arXiv:2603.02149v2 Announce Type: replace Abstract: Volume denoising is a foundational problem in computational imaging, as many 3D imaging inverse problems face high levels of measurement noise. Inspired by the strong 2D image denoising properties of Field of Junctions (ICCV 2021), we propose a novel, fully volumetric 3D Field of Junctions (3D FoJ) representation that optimizes a junction of 3D wedges that best explain each 3D patch of a full volume, while encouraging consistency between overlapping patches. In addition to direct volume denoising, we leverage our 3D FoJ representation as a structural prior that: (i) requires no training data, and thus precludes the risk of hallucination, (ii) preserves and enhances sharp edge and corner structures in 3D, even under low signal to noise ratio (SNR), and (iii) can be used as a drop-in denoising representation via projected or proximal gradient descent for any volumetric inverse problem with low SNR. We demonstrate successful volume reconstruction and denoising with 3D FoJ across three diverse 3D imaging tasks with low-SNR measurements: low-dose X-ray computed tomography (CT), cryogenic electron tomography (cryo-ET), and denoising point clouds such as those from lidar in adverse weather. Across these challenging low-SNR volumetric imaging problems, 3D FoJ outperforms the evaluated classical denoisers, untrained neural denoisers, and denoisers trained only on noisy examples. Code is available at https://github.com/voilalab/3D-Field-of-Junctions.
Science Journals
arXiv:2512.05758v5 Announce Type: replace Abstract: Ferroelectric ordering in polar liquids has been observed in numerical simulations and liquid-crystal experiments. Within mean-field framework, this behaviour remains associated with sample-shape dependent, surface contribution to the free energy, which does not vanish in the thermodynamic limit due to the long-range nature of dipolar interaction. Yet, numerical simulations performed under conducting periodic boundary conditions, for which the surface contribution vanishes, still exhibit ferroelectric order, pointing to an intrinsic bulk origin of the transition. Moving beyond the mean-field approximation, Kirkwood seminal study of the dielectric properties of polar liquids emphasized the role of hindered dipolar rotation in shaping the corresponding pair correlations. In Kirkwood analysis, hindered rotation stems from the mean force between nearest-neighbor dipoles, placing the focus on local structure. Introducing a different perspective while retaining the central role of hindered dipolar rotation in the onset of ferroelectricity, the present study establishes, as an original finding, that annealed averaging of dipolar interaction over positional disorder generates hindered dipolar rotation favoring dipole alignment, and able to drive a ferroelectric phase transition. As a result, unlike approaches centered on local structure, ferroelectricity emerges not in spite of the liquid nature, but because of it. This ferroelectric phase transition is intrinsic to the bulk. Annealed averaging over positional disorder generates an effective dipolar interaction that is shorter-ranged than the bare potential, analogous to the Keesom interaction where screening arises from annealed dipolar disorder. Derived within classical density functional theory, these findings are exact in the infinite-dimensional limit and remain valid within the optimized cluster expansion for dimensions greater than two.
arXiv:2603.02172v2 Announce Type: replace Abstract: We introduce TerraDiT, a diffusion transformer designed for text-to-satellite image generation with point-based control. Existing controlled satellite image generative models often require pixel-level maps that are time-consuming to acquire, yet semantically limited. To address this limitation, we introduce a novel point-based conditioning framework that controls the generation process through the spatial location of the points and the textual description associated with each point, providing semantically rich control signals. This approach enables flexible, annotation-friendly, and computationally simple inference for satellite image generation. To this end, we introduce an adaptive local attention mechanism that effectively regularizes the attention scores based on the input point queries. We systematically evaluate various domain-specific design choices for training TerraDiT, including the selection of satellite image representation for alignment and geolocation representation for conditioning. Our experiments demonstrate that TerraDiT achieves impressive generation performance, surpassing the state-of-the-art remote sensing generative models. Our models, dataset, and code are available at https://github.com/mvrl/TerraDiT.
arXiv:2603.02491v3 Announce Type: replace Abstract: As artificial agents become increasingly capable, what internal structure is necessary for an agent to act competently under uncertainty? Classical results show that optimal control can be implemented using belief states or world models, but not that such representations are required. We prove quantitative "selection theorems" showing that strong task performance (low average-case regret) forces world models, belief-like memory and -- under task mixtures -- persistent regime-tracking variables resembling functional primitives of emotion, along with informational modularity under block-structured tasks. Our results cover stochastic policies, partial observability, and evaluation under task distributions, without assuming optimality, determinism, or access to an explicit model. Technically, we reduce predictive modeling to binary "betting" decisions and show that regret bounds limit probability mass on suboptimal bets, enforcing the predictive distinctions needed to separate high-margin outcomes. In fully observed settings, this yields approximate recovery of the interventional transition kernel; under partial observability, it implies necessity of predictive state and belief-like memory, addressing an open question in prior world-model recovery work.
arXiv:2603.03143v2 Announce Type: replace Abstract: Leveraging the priors of 2D diffusion models for 3D editing has emerged as a promising paradigm. However, multi-view consistency remains challenging in edited results, and the extreme scarcity of paired 3D-consistent editing data makes supervised fine-tuning (SFT) impractical, despite its effectiveness for editing tasks. In this paper, we observe that, while generating multi-view consistent 3D content is highly challenging, verifying 3D consistency is tractable, naturally positioning reinforcement learning (RL) as a feasible solution. Motivated by this, we propose RL3DEdit, a single-pass framework driven by RL optimization with novel rewards derived from the 3D foundation model, VGGT. Specifically, we leverage VGGT's robust priors learned from massive real-world data, feed the edited images into it, and utilize the output confidence maps and pose estimation errors as reward signals, effectively anchoring the 2D editing priors onto a 3D-consistent manifold via RL. Extensive experiments demonstrate that RL3DEdit achieves stable multi-view consistency and outperforms state-of-the-art methods in editing quality with high efficiency. To promote the development of 3D editing, we will release the code and model.
arXiv:2603.03305v2 Announce Type: replace Abstract: Large language models (LLMs) are increasingly used to generate executable outputs, JSON objects, and API calls, where a single syntax error can make the output unusable. Constrained decoding enforces validity token-by-token via masking and renormalization, but it can distort generation when the model assigns low probability mass to valid continuations, pushing decoding toward locally valid yet semantically incorrect trajectories. We propose \emph{Draft-Conditioned Constrained Decoding (DCCD)}, a simple two-step, training-free inference procedure that decouples semantic planning from structural enforcement: an unconstrained draft is generated first, and constrained decoding is then applied, conditioned on this draft, to guarantee validity. We analyze DCCD through a KL-projection view, showing that draft conditioning increases feasible mass and reduces the cumulative "projection tax" induced by hard constraints, with an optional best-of-$K$ draft selection. Across structured reasoning benchmarks, DCCD improves strict structured accuracy by up to +24 percentage points over standard constrained decoding (e.g., 15.2\% to 39.0\% on GSM8K with a 1B model), and enables smaller model pairs to match or exceed much larger constrained baselines, yielding substantial gains in parameter efficiency.
arXiv:2603.03458v2 Announce Type: replace Abstract: The El Ni\~no-Southern Oscillation (ENSO) is a fluctuation in sea surface temperature and pressure across the equatorial Pacific Ocean with a period of 2-7 years. As the largest mode of interannual variability on Earth, ENSO shapes global weather and climate patterns ranging from monsoons in southern Asia to hurricanes in the Atlantic and droughts in South America. Predicting and understanding ENSO's response to greenhouse warming is essential for mitigating the impacts of climate change, yet model ensemble projections are expensive to generate across emission scenarios and remain incompletely understood. Here, we use a hierarchy of models to explain the transient rise and subsequent fall of ENSO strength under greenhouse warming, then develop an efficient and accurate method for predicting ENSO variability in any emissions scenario. Beginning with an East Pacific energy budget, we quantitatively show how enhanced upper-ocean stratification strengthens ENSO and how a slowing Walker circulation and stronger surface flux damping eventually weaken it. This leads to a linear model that predicts the evolution of ENSO variability from only East Pacific temperature and stratification. We further show that subsurface warming, and therefore stratification, is connected to surface warming with a lag, enabling us to create a lag-linear model that explains $\sim$90\% of simulated changes in ENSO variability from only global mean surface temperature and its history. Once calibrated, this efficient predictor can project ENSO strength without running a full climate model, and allows an analytic solution for the timing and magnitude of peak ENSO variability. We find that the ratio of an ocean subsurface adjustment timescale to the warming timescale strongly alters peak ENSO amplitude, meaning that faster emissions lead to larger ENSO variability even with identical total emissions.
arXiv:2603.05002v3 Announce Type: replace Abstract: The Edge of Stability (EoS) is a phenomenon where the sharpness (largest eigenvalue) of the Hessian approaches and then hovers near the stability threshold $2/\eta$ during gradient descent (GD) with step size $\eta$. Despite (apparently) violating classical smoothness assumptions, EoS has been widely observed in deep learning, but its theoretical foundations remain incomplete. We provide an interpretation of EoS through the lens of Directional Smoothness [Mishkin et al., 2024]. This interpretation naturally extends to non-Euclidean norms, which we use to define generalized sharpness under an arbitrary norm. Our generalized sharpness measure includes previously studied vanilla GD and preconditioned GD as special cases, as well as methods for which EoS has not been studied, such as $\ell_{\infty}$-descent, Block CD, Spectral GD, and their normalized versions. Through experiments on neural networks, we show that non-Euclidean GD with our generalized sharpness also exhibits progressive sharpening followed by oscillations around or above the threshold $2/\eta$. Practically, our framework provides a geometry-aware spectral diagnostic that can be applied across a broad class of non-Euclidean gradient methods.
arXiv:2512.16413v2 Announce Type: replace Abstract: Current token-sequence-based Large Language Models (LLMs) struggle to directly process 3D Boundary Representation (B-rep) models that contain complex geometric and topological information. To this end, we propose BrepLLM, the first multimodal framework that enables LLMs to directly parse and reason over raw B-rep data. BrepLLM adopts a two-stage training pipeline: cross-modal alignment pre-training and two-stage LLM fine-tuning. In the first stage, we design an adaptive UV sampling strategy to convert B-reps into graph representations that integrate geometric and topological information. Subsequently, we construct a hierarchical BrepEncoder to extract features from geometric elements (faces and edges) and topology, generating a global token and a sequence of node tokens. Then, via contrastive learning, we conduct an initial alignment between this global token and the text embeddings of a frozen CLIP text encoder (ViT-L/14). In the second stage, we integrate the pre-trained BrepEncoder into the LLM and employ a two-stage progressive strategy to align the sequence of node tokens: (1) training an MLP-based semantic mapping network that utilizes the prior knowledge of a 2D-VLM to align the B-rep representation to the 2D visual semantic space; (2) utilizing LoRA for parameter-efficient fine-tuning of the Q-Former and the LLM backbone network to achieve the final 3D-language generation capability. Furthermore, we construct the Brep2Text dataset, which contains 269,444 B-rep and text question-answer pairs. Experiments demonstrate that BrepLLM achieves SOTA performance on 3D object classification and captioning tasks. The project page is available at https://user-deng.github.io/BrepLLM/.
arXiv:2511.18639v2 Announce Type: replace Abstract: We propose a novel framework for developing, analyzing, and validating reductions between NP-complete problems. Powered by the SAT-based constraint solver URSA, our methodology introduces several distinct features that set it apart from other related approaches. The proposed workflow effectively bridges the crucial gap between informal, high-level reduction descriptions and formalized mathematical proofs. By supplementing rather than replacing human intuition, this interactive methodology serves as an aid for exploring relationships between NP-complete problems.
arXiv:2603.05377v3 Announce Type: replace Abstract: Open-world navigation requires robots to make decisions in complex everyday environments while adapting to flexible task requirements. Conventional navigation approaches often rely on dense 3D reconstruction and hand-crafted goal metrics, which limits their generalization across tasks and environments. Recent advances in vision-language navigation (VLN) and vision-language-action (VLA) models enable end-to-end policies conditioned on natural language, but typically require interactive training, large-scale data collection, or task-specific fine-tuning with a mobile agent. We formulate navigation as a sparse subgoal identification and reaching problem and observe that providing visual anchoring targets for high-level semantic priors enables highly efficient goal-conditioned navigation. Based on this insight, we select visual frontiers as semantic anchors and propose OpenFrontier, a navigation framework that requires no task-specific training or fine-tuning and seamlessly integrates diverse vision-language prior models. OpenFrontier enables efficient navigation with a lightweight system design, without dense 3D semantic mapping, task-specific policy training, or model fine-tuning. We evaluate OpenFrontier across multiple navigation benchmarks and demonstrate strong zero-shot performance, as well as effective real-world deployment on a mobile robot.
arXiv:2602.04641v3 Announce Type: replace Abstract: An All-Path Reachability predicate over an object set is a pair of a source set and a target set, which are subsets of the object set. APR predicates have been defined for Abstract Reduction Systems and then extended to Logically Constrained Term Rewrite Systems as pairs of constrained terms that represent sets of terms modeling configurations, states, etc. An APR predicate is partially valid w.r.t. a rewrite system if every finite maximal reduction sequence of the system starting from any element in the source set includes an element in the target set. Partial validity of APR predicates w.r.t. ARSs is defined by means of two inference rules, which can be considered a proof system to construct (possibly infinite) derivation trees for partial validity. On the other hand, a proof system for LCTRSs consists of four inference rules, leaving a gap between the inference rules for ARSs and LCTRSs. In this paper, we revisit the framework for APR analysis and adapt it to verification of not only safety but also liveness properties. To this end, we first reformulate an abstract framework for partial validity w.r.t. ARSs so that there is a one-to-one correspondence between the inference rules for partial validity w.r.t. ARSs and LCTRSs. Secondly, we show how to apply APR analysis to safety verification. Thirdly, to apply APR analysis to liveness verification, we introduce a novel stronger validity of APR predicates, called total validity, which requires not only finite but also infinite execution paths to reach target sets. Finally, for a partially valid APR predicate with a cyclic-proof tree, we show that the acyclicity of the proof graph obtained from the cyclic-proof tree is a necessary and sufficient condition for total validity. The condition implies that if there exists a cyclic-proof tree for an APR predicate, the proof graph of which is acyclic, then the APR predicate is totally valid.
arXiv:2603.05905v2 Announce Type: replace Abstract: Small object detection in unmanned aerial vehicle (UAV) imagery is challenging because high-altitude viewpoints produce severe scale variation, weak structural cues, and tight computational budgets. Existing lightweight detectors usually fuse multi-scale features after downsampling, where boundary and texture details have already been attenuated and heterogeneous feature streams may be spatially misaligned. To address these issues, we propose CollabOD, a collaborative detection framework that preserves structural details, aligns cross-path features before fusion, and keeps the detection head lightweight at inference time. CollabOD combines a Dual-Path Fusion Stem, a Dense Aggregation Block, a Bilateral Reweighting Module, and a Unified Detail-Aware Head to strengthen localization-oriented representation while limiting extra computation. On VisDrone, CollabOD obtains 52.4 AP50, 30.8 AP75, and 29.9 AP50:95 with 65.5 GFLOPs; on UAVDT it reaches 31.2 AP50 and 17.4 AP50:95; and on AI-TOD it reaches 45.4 AP50 and 20.0 AP50:95 at 137 FPS. The code is available at: https://github.com/Bai-Xuecheng/CollabOD.
arXiv:2602.09305v2 Announce Type: replace Abstract: Large Language Models (LLMs) demonstrate transformative potential, yet their reasoning remains inconsistent and unreliable. Reinforcement learning (RL)-based fine-tuning is a key mechanism for improvement, but its effectiveness is fundamentally governed by reward design. Despite its importance, the relationship between reward modeling and core LLM challenges--such as evaluation bias, hallucination, distribution shift, and efficient learning--remains poorly understood. This work argues that reward modeling is not merely an implementation detail but a central architect of reasoning alignment, shaping what models learn, how they generalize, and whether their outputs can be trusted. We introduce Reasoning-Aligned Reinforcement Learning (RARL), a reasoning-centric taxonomic perspective that organizes diverse reward paradigms for multi-step reasoning. Within this perspective, we present a taxonomy of reward mechanisms, analyze reward hacking as a pervasive failure mode, and examine how reward signals unify challenges ranging from inference-time scaling to hallucination mitigation. We further critically evaluate existing benchmarks, highlighting vulnerabilities such as data contamination and reward misalignment, and outline directions for more robust evaluation. By integrating fragmented research threads and clarifying the interplay between reward design and fundamental reasoning capabilities, this work provides a foundational roadmap for building reasoning models that are robust, verifiable, and trustworthy.
arXiv:2602.10233v2 Announce Type: replace Abstract: LLM-guided evolutionary computation, most notably AlphaEvolve, has been remarkably successful in discovering novel mathematical constructions by solving challenging optimization problems. The standard approach is to evolve a monolithic program that directly outputs a candidate solution. We present ImprovEvolve, an algorithmic alternative that drastically reduces cognitive load on the LLM. Instead of prompting the model for an end-to-end optimizer, we evolve a program with three specialized operators of initialization, local improvement, and perturbation. We then approach the optimum by iteratively applying local improvements and intensity-scheduled perturbations, effectively driving a basin-hopping search with LLM-evolved subroutines. For hexagon in hexagon packing, ImprovEvolve discovers new state-of-the-art packings of 11, 12, 15, and 16 hexagons, and additionally for 14, 17, and 23 hexagons after minimal expert tuning of the generated code. For the second autocorrelation inequality, the evolved and human-scaled program pushes the lower bound from 0.96102 to 0.96258. For spherical codes, the ImprovEvolve program lowers the best-known maximum cosine for the majority of 90 randomly chosen diverse state-of-the-art spherical codes, achieving relative improvements of up to 2.4%.
arXiv:2512.07287v3 Announce Type: replace Abstract: As intents unfold and environments change, multi-turn agents face continuously shifting decision contexts. Although reusing past experience is intuitively appealing, existing approaches remain limited: full trajectories are often too context-specific to transfer, while tool-level reuse ignores the surrounding context and environment. In this paper, we introduce a hybrid episodic-procedural memory strategy (H-EPM) that enables experience-induced self-evolution of multi-turn tool-use policies by adaptively reusing partially overlapping successful experiences during both inference and training. Inspired by human episodic-procedural integration, we construct a tool graph from accumulated trajectories, where recurring tool-to-tool dependencies capture procedural routines and each edge is augmented with compact episodic summaries of relevant context. At inference time, the agent dynamically balances episodic recall for contextual reasoning with procedural execution for routine steps. Beyond inference, H-EPM introduces a memory-guided reinforcement learning paradigm that directly addresses a core challenge in multi-turn agent reinforcement learning, namely ineffective exploration over long trajectories. By biasing exploration toward historically successful tool transitions, H-EPM learns a stronger policy that generalizes at inference time without relying on domain-specific experience collection. Experiments show that H-EPM consistently delivers substantial inference-time gains over strong baselines across multi-turn tool-use benchmarks, reaching improvements of up to fifty percent. It also improves reinforcement learning policy performance, achieving gains of up to forty percent on out-of-distribution tasks.
arXiv:2512.07854v2 Announce Type: replace Abstract: Traffic forecasting task is significant to modern urban management. Recently, there is growing attention on large-scale forecasting, as it better reflects the complexity of real-world traffic networks. However, existing models often exhibit quadratic computational complexity, making them impractical for large-scale real-world scenarios. In this paper, we propose a novel framework, Spatio-Temporal Hierarchical Mixer (HieraMix), which leverages an all-MLP architecture for efficient and effective large-scale traffic forecasting. HieraMix employs a hierarchical spatiotemporal mixing block to extract multi-resolution features through bottom-up aggregation and top-down propagation. Furthermore, an adaptive region mixer generates transformation matrices based on regional semantics, enabling our model to dynamically capture evolving spatiotemporal patterns for different regions. Extensive experiments conducted on four large-scale real-world datasets demonstrate that the proposed method not only achieves state-of-the-art performance but also exhibits competitive computational efficiency.
arXiv:2603.06168v3 Announce Type: replace Abstract: Semantic segmentation across visual modalities such as 3D point clouds and panoramic images remains a challenging task, primarily due to the scarcity of annotated data and the limited adaptability of fixed-label models. In this paper, we present JOPP-3D, an open-vocabulary segmentation framework that jointly leverages panoramic and point cloud data to enable language-driven scene understanding. We convert RGB-D panoramic images into their corresponding wide field-of-view tangential perspectives and 3D point clouds, then use these modalities to extract and align foundational vision-language features. This allows natural language querying to generate semantic masks on both input modalities. Experimental evaluation on the Stanford-2D-3D-s and ToF-360 datasets demonstrates the capability of JOPP-3D to produce coherent and semantically meaningful segmentations across panoramic and 3D domains. Our proposed method achieves a significant improvement compared to the SOTA in open and closed vocabulary 2D and 3D semantic segmentation.
arXiv:2603.08057v2 Announce Type: replace Abstract: Programming by demonstration (PbD) makes robot programming accessible to non-experts, but scaling it to real-world variability remains a challenge for current teaching frameworks, especially when a robot must select suitable task variants online from visual input. We present See & Switch, an interactive teaching-and-execution framework that represents tasks as graphs of skill parts connected by decision states, enabling conditional branching during replay. Its vision-based Switcher uses eye-in-hand images to select the appropriate successor skill part and detect novel situations that require new demonstrations. The framework supports recovery demonstrations during execution through kinesthetic teaching, joystick control, and hand gestures. We evaluate See & Switch on three dexterous manipulation tasks with 8 novice users, collecting approx. 900 real-robot execution rollouts. To isolate visual decision performance from timing errors during decision states, we evaluate the Switcher offline using user-gated decision state windows. In the evaluation within the decision state windows, the method achieves up to 90.6% branch-selection accuracy and detects anomalies with >90% accuracy in 47 of 79 decision states, demonstrating reliable switching based on visual input for conditional robot-skill programming. We provide all code and experiment data at http://imitrob.ciirc.cvut.cz/publications/seeandswitch.
arXiv:2602.12155v2 Announce Type: replace Abstract: Post-training of flow matching models-aligning the output distribution with a high-quality target-is mathematically equivalent to imitation learning. While Supervised Fine-Tuning mimics expert demonstrations effectively, it cannot correct policy drift in unseen states. Preference optimization methods address this but require costly preference pairs or reward modeling. We propose Flow Matching Adversarial Imitation Learning (FAIL), which minimizes policy-expert divergence through adversarial training without explicit rewards or pairwise comparisons. We derive two algorithms: FAIL-PD exploits differentiable ODE solvers for low-variance pathwise gradients, while FAIL-PG provides a black-box alternative for discrete or computationally constrained settings. Fine-tuning FLUX with only 13,000 demonstrations from Nano Banana pro, FAIL achieves competitive performance on prompt following and aesthetic benchmarks. Furthermore, the framework generalizes effectively to discrete image and video generation, and functions as a robust regularizer to mitigate reward hacking in reward-based optimization. Code and data are available at https://github.com/HansPolo113/FAIL.
arXiv:2602.12418v2 Announce Type: replace Abstract: Jailbreak attacks remain a persistent threat to large language model safety. We propose Context-Conditioned Delta Steering (CC-Delta), an SAE-based defense that identifies jailbreak-relevant sparse features by comparing token-level representations of the same harmful request with and without jailbreak context. Using paired harmful/jailbreak prompts, CC-Delta selects features via statistical testing and applies inference-time mean-shift steering in SAE latent space. Across four aligned instruction-tuned models and twelve jailbreak attacks, CC-Delta achieves comparable or better safety-utility tradeoffs than baseline defenses operating in dense latent space. In particular, our method clearly outperforms dense mean-shift steering on all four models, and particularly against out-of-distribution attacks, showing that steering in sparse SAE feature space offers advantages over steering in dense activation space for jailbreak mitigation. Our results suggest off-the-shelf SAEs trained for interpretability can be repurposed as practical jailbreak defenses without task-specific training.
arXiv:2603.08284v2 Announce Type: replace Abstract: This study introduces a systematic and optimised methodology for designing Linear Variable Differential Transformer (LVDT) sensors and Voice Coil (VC) actuators, tailored for high-precision applications such as gravitational wave detectors and particle accelerators. Unlike prior studies, which focus primarily on industrial-grade LVDT design frameworks or isolated parameter studies, this work addresses the specific challenges of achieving both enhanced sensor response and actuation force within strict geometric and thermal constraints. Using a custom-developed simulation pipeline based on Finite Element Method Magnetics (FEMM), we evaluate the influence of key design parameters such as coil dimensions, radial gaps, and coil wire diameter on performance metrics such as response and linearity. The novelty of this work lies in its systematic exploration of design trade-offs, such as maximising performance while minimising heat dissipation, and its applicability to high-precision environments. In this work, particular emphasis is placed on the combination of the LVDT and VC functionalities in one unified sensor-and-actuator system designed for gravitational wave detectors. In addition, the methodology and simulation results are validated with experimental measurements of an optimised design, demonstrating a 2.8-fold increase in LVDT response and a 2.5-fold increase in VC actuation force compared to the initial configuration while preserving LVDT linearity and VC force stability. This work represents a significant advance over existing methodologies by offering a structured, scalable design process.
arXiv:2511.15688v2 Announce Type: replace Abstract: We study a self-consistent approach to introduce higher-order effects in a branching process model of complex contagion on clustered networks. Branching processes operate over an infinite population such that they never circle back and interact with previously exposed parts of the system. This infinite, treelike, structure makes it tricky to account for complex contagion mechanisms such as group effects, peer pressure, or social reinforcement where multiple exposures interact in synergistic ways. Here we present a self-consistent solution that accounts for local group structure and global cycles where the process can feedback on itself. This allows us to distinguish multiple exposures that stem from a single transmission chain, from those occurring at the intersection of different transmission chains. We find that only the latter mechanism can give rise to a discontinuous phase transition in the size of global cascades, which is a defining feature of complex contagions. Group effects alone, without long cycles, produce standard continuous phase transitions.
arXiv:2511.16023v3 Announce Type: replace Abstract: We will be exploring a generalization of real time scheduling problem sometimes called the real time throughput maximization problem. Our input is a sequence of jobs specified by their release time, deadline and processing time. We assume that jobs are announced before or at their release time. At each time step, the algorithm must decide whether to schedule a job based on the information so far. The goal is to maximize the value of the sum of the processing times of jobs that finish before their deadline, this is often called real time throughput with proportional weights. We extend this problem by defining a notion of \(t\)-advance-notice, a measure of how far in advance each job is announced relative to their processing time. We show that there exists a class of algorithms \(\tau-\textsc{Persist}\) parametrized by some value \(\tau\in [1,\infty)\). If an input sequence has \(t\)-advance-notice, \(\tau-\textsc{Persist}\) is \(\frac{\tau - 1}{\tau^2 +\tau - 1}\)-competitive. In particular, we show that for any \(t \leq \frac{1}{2}\), there is an algorithm that achieves \(\frac{t-t^2}{1+t-t^2}\)-competitiveness and for any \(t \geq \frac{1}{2}\), there is an algorithm that achieves \(\frac{1}{5}\)-competitiveness. We also give an upper bound of any algorithm that relies on input sequences having \(t\)-advance-notice. We show that the competitive ratio of any algorithm can be at most \(\frac{t}{2t+1}\) against input sequences that have \(t\)-advance-notice. In particular, we show that regardless of how much advance-notice is given, no algorithm can reach \(\frac{1}{2}\)-competitiveness.
arXiv:2602.13562v2 Announce Type: replace Abstract: While reasoning models have achieved remarkable success in complex reasoning tasks, their increasing power necessitates stringent safety measures. For safety alignment, the core challenge lies in the inherent trade-off between safety and utility. However, prevailing alignment strategies typically construct CoT training data with explicit safety rules via context distillation. This approach inadvertently limits reasoning capabilities by creating a rigid association between rule memorization and refusal. To mitigate the safety-utility trade-off, we propose the Adaptive Safe Context Learning~(ASCL) framework to improve the reasoning given proper context. ASCL formulates safety alignment as a multi-turn tool-use process, empowering the model to autonomously decide when to consult safety rules and how to generate the ongoing reasoning. Furthermore, to counteract the preference for rule consultation during RL, we introduce Inverse Frequency Policy Optimization~(IFPO) to rebalance advantage estimates. By decoupling rule retrieval and subsequent reasoning, our method achieves higher overall performance compared to baselines. Our code is publicly available at https://github.com/ybwang119/ASCL.