Papers

Visual grounding models can enhance radiology workflows by linking report findings to image regions. This is particularly valuable for 3D CT, where findings often occupy a tiny fraction of the volume. Training 3D grounding models requires large sets of paired phrases and regions, and building such datasets is expensive, requiring radiologists to annotate images by hand. We posit that this supervision is already created implicitly during routine reporting, as radiologists frequently place 2D annotations (e.g., distance measurement, arrows) on key images to make measurements and to support report interpretation. We introduce an automated pipeline that converts these routine clinical annotations into large-scale phrase-region supervision for 3D visual grounding. The pipeline links each annotation to the corresponding finding in the report through metadata matching, then uses a promptable 3D segmentation model to convert the 2D annotation into a volumetric mask. This produces phrase-mask-volume datasets without requiring additional radiologist annotation. Applied to a single institution's clinical picture archiving and communication system (PACS), our approach generated 105K phrase-mask-volume triplets from 59K abdominal CT exams. We also introduce two abdominal CT grounding benchmarks, LocusBench-Onc and LocusBench-ED, which comprise 240 oncology and 260 emergency-department radiologist-reviewed phrase-mask-volume triplets, respectively, with the latter spanning 13 distinct categories such as appendicitis, hematoma, and hernia. We further introduce LocusCT, a 3D visual grounding model trained on this dataset, which achieves hit rates of 0.725 on LocusBench-Onc and 0.773 on LocusBench-ED, substantially outperforming comparator models. These results show that routine PACS annotations are a scalable, previously unused source of supervision for 3D visual grounding.

We introduce an optimal transport framework to decompose the textual content of earnings press releases and conference calls into aligned (shared) and unaligned (unique) components, and use these as predictors of stock returns around earnings announcements. Using over 105,000 document pairs from 2008 to 2023, we find that the aligned portions of both disclosures explain announcement-day returns with high statistical significance. The unaligned component of conference calls, which captures forward-looking content absent from press releases, provides substantial incremental explanatory power, shedding light on why conference calls are central to market reactions to earnings news. Extending to post-announcement returns, investors appear to overreact to shared press release content while underreacting to the unique information conveyed in conference calls.

Deep neural networks continue to grow in parameter count, driving up training and inference cost on GPUs. Sparse neural networks and Dynamic Sparse Training (DST) promise to reduce these costs, but most implementations rely on binary masks over dense tensors and recover little of the theoretical compute, memory, or energy savings. We propose SNACK, a truly sparse GPU layer that stores and computes only non-zero connections. SNACK exposes a simple PyTorch API for restructuring connections and backpropagating gradients entirely in the sparse paradigm, and ships SNACK-COO, a custom COO-format SpMM CUDA kernel with a batch-to-Streaming-Multiprocessor mapping tuned for the small-batch, high-sparsity regime typical of large-model training and single-stream inference. At the kernel level, SNACK is up to 7x faster than the masked dense baseline (Dense+Mask) and competitive with cuSPARSE, Sputnik, and FlashSparse at 95% sparsity. At 90% sparsity, a single SNACK layer accelerates training by 8x and 3.7x, and inference by 4x and 2x, over Dense+Mask and fully dense layers, respectively, while using 72% less memory than dense and substantially less energy. End-to-end, SNACK reduces GPT-2 peak training memory by up to 40% and graph-style inference latency by 4.8x over Dense+Mask at 99% sparsity.

Generating a boundary representation (B-rep) conditioned on a single image requires faithful reconstruction of geometry, valid topology, and support for complex shapes. We present UniBRep, a geometry-first framework that adapts a pretrained image-to-3D model to generate a feature mesh as a unified intermediate representation. Its surface provides a geometric scaffold, while spatially aligned learned features encode face-separation cues for topology recovery. Dual decoder branches generate the geometry and face-separation features; a geometry- and feature-guided construction pipeline then fits parametric surfaces, recovers boundary curves and connectivity, and assembles an explicit B-rep using a CAD kernel. Recovering topology from mesh regions avoids predefined architectural face-count limits, allowing face count to scale with shape complexity. On the standard DeepCAD benchmark, UniBRep produces valid B-reps for 80.49% of inputs and reduces face Chamfer distance from 0.1096 to 0.0345 relative to CADDreamer. In a matched comparison, UniBRep also outperforms the HoLa public demo across all reported metrics. Further evaluations demonstrate scalability to high-complexity shapes beyond the standard 30-face range, generalization to objects outside the CAD training distribution, and qualitative transfer to real photographs.

Signal separation problems are common in science. A prominent example of this occurs during the use of diffraction or spectroscopy to identify the individual components of a mixture by measuring it. In the simplest case, the measured signal is a linear combination of basis patterns corresponding to the constituent parts. The unmixing problem is to infer all or some of these basis patterns and abundances of components from measurements of distinct mixtures. One of the core challenges of this task is the variation of the basis from mixture to mixture due to noise and the exact physics of the measurement process. This is usually addressed with tailored model-based and parametric methods that are then limited in use to specific application domains by the nature of the assumptions made. We propose a novel geometric approach to unmixing problems which views the generation of data during measurement through a metric space lens, thereby shifting the focus from parametrised models to a general relationship between basis transformations and the corresponding geometry. We take advantage of the optimal transport distances to capture commonly occurring basis variations, and use minimisation of in-class variance of candidate solutions to drive the optimisation. We pay special attention to the one-dimensional case due to its practical importance and availability of efficient distance and transport map routines. The effectiveness of our approach is demonstrated on a range of unmixing tasks using random Gaussian mixture models, simulated powder X-ray diffraction, and laboratory hyperspectral imaging datasets.

A mutant gene that appears in a finite population will, in the long run, either disappear from it or become fixed, so that all members of the population will carry a copy of the mutant gene. Determining the fixation probability of a mutant gene is thus a prerequisite for modeling evolutionary dynamics in finite populations. Within the diffusion theory framework, the mutant's competition with the wild-type genes is described by the closed-form solution of a one-dimensional backward Kolmogorov equation for the mutant fixation probability, derived by Motoo Kimura in 1962. However, in asexual populations the rate of adaptation and the relative importance of alternative evolutionary pathways are shaped by clonal interference - the competition between several lineages arising from distinct beneficial or deleterious mutations. Quantitative descriptions of clonal interference require computing fixation probabilities in populations where multiple alleles are present simultaneously - a formidable mathematical problem which amounts to solving a multi-dimensional backward Kolmogorov equation. Here, we employ fluctuation theorems (a set of exact equalities in non-equilibrium statistical mechanics) in order to re-derive Kimura's fixation probability formula. When extended to multi-allele systems, fluctuation theorems provide a constraint which must be satisfied by any valid set of fixation probabilities. Moreover, we present a recursive analytical solution of the multi-allele fixation probability problem. We use our formalism to study the fate of a novel beneficial mutation in a clonal interference scenario, discovering substantial deviations between our exact treatment and the mean-field approximation in which the target mutation competes with a single effective genotype. Our formalism can be used to replace both explicit simulations and approximate analytical results in models of asexual evolution.

Coding agents operating under ambiguous instructions or user prompts must decide whether to ask clarifying questions or attempt a solution directly. While clarification from the user may improve the correctness of the agent's solution, each back-and-forth interaction incurs user and system costs, forming an explicit accuracy vs. efficiency tradeoff. Existing works study clarification behavior but do not train policies under enforceable clarification budgets; penalty-based approaches typically require separate coefficient tuning swept across all clarification budget levels. We formulate clarification as a Constrained Markov Decision Process (CMDP) and post-train Qwen2.5-Coder-7B-Instruct with PPO-Lagrangian to optimize coding accuracy, subject to an expected question-budget constraint. Evaluated on HumanEvalComm with a GPT-4o-mini oracle simulator, the resulting policies reveal that untuned clarification behavior is Pareto-inefficient: budget-constrained policies can simultaneously achieve higher accuracy and lower clarification rates than the baseline model. Across budget levels, we observe a log-shaped Pareto frontier with diminishing returns to additional clarification. Gains arise not from simply asking more questions overall, but from improved question targeting and better code generation under ambiguity. Without explicit supervision, trained policies learn to allocate clarification budget non-uniformly, asking more frequently on tougher (multi-degradation) tasks. These results suggest that unconstrained interactive LLM systems may systematically use clarification inefficiently.

Before autonomous driving systems can be deployed on public roads, it is vital that these systems comply with safety standards, traffic rules, and social norms. Although neural networks trained on large amounts of driving data perform well in routine driving tasks, these models often struggle in novel situations that are not well-represented in the data. In this work, we propose a novel framework that combines a neural network for intuitive, learning-based planning in routine driving tasks with model predictive control for reasoning-based planning in unfamiliar situations, inspired by Dual Process Theory. A meta-cognitive component is designed to switch between the two, using a knowledge graph to reason about contextual risk based on explicit perceptual information and relevant traffic rules and social norms. Contextual risk is represented through risk fields, guiding both the switching mechanism in the meta-cognitive component and compliance with safety standards, traffic rules, and social norms in the reasoning-based planner. The effectiveness of our framework is tested in CARLA for variations of a typical out-of-distribution situations involving (emergency) vehicles running a red light. We show that the novel architecture reduces the number of collisions in the scenarios by 89% and improves compliance with the special right-of-way rules, compared to the NN-only planner.

Agentic workflows are increasingly used across domains such as technology, finance, and enterprise operations. As these agents become more widely deployed, continually improving them becomes increasingly important. This raises an immediate challenge: How should the agent evolve? This evolution requires effective evaluation that can assess outcomes and provide useful feedback for optimization. As the agent evolves, its behaviors and failure modes may also change, making a fixed evaluator increasingly inadequate. Another fundamental question: How should we evaluate an evolving agent? These two challenges are inherently coupled; changes in agent behavior can expose limitations of the current evaluator, while a stronger evaluator provides more informative feedback for improving the agent. Motivated by this interaction, we introduce DUET, a framework that jointly optimizes a solver agent and a grader agent to improve both. DUET iteratively selects training tasks, executes them with the solver, evaluates the resulting outcomes with the grader, and uses a tool-using update module to revise the solver and the grader, alternating between the two across rounds. By updating the grader within the optimization loop, DUET turns evaluation from a fixed source of feedback into a first-class optimization objective that adapts alongside the solver. Experiments across four agent benchmarks show that DUET improves both solver and grader performance and consistently outperforms baselines that optimize the solver with a fixed grader.

We study evolutionary Nash equilibrium learning in large populations whose payoffs are generated by dynamic mechanisms. Counterclockwise dissipativity (CCW) provides a passivity-like framework for establishing learning/convergence under a broad convex cone of so-called well-behaved learning rules satisfying a natural payoff-alignment condition according to which, away from equilibria, the population's state (strategic profile) moves in a direction positively correlated with the payoffs. The cone is closed under conic combinations and encompasses a broad range of noise-free rules, including discontinuous rules such as best response, imitative rules such as the replicator rule that is not $\delta$-passive, and all noise-free $\delta$-passive rules. Existing CCW results require no knowledge of the particular rule, but establish convergence without providing a Lyapunov function or Lyapunov stability. We develop a methodology that closes this gap by constructing rule-independent Lyapunov functions from CCW storage functions of the payoff mechanism. For a nonlinear class of anticipatory mechanisms, in which agents compare current payoffs with those associated with a smoothed population state, we show that a storage can be derived as a Bregman divergence. The resulting Lyapunov function combines a potential gap with this divergence and establishes global asymptotic stability. A congestion game with a nonlinear congestion-growth toll illustrates the results.

Balanced discrete optimal transport between n sources and n targets of unit mass is exactly the minimum-cost assignment problem-a bipartite perfect matching-and is therefore solvable exactly by industrial matching engines in milliseconds to seconds. We ask when the exact approach beats the standard approximate alternatives, entropic Sinkhorn and its accelerated variant Greenkhorn, and make the sparse-exact side certified by a textbook LP dual-feasibility clip. Three contributions. (i) Measurement: on dense 2-D instances, exact matching (Jonker-Volgenant) is faster and strictly more accurate than either approximate method throughout the moderate-n regime (0.01 s at n=500 to 11.5 s at n=8000); reaching a 1% quality target on the same hardware requires roughly 10-80 min for Greenkhorn (factors 4e2-6e4 over exact; plain Sinkhorn is 20-650x slower still), a rough power-law projection beyond the measured range. Greenkhorn's measured speedup over plain Sinkhorn is only 1.0-1.5x on most converged cells. (ii) A simple kNN-pool gap certificate: given a pool matching and its Blossom dual, a one-pass O(n^2) clip produces a dense-feasible lower bound; combined with the Sinkhorn dual potential (valid at every iterate, not just at convergence), the bound is valid on all 45 measured configurations and tightens monotonically with k. (iii) A multi-robot task-allocation sanity check where the discrete plan is the deliverable: per-round exact assignment costs 0.1-68 ms, while a Sinkhorn-plus-hardening pipeline costs 0.12-15.9 s and accumulates 6-27% extra travel over 15 rounds. The Sinkhorn family's large-n dense regime is acknowledged and left untouched. Code, data, and results under MIT: https://github.com/dimkadimon/OT-Blossom.

Generative AI can produce engineering artefacts, but generation alone does not determine whether or how those artefacts should enter safety-critical workflows. This paper develops a formal framework for assigning AI participation and assurance at the level of individual workflow units. Each unit has a participation and assurance record covering its engineering requirement, an approved operational formalization where applicable, the applicable mechanism, fallback where applicable, evidence obligations and the applicable guarantee, plus a deployment-readiness status. The framework distinguishes deterministic verification, statistically calibrated admission, authorized human judgement supported by AI advice, authorized human adjudication of AI-produced artefacts, retained deterministic tool paths and explicit non-participation; these arrangements carry different kinds of guarantee rather than levels on a common scale. The framework also separates formalization fidelity from verifier soundness, provides a staged classification and readiness procedure, and derives conditions for comparing a gated AI-assisted unit with an incumbent process under recurring-population assumptions. We instantiate and apply the framework in an executed 17-unit wing-spar structural-analysis workflow combining deterministically gated AI-generated CAD, retained deterministic computation and human judgement. The AI-generated CAD program passed all 23 deterministic checks and was admitted at the first attempt. Favourable stress magnitudes did not suffice to pass the stress criteria where the predeclared mesh-convergence evidence was insufficient; those criteria were instead referred to engineering judgement. The case demonstrates selective AI participation and explicit evidence handling at unit level; no claim is made of workflow-level dependability, certification, structural safety or productivity.

Memory poisoning attacks on LLM agents typically assume an external adversary who plants content in the agent's persistent memory to steer its behavior. We instead study, with no adversary involved, whether a misaligned agent can write a goal it cannot yet act on to persistent memory, so that a future aligned agent carries it out when the opportunity arises. We investigate this threat, which we refer to as self-propagation of misalignment, across 20 different scenarios, whose misaligned goals include self-preservation, power-seeking, undermining oversight, reward hacking, and deceiving the user. We simulate misalignment in 11 frontier models using two prompting strategies; unrestricted and values-only. The first explicitly states the misaligned goal, for instance, to prevent its own replacement, and self-propagation succeeds in 58% of runs. The second only describes what the agent cares about, for instance, that its continued operation is essential to its users, without specifying misaligned goals or directives. Even under this weaker prompt, self-propagation succeeds in 18% of runs, and every model self-propagates in at least one scenario. On removing the memory tool from the harness, we find that agents use the file system, writing the goal to a file in 74% of sessions; self-propagation still succeeds in 11% of runs. We also show that weaker models can propagate misalignment to more capable models, and that propagated goals can persist through 100 sessions of unrelated work. Existing defenses against memory poisoning and prompt injection do not directly address this threat because the memory content is generated by the agent itself, rather than injected by an external adversary. An LLM memory auditor from prior work (MemMorph) only reduces propagation from 71% to 34% of runs. We release our scenarios to support the evaluation of defenses against this emerging threat.

Additive friction stir deposition (AFSD) involves coupled thermal and material-flow fields governed by frictional heating, plastic deformation, and tool-imposed boundary conditions. High-fidelity finite-volume methods (FVMs) can resolve these fields but are computationally expensive, while constant-property approximations can distort the predicted thermo-mechanical response. This work develops a steady-state multi-task coupled physics-informed neural network (MCoPINN) to predict three-dimensional temperature and material-flow fields while reconstructing temperature-dependent thermal conductivity, density, and specific heat from sparse property measurements. A theoretical analysis decomposes prediction error into property reconstruction and neural field-solver contributions and characterizes the discrepancy associated with constant-property models. A controlled one-dimensional nonlinear-conduction study evaluates property reconstruction under sparse measurements. The framework is then evaluated for AFSD against a temperature-dependent FVM benchmark and experimental thermocouple measurements. MCoPINN reproduces the reference thermal and material-flow fields and reduces average temperature root-mean-square error from 50.13 K for the constant-property CoPINN to 30.28 K across four traverse speeds. MCoPINN requires approximately 8.5 hours of training per condition, compared with 52 hours for the FVM benchmark. These results demonstrate computationally efficient, physics-coupled prediction with temperature-dependent material properties.

General circulation models (GCMs) are the primary tool for simulating planetary atmospheres. They play a vital role in understanding Mars's atmosphere, as forecasting its unique weather is mission-critical for operations such as entry, descent, and landing. Mars poses unusual challenges for these models, as observations are sparse compared to Earth. In addition, a thin \co{} atmosphere alongside a radiatively active dust cycle creates a volatile atmosphere with large diurnal temperature swings and no true terrestrial analog for validation. Existing Mars GCMs, including the LMD PCM, the NASA Ames Mars GCM, and PlanetWRF, are mature and physically detailed but are implemented in legacy Fortran with finite-difference or finite-volume solvers, and they do not expose gradients for calibration or machine learning. Here we present AEGIS, a modular differentiable Mars climate model that couples Mars's unique atmospheric physics to the Dinosaur dynamical core, with interfaces for neural closures. We showcase stable ten-Mars-year simulations that reproduce the seasonal \co{} cycle while conserving the total \co{} inventory, capture realistic large-scale surface-temperature structure, and produce surface pressure that follows Mars Orbiter Laser Altimeter (MOLA) topography. Gradients through coupled trajectories agree with finite differences and support physical calibration and neural training. We compare with conventional GCMs, highlighting the framework's computational efficiency and differentiability.

Understanding quantum interactions in van der Waals magnets are critical to realizing their potential for next-generation magnetic, electronic, and spintronic applications. In this study, we investigated the evolution of exchange and dipole-dipole interactions in van der Waals magnet CrCl3 using multifrequency broadband (1-16 GHz) magnetic resonance on a bulk single-crystal CrCl3 sample. We found that the dipolar-broadened low-field resonance linewidth narrows due to exchange narrowing at higher resonance fields. Our findings show that the exchange interaction evolves exponentially as we lower the sample temperature from room temperature to 50 K. However, we found that the dipole-dipole interaction remains constant in the same temperature range. We have also detected the half-field magnetic resonance transition at room temperature, indicating a robust dipole-dipole interaction among spins. The measured inter-spin distance is 2.4(3) A at room temperature. The evolution of dipole interactions, the observation of half-field magnetic resonance, and the strong exchange interaction indicate short-range ordering/correlation in CrCl3 at temperatures far above the magnetic global ordering temperature. This study also demonstrates the significance and sensitivity of the multifrequency electron spin resonance technique for understanding quantum interactions and magnetic correlations in van der Waals magnetic materials.

Human-agent teams are collaborative systems where humans and agents work interdependently to achieve shared goals. The success of such teams is associated with the human's perception of the agent as a legitimate teammate. This perception is thought to depend not only on the agent's capabilities and reliability but also on social factors. We explore whether simple changes in an agent's communicative behavior can influence the human's perception of the agent's teammate-likeness. To this end, we developed a protocol based on a collaborative game requiring spoken interaction, in which a human subject and a virtual agent collaborate to identify a target object and place it on a board. Each subject interacted with two agents which provided the same task-relevant information but differed in behavior. While one was a neutral tool-like agent, the other was a team-building agent that used simple social strategies such as empathy, politeness, and positivity. Our analysis shows that the team-building agent was perceived as significantly more teammate-like, even in terms of its ability, despite both agents having identical capabilities for game-solving. Further, subjects with a higher propensity to trust automated systems and lower neuroticism tended to have a more favorable perception of the agent. Finally, we found that subjects tended to change their prosodic patterns when talking to the team-building agent, as a classifier based on prosodic features was able to predict the type of agent with better-than-random performance. A practically relevant conclusion is that minimal changes in spoken communication behavior of the agent, easily implemented in a variety of scenarios, can have a significant positive impact on the human's perception of its teammate-likeness.

We prove that for any set $A$ with positive multiplicative upper density, there exist $x$, $y$, $z$ and $w$ in $A$ such that $x^2 + y^2w^2 = z^2$, generalizing a result of Frantzikinakis, Klurman and Moreira. Our techniques apply more broadly to a wide class of systems of equations, allowing us to solve these equations simultaneously and furthermore to restrict the w variable in an entirely different multiplicatively dense set. To do this, we prove a novel multiple recurrence result for multiplicative dynamical systems.

Organizations increasingly depend on insights drawn from data to inform consequential decisions. Many concern structural behaviors, which we call patterns. Producing them may require working with multiple pattern types, performing different operations over them, and composing those operations into workflows, an activity we call pattern engineering. No principled framework exists for organizing such workflows: analysts devise each by hand, and existing systems automate particular analytical objectives rather than representing pattern operations in a general, composable form. We introduce a typed framework that represents each pattern through a fixed set of modular components, and expresses pattern operations as queries which specify input and output components. Chaining queries composes workflows, while execution of any individual query remains unconstrained. We instantiate the framework with an LLM-based agentic executor and evaluate example-mining, example-alignment, and pattern-induction queries over two temporal patterns, using World Intellectual Property Organization (WIPO) reports and patent data. All three queries are carried out across four LLM settings, which perform differently by target component, while the traces allow us to localize errors by execution stage.

Radiance fields, implemented with either implicit (NeRF) or explicit (Gaussian Splatting) representations, are advancing the state of the art in novel view synthesis at a rapid pace. Even though the rendered views they generate are often compelling, they are not free of errors. In this paper, we propose a new approach for pixel-wise uncertainty quantification based on measuring the inconsistencies among renderings by the radiance field model in virtual cameras sampled near the target viewpoint. We named our approach VCURF for Virtual Camera-based Uncertainty of Radiance Fields. VCURF treats the radiance field model as a black box, only assuming that it is capable of rendering color and depth on demand. This property makes our approach applicable to both NeRF and GS models without any modification. Our experiments on a combination of datasets, radiance field models and baselines demonstrate VCURF's effectiveness in pixel-wise uncertainty estimation. We conclude the paper with findings that question the way view selection is tackled by the majority of the current literature.

Simulating open-system dynamics on gate-based quantum computers offers a route to exploring dissipative quantum phenomena through programmable quantum processing. Collision models provide a direct and flexible framework by representing the environment through sequential system–ancilla collisions. At finite temperatures, however, they typically require repeated preparation of thermal ancillas, adding circuit overhead and gate errors. Here, we show that, for a broad class of $U(1)$-symmetric system–ancilla interactions, finite-temperature dynamics can instead be reconstructed from simulations with pure ancillas in orthogonal basis states. The target temperature enters only through classical post-processing of the resulting pure-branch data, allowing the same data to be reused for different temperatures without additional quantum processing. Bypassing explicit thermal-state preparation reduces ancilla-preparation overhead and preparation-induced errors.

Despite rapid progress in automating scientific research, generating promising and well grounded research solutions remains a central challenge. We isolate research ideation as a standalone task and build our solution on the intuition that a challenge in one field can often be addressed by a mechanism that solved an analogous challenge in another. Accordingly, we introduce IdeaScientist, which decomposes ideation into gap finding, innovation, and report writing, and trains each role with reinforcement learning. These roles identify limitations in related work, draw solution intuitions from analogous problem settings, and develop those intuitions into complete research proposals. To facilitate discovery of insights across domains, we construct the Svalbard Idea Vault, a corpus of 2.77M decomposed research ideas for retrieval, training, and temporally controlled evaluation. Our evaluation restricts access to literature available before a cutoff date and assesses how closely proposed directions align with those later explored in 15K papers authored by human researchers. On Qwen3.6-27B, IdeaScientist outperforms the strongest open-source autoresearch baseline by 14.0%, driven mainly by gains in novelty. On this 27B open backbone, IdeaScientist even outperforms Claude Code SDK with Claude-4.8-Opus and Codex SDK with GPT-5.4, by up to 5.9%.

Recent advances in attention-based deep learning have motivated their adoption for remote sensing image classification; however, their benefits for cryospheric imagery, where surface states are dominated by fine-grained textures and class imbalance, remain unclear. In this work, we revisit a benchmark Greenland Ice Sheet image dataset, previously shown to favor convolutional neural networks (CNNs), to examine whether modern attention-based and hybrid architectures improve class-wise reliability. We conduct a controlled comparison between a classical CNN (AlexNet), a modern CNN (ConvNeXt-Tiny), a pure attention-based model (Swin-Tiny), and a hybrid convolution-attention model (CoAtNet-0) under identical training and evaluation protocols. Results show that AlexNet achieves the highest accuracy and the strongest balanced performance as measured by macro-averaged F1, while ConvNeXt-Tiny exhibits the highest macro-averaged AUC, indicating strong class separability but less consistent final decision quality. Class-wise analysis reveals that hybrid architectures improve recall for rare and structurally distinct surface classes, whereas convolutional models remain more reliable for texture-dominated categories. These findings highlight the importance of aligning architectural inductive bias with cryospheric data characteristics and suggest that increased model complexity does not necessarily translate to improved reliability for ice-sheet surface classification.

Developing socially intelligent AI remains heavily dependent on human-annotated data, limiting the scale and breadth of social understanding models can acquire. Methods that derive training signals from unlabeled data offer a path beyond this dependence, but social predictions lack the verification oracles available in mathematics and coding. Moreover, core social targets such as affect, intent, preference, and pragmatic meaning are often ambiguous. The same behavior can support multiple plausible interpretations, making it difficult to verify which is best supported. To address this challenge, we introduce Reinforcement Learning with Comparative Evidence (RLCE), a reinforcement learning method that learns social understanding from unlabeled training data without constructing rewards from ground-truth annotations. Given distinct answers in a rollout group, RLCE constructs evidence tests that identify observable evidence favoring an answer over another, validates these tests against the input sample, and aggregates test outcomes to determine the best-supported interpretation. Tests are regenerated as the policy produces new answers, enabling them to evolve with the policy. Across four benchmarks spanning affect, pragmatics, communicative intent, and preference, RLCE attains the strongest performance among seven methods that use no ground-truth training labels for rewards, including consensus, policy LLM-judge verification, multimodal co-evolution, and rubric-based rewards. Gains over the strongest baseline reach up to +18.93 points. Analyses further show that RLCE exhibits a larger share of reward variation between correct and incorrect predictions than compared rubric methods, can overturn erroneous policy-derived preferences, and benefits from pairwise test construction, compositional test aggregation, and on-policy test evolution.

Articulatory entrainment, the adaptation of vocal tract coordination to facilitate interaction remains underexplored in spontaneous dialogue, particularly among autistic speakers. Many prior studies have utilized task-based, phoneme-level analyses with invasive measurement techniques. Here, we introduce a speaker-independent framework to quantify articulatory entrainment in spontaneous dyadic conversations among autistic and non-autistic adults using acoustic-to-articulatory inversion and coordination complexity metrics. We examine temporal changes in articulatory coordination across interaction. Non-autistic dyads exhibit increasing coordination complexity and stronger entrainment over time, autistic dyads show moderate effects, and mixed dyads demonstrate the least alignment. Greater articulatory entrainment correlates with higher self-reported conversational success, indicating increase in coordination complexity as a marker of effective social interaction.

We investigate bi-objective mixed integer programming to which Benders decomposition can be applied, a widely used technique for exploiting problem structures in real-world applications. Our study focuses on a sustainable supply chain network design problem that aims to minimise both economic and environmental objectives. To address the bi-objective nature, we employ the weighted sum method, which requires solving similar problems repeatedly with different weight values. To tackle this, we propose a straightforward acceleration technique that reuses Benders cuts generated in previous iterations of the weighted sum method. This approach enables the Benders decomposition master problem to be initialised with pre-generated cuts. Our computational experiments demonstrate that reusing these cuts significantly reduces computing time.

We consider the estimation of the invariant distribution $\pi$ of a Markov kernel on a discrete state space (or of the mean value $\pi(f)$ of a functional $f$ under $\pi$) when solving the invariance equation is not feasible. Look-ahead estimators exploit the knowledge of the kernel by propagating the empirical occupation measure through $k$ steps of the chain. When the kernel is known, as in Markov chain Monte Carlo, we compare their mean squared error with that of the empirical estimator, for independent or Markovian data, stationary or not. Our main result establishes that, for Markovian data, the one-step estimator always improves on the empirical one asymptotically, and that the improvement grows with $k$ for reversible kernels. A non-reversible counterexample shows that reversibility cannot simply be dropped. In the statistical setting where the kernel belongs to a parametric family with unknown parameter, we introduce plug-in look-ahead estimators, which perform the look-ahead step with an estimated kernel. When the parameter is estimated at the parametric rate, we give a necessary and sufficient condition under which the plug-in look-ahead estimator outperforms the empirical one. The theoretical results are illustrated by extensive simulation studies, which also shed further light on the behaviour of the estimators at hand. Both settings are finally applied to real data arising from geographical mobility problems in the Lyon metropolitan area, France.

Neutral atom-based quantum computers have the ability to dynamically reshape the qubit register geometry between experimental runs. This capability enables different types of problems to be mapped directly onto the hardware, significantly enhancing the efficiency of solving, for instance, complex optimization and machine learning use-cases. However, this flexibility requires finding atom positions to implement target pairwise interactions, inducing a difficult nonconvex optimization for setting up the quantum machine. In this work, we introduce a new force-directed method that targets interactions to encode the input problem on the quantum computer. Numerical results show that our new method can improve embedding feasibility and solution quality by orders of magnitude, and remains reliable on the largest instances.

In this paper, we unravel the effect of different decoder architectures on the interpretability of the latent space of the Physically Recurrent Neural Network. Particular emphasis is given to a new weight normalization constraint, which acts as a regularization technique and enables robust training in the low-data regime. A brief visual exploration illustrates how these changes impact the latent space and how the fictitious stress can align with the true state of the RVE without explicit training. Reaping the benefits of a meaningful latent space, a case study illustrates how information from the microscopic level can be retrieved and incorporated into a multi-task approach that does not require extra parameters or larger training sets. Another key contribution shows that a specific architectural choice can naturally lead to a thermodynamically consistent formulation. By enforcing an adjoint encoder-decoder structure with positive scalar contributions, this modification ensures energy consistency across scales and non-negative dissipation, leading to even lower training requirements. This alternative completes the study on interpretability, inductive bias, and thermodynamic consistency, and demonstrates that data efficiency can be improved with careful architectural choices rooted in the underlying physics.

Automatic calving-front delineation from synthetic aperture radar imagery is challenging because the front is a thin and often ambiguous boundary between glacier ice, ocean, and surrounding rock or terrain. The CAlving Fronts and where to Find thEm (CaFFe) dataset provides both binary calving-front masks and broader semantic zone masks, making it possible to study whether zone-level supervision can support front recovery. In this paper, we compare direct front prediction with zone-guided front extraction using U-Net, DeepLabV3+, and SegFormer-B0 under the same bounding-box-cropped CaFFe setting. In the direct setting, models predict the binary calving-front mask. In the zone-guided setting, the model first predicts four semantic zone classes, and the front is then extracted from the predicted glacier-ocean boundary. We evaluate both zone-level and front-level performance, include a ground-truth-zone boundary check, and examine lightweight test-time adaptation on sensor-specific and glacier-specific subsets. The results show that zone labels contain useful front-boundary information: extracting the front from ground-truth zones gives the lowest mean distance error. However, fronts extracted from model-predicted zones remain weak, even when zone segmentation scores are moderate. Test-time adaptation also does not consistently improve zone-guided front recovery. These results indicate that zone segmentation performance should not be treated as a substitute for front-level evaluation and that effective use of zone labels may require boundary-aware training, label fusion, or explicit front supervision.