arXiv:2602.16793v2 Announce Type: replace Abstract: In the past year, custom and unreleased math reasoning models reached gold medal performance on the International Mathematical Olympiad (IMO). Similar performance was then reported using large-scale inference on publicly available models but at prohibitive costs (e.g., 3000 USD per problem). In this work, we present an inference pipeline that attains best-in-class performance on IMO-style math problems at an average inference cost orders of magnitude below competing methods while using only general-purpose off-the-shelf models. Our method relies on insights about grader failure in solver-grader pipelines, which we call the Cognitive Well (iterative refinement converging to a wrong solution that the solver as well as the pipeline's internal grader consider to be basically correct). Our pipeline addresses these failure modes through conjecture extraction, wherein candidate lemmas are isolated from generated solutions and independently verified alongside their negations in a fresh environment (context detachment). On IMO-ProofBench Advanced (PB-Adv), our pipeline achieves 67.1 percent performance using Gemini 3.0 Pro with an average cost per question of approximately 31 USD. At the time of evaluation, this represented the state-of-the-art on PB-Adv among both public and unreleased models, and more than doubles the success rate of the next best publicly accessible pipeline, all at a fraction of the cost.
Science Journals
arXiv:2606.15420v1 Announce Type: new Abstract: A constitution tells a language model what to value, but little tells us whether it does. Adherence is judged from outputs, and output evidence is most fragile on value conflicts, where what matters is not which value a model mentions but which one it is willing to sacrifice. We provide evidence that this arbitration can be read from activations in a structured margin readout. We introduce Constitutional Value Potentials (CVP). For each value we learn a scalar potential from the hidden state: an internal pressure to preserve that value, supervised not by the prompt but by an independent judge's verdict on which value the model's own response actually preserved. The signed difference of two potentials is a priority margin. A constitutional clause becomes the claim that a margin stays positive, and a single monitor score flags when it does not. The monitor predicts conflict violations with AUROC up to 0.95, beats a strong hidden-state probe, and generalizes to held-out synthetic conflicts across three Qwen2.5 scales. The signal appears as the answer begins, from the prompt tail and first response token. Read this early, the same signal reveals whether an adversarial priority hack has actually pushed the model toward a violation, rather than only whether the prompt looks adversarial. The same directions also support intervention tests: under selected steering settings, moving along a value direction shifts judged trade-offs in the intended direction. Together, these results suggest that some constitution-relevant priorities are accessible as activation-space margins, rather than only as output behavior.
arXiv:2606.16991v1 Announce Type: new Abstract: Multiphasic contrast-enhanced CT (CECT) is widely used for abdominal lesion characterization, yet it carries inherent risks of contrast-induced nephropathy, escalates acquisition burden, and heavily contributes to radiologist workload. To address these challenges, we introduce a novel multi-center benchmark for multi-organ abdominal disease diagnosis and automated radiology report generation, which learns to synthesize contrast-enhanced findings from single-phase non-contrast CT (NCCT). To support this, we curated a large-scale dataset of paired NCCT-CECT studies and their corresponding contrast-enhanced radiology reports from two centers, partitioned into internal sets and an external validation cohort. Under a unified evaluation protocol, we benchmarked five contemporary deep learning architectures encompassing chest-specific, abdomen-specific, and general-purpose multimodal domains. Extensive experiments demonstrate that NCCT retains diagnostic signals, achieving an average multi-organ AUC of 69.1% on the internal cohort and 63.1% on the external cohort, respectively. By releasing this dataset and standardized benchmark publicly, this study aims to catalyze future research into safer, resource-efficient, and globally accessible contrast-free abdominal imaging workflows. Code is available at: https://github.com/xmed-lab/TriALS-Report.
arXiv:2606.16994v1 Announce Type: new Abstract: Topological semimetals have been proposed as electrocatalytic platforms because their surface states may participate in adsorbate bonding and interfacial charge response. Here we examine this role for hydrogen chemisorption on NbP(001) by combining density functional theory, Wannier-based surface spectral functions, orbital projections, and projected crystal orbital Hamilton population analysis. An SOC on/off comparison allows us to examine two electronic regimes of the same surface: including nodal-line-derived surface states in the non-SOC limit and Weyl Fermi arcs in the SOC case. We find that SOC changes the hydrogen adsorption free energy only weakly, showing that the equilibrium adsorption descriptor is largely insensitive to the SOC-driven reorganization of the surface states. Momentum- and energy-resolved bonding analysis assigns the main H-surface stabilization to occupied local H-Nb hybridized states below the Fermi level. At $E_F$, the Fermi arcs provide the dominant hybridization channels that contribute to the state-resolved bonding pattern and transfer spin-textured character to the adsorbate, producing a finite H-projected current-induced spin polarization. Thus, NbP demonstrates that topological surface states can play a minor role in adsorption thermodynamics while governing Fermi-level hybridization and adsorbate-local current-induced spin polarization.
arXiv:2602.07687v2 Announce Type: replace Abstract: We present a low-rank Koopman operator formulation for accelerating deformable subspace simulation. Using a Dynamic Mode Decomposition (DMD) parameterization of the Koopman operator, our method learns the temporal evolution of deformable dynamics and predicts future states through efficient matrix evaluations instead of sequential time integration. This yields log-linear scaling in the number of time steps and allows large portions of the trajectory to be skipped while retaining accuracy. The resulting temporal efficiency is especially advantageous for optimization tasks such as control and initial-state estimation, where the objective often depends largely on the final configuration. To broaden the scope of Koopman-based reduced-order models in graphics, we introduce a discretization-agnostic extension that learns shared dynamic behavior across multiple shapes and mesh resolutions. Prior DMD-based approaches have been restricted to a single shape and discretization, which limits their usefulness for tasks involving geometry variation. Our formulation generalizes across both shape and discretization, which enables fast shape optimization that was previously impractical for DMD models. This expanded capability highlights the potential of Koopman operator learning as a practical tool for efficient deformable simulation and design.
arXiv:2602.17990v2 Announce Type: replace Abstract: Multi-agent LLM systems that generate structured workflows from natural-language requests are now deployed in production across cloud automation, DevOps, and enterprise process orchestration. Operating such systems exposes a recurring change-management problem. Routine updates, such as re-running the same input, swapping the underlying LLM, or refactoring an agent's prompt or orchestration code, frequently produce workflows that differ substantially from previously validated references. Engineers are then left without a principled way to decide whether a change is safe to ship. Automatic workflow evaluation is the natural tool for answering this question. In practice, however, metric scores are poorly calibrated, and a numeric change rarely communicates the severity of the underlying degradation. We introduce WorkflowPerturb, a controlled benchmark for studying workflow evaluation metrics by applying realistic, graded perturbations to golden workflows. WorkflowPerturb contains 4,973 golden workflows and 44,757 perturbed variants across three perturbation types (Missing Steps, Compressed Steps, and Description Changes), each applied at severity levels of 10%, 30%, and 50%. We benchmark multiple metric families and analyze their sensitivity and calibration using expected score trajectories and residuals. Our results characterize systematic differences across metric families and support severity-aware interpretation of workflow evaluation scores in change-management settings. Our dataset will be released upon acceptance.
STEPS: Semantic Contract-Guided Scheduling for LLM-Assisted Natural Language-Driven Edge AI Services
arXiv:2606.09537v2 Announce Type: replace Abstract: Edge user/service scheduling has become a cornerstone of distributed AI systems, determining where and how AI services are executed under limited communication and computing resources. Existing edge scheduling frameworks usually assume that service requirements are given as numerical constraints, such as latency bounds or energy budgets. In practice, users often express service expectations through ambiguous and context-dependent natural language, creating a gap between user intent and scheduling decisions. To bridge this semantic-to-optimization gap, we propose semantic contract-guided edge potential scheduling (STEPS), a natural language-driven scheduling framework that introduces semantic contracts as executable interfaces between user-side semantics and edge-side decision making. In STEPS, a large language model (LLM)-assisted parser interprets natural language requests and extracts semantic service requirements with confidence scores, which are converted into service requirements and semantic uncertainty. Based on this information, STEPS formulates edge scheduling as a contract-guided potential game that jointly determines execution-node selection, computing-resource provisioning, and bandwidth allocation. STEPS further uses feedback signals to support adaptive scheduling under evolving service and network conditions. We characterize the exact potential game structure, establish the existence of a pure-strategy Nash equilibrium, and prove convergence and stability properties of the scheduling and adaptation processes. Extensive experiments show that STEPS improves semantic contract fulfillment, reduces contract-guided service loss, and maintains robust adaptation under ambiguous natural language requests in non-stationary networked AI environments.
arXiv:2606.15431v1 Announce Type: new Abstract: This paper presents an implemented CARLA-VISSIM co-simulation framework for an urban corridor comprising approximately fifteen connected intersections centered on Martin Luther King Jr. Boulevard in Chattanooga, Tennessee. The system integrates CARLA 0.10.0 Unreal Engine 5 with PTV VISSIM 2026 through a bidirectional, step-synchronized interface that couples VISSIM's microscopic vehicle, pedestrian, and signal-controller logic with CARLA's high-fidelity 3D rendering. A LiDAR-derived elevation model and RoadRunner-based High Definition (HD) map provide terrain-accurate road geometry deployed consistently across both simulators. The framework incorporates explicit actor ownership, mirrored lifecycle management, coordinate reconciliation, and a latest-state-per-actor update policy, enabling stable interaction between VISSIM-controlled traffic and a CARLA-controlled ego vehicle. A corridor-scale case study demonstrates consistent traffic-signal mirroring, synchronized vehicle-pedestrian interactions, and stable mixed-authority operation under peak loads of approximately 100 vehicles and 100 pedestrians. The deployment captures the interaction of the five signalized intersections along MLK Street and their connecting upstream and downstream intersections, revealing synchronization challenges unique to multi-intersection corridors. Results indicate that this MLK-centered corridor provides an effective testbed for verifying cross-simulator consistency and that the proposed architecture supports reliable, perception-ready co-simulation for corridor-level traffic studies.
arXiv:2603.07920v2 Announce Type: replace Abstract: All-weather autonomy is critical for autonomous driving, which necessitates reliable localization across diverse scenarios. While LiDAR place recognition is widely deployed for this task, its performance degrades in adverse weather. Conversely, radar-based methods, though weather-resilient, are hindered by the general unavailability of radar maps. To bridge this gap, radar-to-LiDAR place recognition, which localizes radar scans within existing LiDAR maps, has garnered increasing interest. However, extracting discriminative and generalizable features shared between modalities remains challenging, compounded by the scarcity of large-scale paired training data and the signal heterogeneity across radar types. In this work, we propose RLPR, a robust radar-to-LiDAR place recognition framework compatible with single-chip, scanning, and 4D radars. We first design a dual-stream network to extract structural features that abstract away from sensor-specific signal properties (e.g., Doppler or RCS). Subsequently, motivated by our task-specific asymmetry observation between radar and LiDAR, we introduce a two-stage asymmetric cross-modal alignment (TACMA) strategy, which leverages the pre-trained radar branch as a discriminative anchor to guide the alignment process. Experiments on four datasets demonstrate that RLPR achieves state-of-the-art recognition accuracy with strong zero-shot generalization capabilities.
arXiv:2603.08999v3 Announce Type: replace Abstract: Large language models (LLMs) can achieve strong reasoning performance through chain-of-thought (CoT) reasoning, yet they often generate unnecessarily long reasoning paths that incur high inference cost. Self-consistency-based approaches push accuracy higher still, but they require sampling and aggregating multiple reasoning trajectories, leading to substantial computational overhead. In this paper, we introduce a confidence-aware selective sampling framework that, at inference time, analyzes a single reasoning trajectory to adaptively determine whether to rely on that trajectory alone or trigger multi-path sampling. The framework uses trajectory-level numeric features and sentence-level linguistic features extracted from reasoning states to guide selective multi-path reasoning. We train it on MedQA and evaluate it in-domain on MedQA and under calibration-only transfer on MathQA, MedMCQA, and MMLU, without further fine-tuning. Experimental results show that the proposed framework maintains comparable performance to full and efficient multi-path reasoning baselines, with accuracy changes of $-0.41 \pm 0.58$ and $-0.31 \pm 0.58$ percentage points, respectively, while reducing token usage by $71.7 \pm 5.0%$ and $36.6 \pm 9.1%$. These findings demonstrate that reasoning trajectories contain rich signals for uncertainty estimation, enabling a simple, transferable mechanism to balance accuracy and efficiency in LLM reasoning.
Automated Digital Twin Construction for Highway Scenarios Using LiDAR Point Clouds and OpenStreetMap
arXiv:2606.16570v1 Announce Type: new Abstract: Accurate road environment modeling is fundamental to the simulation and validation of automated driving systems. However, constructing road maps in standardized formats such as ASAM OpenDRIVE from real-world sensor data remains a time-consuming and costly process. Mobile mapping LiDAR captures accurate lane-level geometry but is confined to the driven corridor, while OpenStreetMap (OSM) provides broad road network topology but lacks geometric precision at the lane level. To address this, an automated workflow is proposed to fuse LiDAR point clouds with OSM data to generate georeferenced ASAM OpenDRIVE maps of highway environments, requiring minimal manual intervention. The pipeline reconstructs mainline roads from LiDAR-derived measurements and infers ramp geometry and topology from the OSM road graph, enabling complete highway interchange modeling without full sensor coverage. Experiments demonstrate a mean lateral RMSE of 0.740 m, and the generated maps are directly usable in mainstream simulation platforms including IPG CarMaker and Esmini. These results validate the effectiveness of combining measurement-derived geometry with map-derived topology for automated OpenDRIVE digital twin generation. The project code is available at https://github.com/ftgTUGraz/opendrive-digital-twin-generator
arXiv:2602.22422v2 Announce Type: replace Abstract: Smooth-basis models such as Chebyshev polynomial regressors and radial basis function (RBF) networks are well established in numerical analysis. Their continuously differentiable prediction surfaces suit surrogate optimisation, sensitivity analysis, and other settings where the response varies gradually with inputs. Despite these properties, smooth models seldom appear in tabular regression, where tree ensembles dominate. We ask whether they can compete, benchmarking models across 55 regression datasets organised by application domain. We develop an anisotropic RBF network with data-driven centre placement and gradient-based width optimisation, a ridge-regularised Chebyshev polynomial regressor, and a smooth-tree hybrid (Chebyshev model tree); all three are released as scikit-learn-compatible packages. We benchmark these against tree ensembles, a pre-trained transformer, and standard baselines, evaluating accuracy alongside generalisation behaviour. The transformer ranks first on accuracy across a majority of datasets, but its GPU dependence, inference latency, and dataset-size limits constrain deployment in the CPU-based settings common across applied science and industry. Among CPU-viable models, smooth models and tree ensembles are statistically tied on accuracy, but the former tend to exhibit tighter generalisation gaps. We recommend routinely including smooth-basis models in the candidate pool, particularly when downstream use benefits from tighter generalisation and gradually varying predictions.
arXiv:2603.10574v2 Announce Type: replace Abstract: The crustal architecture of southern Ghana remains inadequately resolved despite its tectonic significance and resource potential. Existing geological and geophysical studies provide only broad constraints on crustal composition, lacking the resolution to accurately define sediment-basement interfaces or intra-crustal stratigraphy. To address these limitations, we employ single-station ambient noise autocorrelation (SSANA) on continuous waveform data from the Ghana Digital Seismic Network (GHDSN). We extract P-wave reflectivity responses using a processing sequence that involves data pre-processing, Phase Cross-Correlation (PCC) for robust noise correlation, and phase-weighted stacking (PWS) of the derived autocorrelograms. This procedure yields a two-way travel-time (TWT) function representing the zero-offset P-wave reflection response beneath each station, enabling high-resolution imaging of the stratified crustal column. To facilitate depth conversion, we develop an enhanced one-dimensional crustal velocity model for the region. Using a compiled dataset of local earthquake P- and S-wave arrival times from the GHDSN and an additional station in Cote d'Ivoire, we perform a joint inversion via a grid-search algorithm to derive a regional 1D velocity structure. Our results provide new constraints on the depth and configuration of the Paleozoic basement beneath the Voltaian Basin, demonstrating the efficacy of ambient noise autocorrelation for crustal imaging in sparsely instrumented regions. We also present an updated seismicity catalog, relocated using the new velocity model, and analyze the spatial clustering of seismicity in southern Ghana. This study highlights the utility of passive seismic methods for elucidating crustal structure and evaluating resources in intraplate West Africa and analogous Precambrian terrains.
arXiv:2606.11751v2 Announce Type: replace Abstract: Multi-turn image editing is essential for iterative design, yet current models often struggle with identity drift and error accumulation over successive steps. While existing research leverages video priors for consistency, their reliance on bidirectional attention is fundamentally misaligned with the causal, sequential nature of interactive editing. In this paper, we propose AnchorEdit, the first autoregressive (AR) diffusion-based framework designed specifically for high-resolution, long-term multi-turn editing. AnchorEdit bridges the gap between video priors and causal inference through a three-stage training curriculum: identity-preserving sing-turn pretraining, causal AR forcing fine-tuning with a novel self-rollout strategy to mitigate exposure bias, and consistency distillation for efficient 4-step generation. During inference, we introduce a memory mechanism to anchor the initial subject identity and ensure stable extrapolation across extended editing trajectories. To evaluate performance, we provide a new high-resolution multi-turn editing benchmark designed to stress-test long-horizon stability. Extensive experiments demonstrate that AnchorEdit achieves state-of-the-art results, maintaining exceptional subject fidelity and instruction following even over 10+ interaction rounds.
arXiv:2606.09322v2 Announce Type: replace Abstract: Compressibility and wall heat transfer change the inner scaling of wall turbulence through the mean density and viscosity fields. Most semi-local transformations are applied after a wall-normal profile has been selected. Here the transformed coordinate and transformed velocity are defined by wall-anchored field equations before any profile is extracted. Wall density and friction velocity enter as boundary data for auxiliary field equations; the resulting reference density and friction velocity, together with the local density and viscosity, set the viscous scaling in the wall layer. The local logarithmic density contrast is introduced as a bounded correction to the coordinate-stretching factor. For weak density variation the corrected stretch follows the local inverse kinematic-viscosity variation, while the bounded form limits the influence of finite density contrasts. The transformed coordinate $Y^+$ and transformed velocity $\bm{U}^+$ are obtained from field equations using the corrected stretching and the mean viscous shear. In the constant-property limit the density contrast vanishes, $Y^+$ reduces to the ordinary wall coordinate and each transformed velocity component reduces to the conventional mean velocity component in wall units. For channel flows, the field equations are solved with wall boundary data and the extracted profiles retain their own wall origins. A cooled shock/boundary-layer interaction examines the response when the wall density and friction velocity vary in the streamwise direction. Across the cooled high-speed boundary layers considered here, the bounded density correction narrows the inner- and buffer-layer profile spread...
arXiv:2606.09365v3 Announce Type: replace Abstract: Medical agent systems are increasingly expected to support interactive clinical decision making rather than only static question answering. In such settings, effective agents must reuse prior experience across evolving cases, yet existing memory mechanisms often retain raw historical traces that are redundant, noisy, and difficult to govern. More importantly, they rarely distinguish which memories are truly useful for future reasoning. This limits their ability to accumulate compact and reliable experience for long-horizon clinical reasoning. To close this gap, we propose SkeMex, a post-deployment self-evolution framework that improves medical agents through a skill-based memory without updating model weights. SkeMex distills informative interaction trajectories into structured skills that encode reusable procedural knowledge, and organizes them into a multi-branch repository spanning general, task-specific, and action-level experience. To determine which memories should be reused and retained, SkeMex estimates context-dependent utility from environment feedback and uses it to guide value-aware retrieval and repository governance. A closed-loop ``Read--Write--Assess--Govern" lifecycle further supports continual evolution by writing new skills, updating utilities, promoting useful memories, and removing harmful entries. Experiments across diverse clinical tasks show that SkeMex consistently outperforms representative memory-based agents in both offline and online settings. It also generalizes across model backbones and supports transferable skill memory. All data and code will be released publicly.
arXiv:2606.12707v2 Announce Type: replace Abstract: We study a two-node stochastic resource system operating over a finite horizon. Each node experiences uncertain supply and demand and is equipped with finite storage. The objective is to ensure that resource levels remain within prescribed limits with high probability. To this end, we formulate a chance-constrained capacity-design problem in which resources can be exchanged through a capacity-limited transport link. We characterize the minimum storage required at each node, derive the optimal transport policy, and quantify the trade-off between storage and transport capacities. Our results show the existence of a critical transport-capacity threshold that enables full risk pooling between the nodes. Moreover, this threshold decreases with the operating horizon, implying that full-pooling performance can be achieved with progressively smaller transport capacity over longer horizons.
arXiv:2603.22766v2 Announce Type: replace Abstract: As AI systems increasingly mediate negotiations, understanding how the number of negotiated issues impacts human performance is crucial for maintaining human agency. We designed a human-AI negotiation case study in a realistic property rental scenario, varying the number of negotiated issues; empirical findings show that without support, performance stays stable up to three issues but declines as additional issues increase cognitive load. To address this, we introduce a novel uncertainty-based visualization driven by Bayesian estimation of agreement probability. It shows how the space of mutually acceptable agreements narrows as negotiation progresses, helping users identify promising options. In a within-subjects experiment (N=32), it improved human outcomes and efficiency, preserved human control, and avoided redistributing value. Our findings surface practical limits on the complexity people can manage in human-AI negotiation, advance theory on human performance in complex negotiations, and offer validated design guidance for interactive systems.
arXiv:2603.22376v2 Announce Type: replace Abstract: We present an AI Co-Scientist framework that closes the research loop for the production search-ranking system of a large online travel platform -- pairing LLM agents with direct cloud-compute access so that idea generation, code implementation, GPU experimentation, and result analysis iterate end-to-end with a human scientist in the loop. The framework uses a hybrid agent architecture: single-LLM agents handle routine work, while multi-LLM consensus (GPT-5.2, Gemini Pro 3, Claude Opus 4.5) is invoked for higher-stakes decisions. On the production ranking task, a human-designed transformer baseline (V2) yielded $+0.118\%$ over a pre-transformer baseline (V1); the AI Co-Scientist's automated loop on top of V2 contributed an additional $+0.083\%$, for a combined $+0.201\%$ offline gain delivered in roughly one extra week of wall-clock time (single-run numbers; statistical limits discussed in the paper). The most useful AI proposals -- unified long-sequence layouts, slot-type embeddings, and multi-phase learning-rate schedules -- are standard practice in NLP and Vision but were absent from our production stack, suggesting that LLM agents can serve as cross-disciplinary connectors for ranking teams. We also report deployment context, negative results, and lessons learned.
arXiv:2606.16190v1 Announce Type: new Abstract: Embedded devices from wildlife monitoring stations to clinical wearables require local AI inference due to latency, communication, or privacy constraints. Optimizing models for heterogeneous microcontrollers (MCUs) requires simultaneously satisfying hard physical constraints on memory, power, and temperature while preserving accuracy, a multidimensional optimization that is today performed manually by experts. We ask whether an LLM agent can autonomously navigate this complex, multi-turn pipeline guided by real hardware feedback, and introduce a hardware-in-the-loop agent arena in which the agent iteratively refines both model and firmware -- compiling, flashing, and measuring on real hardware -- to enable closed-loop optimization. Frontier models, including Claude Opus 4.7 and Gemini 3.1 Pro, fail entirely without hardware feedback (0% deployment success), whereas our hardware-in-the-loop formulation achieves the first successful deployment within three iterations and can surpass human expert results within seven. This agentic co-optimization achieves 250x compression for vision models with <3.3% accuracy loss and 400x for audio with <6% Feature Error Rate loss, enabling battery-free operation on a commercial MCU via solar harvesting. We demonstrate practical impact in two real-world systems: an elk-detection camera trap (96.7% accuracy) and a phonetic-transcription wearable (8.44% FER) for child development research.
arXiv:2606.15455v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) has become a key approach for enhancing the reasoning abilities of large language models. However, RLVR often suffers from \emph{diversity collapse}: Pass@$1$ improves while high-$k$ Pass@$k$ degrades, which is viewed as a narrowing of the model's reasoning boundary. We formalize this diversity collapse through the lens of \emph{overtraining}: once a problem's contribution to the reference metric has effectively saturated, further updates no longer expand what the model can solve but still concentrate probability mass on the trajectories favored by on-policy sampling. Under a standard setup with few rollouts per problem, even a single observed success places a problem in a nearly saturated regime for high-$k$ Pass@$k$, so most updates in standard RLVR are overtraining from the boundary perspective. This perspective also suggests a reading of whether RLVR can expand the model's reasoning abilities beyond the base model: since RLVR is structurally biased against high-$k$ Pass@$k$, its aggregate decline does not by itself mean that no new reasoning gains occurred. Interventionally, restricting updates to problems with zero observed success lifts Pass@$256$ above the base model on difficult benchmarks; observationally, a non-trivial fraction of initially unsolvable problems become solvable during standard RLVR training. Building on these findings, we propose \emph{Bayesian Boundary Gating} (BBG), which redirects optimization away from overtraining by estimating each problem's marginal contribution to the reasoning boundary. Across multiple reasoning benchmarks, BBG improves average Pass@$k$ across a wide range of $k$.
arXiv:2606.15457v1 Announce Type: new Abstract: 3D FLAIR MRI is widely recommended as one of the standard MRI sequences for brain imaging in multiple sclerosis (MS), but publicly available MS datasets remain relatively small and vary across scanners, acquisition protocols, and lesion patterns. This scarcity and variability hinder the development of robust neuroimaging machine learning models and are particularly challenging for generative models that aim to synthesize images while preserving small, sparse lesions. We propose Lesion-DDPM, a 3D conditional diffusion framework for lesion-aware FLAIR synthesis that incorporates multi-level anatomical mask injection together with a lesion-weighted reconstruction loss to emphasize lesion voxels while maintaining global brain structure. Using a curated subset of the MSLesSeg dataset, we compare Lesion-DDPM with representative state-of-the-art GAN- and diffusion-based models, assessing both image-generation metrics and downstream 3D U-Net segmentation. In our experiments, Lesion-DDPM achieved the lowest lesion-region reconstruction error among all methods. In a downstream 3D U-Net lesion segmentation task, a model trained only on Lesion-DDPM-generated scans and evaluated on real MRIs reached a Dice score of 0.616 compared with 0.569 for the best competing synthetic dataset. When Lesion-DDPM images were added to the real training set, the Dice score further increased to 0.685.
arXiv:2603.01371v2 Announce Type: replace Abstract: Precise spatial fidelity in Image-to-3D multi-instance generation is critical for downstream real-world applications. Recent work attempts to address this by fine-tuning pre-trained Image-to-3D (I23D) models on multi-instance datasets, which incurs substantial training overhead and struggles to guarantee spatial fidelity. In fact, we observe that pre-trained I23D models already possess meaningful spatial priors, which remain underutilized as evidenced by instance entanglement issues. Motivated by this, we propose TIMI, a novel Training-free framework for Image-to-3D Multi-Instance generation that achieves high spatial fidelity. Specifically, we first introduce an Instance-aware Separation Guidance (ISG) module, which facilitates instance disentanglement during the early denoising stage. Next, to stabilize the guidance introduced by ISG, we devise a Spatial-stabilized Geometry-adaptive Update (SGU) module that promotes the preservation of the geometric characteristics of instances while maintaining their relative relationships. Extensive experiments demonstrate that our method yields better performance in terms of both global layout and distinct local instances compared to existing multi-instance methods, without requiring additional training and with faster inference speed.
arXiv:2603.24819v2 Announce Type: replace Abstract: We propose Weak and Entropy PINNs (WE-PINNs) for the approximation of entropy solutions to nonlinear hyperbolic conservation laws. Standard physics-informed neural networks enforce governing equations in strong differential form, an approach that becomes structurally inconsistent in the presence of discontinuities due to the divergence of strong-form residuals near shocks. The proposed method replaces pointwise residual minimization with a space--time weak formulation derived from the divergence theorem. Conservation is enforced through boundary flux integrals over dynamically sampled space--time control volumes, yielding a mesh-free control-volume framework that remains well-defined for discontinuous solutions. Entropy admissibility is incorporated in integral form to ensure uniqueness and physical consistency of the weak solution. The resulting loss functional combines space--time flux balance and entropy inequalities, without resorting to dual-norm saddle-point formulations, auxiliary potential networks, or fixed discretization meshes. This makes the proposed method remarkably easy to implement, requiring only a simple standard neural network architecture. We establish a rigorous convergence analysis linking the network's loss function to the $L^1$ error towards the entropy solution, providing the first explicit $L^1$ convergence rate for a mesh-free control-volume PINN formulation via the Bouchut-Perthame framework for scalar conservation laws. Numerical experiments on the Burgers equation, the shallow water equations, and the compressible Euler equations demonstrate accurate shock resolution and robust performance in both smooth and shock-dominated regimes.
arXiv:2606.13127v2 Announce Type: replace Abstract: Multi-camera tracking with overlapping fields of view typically relies on centralized fusion, which creates computational bottlenecks that prevent deployment at scale. We present MV3DT, a fully distributed framework for real-time multi-view 3D tracking that achieves accurate identity propagation and occlusion recovery through peer-to-peer coordination, eliminating the need for central aggregation. Each camera node executes a lightweight modular pipeline comprising monocular 3D perception, distributed multi-view association, and collaborative fusion via lightweight messaging. MV3DT achieves 96.5% IDF1, 93.1% MOTA, and 94.6% MOTP on WILDTRACK, competitive with state-of-the-art centralized methods, and unprecedented 41.7% IDF1 and 50.9% MOTA on SCOUT while demonstrating superior scalability: sustaining 30 FPS on 100 cameras with <10ms inter-camera latency and only 2.2% communication overhead. MV3DT operates in a zero-shot regime given camera calibrations, requiring no scene-specific learning and making it directly deployable in new environments. These results establish MV3DT as a practical solution for real-time multi-view tracking in large-scale overlapping camera networks.