrium defect dynamics and plasticity requires improving defect accuracy without degrading the near-equilibrium elastic and vibrational response. Here we achieve this balance for the aluminium MACE potential QP17 by integrating conditional readout uncertainty, iterative full-parameter updates and protocol-compatible relative-energy anchoring. Sensitivities over 144 linear readout weights guide acquisition, with within-batch covariance updates and evolving neural representations. After four active-learning rounds (32 acquired configurations and 400 gradient updates), force-uncertainty acquisition reduces mean development force error from 22.43 to 21.52 meV/Angstrom, outperforming random sampling (22.70), geometry-diversity selection (22.46) and uniform training across all 104 candidate labels (22.71) under the same update budget. All three force-uncertainty runs satisfy bulk-mechanical retention criteria: historical relative-energy error decreases from 0.375 to 0.207 meV/atom, while harmonic phonon root-mean-square drift remains below 0.05 THz. Matched objective ablations reveal that relative-energy anchoring suppresses harmonic drift. Out-of-sample evaluations across planar faults, twin boundaries and grain boundaries demonstrate transferable gains in defect energies and deviatoric stresses, even though virial stresses are never fitted. A closed-loop molecular dynamics and density functional theory (MD-DFT) pilot confirms that automated acceptance gating intercepts updates that violate historical elastic-group energy or vibrational tolerances. Together, these components establish a retention-constrained fine-tuning protocol for improving plastic-deformation observables while preserving the tested bulk mechanical response.

JEV decision models pair a language model backbone with a lightweight scoring head to select among natural-language options. Direct selection compares candidate answers within one request, whereas entailment reranking evaluates each candidate in a separate natural language inference (NLI) request over fixed relation options. We examine whether the readout and adaptation choices used for direct selection remain effective for cross-request reranking. We combine a mean-span readout with selective backbone adaptation. Mean-span pooling replaces the option-ending pointer with an average of the contextualized states inside each NLI option, without adding head parameters. Selective adaptation retains low-rank updates in attention and gated DeltaNet projections while removing feed-forward adapters. On MC11, which averages accuracy over eleven reranking configurations from seven sources, selective adaptation improves NeoHorse-1-4B accuracy over the all-module baseline while cutting trainable parameters by more than half, and mean-span pooling yields further consistent gains at the same parameter count.

Standard reinforcement learning (RL) algorithms based on Markov Decision Processes (MDPs) are vulnerable to sim-to-real gaps. To address this issue, distributionally robust RL algorithms based on robust MDPs (RMDPs) have been proposed, for ensuring that agents can perform reliably to unpredictable conditions in real-world environments. Despite of the rich literature, \revision{an efficient model-free method is still lacking} for robust RL, particularly for Markovian data, which is more relevant in practical applications. In this work, we take a step towards settling this problem by proposing a sample efficient (asynchronous Q-learning) algorithm for robust RL. In particular, consider a $γ$-discounted infinite-horizon RMDP with state space $\mathcal{S}$ and action space $\mathcal{A}$, where the uncertainty set is specified by the total variation (TV) distance with uncertainty level $σ$. Assuming access to a single trajectory of Markovian samples induced by a behavior policy and the nominal transition kernel, the sample complexity of the proposed Q-learning algorithm --- to achieve an entrywise $\varepsilon$-approximation of the optimal Q-function --- is bounded by $$ \frac{1}{μ_{\mathsf{min}}\varepsilon^2(1-γ)^3\max\{1-γ,σ\}} + \frac{t_{\mathsf{mix}}}{μ_{\mathsf{min}}(1-γ)} $$ up to logarithmic factors. Here, $t_{\mathsf{mix}}$ and $μ_{\mathsf{min}}$ denote the mixing time and the minimum state-action occupancy probability of the sample trajectory, respectively. This result is nearly minimax optimal, with an additional factor of $\frac{1}{1-γ}$. Moreover, it achieves a significant improvement over existing bounds.

Aviation authorities hold decades of accident reports whose narratives record how investigators reconstructed each accident. Computational work on these archives has mostly sorted reports into fixed taxonomies. In this work, we propose a hybrid framework in which a text-based corpus identifies the leading factors of a failure mode. Physics-based models then quantify how likely that failure is at a stated operating point. Ordered causal chains extracted from 56,202 American accident reports with a local large language model aggregate into a causation knowledge graph. On an evaluation sample of 2,000 reports, the chains recover 90.8% of investigator-flagged causes. Each physical factor whose drivers the record carries receives an occurrence model calibrated to the accident population. A leaky noisy-OR fitted to the corpus combines the factors into a failure mode probability that supports intervention. For engine power loss, the carburetor icing model separates 1,205 icing accidents from other accidents using temperature and dewpoint alone, reaching 0.61 in area under the receiver operating characteristic curve against a density ratio ceiling of 0.64. Routine weather hours give no valid control, because flying concentrates in fair weather. On later years the physics adds 0.03 to that area for occurrence and nothing for severity, so its value lies in interpretability. On four named accidents, abduction gives a probability of necessity of 0.29 to 0.50 for the heat omission, under monotonicity. The Fast Causal Inference algorithm does not license the largest backdoor adjustment on the learned networks.

Robot corrections allow people to adjust a robot's behavior, but a single correction changes many features. Also, preferences about object attributes and relationships between objects cannot be expressed by moving the robot. We propose environment edits as a complementary form of feedback. The robot's motion is held constant as the human edits the scene, so the only explanation for the edit is the moved object. We introduce Feature Learning from Environmental and Robot-Relational Edit Traces (FERRET), which learns preferences from both robot and environment corrections. We show that environment edits are able to disambiguate features and learn preferences that robot corrections alone cannot.

In this work, we present a first-principles approach to investigate phonon-assisted shift current in extended systems. Our method combines the special displacement technique with real-time simulations based on the dynamical Berry phase. Merging these two frameworks enables direct access to non-linear optical responses at finite temperatures, while naturally incorporating phonon scattering and phonon-assisted electronic transitions into the time-dependent dynamics. This allows us to capture temperature-dependent features and lattice fluctuations that are essential for a realistic description of non-linear photocurrent generation. As a representative example, we apply our approach to the shift-current response of a monolayer of hexagonal AlN (h-AlN). This system provides an ideal testing ground because it features an indirect band gap and lacks inversion symmetry. Moreover, h-AlN layers of different thicknesses have been fabricated and characterized in a variety of experimental settings, opening the concrete possibility of testing our predictions. Our results demonstrate the relevance of phonon-mediated processes and the importance of temperature in determining the magnitude and spectral features of the shift current in layered materials.

One of the main challenges of federated learning is heterogeneous data split among the clients, leading to biased local updates called client drift. The SCAFFOLD algorithm (Karimireddy et al., 2020) corrects client drift with control variates, but its relation to the similarity of client data is not yet fully understood: the original similarity-based guarantees cover only quadratics with a stronger second-order similarity assumption, and a follow-up weakly convex analysis (Luo et al., 2025) obtains a $\sqrt{Lδ}$ dependence for $L$-smooth losses with averaged Hessian similarity $δ$. We analyze SCAFFOLD with minibatch-based control variates and establish linear dependence on Hessian similarity beyond quadratics, for both convex and nonconvex objectives. In particular, for convex objectives with exact gradients, $R$ SCAFFOLD rounds with $K$ local steps lead to an objective gap of order $(L/K+δ)D_0^2/R$, where $D_0$ is the initial distance to a minimizer. When $K=Θ(L/δ)$, it improves the minibatch SGD rate by a factor $Θ(L/δ)$. For nonconvex objectives, we prove an analogous stationarity guarantee under uniform Hessian similarity for a client-selection variant. In both settings, we extend these guarantees to stochastic gradients and prove matching lower bounds for the analyzed methods with constant local stepsizes. We also show that, under partial participation, even identical clients need not obtain the full communication benefit of additional local steps.

Agile quadruped navigation in dynamic environments demands trajectories that are both dynamically feasible and provably collision-free. Most real-time planners achieve tractability through a convex reduction that requires simplified kinematic models, discarding motion that the robot can readily execute, and certify safety only at discretization points, leaving open possible inter-sample collisions with moving obstacles. We present a real-time motion planning framework that retains the fully-actuated $SE(2)$ model of a quadruped, with its anisotropic, heading-coupled velocity bounds, and accepts the resulting nonconvexity. By parameterizing trajectories as composite Bernstein polynomials, we exploit the convex hull property to certify collision avoidance and the input bounds over the entirety of each planning horizon. Solved at online replanning rates, the optimization runs in closed loop, commanding body velocities to the locomotion controller. We validate the approach in simulation against ablations and external baselines, and in hardware experiments on a quadruped navigating among unknown moving obstacles, recovering both lateral and heading-coupled motion that simplified models exclude.