Papers

Vision-Language-Action (VLA) models are a class of generalist robot policies that map camera images and language instructions directly to robot actions. While promising, these models remain slow at test time, particularly for long-horizon tasks that require many queries to the policy. Recent efforts reduce VLA latency by distilling smaller models, overlapping asynchronous action chunks, or pairing the VLA with a fast low-level policy, but still run a learned policy for the entire task. In contrast to VLA, classical motion planners quickly find collision-free motions, but require an explicit goal and have no semantic understanding of the task. In this work, we present SkipVLA, a hybrid policy that combines a pretrained VLA with a classical motion planner, using the planner for free-space motion and querying the VLA only for contact-rich skills such as grasping and placing. SkipVLA reuses the frozen vision-language backbone of the VLA to predict a target pose for each planned motion, and learns this predictor without additional demonstrations introduced into the system by using what was already learnt by the large VLA. We evaluate SkipVLA with three VLAs on 13 LIBERO tasks in simulation and three pick-and-place tasks on a physical 6-DoF YAM arm, demonstrating up to 2.5x faster task completion and significantly lower energy consumption while achieving the same task success rate.

For every dimension $m\geq 3$, we show the existence of an open set $U\subset \mathbb R^{m+1}$, a smooth Riemannian metric $g$ on it, and an integral $m$-dimensional stationary varifold $V$ in $(U,g)$ with the following properties. $V$ has a flat tangent plane of multiplicity $2$ at an interior point $p$, it is a smooth immersed minimal surface in $U\setminus \{p\}$, and it has infinite topology in any neighborhood of $p$. The construction starts from an idea of three of the authors, who in \cite{DHS} tried to build a similar example with $C^{m-1, \alpha}$ regularity for $g$. That previous attempt contained an important error, which has been overcome using a crucial suggestion of gpt-astra-6.

A vision-language-action (VLA) policy with a flow-matching action expert generates each action chunk (a short command sequence) by integrating a learned velocity field; once its weights are fixed, the success or failure of an earlier rollout cannot change the chunk generated now. Concurrent test-time methods give a frozen policy such an input from retrieved successes, a learned critic, a verifier, or a dynamics model, but none uses the robot's own failed rollouts as negative evidence with nothing but a terminal outcome bit. We introduce TraceFlow, a progress-aligned guidance field that turns the action densities of retrieved successful and failed rollouts into a bounded correction to a frozen flow-matching action expert, using one terminal outcome bit per rollout and no other label. Its TraceBank stores traces, time-ordered state-action records with a terminal label, starts from the target-task training traces, and later admits the deployed robot's own rollouts. On an ordered real-robot packing task the base completes 21 of 50 trials in order, TraceFlow 39, and one stacking round without any weight update 47, with wrong-sequence episodes falling from 20 to 0. In simulation the gain is selective: with per-suite selected settings, TraceFlow raises RoboMemArena Sequence from 78.92% to 91.50% task success and Transferring from 54.41% to 62.00% at stacking round 2, leaves the 26-task aggregate unchanged, lowers Counting and Occlusion by 1.12 and 1.42 points, and changes LIBERO-Plus (Long) by +1.27 points (p = 0.0733). Stacking gains are finite, every branch peaking before round ten, and the bank's success-to-failure ratio predicts no retrieval allocation. Project page: https://zhangjiaxuan-xuan.github.io/TraceFlow/

Hydraulic transport in compacted bentonite is diffusion-controlled and governed by the hydration and microstructure of sodium montmorillonite (Na-MMT). Experiments cannot resolve how platelet interactions govern pore structure, transport and stiffness, while existing coarse-grained models smooth hydration oscillations or require manual corrections. We develop a tabulated potential combining Morse interactions between platelet centre and edge sites with a Gaussian process regression correction trained on all-atom potentials of mean force. It captures the hydration-induced complexity of the potential-of-mean-force profiles, including the three-water (3-W) hydration minimum and transfers across geometries, layer-charge variants and unseen configurations. Applied to monodisperse and polydisperse Na-MMT assemblies at dry densities of 0.8-1.3 g cm^-3, the model captures the 3-W to 1-W transition, loss of non-interlayer porosity and evolution of pore structure, random-walk tortuosity, diffusion and stiffness. Predicted diffusion agrees with compacted Na-bentonite measurements.

Does quantum mechanics allow for deterministic amplification schemes that simultaneously behave uniformly for all angles of bosonic phase space (i.e. phase preserving) and are nonlinear in the underlying field operators? Would such a scheme have some utility to quantum information science? In this paper we answer both questions with a resounding yes. Such amplifiers exist, but their form is highly constrained compared to the more general set of noise limited nonlinear amplifiers previously studied in the literature. This enables us to characterize the entire class of bosonic noise-limited nonlinear phase preserving amplifiers, and identify their utility: namely, the optimal amplification of Yurke-Stoler cat states and Kerr kitten states.

We study Fourier extension from compact convex $C^2$ planar curves by organizing the mass $|f|^2\,\mathrm d\sigma$ through the Gauss map. The pushforward measure \[ \nu_f= N_\#\bigl(|f|^2\,\mathrm d\sigma\bigr) \] records its distribution in normal directions. The turning measure $\mathrm d\mu_\kappa=\kappa\,\mathrm d\sigma$ determines an intrinsic terminal tangential scale $r_R(\xi)$ through \[ r_R(\xi)\, \mu_\kappa\bigl(B_\Gamma(\xi,r_R(\xi))\bigr) \asymp R^{-1}, \] and hence a position-dependent angular resolution $\rho_R(\xi)=R^{-1}/r_R(\xi)$. Under a doubling hypothesis on $\mu_\kappa$, $r_R$ is, up to structural constants, the largest scale on which the curve can be linearized to precision $R^{-1}$. These scales define a normal-direction energy, and the associated terminal decomposition and transverse geometry yield local $L^4$ Fourier extension estimates and weighted variants. For the monomial curves $\gamma_k(t)=(t,t^k)$, $k\geq3$ real, the terminal scales are explicit and the energy admits a multiscale representation in terms of angular correlations of $\nu_f$. Under an $s$-dimensional Frostman condition on $\nu_f$, this yields a growth diagram with critical threshold \[ s_c(k)=\frac{k-2}{3k-4}, \] separating the flat-point and nondegenerate regimes. The resulting rates, including the critical logarithmic correction, are sharp at the energy level.

This work presents an all-fiber broadband wavelength-division multiplexing architecture using 1x7 heterogeneous photonic lanterns (PLs), specifically PL-(6+1) and PL-(4+3). By combining mismatched single-mode cores, namely SM450 and SMF28, without rigid geometric constraints, these devices break system degeneracy and enable passive, wavelength-selective spatial routing. Experimental characterization reveals distinct dual-regime performance across a wide spectral range. In the visible band (350-740 nm), the devices function as multi-channel broadband collectors, effectively guiding light across all channels without spatial separation. Conversely, in the H-band (1520-1610 nm), core asymmetries lift mode degeneracy and maximize propagation constant (beta) matching conditions, thereby enabling passive spectral sorting. These results confirm that unconstrained, free-packaged heterogeneous fiber bundles can meet adiabatic transition criteria, showcasing their potential for next-generation astronomical facilities that require simultaneous multi-wavelength capabilities to feed separate, band-optimized spectrographs.

Physics-informed neural networks and finite element methods provide two different paradigms for the numerical approximation of partial differential equations: the former are commonly trained by minimizing pointwise strong residuals, whereas the latter are naturally built from weak variational formulations and the finite-dimensional systems obtained after discretization. In this work, we introduce a common framework based on the discretization of functional Gauss–Newton problems by finite families of linear measurements. We show that, through an appropriate duality pairing, the linear measurements can be represented by test functions. The resulting Gauss–Newton system is then precisely a Petrov–Galerkin discretization of the linearized functional problem. This perspective recovers pointwise collocation and natural-gradient constructions as particular cases, while making the choice of test functions an explicit algorithmic design choice. We specialize this framework to elliptic problems, where it naturally leads to weak residual formulations and to a hybrid finite element–neural construction acting on complementary approximation spaces. Numerical experiments support the proposed framework and demonstrate the effectiveness of weak Gauss–Newton formulations and hybrid finite element–neural approximations.

Accurate positioning is increasingly required in indoor and dense urban environments; nonetheless, satellite-based systems are not always available, and standardized 5G localization solutions remain difficult to deploy with commercial devices. This paper presents RUN-O-RAN, an O-RAN-native framework that enables network-centric uplink localization using standard Sounding Reference Signal (SRS) transmissions from commercial 5G devices. RUN-O-RAN xApp coordinates serving and neighboring base stations, enabling non-serving base stations to retrieve SRS-based uplink timing measurements that would be unavailable in conventional RAN deployments, without modifying the UE or existing 3GPP signaling procedures. The framework combines cooperative SRS collection, first-path time-of-arrival estimation, timing-advance tracking, clock-drift compensation, and multi-anchor position estimation into a complete network-side localization service. Experimental evaluation over $150,000$ SRS transmissions validates the proposed framework, achieving meter-level localization under diverse propagation conditions while revealing the impact of anchor geometry and multipath on positioning accuracy. These findings demonstrate that cooperative SRS-based localization can be realized within the O-RAN ecosystem without modifying commercial UEs, providing a practical foundation for future network-native ISAC positioning services.

Electron-phonon interactions renormalize phonon frequencies and lifetimes and are central to the dynamical properties of solids. While these effects are usually described within linear electron-phonon coupling, the role of nonlinear electron-phonon interactions for phonon properties remains largely unexplored. In this work, we study phonon renormalization arising from the long-range linear one-electron-one-phonon and the nonlinear one-electron-two-phonon interactions within a diagrammatic framework. We derive the corresponding self-energy diagrams, which depend on the chemical potential and temperature, and evaluate them from first principles for the two polar semiconductors LiF and KTaO$_3$. In both materials, the two interaction channels renormalize the phonon spectrum in qualitatively distinct ways. The linear contribution is sharply localized near the Brillouin-zone center, whereas the nonlinear process couples an incoming phonon to other branches throughout the spectrum. As a result, it renormalizes phonons across the entire Brillouin-zone, with a pronounced temperature dependence governed by the thermal occupation of those branches. This behavior provides a clean experimental signature of the one-electron-two-phonon coupling. While the nonlinear phonon renormalization is small in LiF, it is somewhat larger in KTaO$_3$, which we attribute to its greater number of thermally populated phonon branches at room temperature. Our results establish a general framework to assess nonlinear electron-phonon effects on the phonon properties in materials with stronger lattice fluctuations, including soft semiconductors such as lead-halide perovskites.

An assistant watching egocentric video should notice a mistake from past frames alone, before the next step begins, and keep working once the person recovers. A mistake changes the state of the work, so every later step has to be read against what was done rather than against the plan. The first-mistake protocol that current online methods report on cuts each recording at its first mistake, so a fixed-time rule that never looks at the video is right on every case. We evaluate on complete trials, where mistakes and recoveries arise naturally, under a validation false-alarm budget and against controls that use timing alone. PROVIA keeps two records apart: a factual state, a learned summary of the steps each actor performed, mistakes included, and the accepted progress, an exact posterior over the state of an automaton induced from correct demonstrations by Bayesian state merging and over the execution status of each actor. Procedure-state transitions occur only in the correct-status branch; the mistake and correction branches retain the source state. A sequential test turns the per-frame mistake probability into alarms. With one filter and one optimization rule, PROVIA ranks mistakes best among the evaluated controlled baselines on CaptainCook4D, IndustReal, HoloAssist and IMPACT-ego. At a validation budget of 0.1 false alarms per minute it recalls .154 against .128 on CaptainCook4D and .034 against .015 on HoloAssist, where it leads at every budget. The pipeline runs at 58-70 frames per second. The source code is available at https://github.com/Kratos-Wen/PROVIA.

AI companions can purportedly adopt racial personas, raising questions about how they represent identity and how users interpret these portrayals. We combined an algorithmic audit of race-coded AI personas with interviews with 12 companion users who interacted with a probe. Our audit revealed systematic differences, such as Asian-coded male personas receiving higher submissiveness scores than White counterparts, and Black, Hispanic, and Indigenous male personas receiving higher aggression scores than their White counterparts in open-weight models. Interviews revealed that participants envisioned AI companions as offering cultural familiarity and outside perspectives, but differed in which portrayals they considered meaningful or stereotypical. Some rejected overt racial signaling while still expecting culturally distinctive responses. Triangulating these findings with theory, we highlight how social norms and cultural expectations complicate efforts to support meaningful racial representation without reproducing stereotypes. We discuss how companion personalization should be evaluated beyond user satisfaction to account for broader representational harms.

We show that unless P = NP, there cannot be a polynomial-time (or even pseudo-polynomial-time) algorithm for output feedback stabilization of a linear dynamical system with a linear controller. This settles one of the best-known open problems in control theory. The result holds in both continuous and discrete time. We also present a family of stabilizable linear dynamical systems for which no polynomial-time algorithm can write down a stabilizing controller in its standard representation.

Data rates from current radio telescopes are of $\sim$O(10)PB/y and we expect a growth of an order of magnitude for future radio telescopes, e.g. SKA. Coping with these data flows requires a paradigm change in data management: building pipelines with intense I/O between their modules (offline mode) is becoming unsustainable, leading to a transition to a real-time paradigm, where data are moved to RAM, e.g.,through PSRDADA ring buffers. Our objective is to obtain a compact pipeline to efficiently manage data in the real-time flow, while minimizing HPC resources. We employ Heimdall, an open source detection pipeline that performs incoherent de-dispersion, to detect single pulses in radio astronomy data, such as FRBs, astrophysical signals whose origin has not been completely explained. Heimdall currently processes a single beam on a single GPU and can operate both in offline and real-time modes, reading in input either a SIGPROC filterbank file or a PSRDADA ring buffer. We run a test pipeline with data collected from 16 cylinders of the NS arm of the Northern Cross radio telescope near Medicina, both in offline and real-time modes, to compare the obtained results and performance. To corroborate these tests, we employ different input data and explore the problem's parameter space. PSRDADA and SIGPROC results generally agree within 3% for the most important quantities characterizing a FRB (TOA and DM) and within 10% for the SNR. Execution times in PSRDADA mode stay in the real-time flow for most runs from data containing true FRBs but they are still generally comparable with SIGPROC mode in this testbed setup. This test provides an important preliminary comparison in terms of accuracy and performance: this work is functional and preparatory for the next integration of the beamformer with the Heimdall pipeline, having the same logical structure as the test pipeline, work already in progress.

Allen's interval algebra is a qualitative calculus for temporal relations, but its thirteen base relations are crisp predicates over exact interval boundaries. This is inadequate for temporal information from language, perception, databases, or uncertain histories, where times, durations, and boundaries are uncertain and expressions such as "just before" or "roughly during" have graded meaning. We develop the probabilistic Allen algebra (PAA): a generative and complete extension in which relation probabilities are derived from distributions over interval boundaries rather than assigned as scores. Time points are Gaussian; intervals have Gaussian midpoints and truncated-Gaussian durations. Every relation is a boundary-ordering predicate in one common probability space: point-point relations reduce to error functions, and point-interval and interval-interval relations to multivariate Gaussian orthant probabilities induced by linear inequalities. Contact relations (meets, starts, finishes, equals) receive positive measure through a tolerance band, and under a single tolerance the thirteen relations form a true partition that recovers crisp Allen as the tolerance vanishes. The construction derives Allen's taxonomy rather than positing it: coarse predicates such as precedence, overlap, and containment are unions of leaves whose probabilities are leaf sums, and this hierarchy is preserved as intervals collapse to points and thirteen relations reduce to five and then three. Each relation further decomposes into correlation-aware temporal primitives in the spirit of CIDOC CRM. The algebra is scale-invariant and separates graded expressions such as "shortly before" from contact relations. All results are Monte-Carlo validated and shipped as an open, tested Python package.

Text-guided image editing must introduce the requested changes while preserving unrelated source content. In training-free editing, diffusion editors often use spatial controls whose inaccuracies can leave edits incomplete or alter unrelated regions. Causal autoregressive editors face a further constraint: their fixed decoding order limits revision of earlier decisions. As the first to explore training-free image editing with Generative Refinement Networks (GRN), we observe that its refinement process is inherently suitable for editing and offers a promising way to address these limitations. Motivated by this observation, we introduce RefineEdit, a training-free prompt-to-prompt image editing framework built on the GRN. Our key idea is to couple edit localization with content generation through the global refinement of binary image codes, allowing editing evidence to be revised as the image evolves. More specifically, RefineEdit combines bit routing with two stabilization mechanisms: adaptive spatial freezing and finite bit locking. Bit routing starts from an intermediate source state and uses signed probability differences between the two branches to identify editable positions and bits. It directs selected bits toward editing refinement while anchoring the rest to the evolving source trajectory. Adaptive spatial freezing limits unnecessary expansion of the editing region, while finite bit locking maintains recent bit activations to support continued editing. The overall framework requires no additional training, external masks, or attention control. Across nine editing categories of PIE-Bench, RefineEdit achieves the best background-preservation scores in PSNR, LPIPS, MSE, and SSIM, together with the highest whole-image and edited-region CLIP scores among the evaluated methods. Code is available at https://github.com/mura1n/RefineEdit.

Phase reduction is a powerful technique to obtain phase models from high dimensional oscillatory systems. Starting from pairwise interactions, a first order approximation yields Kuramoto and Winfree phase models, while nonpairwise interactions emerge at the second order. Recently, phase models have been extended to account for nonpairwise interactions, notably, the higher-order Kuramoto model. Is there an intrinsic difference between physical and emergent nonpairwise interactions? And can we make use of the former to even out the effects of the latter? In this work, we exploit a recently developed parametrization method to compute the phase reduction and answer these questions. After a comparison of the new method with the classic Kuramoto-style phase reduction, we derive the network motifs of the emergent nonpairwise interactions and compare them with the physical once. Lastly, we exploit the motifs to adopt a coupling design approach. Our framework paves the way for further exploitations of physical nonpairwise interactions for applications in synchronization engineering.

We study the superconformal index of four-dimensional $\mathcal{N}=2$ $\mathrm{SU}(N)$ elliptic models through the Bethe Ansatz approach, where the index is evaluated as a sum of vacua arising from a set of transcendental equations. Starting from the known discrete solutions of these equations, that we refer to as progenitors, we show that there exist large classes of new descendant solutions. We then show that the solutions can be organized into orbits of the S-duality group, mimicking its action on the massive vacua of the $\mathcal{N}=1^*$ deformations of the elliptic spin Calogero-Moser models. In this way we generalize the results already obtained for $\mathcal{N}=4$ $\mathrm{SU}(N)$ SYM, where a relation between the discrete solutions to the Bethe Ansatz Equations and the massive vacua of its $\mathcal{N}=1^*$ massive deformation was conjectured. We conclude our analysis by determining the contributions of the new descendant solutions to the index.

Search advertising connects user intent with commercial content and plays a critical role in platform monetization. Recent systems typically align pretrained generative models with a single business reward, such as eCPM, or use naive reward fusion for preliminary multi-objective alignment. However, an ideal search advertising system must jointly account for heterogeneous objectives, including relevance, click propensity, and commercial value, to balance user experience and business value while mitigating globally suboptimal performance caused by gradient competition. We propose UniPolicy, an objective-aware multi-policy alignment framework. UniPolicy combines objective-specific prefix tokens, sparse MoE-LoRA routing, and objective-specific residual FFNs to hierarchically decouple parameters within a shared backbone, providing differentiated parameter and policy-expression spaces for different business objectives. It further constructs pairwise preferences from multi-stage behavioral feedback, supplementing the relative preference information in exposed-but-unclicked samples and strengthening the relative advantage of clicked candidates in the generation distribution. At inference, UniPolicy supports parallel, business-customizable multi-policy beam search, flexibly allocating candidate quotas across objectives under a fixed retrieval budget. Large-scale offline experiments show that UniPolicy delivers balanced improvements across multiple metrics while preserving retrieval quality, outperforming single-objective reinforcement learning and naive reward-fusion baselines. In a 7-day online A/B test on a real search advertising system, UniPolicy improves CTR by 0.71%, RPS by 1.58%, and advertising revenue by 1.32%, while maintaining stable serving latency.

Autonomous recovery of a micro unmanned aerial vehicle (UAV) onto a moving airborne carrier enables reusable deploy-mission-recover operation, but couples long-range rendezvous, close-range perception, carrier motion, aerodynamic interaction, and a discontinuous contact event. This paper presents an RTK-vision-guided reinforcement-learning framework in which a child UAV is physically transported by a larger carrier, takes off from the carrier while airborne, executes an independent sortie, returns to the carrier's current position, redocks, and subsequently descends with the carrier. Both vehicles carry RTK-GNSS, and the carrier continuously shares its navigation state with the child. Near the recovery deck, RTK remains active while a downward-facing camera with a fiducial marker detector provides marker-relative alignment cues. A proximal policy optimization (PPO) policy governing the terminal recovery phase is trained in a physics-based MuJoCo simulation environment with explicit sensor noise models, an aerodynamic disturbance surrogate, and marker-latency randomization, then transferred to hardware. PX4 retains low-level stabilization, and a deterministic safety gate authorizes descent independently of the learned policy. The PPO checkpoint achieves 99.55% success over 2,000 held-out randomized terminal episodes, compared with 78.4% for a tuned PD baseline under identical conditions, with a median planar terminal error of 6.62 cm. Across 14 outdoor trials, the full mission succeeds in 13 trials (92.9%), spanning both near-region recovery and recovery after the carrier translates away from the release point. The results demonstrate a complete autonomous aerial deployment-and-recovery cycle rather than an isolated landing maneuver, establishing a practical basis for reusable carrier-child operation in inspection, surveillance, and mobile-logistics applications.

In this paper, we use optimal mass transportation to derive an isoperimetric inequality and an interior Michael-Simon-Sobolev inequality for weakly twice differentiable integral varifolds of dimension and codimension at least $2$, with locally bounded first variation and $L^2_{loc}$ mean curvature. This generalizes a result of Brendle and Eichmair. In both inequalities the corresponding constants are optimal in codimension 2. As part of this process we also generalize Alexandrov's theorem for convex functions to this setting. This necessitates the use of measure and distribution-valued solutions to certain geometric differential operators, in addition to the theory of $L^p$ Taylor expansions started by Calderón and Zygmund.

In this paper, we study a heat-wave fluid-structure interaction system posed on a periodic strip geometry and investigate whether higher-order $H^s$-estimates for arbitrarily large values of $s$ can be established for the coupled system despite the inherent mismatch of parabolic and hyperbolic regularity. We establish global existence of weak solutions and Sobolev regularity up to the $H^3$-level, which, to the best of our knowledge, is the highest regularity currently established for this system. We further show that, even for arbitrarily regular initial data, neither the semigroup approach through higher-order generator domains $D(A^k)$ with $k\in\mathbb{N}$ nor the PDE-based tangential-normal recovery procedure yields closed estimates beyond $H^3$ due to the complexities of the coupled PDE system.

We introduce new coordinates that extend the Kerr–Newman–Bertotti–Robinson spacetime consistently with the previous reciprocal continuation. In these coordinates, the two points at $\Omega=0$ are resolved into complete Bertotti–Robinson null infinities. The extended spacetime describes two black holes sharing a common exterior of nontrivial topology, together with a naked ring singularity. In the regular static limit, the new coordinates globally identify the spacetime with the balanced Alekseev–García geometry.

Large language model responses are non-deterministic, so failures in LLM agents are hard to reproduce: a failure depends on inference that is not bitwise reproducible, on tools that read changing state, and on a multi-step trajectory that a re-run rarely repeats. Record-and-replay makes a run reproducible, but existing agent tooling records runs only to trace or score them, not to test a code change against them. We present Chronicle, which records an agent run at its non-deterministic boundaries as immutable envelopes and replays it from the record. Its central operation, cut-point replay, serves a chosen subset of boundaries from the record and executes the complementary subset live with new code, turning a recorded incident into a regression test that runs in continuous integration. On a benchmark of 6 recorded failures with simulated model boundaries, recording adds 23 {\mu}s per crossing (0.008% of an assumed 300 ms model call), full replay issues zero model calls and is bit-stable across 20 repetitions, and cut-point tests fail on faulty code and pass on guarded and benign changes for all 6 incidents. In a mutation study of the guarded tools, cut-point tests catch every mutant that lets the recorded unsafe action through, while a baseline that stubs every boundary, using the same assertion, catches none. Chronicle and the benchmark are publicly available at https://github.com/theagentplane/chronicle.

A robot sent to a named gas leak must preserve gas identity, estimate the source, and navigate to the resulting goal. We present SmellDiffusion, a simulation pipeline that represents species-specific gas zones in an open-vocabulary olfactory scene graph and shares the selected goal between classical and diffusion planners. Its key components are a peak-local geometric gate for selective source correction and diffusion-based, gas-guided trajectory generation. Among 424 unique source-wind configurations in solved flow, 28 have a concentration peak displaced more than 0.5m from the source. A source-independent geometric gate, calibrated only on the training split and evaluated at the observed peak, detects 9 of 10 held-out displacements at 0.64 precision. Gating a precomputed forward-matching correction reduces mean error on the displaced cases from 1.468m to 0.592m (60%), using matching for only 14/204 cases. All-case mean error falls from 0.205m to 0.180m. All planners receive the same scene-graph source estimate as their goal. In a controlled comparison, best-of-ten diffusion achieves mean gas exposure comparable to gas-guided A* (0.0476 versus 0.0455). A single diffusion proposal takes 41.7ms, compared with 72.3ms for gas-guided A*, although best-of-ten sequential sampling increases total runtime. Plain A* also reaches the same goal and remains the fastest and shortest-path method. Six matched Gazebo runs give mean robot-to-source errors of 0.39m for A* and 0.31m for diffusion.

Recent single-view feed-forward 3D Gaussian Splatting (3DGS) generation predicts a fixed number of Gaussians per camera ray, introducing severe spatial redundancy. Most existing compaction strategies target multi-view setups to exploit cross-view consistency and are incompatible with single-image models. Instead of retraining the base feed-forward network to directly output compact representations, our insight is to keep the base models frozen and apply post-hoc pruning and recurrent refinement to the generated Gaussians. Consequently, we propose a backbone-agnostic compaction pipeline for single-view feed-forward 3DGS that couples an importance-score-based pruning mechanism with a trainable, lightweight recurrent refinement module, which iteratively updates the surviving primitives to restore image quality. Our results demonstrate seamless integration with existing baselines while preserving novel-view rendering fidelity and achieving high memory reduction. Furthermore, our method supports flexible inference-time keep ratios for application needs.

We study single-field slow-roll inflation in the context of Palatini gravity for the class of non-minimally coupled $\xi$-attractors, i.e.,\ models where the same function $f(\phi)$ fixes both the non-minimal coupling, $1+\xi f(\phi)$, and the inflationary potential, $V(\phi) = M^4 f(\phi)^2$. For an arbitrary $f(\phi)$, the number of $e$-folds is, at leading order, independent of $\xi$. Thanks to this, we provide an immediate mapping between the observables of the non-minimally coupled setup and those of the minimally coupled one. In the strong coupling limit, the tensor-to-scalar ratio is, as usual, suppressed, while the scalar spectral index is shifted towards larger values, with the magnitude of the shift depending only on the tensor-to-scalar ratio associated with the original potential $V(\phi)$.

Fingerprint recognition in children under four is of interest for longitudinal identity applications, but research in this age range is constrained by the limited availability and sensitivity of real fingerprint data. Synthetic data may provide a useful complementary resource if generated samples are carefully evaluated for biometric quality, similarity to the real training data, and identity diversity. This paper presents an evaluation and selection protocol for synthetic fingerprints generated from a small pediatric dataset. The protocol was applied to 32,000 candidates produced by an age-conditioned generator fine-tuned with 205 fingerprints from nine children. Candidates were evaluated using NFIQ 2, NBIS minutiae extrac-tion and matching, fingerprint-pattern classification, similarity to the complete real reference set, and pairwise similarity among retained synthetic samples. After candidate filtering and a final symmetric pairwise verification, 1,985 synthetic fingerprints remained. The youngest age group continued to produce retained samples within the sampling budget, whereas the oldest original age group produced 24 retained prints. A data-derived two-group age representation improved pattern-distribution agreement for the younger group. Repeated renderings of fixed synthetic iden-tities produced minutiae-based mated scores comparable to the real mated scores, although a DINOv2 texture representation showed substantially greater within-generator similarity. The results indicate that synthetic pediatric fingerprints can support controlled research use, but their evaluation should include both real-to-synthetic similarity and synthetic-to-synthetic diversity, and conclusions should remain specific to the matchers and measurements used.

Autonomous underwater vehicles (AUVs) operating beyond reliable communications must recover from failures without human intervention. We investigate an architecture in which conventional deterministic layered control autonomy manages normal operations, while an invokable large language model (LLM) serves as a diagnostic and recovery planner when onboard anomaly detection identifies performance outside expected limits. Because language models are stochastic, rigorous evaluation requires ensemble testing rather than individual demonstrations. We present a closed-loop simulation architecture that couples real-time C vehicle software with a higher-level orchestration layer for physics-based fault injection, structured prompting, language-model interaction, mission file generation, validation, execution, and LLM-judge scoring. The framework, which we call SPAR (Simulation Platform for AUV Recovery), supports evaluation across fault realizations, prompt structures, reasoning models, and mission conditions. We vary these for a mass-shift fault over 480 SPAR trials, evaluating a frontier model and three off-the-shelf locally deployable LLMs. Model choice dominates diagnosis: the frontier model places the CG-shift mechanism in its top three hypotheses in 85-90% of trials, versus 60-78% for the best local model. Reasoning analysis indicates that local-model success is associated with following the complete diagnostic procedure, whereas weaker models often commit prematurely to elevator failure even though the actuator tracks its command. Diagnosis and operational decision performance do not appear to be coupled in this dataset. The contributions are an architecture extending unanticipated-fault recovery from detection to mitigation and an ensemble methodology for evaluating LLM-assisted mission management on low-power AUVs.

In this paper we propose a geometrically motivated formulation of gravity in an invariant Weyl–Integrable space-time (IWIST), in which the background geometry is not imposed a priori but is dynamically determined through the Palatini variational principle. For a scalar field non-minimally coupled to gravity action, the resulting compatibility condition defines a Weyl–Integrable geometry which is preserved by a local Weyl transformation of the metric and the geo\-me\-tri\-cal scalar field. We construct an action invariant under both diffeomorphisms and Weyl transformations and introduce a Weyl-covariant variational procedure based on an invariant extension of the divergence theorem. The corresponding field equations are obtained for the metric, the Weyl scalar, and the Weyl gauge field. We further formulate the theory in the Einstein–Riemann frame, where the effective metric is Riemannian and the scalar field is re\-inter\-pre\-ted as a phy\-si\-cal degree of freedom of geometric origin. Finally, we propose a geometrical mechanism for the explicit breaking of Weyl symmetry by introducing a coupling between the Weyl gauge field and a current constructed from the scalar sector. The resulting theory remains diffeomorphism covariant while departing from local Weyl invariance. The Einstein–Riemann representation of the broken theory contains an additional generally covariant interaction, making the Weyl-invariant model a particular limit of a more general gravitational theory.