History-dependent sampling can reduce long-run Monte Carlo variance by discouraging redundant revisits, but existing schemes typically encode history through empirical measure on finite state spaces, which is infeasible in high-dimensional discrete configuration spaces or ill-posed in continuous domains. We propose Score-Repellent Monte Carlo (SRMC) framework that summarizes trajectory history by a running average of score evaluations in $\mathbb{R}^d$, where $d$ is the dimension of the score and state representation. This history is converted into a surrogate target through an exponential score tilt, indexed with $α$ that represents the strength of repellence in controlling the magnitude of the history-based repulsion. The surrogate family is normalization-free in the standard MCMC sense, yielding a generic wrapper: at each iteration, any base kernel targeting $π$ can instead be run on the current surrogate $π_{θ_n}$ while the history is updated online. We analyze the coupled evolution of the history recursion and Monte Carlo estimators using stochastic approximation with controlled Markovian noise, establishing almost sure convergence and a joint central limit theorem. We further identify regimes in which the asymptotic covariance decreases as $α$ increases, with scaling $O(1/α)$, extending the near-zero-variance effect of finite-state history-dependent samplers to general state spaces with constant memory. Experiments on continuous targets and discrete energy-based models demonstrate improved estimator variance and mode coverage, while retaining $O(d)$ memory usage and modest per-iteration overhead.

Reinforcement learning for multi-step reasoning with large language models (LLMs) typically relies on sparse terminal rewards, which creates a poorly conditioned credit-assignment problem: the final feedback is propagated uniformly across all intermediate decisions. This leads to high gradient variance, unstable training, and many ineffective updates, ultimately limiting sustained model improvement. We propose a counterfactual-comparison framework for credit assignment. For each input, the framework samples multiple reasoning trajectories and treats their differences as implicit approximations to alternative decisions. This yields an implicit process-level advantage estimator that converts sparse terminal rewards into step-sensitive learning signals. Building on this framework, we introduce Implicit Behavior Policy Optimization (IBPO), which substantially improves training stability and the performance ceiling on mathematical and code-reasoning benchmarks. Our results point to a promising direction for unlocking the reasoning potential of LLMs.

Agentic search enables language models to solve knowledge-intensive tasks by adaptively acquiring external evidence over multiple steps. Reinforcement learning with verifiable rewards (RLVR) has emerged as a widely adopted training paradigm for search agents, yet outcome-only rewards are sparse and provide limited credit assignment for intermediate search actions. Existing process-reward methods therefore seek to densify supervision through proxy signals, external evaluators, or likelihood-based information gain. However, proxy rewards can deviate from the final outcome objective, while fixed evaluators can become stale as the search policy evolves, leading to unreliable process supervision. To address these challenges, we propose OASES, an Outcome-Aligned Search-Evaluation Supervision framework for agentic search. OASES derives outcome-aligned process rewards by evaluating how well each intermediate search state supports answering the original question. It further co-trains the search policy and the state evaluator on policy, allowing the evaluator to adapt to evolving search behavior and provide more reliable process rewards. Experiments on five multi-hop QA benchmarks show that OASES consistently outperforms strong RL baselines, with further analyses confirming the benefits of outcome-aligned process rewards and search-evaluation co-training.

Monocular cameras are attractive for robotic perception due to their low cost and ease of deployment, yet achieving reliable real-time spatial understanding from a single image stream remains challenging. While recent multi-task dense prediction models have improved per-pixel depth and semantic estimation, translating these advances into stable monocular mapping systems is still non-trivial. This paper presents M2H-MX, a real-time multi-task perception model for monocular spatial understanding. The model preserves multi-scale feature representations while introducing register-gated global context and controlled cross-task interaction in a lightweight decoder, enabling depth and semantic predictions to reinforce each other under strict latency constraints. Its outputs integrate directly into an unmodified monocular SLAM pipeline through a compact perception-to-mapping interface. We evaluate both dense prediction accuracy and in-the-loop system performance. On NYUDv2, M2H-MX-L achieves state-of-the-art results, improving semantic mIoU by 6.6% and reducing depth RMSE by 9.4% over representative multi-task baselines. When deployed in a real-time monocular mapping system on ScanNet, M2H-MX reduces average trajectory error by 60.7% compared to a strong monocular SLAM baseline while producing cleaner metric-semantic maps. These results demonstrate that modern multi-task dense prediction can be reliably deployed for real-time monocular spatial perception in robotic systems.

Reinforcement learning with verifiable rewards (RLVR) has emerged as an effective post-training paradigm for improving the reasoning capabilities of large language models. However, existing group-based RLVR methods often suffer from severe sample inefficiency. This inefficiency stems from reliance on point estimation of rewards from a small number of rollouts, leading to high estimation variance, variance collapse, and ineffective utilization of generated responses. In this work, we reformulate RLVR from a statistical estimation perspective by modeling rewards as samples drawn from a policy-induced distribution and casting advantage computation as the problem of estimating the reward distribution from finite data. Building on this view, we propose Discounted Beta-Bernoulli (DBB) reward estimation, which leverages historical reward statistics for the non-stationary distribution. Although biased, the resulting estimator exhibits reduced and stable variance, theoretically avoids estimated variance collapse, and achieves lower mean squared error than standard point estimation. Extensive experiments across six in-distribution and three out-of-distribution reasoning benchmarks demonstrate that GRPO with DBB consistently outperforms naive GRPO, achieving average Acc@8 improvements of 3.22/2.42 points in-distribution and 12.49/6.92 points out-of-distribution on the 1.7B and 8B models, respectively, without additional computational cost or memory usage.

A foundational principle in cognitive science holds that intelligent agents do not learn by storing experiences as isolated instances, but by forming abstract schemas that capture relational structure shared across situations. Even though this claim is well supported by behavioral and neuroimaging studies, its role as a computational training signal in language models remains underexplored. We target this gap in the setting of non-stationary language model training, asking does biasing learning toward structural abstraction reduce catastrophic interference and improve relational generalization as predicted by human results? To study this question, we introduce Abstraction-Augmented Training (AAT), a lightweight loss-level modification that jointly optimizes over concrete instances and their structural abstractions, and two benchmarks, the Relational Cycle Benchmark (RCB) and the Narrative Abstraction Benchmark (NAB). These resources operationalize core cognitive constructs: entity masking as a computational analog of relational alignment, and proverbs as vehicles for implicit abstract meaning that must be inferred across surface-dissimilar situations. Our empirical results demonstrate that AAT consistently reduces forgetting and improves generalization in a pattern that aligns with cognitive predictions for schema-based learning. Beyond the practical implications for continual learning, these results offer preliminary computational evidence that structural abstraction is a signal for stable learning in non-stationary environments.

Unified multimodal models share a language model backbone for both understanding and generating images. Can DPO align both capabilities simultaneously? We present the first systematic study of this question, applying DPO to Janus-Pro at 1B and 7B parameters under seven training strategies and two post-hoc methods. The central finding is negative: generation quality resists DPO alignment across all tested conditions on this architecture. No method improves generation CLIPScore at 7B (|Delta| < 0.2, p > 0.5 at n=200 per seed, 3 seeds); at 1B, all methods degrade generation, and the result holds across preference data types (real-vs-generated and model-vs-model) and the data volumes tested (150-288 pairs). Gradient analysis reveals why: understanding and generation gradients are near-orthogonal (cos ~ 0) with ~11-14x magnitude imbalance driven by VQ token count asymmetry (576 generation tokens vs. ~30-100 text tokens). This imbalance is the dominant interference mechanism in multi-task DPO; magnitude-balancing yields directionally positive understanding deltas (+0.01-0.04 VQA, though individually not significant), but the generation gap persists regardless. We identify discrete VQ tokenization as a likely structural bottleneck -- supported by the generation DPO loss converging to ln(2) -- and provide practical guidance for practitioners working with VQ-based unified models.

An Automatic License Plate Recognition (ALPR) system constitutes a crucial element in an intelligent traffic management system. However, the detection of Bangla license plates remains challenging because of the complicated character scheme and uneven layouts. This paper presents a robust Bangla License Plate Recognition system that integrates a deep learning-based object detection model for license plate localization with Optical Character Recognition for text extraction. Multiple object detection architectures, including U-Net and several YOLO (You Only Look Once) variants, are compared for license plate localization. This study proposes a novel two-stage adaptive training strategy built upon the YOLOv8 architecture to improve localization performance. The proposed approach outperforms the established models, achieving an accuracy of 97.83% and an Intersection over Union (IoU) of 91.3%. The text recognition problem is phrased as a sequence generation problem with a VisionEncoderDecoder architecture, with a combination of encoder-decoders evaluated. It was demonstrated that the ViT + BanglaBERT model gives better results at the character level, with a Character Error Rate of 0.1323 and Word Error Rate of 0.1068. The proposed system also shows a consistent performance when tested on an external dataset that has been curated for this study purpose. The dataset offers completely different environment and lighting conditions compared to the training sample, indicating the robustness of the proposed framework. Overall, our proposed system provides a robust and reliable solution for Bangla license plate recognition and performs effectively across diverse real-world scenarios, including variations in lighting, noise, and plate styles. These strengths make it well suited for deployment in intelligent transportation applications such as automated law enforcement and access control.

Surface manipulation tasks require robots to generate trajectories that comprehensively cover complex 3D surfaces while maintaining precise end-effector poses. Existing ergodic trajectory optimization (TO) methods demonstrate success in coverage tasks, while struggling with point-cloud targets due to the nonconvex optimization landscapes and the inadequate handling of SE(3) constraints in sampling-as-optimization (SAO) techniques. In this work, we introduce a preconditioned SE(3) Stein Variational Gradient Descent (SVGD) approach for SAO ergodic trajectory generation. Our proposed approach comprises multiple innovations. First, we reformulate point-cloud ergodic coverage as a manifold-aware sampling problem. Second, we derive SE(3)-specific SVGD particle updates, and, third, we develop a preconditioner to accelerate TO convergence. Our sampling-based framework consistently identifies superior local optima compared to strong optimization-based and SAO baselines while preserving the SE(3) geometric structure. Experiments on a 3D point-cloud surface coverage benchmark and robotic surface drawing tasks demonstrate that our method achieves superior coverage quality with tractable computation in our setting relative to existing TO and SAO approaches, and is validated in real-world robot experiments.

Advances in vision-language models (VLMs) have achieved remarkable success on complex multimodal reasoning tasks, leading to the assumption that they should also excel at reading analog clocks. However, contrary to this expectation, our study reveals that reading analog clocks in real-world environments remains a significant challenge for state-of-the-art VLMs. Existing analog clock datasets are largely synthetic or planar with limited stylistic diversity and minimal background context, failing to capture the visual variability of real-world scenes. As a result, VLMs trained on such data exhibit weak spatiotemporal reasoning, frequently confusing the hour and minute hands and struggling under common visual conditions such as occlusion, lighting variation, and cluttered backgrounds. To address this issue, we introduce TickTockVQA, a human-annotated dataset containing analog clocks in diverse real-world scenarios. TickTockVQA provides explicit hour and minute annotations, and includes an AM/PM tag when it is inferable from the visual context. Furthermore, we propose Swap-DPO, a direct preference optimization-based fine-tuning framework to align model reasoning toward accurate time interpretation. Experimental results demonstrate that our approach substantially enhances clock reading accuracy and robustness under real-world conditions, establishing a foundation for future research on spatiotemporal reasoning and visual understanding in VLMs.

Monte Carlo rendering and modern generative models both transform uncertain states into structured images, yet they are usually studied as separate processes. We introduce Monte Carlo Transport Scheduling, a framework that treats progressive path tracing as a continuous sampling-driven transport process. Our key observation is that the renderer already produces physically valid states along this process: nested Monte Carlo estimates trace a refinement trajectory whose natural time coordinate follows from sampling variance. This view leads to a continuous training framework that learns from real render endpoints rather than synthetic interpolants, preserving the statistical structure of Monte Carlo estimation while enabling arbitrary-step neural refinement. We evaluate the framework on a controlled rendering benchmark designed to separate transport difficulty from scene context, and show that it yields stable render refinement, supports continuous stopping between rendering states, and transfers as a physical prior for frozen generative samplers. These results suggest a common continuous-time substrate for rendering and generation, where Monte Carlo sampling provides both the physical states and the supervision for learning image transport.

Recent advances in large language model (LLM) have empowered autonomous agents to perform multi-turn interactions with tools and environments. However, scaling such agent training is limited by the lack of diverse and reliable environments. In this paper, we propose Agent World Model (AWM), a fully synthetic environment generation pipeline. Using this pipeline, we scale to 1,000 environments covering everyday scenarios, in which agents can interact with rich toolsets and obtain high-quality observations. Notably, these environments are code-driven and backed by databases, providing more reliable and consistent state transitions than environments simulated by LLMs. Moreover, they enable more efficient agent interaction compared with collecting trajectories from realistic environments. To demonstrate the effectiveness of this resource, we perform large-scale reinforcement learning for multi-turn tool-use agents. Thanks to the fully executable environments and accessible database states, we can also design reliable reward functions. Experiments on three benchmarks show that training exclusively in synthetic environments, rather than benchmark-specific ones, yields strong out-of-distribution generalization. The code is available at https://github.com/Snowflake-Labs/agent-world-model.

Constraint Answer Set Programming (CASP) is a hybrid paradigm that enriches Answer Set Programming (ASP) with numerical constraint processing, a crucial requirement for many real-world applications. However, the specification of constraints in most CASP solvers aligns more closely with the expressiveness and semantics of the numerical back-end than the ASP paradigm. In the latter, numerical attributes are represented as predicates, which allows declaring default values, leaving the attribute undefined, making non-deterministic assignments with choice rules, or using aggregated values. In CASP, most (if not all) of these features are lost once we switch to a constraint-based representation of those same attributes. In this paper, we present the flingo language (and tool) that incorporates the aforementioned expressiveness within numerical constraints, and we illustrate its use with several examples. Based on previous work that established its semantic foundations, we also present a translation from the newly introduced flingo syntax to regular CASP programs following the clingcon input format.

While large language model (LLM) multi-agent systems achieve superior reasoning performance through iterative debate, practical deployment is limited by their high computational cost and error propagation. This paper proposes AgentArk, a novel framework to distill multi-agent dynamics into the weights of a single model, effectively transforming explicit test-time interactions into implicit model capabilities. This equips a single agent with the intelligence of multi-agent systems while remaining computationally efficient. Specifically, we investigate three hierarchical distillation strategies across various models, tasks, scaling, and scenarios: reasoning-enhanced fine-tuning; trajectory-based augmentation; and process-aware distillation. By shifting the burden of computation from inference to training, the distilled models preserve the efficiency of one agent while exhibiting strong reasoning and self-correction performance of multiple agents. They further demonstrate enhanced robustness and generalization across diverse reasoning tasks. We hope this work can shed light on future research on efficient and robust multi-agent development. Our code is at https://github.com/AIFrontierLab/AgentArk.

Enabling robots to perform novel manipulation tasks from natural language instructions remains a fundamental challenge in robotics, despite significant progress in generalized problem solving with foundational models. Large vision and language models (VLMs) are capable of processing high-dimensional input data for visual scene and language understanding, as well as decomposing tasks into a sequence of logical steps; however, they struggle to ground those steps in embodied robot motion. On the other hand, robotics foundation models output action commands, but require in-domain fine-tuning or experience before they are able to perform novel tasks successfully. At its core, there still remains the fundamental challenge of connecting abstract task reasoning with low-level motion control. To address this disconnect, we propose Language Movement Primitives (LMPs), a framework that grounds VLM reasoning in Dynamic Movement Primitive (DMP) parameterization. Our key insight is that DMPs provide a small number of interpretable parameters, and VLMs can set these parameters to specify diverse, continuous, and stable trajectories. Put another way: VLMs can reason over free-form natural language task descriptions, and semantically ground their desired motions into DMPs -- bridging the gap between high-level task reasoning and low-level position and velocity control. Building on this combination of VLMs and DMPs, we formulate our LMP pipeline for zero-shot robot manipulation that effectively completes tabletop manipulation problems by generating a sequence of DMP motions. Across 31 real-world manipulation tasks, we show that LMP achieves 65% task success as compared to 35% for the best performing baseline. See videos at our website: https://collab.me.vt.edu/lmp

Neuro-symbolic systems aim to combine the expressive structure of symbolic logic with the flexibility of neural learning; yet, generative models typically lack mechanisms to enforce declarative constraints at generation time. We propose Logic-Guided Vector Fields (LGVF), a neuro-symbolic framework that injects symbolic knowledge, specified as differentiable relaxations of logical constraints, into flow matching generative models. LGVF couples two complementary mechanisms: (1) a training-time logic loss that penalizes constraint violations along continuous flow trajectories, with weights that emphasize correctness near the target distribution; and (2) an inference-time adjustment that steers sampling using constraint gradients, acting as a lightweight, logic-informed correction to the learned dynamics. We evaluate LGVF on three constrained generation case studies spanning linear, nonlinear, and multi-region feasibility constraints. Across all settings, LGVF reduces constraint violations by 59-82% compared to standard flow matching and achieves the lowest violation rates in each case. In the linear and ring settings, LGVF also improves distributional fidelity as measured by MMD, while in the multi-obstacle setting, we observe a satisfaction-fidelity trade-off, with improved feasibility but increased MMD. Beyond quantitative gains, LGVF yields constraint-aware vector fields exhibiting emergent obstacle-avoidance behavior, routing samples around forbidden regions without explicit path planning.

Mobile Graphical User Interface (GUI) World Models (WMs) offer a promising path for improving mobile GUI agent performance at train- and inference-time. However, current approaches face a critical trade-off: text-based WMs sacrifice visual fidelity, while the inability of visual WMs in precise text rendering led to their reliance on slow, complex pipelines dependent on numerous external models. We propose a novel paradigm: visual world modeling via renderable code generation, where a single Vision-Language Model (VLM) predicts the next GUI state as executable web code that renders to pixels, rather than generating pixels directly. This combines the strengths of both approaches: VLMs retain their linguistic priors for precise text rendering while their pre-training on structured web code enables high-fidelity visual generation. We introduce gWorld (8B, 32B), the first open-weight visual mobile GUI WMs built on this paradigm, along with a data generation framework (gWorld) that automatically synthesizes code-based training data. In extensive evaluation across 4 in- and 2 out-of-distribution benchmarks, gWorld sets a new pareto frontier in accuracy versus model size, outperforming 8 frontier open-weight models over 50.25x larger. Further analyses show that (1) scaling training data via gWorld yields meaningful gains, (2) each component of our pipeline improves data quality, and (3) stronger world modeling improves downstream mobile GUI policy performance.

Background: As of 2026, Large Language Models (LLMs) demonstrate expert-level medical knowledge. However, deploying them as autonomous "Clinical Agents" remains limited. Current Electronic Medical Records (EMRs) and standards like FHIR are designed for human review, creating a "Context Mismatch": AI agents receive fragmented data and must rely on probabilistic inference (e.g., RAG) to reconstruct patient history. This approach causes hallucinations and hinders auditability. Methods: We propose MedBeads, an agent-native data infrastructure where clinical events are immutable "Beads"--nodes in a Merkle Directed Acyclic Graph (DAG)--cryptographically referencing causal predecessors. This "write-once, read-many" architecture makes tampering mathematically detectable. We implemented a prototype with a Go Core Engine, Python middleware for LLM integration, and a React-based visualization interface. Results: We successfully implemented the workflow using synthetic data. The FHIR-to-DAG conversion transformed flat resources into a causally-linked graph. Our Breadth-First Search (BFS) Context Retrieval algorithm traverses relevant subgraphs with O(V+E) complexity, enabling real-time decision support. Tamper-evidence is guaranteed by design: any modification breaks the cryptographic chain. The visualization aids clinician understanding through explicit causal links. Conclusion: MedBeads addresses the "Context Mismatch" by shifting from probabilistic search to deterministic graph traversal, and from mutable records to immutable chains, providing the substrate for "Trustworthy Medical AI." It guarantees the context the AI receives is deterministic and tamper-evident, while the LLM determines interpretation. The structured Bead format serves as a token-efficient "AI-native language." We release MedBeads as open-source software to accelerate agent-native data standards.

Diagnosing failure patterns in Deep Research Agents (DRAs) remains a critical challenge. Existing benchmarks predominantly rely on end-to-end evaluation, obscuring intermediate hallucinations that accumulate throughout the research trajectory. To bridge this gap, we propose a shift from outcome-based to processaware evaluation by auditing hallucinations in the full plan-search-summarize trajectory. We introduce the PING Taxonomy, which categorizes DRA hallucinations into four complementary types: Propagation, Intent, Noiseinduced, and Grounding. We further instantiate this taxonomy into a fine-grained evaluation framework that decomposes trajectories into atomic actions, claims, and sub-queries for rigorous verification. Leveraging this framework to isolate 100 distinctively hallucinationprone tasks including adversarial scenarios, we curate DeepHalluBench. Experiments on six representative DRAs show that, on our hallucination-prone stress-test set, all evaluated systems still exhibit non-negligible reliability gaps. Furthermore, our diagnostic analysis traces these failures to systemic deficits, especially hallucination propagation and cognitive biases, providing actionable insights for future architectural optimization. Code and data are available in https://github.com/yuhao-zhan/DeepHalluBench.

Deep time series models are vulnerable to noisy data ubiquitous in real-world applications. Existing robustness strategies either prune data or rely on costly prior quantification, failing to balance effectiveness and efficiency. In this paper, we introduce DropoutTS, a model-agnostic plugin that shifts the paradigm from "what" to learn to "how much" to learn. DropoutTS employs a Sample-Adaptive Dropout mechanism: leveraging spectral sparsity to efficiently quantify instance-level noise via reconstruction residuals, it dynamically calibrates model learning capacity by mapping noise to adaptive dropout rates - selectively suppressing spurious fluctuations while preserving fine-grained fidelity. Extensive experiments across diverse noise regimes and open benchmarks show DropoutTS consistently boosts superior backbones' performance, delivering advanced robustness with negligible parameter overhead and no architectural modifications. Our code is available at https://github.com/CityMind-Lab/DropoutTS.

Multimodal fusion is often treated as an optimization-balancing problem, where training signals are adjusted to prevent one modality from dominating the others. However, balanced optimization does not fully determine the geometry of intermediate representations. Supervised multimodal models may still learn low-diversity modality-specific embeddings or allow paired cross-modal observations to drift excessively apart, weakening both unimodal robustness and multimodal fusion. We introduce \regName, a lightweight plug-and-play geometric regularization framework for multimodal representation learning. Rather than enforcing rigid cross-modal alignment, \regName follows a bounded-agreement principle: preserve modality-specific diversity while softly constraining only the portion of paired cross-modal drift that exceeds an admissible agreement band. Operationally, \regName combines a dispersion term that mitigates spectral concentration with an agreement-band anchoring term that controls excessive paired drift, requiring no architectural modification or inference-time overhead. Experiments across audio-visual, image-text, and RF-based benchmarks show that \regName consistently improves multimodal performance and often strengthens unimodal representations. These results suggest that explicitly regulating representation geometry is an effective complement to optimization balancing, and provide evidence that geometry-aware regularization can improve multimodal learning across diverse architectures and domains.

We rigorously study the thermodynamic limit of deep neural networks (DNNS) and recurrent neural networks (RNNs), assuming that the activation functions are sigmoids. A thermodynamic limit is a continuous neural network, where the neurons form a continuous space with infinitely many points. We show that such a network admits a unique state in a certain region of the parameter space, which depends continuously on the parameters. This state breaks into an infinite number of states outside the mentioned region of parameter space. Then, the critical organization is a bifurcation in the parameter space, where a network transitions from a unique state to infinitely many states. We use p-adic integers to codify hierarchical structures. Indeed, we present an algorithm that recasts the hierarchical topologies used in DNNs and RNNs as p-adic tree-like structures. In this framework, the hierarchical and the critical organizations are connected. We study rigorously the critical organization of a toy model, a hierarchical edge detector for grayscale images based on p-adic cellular neural networks. The critical organization of such a network can be described as a strange attractor. In the second part, we study random versions of DNNs and RNNs. In this case, the network parameters are generalized Gaussian random variables in a space of quadratic integrable functions. We compute the probability distribution of the output given the input, in the infinite-width case. We show that it admits a power-type expansion, where the constant term is a Gaussian distribution.

Neural PDE solvers are increasingly used as learned surrogates for families of partial differential equations, where the key machine learning challenge is not only interpolation on a fixed benchmark distribution but generalization under structured shifts in coefficients, boundary conditions, discretization, and rollout horizon. Yet evaluation is still often dominated by in-distribution test error, making robustness difficult to assess. We introduce a standardized stress-testing framework for neural PDE solvers under deployment-relevant shift. We instantiate it on three representative architectures -- Fourier Neural Operators (FNOs), a DeepONet-style model, and convolutional neural operators (CNOs) -- across five qualitatively different PDE families: dispersive, elliptic, multi-scale fluid, financial, and chaotic systems. Across 750 trained models, we measure robustness using baseline-normalized degradation factors together with spectral and rollout diagnostics. The resulting comparisons reveal that strong in-distribution accuracy does not reliably predict robustness, and that failure patterns depend jointly on architecture and PDE family. Our results provide a clearer basis for evaluating robustness claims in neural PDE solvers and suggest that function-space generalization under structured shift should be treated as a first-class evaluation target.

Although the Gradient Boosted Decision Trees (GBDTs) dominate industrial tabular applications, upgrading legacy models in high-concurrency production environments still faces prohibitive retraining costs and systemic risks. To address this problem, we present NSR-Boost, a neuro-symbolic residual boosting framework designed specifically for industrial scenarios. Its core advantage lies in being ``non-intrusive''. It treats the legacy model as a frozen model and performs targeted repairs on "hard regions" where predictions fail. The framework comprises three key stages: First, finding hard regions through residuals, then generating interpretable experts by generating symbolic code structures using Large Language Model (LLM) and fine-tuning parameters using Bayesian optimization, and finally dynamically integrating experts with legacy model output through a lightweight aggregator. Experimental results demonstrate that the framework significantly outperforms state-of-the-art (SOTA) baselines across six public datasets and one private dataset. More importantly, we report the successful deployment of NSR-Boost within the core financial risk control system of Qfin Holdings, where empirical results on real-world online traffic exhibit superior performance improvements and a significant reduction in the bad rate. In conclusion, it effectively captures long-tail risks missed by traditional models and offers a safe, low-cost evolutionary paradigm for industry.

Group Relative Policy Optimization (GRPO) is widely used for critic-free Large Language Model (LLM) post-training, but its KL regularization is usually implemented as a local loss-side token penalty. We show that this misses the policy-gradient signal induced by autoregressive KL regularization. Unlike standard KL-regularized Reinforcement Learning (RL) objectives, GRPO's group normalization induces a non-linear prompt-level utility; for binary verifier rewards, this utility is $2\arcsin\sqrt p$. As a result, reward and KL cannot be fused before normalization without changing the implicit objective. We derive the on-policy gradient of GRPO-style objectives with token-wise $f$-divergence regularization. The reward term recovers the standardized GRPO advantage, while the regularizer term includes a causal future-regularization return-to-go omitted by local KL losses. For reverse KL, this yields a simple future KL correction: add a reverse cumulative sum of per-token log ratios after advantage construction. The resulting method, Future-KL Regularized Policy Optimization (FRPO), requires no critic or extra model passes. On mathematical reasoning tasks, FRPO improves pass@16 in our main large-model setting while maintaining higher entropy and lower policy drift than conventional loss-side KL baselines.

Multimodal decentralized federated learning (DFL) must support collaboration among agents that hold different modality subsets and often different model components, while operating over peer-to-peer (P2P) overlays without a coordinating server or a global network view. A key obstacle is that conventional multimodal training often relies on a single shared representation, which implicitly assumes that heterogeneous peers can exchange and aggregate the same model components over the same communication links. In multimodal DFL, this assumption breaks down: uni- and multimodal agents may push incompatible updates through shared overlays, weakening both inter-agent transfer and cross-modal interaction. We present PARSE, a server-free framework that brings partial information decomposition (PID) into multimodal DFL. Each agent splits its latent features into redundant, unique, and synergistic slices ("feature fission"), and performs slice-aware communication over modality-conditioned P2P overlays. During training, agents exchange only the slices that are semantically alignable with their neighbors, according to the modalities and model components they share ("partial alignment"). This design avoids centralized orchestration and gradient-surgery style conflict handling, while remaining compatible with standard DFL constraints and a range of P2P overlay topologies. Across multiple benchmarks and heterogeneous peer mixes, PARSE consistently outperforms task-, modality-, and hybrid-sharing multimodal DFL baselines while keeping per-link payloads bounded. Ablations on fusion choices and split ratios, together with qualitative feature analyses and overlay-topology studies, demonstrate the robustness and communication efficiency of the proposed slice-aware design.

We revisit the problem of constructing interoperable patient digital twins from unstructured electronic health records (EHRs) and argue that the task is better cast not as a cascade of extraction modules but as constrained generation of a valid FHIR bundle. We introduce SG-LLM, a schema-grounded LLM extractor that (i) augments the prompt with candidate SNOMED-CT, RxNorm, and LOINC codes retrieved through a SapBERT index, (ii) decodes under a JSON Schema derived directly from FHIR R4 StructureDefinitions, and (iii) closes a validator-in-the-loop repair stage whose diagnostics are fed back as structured error messages. We argue that the twin's usefulness, not only span-level F1, is the right object of evaluation, and operationalize this with a clinical-utility experiment that measures the gap in 30-day readmission AUROC between classifiers trained on SG-LLM-generated FHIR bundles versus expert-curated ones. On MIMIC-IV and n2c2 2018 Track 2 benchmarks, SG-LLM matches or exceeds strong joint-extraction and vanilla-LLM baselines while producing substantially more valid bundles. Ablations isolate the contributions of retrieval, schema constraint, and the repair loop. All code, prompts, and schemas are released.

Self-destructive behaviors are linked to complex psychological states and can be challenging to diagnose. These behaviors may be even harder to identify within subcultural groups due to their unique expressions. As large language models (LLMs) being deployed across various fields, some researchers have begun exploring their application for detecting self-destructive behaviors. Motivated by this, we investigate self-destructive behavior detection within subcultures using current LLM-based methods. However, these methods have two main challenges: (1) Knowledge Lag: Subcultural slang evolves rapidly, faster than LLMs' training cycles; and (2) Semantic Misalignment: it is challenging to grasp the specific and nuanced expressions unique to subcultures. To address these issues, we propose Subcultural Alignment Solver (SAS), a multi-agent framework that incorporates automatic retrieval and subculture alignment, significantly boosting the performance of LLMs in detecting self-destructive behavior. Our experimental results show that SAS outperforms the current advanced multi-agent framework OWL. Notably, it competes well with fine-tuned LLMs. We hope that SAS will advance the field of self-destructive behavior detection in subcultural contexts and serve as a valuable resource for future researchers.

Multimodal medical large language models have shown substantial progress in chest X-ray interpretation but continue to face challenges in spatial reasoning and anatomical understanding. Although existing grounding techniques improve overall performance, they often fail to establish a true anatomical correspondence, resulting in incorrect anatomical understanding in the medical domain. To address this gap, we introduce AnatomiX, a multitask multimodal large language model for anatomically grounded chest X-ray interpretation. Inspired by the radiological workflow, AnatomiX adopts a two stage approach: first, it identifies anatomical structures and extracts their features, and then leverages a large language model to perform diverse downstream tasks such as phrase grounding, report generation, visual question answering, and image understanding. Extensive experiments across multiple benchmarks demonstrate that AnatomiX achieves superior anatomical reasoning and delivers over 25% improvement in performance on anatomy grounding, phrase grounding, grounded diagnosis and grounded captioning tasks compared to existing approaches. Code and pretrained model are available at https://aneesurhashmi.github.io/anatomix

Generating patent descriptions from scientific papers is challenging due to fundamental rhetorical and structural disparities between the two genres. Existing approaches treat this as surface-level rewriting, failing to capture the hierarchical reasoning and statutory constraints inherent in patent drafting. We propose FlowPlan-G2P, a graph-mediated generation framework that decomposes this transformation into three stages: (1) Concept Graph Induction, extracting technical entities and functional dependencies into a directed graph; (2) Section-level Planning, partitioning the graph into coherent subgraphs aligned with canonical patent sections; and (3) Graph-Conditioned Generation, synthesizing legally compliant paragraphs conditioned on section-specific subgraphs. Experiments on expert-validated benchmarks reveal that standard NLG metrics systematically favor legally non-compliant outputs over valid patent descriptions, motivating our domain-specific evaluation. Under this evaluation, FlowPlan-G2P with an open-weight backbone consistently outperforms vanilla proprietary models, demonstrating that structured decomposition is a stronger determinant of quality than model scale.