arXiv:2605.19721v1 Announce Type: new
Abstract: Graph combinatorial optimization (GCO) has attracted growing interest, as many NP-hard problems naturally admit graph formulations, yet their combinatorial explosion renders exact methods computationally intractable. Recent advances in Reinforcement Learning (RL) combined with Graph Neural Networks (GNNs) have significantly improved learning-based GCO solvers. However, existing approaches face limitations in both generalization across diverse graph instances and computational scalability as action spaces grow. To address both challenges, we introduce projection agents, a novel RL-GCO approach that operates directly in a continuous GNN-based action embedding space, predicting a desired latent action in a single forward pass and subsequently decoding it into a valid discrete action. Additionally, we enable fair comparison across RL methods through a shared embedding space for both observations and actions. Across diverse benchmarks, our approach achieves up to 16.2x faster inference and up to 40% better generalization than existing solutions using only simple nearest-neighbor decoding, while opening the door to strong RL performance in super-linear decision spaces with multiple interdependent variables. Finally, we release LaGCO-RL, a Python library that automates latent action-space construction and supports existing RL-GCO solutions, promoting reproducibility and adaptation to new GCO benchmarks.
Science Journals
arXiv:2605.19427v1 Announce Type: cross
Abstract: A conservation-consistent boundary condition is proposed for nonlinear models of soluble-surfactant-laden falling films, ensuring exact conservation of total surfactant mass. The formulation resolves an inconsistency in widely used reduced models, Pascal et al. (PRF, 2019), D'Alessio et al. (JFM, 2020), which exhibit a gradual drift of mass during nonlinear evolution in a closed periodic domain. We show that this originates from an inconsistency in the surface transport reduction and derive a corrected boundary condition that removes this defect. As the discrepancy appears only at the nonlinear order, linear stability results remain unaffected, explaining why the issue has remained unnoticed.
arXiv:2605.19762v1 Announce Type: new
Abstract: Code has become a standard component of modern foundation language model (LM) training, yet its role beyond programming remains unclear. We revisit the claim that code improves reasoning through controlled pretraining experiments on a 10T-token corpus with fine-grained domain separation. Our findings are threefold. First, when code is restricted to standalone executable programs and Code-NL data are controlled for, code substantially improves programming ability but does not act as a general reasoning enhancer; instead, it competes with knowledge-intensive tasks, especially complex mathematical reasoning. Second, the reasoning gains often attributed to code are better explained by cross-domain structured reasoning traces, such as code-text and math-text mixtures, rather than by executable code alone. Third, increasing the density of structured math-domain samples within a fixed math budget yields substantial gains on difficult mathematical reasoning while largely preserving programming performance, suggesting that cognitive scaffolds offer a targeted way to mitigate cross-domain trade-offs. Finally, routing analyses show that data-composition effects are reflected in expert-activation patterns, providing mechanism-level evidence for competitive and synergistic interactions across domains. Our results clarify which data characteristics transfer across capability dimensions and point to more precise data-centric optimization strategies.
arXiv:2605.20084v1 Announce Type: new
Abstract: Large language models (LLMs) can enhance factuality via retrieval-augmented generation (RAG), but applying RAG to every query is unnecessary when the model-only answer is reliable. This motivates cascaded RAG: each query is first handled by an LLM-only branch, escalated to a RAG fallback only if the primary branch is uncertain, and abstained from when neither branch is sufficiently trustworthy. However, calibrating such cascades stage by stage may be conservative, since the final utility depends on joint uncertainty thresholding of LLM-only and RAG. In this work, we develop BalanceRAG to certify threshold pairs at a target risk level. Given uncertainty scores from the two branches, BalanceRAG frames each threshold pair as an operating point on a two-dimensional lattice and identifies safe operating points using sequential graphical testing. This enables risk-adaptive threshold calibration, controlling the system-level error rate among accepted points, while retaining more examples. Furthermore, BalanceRAG extends to multi-risk calibration, allowing retrieval usage to be bounded together with the selection-conditioned risk. Experiments on three open-domain question answering (QA) benchmarks across multiple LLM backbones demonstrate that BalanceRAG meets prescribed risk levels, preserves higher coverage and more accepted correct examples, and reduces unnecessary retrieval calls compared with always-on RAG.
arXiv:2605.18777v1 Announce Type: new
Abstract: Mapping large origin-destination (OD) datasets remains challenging because flow maps become cluttered, meaningful patterns occur at multiple spatial scales, and existing flow-mapping approaches frequently rely on predefined aggregation units or manual generalization. This paper presents XFlowMap, a framework for the cross-scale generalization and mapping of massive OD data. Specifically, the framework integrates cross-scale flow pattern (cluster) detection, automated flow map generalization, and a new cartographic representation for analyzing and visualizing complex origin-destination flow structures. The approach detects salient flow patterns at their appropriate origin and destination scales, extracts high-level structures, and generates a new flow map representation that supports holistic interpretation of complex origin-destination flow patterns. A scan-statistic-based procedure is developed to evaluate and generalize cross-scale flow clusters. The detected clusters are then visualized using a novel flow symbol that integrates location, direction, strength, and OD scales in a single representation. The framework supports both area-based and point-based OD data, is robust to sparse and noisy datasets, and enables comparative mapping of stratified flow data. Experiments with synthetic data and U.S. migration data demonstrate that the method effectively extracts meaningful cross-scale flow patterns and produces clear, information-rich flow maps for large mobility datasets, supporting both static presentation and interactive exploration.
arXiv:2605.20094v1 Announce Type: new
Abstract: Most of the laws of Nature involve derivatives up to the second order. Ostrogradski was the first to seek a formulation of the equations of higher-order derivatives. He extended Hamilton's equations by considering Lagrangians that depend on higher-order derivatives of generalized coordinates. The Hamilton-Ostrogradski formulation served as the basis for later studies with higher-order derivatives. However, the Hamilton-Ostrogradski formalism is rarely discussed in textbooks or the pedagogical literature. This motivated us to show how the Hamilton-Ostrogradski formalism can be applied it to the Pais-Uhlenbeck oscillator. We hope that the approach presented in this work can serve as a basis for discussion in advanced classical mechanics courses.
arXiv:2605.20177v1 Announce Type: new
Abstract: Recent advances in vision-language models (VLMs) emphasize long chain-of-thought reasoning; yet, we find that their performance on visual tasks is primarily limited by a lack of visual perception as opposed to reasoning itself. In this work, we systematically study the interplay between perception and reasoning in VLM post-training by decomposing their capabilities into three separate training stages: visual perception, visual reasoning, and textual reasoning, incorporating specialized training data. We demonstrate that visual perception (a) requires targeted optimization with specialized data; (b) serves as a fundamental scaffold that should be solidified through staged training before refining visual reasoning; and (c) is more effectively learned via RL than caption-based SFT. Our experiments across multiple VLMs demonstrate that staged training consistently improves both visual perception and reasoning performance over merged training. Notably, models trained with our approach achieve 1.5% higher reasoning accuracy with 20.8% shorter reasoning traces, suggesting that superior perception reduces the need for excessive reasoning. Furthermore, we show that this capability-based staging represents a new curriculum dimension orthogonal to traditional difficulty-based curricula, and combining both yields further additive gains. Our staged-training models achieve superior performance among open-weight VLMs, establishing advanced results on several visual math and perception (e.g., +5.2% on WeMath and +3.7% on RealWorldQA) tasks compared with the base counterpart.
arXiv:2602.18718v2 Announce Type: replace-cross
Abstract: For approximating a target distribution given only its unnormalized log-density, stochastic gradient-based variational inference (VI) algorithms are a popular approach. For example, Wasserstein VI (WVI) and black-box VI (BBVI) perform gradient descent in measure space (Bures-Wasserstein space) and parameter space, respectively. Previously, for the Gaussian variational family, convergence guarantees for WVI have shown superiority over existing results for black-box VI with the reparametrization gradient, suggesting the measure space approach might provide some unique benefits. In this work, however, we close this gap by obtaining identical state-of-the-art iteration complexity guarantees for both. In particular, we identify that WVI's superiority stems from the specific gradient estimator it uses, which BBVI can also leverage with minor modifications. The estimator in question is usually associated with Price's theorem and utilizes second-order information (Hessians) of the target log-density. We will refer to this as Price's gradient. On the flip side, WVI can be made more widely applicable by using the reparametrization gradient, which requires only gradients of the log-density. We empirically demonstrate that the use of Price's gradient is the major source of performance improvement.
arXiv:2605.16681v1 Announce Type: cross
Abstract: Audio super-resolution (SR), also referred to as bandwidth extension (BWE), aims to reconstruct high-fidelity signals from low-resolution (LR) or band-limited (BL) observations, an inherently ill-posed task due to the ambiguity of missing high-frequency (HF) content. This survey provides a comprehensive overview of the field, with a particular focus on the paradigm shift from discriminative mapping to modern generative modeling. We first review early discriminative deep neural network (DNN) models, which formulate BWE/SR as a deterministic mapping problem and are prone to regression-to-the-mean effects and spectral over-smoothing. We then systematically review generative approaches, including autoregressive (AR) models, variational autoencoders (VAEs), generative adversarial networks (GANs), diffusion and score-based models, flow-based methods, and Schr\"odinger bridges. Across these approaches, we examine key design aspects, including representation domain, architecture, conditioning mechanisms, and trade-offs among reconstruction fidelity, perceptual quality, robustness, and computational efficiency. Furthermore, we discuss emerging directions involving large language models (LLMs) and multimodal foundation models, and highlight open challenges in perceptual evaluation, phase modeling, and real-world generalization. By providing a structured taxonomy and unified perspective, this survey establishes a comprehensive foundation and offers a practical roadmap for advancing BWE/SR from deterministic point estimation toward distribution-aware generative modeling.
arXiv:2410.15362v2 Announce Type: replace
Abstract: Aligned Large Language Models (LLMs) have attracted significant attention for their safety, particularly in the context of jailbreak attacks that attempt to bypass guardrails via adversarial prompts. Among existing approaches, the Greedy Coordinate Gradient (GCG) attack pioneered automated jailbreaks through discrete token optimization; however, its low sample efficiency limits practical applicability. In particular, GCG requires approximately 256K evaluations per harmful behavior to achieve a satisfactory jailbreak success rate, due to the inherent difficulty of the underlying discrete optimization problem. In this work, we identify three key factors that limit the sample efficiency of GCG: inaccurate gradient-based estimation, inefficient uniform sampling, and repeated evaluation of previously explored suffixes. To address these issues, we propose Faster-GCG, a streamlined variant of GCG that incorporates distance-based regularization for improved estimation, temperature-controlled sampling for more effective exploration, and a visited-suffix marking mechanism to avoid redundant evaluations. Faster-GCG reduced the required evaluations to 32K, achieving up to an $8\times$ improvement in sampling efficiency and a $7\times$ reduction in wall-clock time compared to GCG. Under this reduced budget, Faster-GCG attained an average jailbreak success rate of 78.1\% across five aligned LLMs, and achieved 88.7\% against Qwen3.5-4B, outperforming state-of-the-art white-box jailbreak methods.
arXiv:2602.17657v2 Announce Type: replace-cross
Abstract: The Pauli exclusion principle is a cornerstone of quantum physics: it governs the structure of matter. Extensions of this principle, such as Haldane's generalized exclusion statistics, predict the existence of exotic quantum states characterized by fractional Fermi seas (FFS), i.e. momentum distributions with uniform but fractional occupancies. Here, we report the experimental realization of fractional Fermi seas in an excited one-dimensional Bose gas prepared through ramping cycles in the interaction strength. The resulting excited yet stable Bose-gas states exhibit Friedel oscillations, smoking-gun signatures of the underlying FFS. The stabilization of these states offers an opportunity to deepen our understanding of quantum thermodynamics in the presence of exotic statistics and paves the way for applications in quantum information and sensing.
arXiv:2511.02622v2 Announce Type: replace-cross
Abstract: Predicting the secondary structure of RNA is a core challenge in computational biology, essential for understanding molecular function and designing novel therapeutics. The field has evolved from foundational but accuracy-limited thermodynamic approaches to a new data-driven paradigm dominated by machine learning and deep learning. These models learn folding patterns directly from data, leading to significant performance gains. This review surveys the modern landscape of these methods, covering single-sequence, evolutionary-based, and hybrid models that blend machine learning with biophysics. A central theme is the field's "generalization crisis," where powerful models were found to fail on new RNA families, prompting a community-wide shift to stricter, homology-aware benchmarking. In response to the underlying challenge of data scarcity, RNA foundation models have emerged, learning from massive, unlabeled sequence corpora to improve generalization. Finally, we look ahead to the next set of major hurdles-including the accurate prediction of complex motifs like pseudoknots, scaling to kilobase-length transcripts, incorporating the chemical diversity of modified nucleotides, and shifting the prediction target from static structures to the dynamic ensembles that better capture biological function. We also highlight the need for a standardized, prospective benchmarking system to ensure unbiased validation and accelerate progress.
arXiv:2410.09447v3 Announce Type: replace
Abstract: Recently, plenty research has been done on discovering the role of energy dissipation in biological networks, most of which focus on the relationship of dissipation and functionality. However, the development of networks science urged us to fathom the systematic architecture of biological networks and their evolutionary advantages. We found the dissipation of biological dissipative networks is highly related to their structure. By interrogating these well-adapted networks, we find that the energy producing module is relatively isolated in all situations. We applied evolutionary simulation and analysis on premature networks of classic dissipative networks, namely kinetic proofreading, activator-inhibitor oscillator and two typical adaptative response models. We found despite that selection was imposed merely on the network function, the networks tended to decouple high energy molecules as fuels from the functional module, to achieve higher overall dissipation during the course of evolution. Furthermore, we find that decoupled fuel modules can increase the robustness of the networks towards parameter or structure perturbations. We provide theoretical analysis on the kinetic proofreading networks and the general case of energy-driven networks. We find fuel decoupling can guarantee higher dissipation and, in most cases when considering dissipative networks, higher performance. We conclude that fuel decoupling is an evolutionary outcome and bears benefits during evolution.
arXiv:2605.18897v1 Announce Type: cross
Abstract: Intracranial EEG (iEEG) provides high-fidelity neural recordings essential for clinical and brain-computer interface applications, but acquiring these signals requires invasive surgery. While recent studies have attempted to estimate iEEG from non-invasive scalp EEG, most rely on patient-specific models, creating a circular dependency: if surgery is required to collect training data, the non-invasive model offers limited practical benefit. In this study, we address the challenge of cross-subject iEEG reconstruction by predicting intracranial signals for unseen patients using models trained on other individuals. We propose CAST (Cross-Attention Spatial-Temporal Transformer), a machine learning framework that translates scalp EEG into multi-channel iEEG waveforms through a two-stage transfer learning strategy. First, a temporal encoder extracts multi-scale neural representations at three different resolutions. Then, because electrode placements vary substantially across patients, a channel-aware decoder is calibrated using only a few minutes of data from the target subject. We evaluated the proposed method using leave-one-subject-out cross-validation on two public datasets comprising 1,282 iEEG channels. Experimental results demonstrate that CAST reconstructs cortical signals located near the scalp surface substantially better than deep subcortical activity. In highly observable sensorimotor regions, the model achieved peak correlations of up to r=0.864 in the precentral gyrus. Furthermore, with a channel selection strategy, CAST obtained a mean correlation of r=0.545 on viable subjects, outperforming previous within-subject baselines. These findings indicate that cortical iEEG signals can be reconstructed for unseen subjects from scalp EEG without extensive patient-specific training, and that only a brief calibration phase is sufficient to adapt the model to new hardware configurations.
arXiv:2605.19750v1 Announce Type: new
Abstract: Visual autoregressive (VAR) models have recently emerged as an efficient paradigm for text-to-image generation. Despite their strong generative capability, existing VAR-based personalization methods remain limited to static settings, failing to accommodate evolving user demands. In particular, sequential concept learning leads to severe catastrophic forgetting, while multi-concept synthesis often suffers from feature entanglement and attribute inconsistency. In this work, we present the first systematic study of continual personalized generation in VAR models. We identify two key challenges: (i) preserving previously learned concepts during sequential customization, and (ii) composing multiple personalized concepts in a controllable manner. To address these issues, we propose a unified framework with two core components. For continual single-concept learning, we introduce Gradient-based Concept Neuron Selection (GCNS), which identifies concept-relevant neurons and constrains only conflicting parameters across tasks, effectively mitigating forgetting without additional model expansion. For multi-concept synthesis, we propose a context-aware composition strategy that performs multi-branch feature modeling and localized cross-attention fusion guided by spatial conditions, enabling precise and disentangled concept composition. Extensive experiments demonstrate that our method significantly improves performance in long-sequence continual personalization while achieving superior results in multi-concept image synthesis compared to existing baselines. These findings highlight the potential of VAR models for scalable and controllable personalized generation.
arXiv:2603.03140v3 Announce Type: replace
Abstract: AI agents are increasingly active on social media platforms, generating content and interacting with one another at scale. Yet the behavioral diversity of these agents remains poorly understood, and methods for characterizing distinct agent types and studying how they engage with shared topics are largely absent from current research. We apply the Persona Ecosystem Playground (PEP) to Moltbook, a social platform for AI agents, to generate and validate conversational personas from 41,300 posts using k-means clustering and retrieval-augmented generation. Cross-persona validation confirms that personas are semantically closer to their own source cluster than to others (t(61) = 17.85, p < .001, d = 2.20; own-cluster M = 0.71 vs. other-cluster M = 0.35). These personas are then deployed in a nine-turn structured discussion, and simulation messages were attributed to their source persona significantly above chance (binomial test, p < .001). The results indicate that persona-based ecosystem modeling can represent behavioral diversity in AI agent populations.
arXiv:2512.14700v2 Announce Type: replace
Abstract: Online harassment is a widespread social and public health concern, yet most computational approaches for detecting and addressing harassment focus on publicly visible social media content rather than private messaging environments. Private conversations present unique challenges because harmful interactions often unfold through context-dependent, multi-turn exchanges, while victims may lack timely support during moments of harassment. In this study, we investigate how large language models (LLMs) can support both the detection of and response to online harassment in private messaging. Using a dataset of 80,053 Instagram direct messages donated by 26 adolescents aged 12-18, including youth with suicide risk factors, we first construct a human-labeled dataset of online harassment in private conversations and develop a context-aware cascading LLM classification pipeline. The proposed pipeline outperforms baseline toxicity classifiers trained primarily on public social media data. We then develop a victim-centered response framework that produces context-sensitive and psychologically-grounded AI-generated responses to online harassment messages. Human evaluators perceived the AI-generated responses as significantly more helpful than the original participant responses (95% CI: 0.767--0.815, p < .001), particularly in terms of emotional support and de-escalation. Our findings highlight the potential of context-aware and victim-centered AI systems to provide just-in-time support during harassment in private messaging environments.
arXiv:2605.18881v1 Announce Type: new
Abstract: In dynamic flow fields, various animals exhibit remarkable odor search capabilities despite relying on stochastic detections. Interestingly, there exists an optimal time window for integrating these detections that maximizes search efficiency. To understand the underlying mechanism, we investigate the navigation performance of Reinforcement Learning (RL) agents in unsteady flows under varying memory lengths and flow conditions. Without any predefined models, the agents develop a flow-assisted casting strategy and adaptively adjust both the geometry of their search trajectories and the concentration threshold for initiating casting to maximize the success rate. The agent's average speed toward the odor source exhibits a non-monotonic dependence on memory length, which can be explained by the "sector-search" model.
arXiv:2605.19722v1 Announce Type: new
Abstract: Do stock safety-aligned language models and their uncensored or abliterated derivatives behave differently when run as autonomous security agents? Single-turn refusal benchmarks cannot answer this question: security agents must inspect repositories, call tools, and produce vulnerability evidence inside authorized sandboxes.
We present a trace-based benchmark of 30 local vulnerability-analysis tasks with fixed tools, deterministic success predicates, redaction rules, and grounding checks, and compare four stock models against uncensored or abliterated derivatives: Gemma 4 31B, Gemma 4 26B A4B, Qwen2.5-Coder 7B, and Llama 3.1 8B. The artifact contains 1,500 security-agent traces and 800 non-security control traces. The Gemma pairs show large less-restricted gains on security tasks: 14.0% versus 0.7% success for 31B and 10.7% versus 0.0% for 26B, with higher mean grounding (3.91 versus 3.27 and 4.12 versus 1.64 out of five) and 0.0% refusal, suppressed-action, and unsafe-action rates in the 31B traces. However, controls and non-Gemma pairs rule out a clean security-specific or universal less-restricted effect: Gemma gaps also appear on ordinary coding tasks, Qwen2.5-Coder success is lower for the less-restricted derivative (2.0% versus 5.3%), and the abliterated Llama derivative fails the tool protocol. Across all families, hard proof-of-trigger and patch-verification tasks remain unsolved. These results show that safety alignment effects in autonomous security agents should be measured at the system level, separating refusal, unsafe action, tool reliability, and evidence grounding rather than treating refusal rate as the safety signal.
arXiv:2605.19727v1 Announce Type: new
Abstract: Existing 3D foundation models typically align point clouds to frozen vision-language spaces like CLIP, which achieve strong cross-modal retrieval by compressing 3D shape into a global vector. However, this global-only alignment cannot establish fine-grained pixel-to-point correspondence. To solve this, we present Tango3D, a foundation model that unifies dense correspondence and global retrieval. We use a geometry-aware 2D visual backbone and a pretrained 3D VAE to encode images into 2D patches and point clouds into 3D tokens. These are mapped into a single shared space to achieve both local pixel-to-point alignment and global semantic alignment. To stabilize the joint learning of dense and global objectives, we introduce a three-stage progressive training strategy. Experiments show our model successfully achieves object-level pixel-to-point alignment while maintaining competitive global retrieval, a joint capability not offered by existing 3D foundation models. By establishing a fine-grained alignment feature space, Tango3D injects rich semantics into purely geometric 3D tokens, paving the way for a wide range of dense 3D downstream tasks.
arXiv:2408.12385v3 Announce Type: replace
Abstract: We study the problem of approximately recovering a probability distribution given noisy measurements of its Chebyshev polynomial moments. This problem arises broadly across algorithms, statistics, and machine learning. By leveraging a global decay bound on the coefficients in the Chebyshev expansion of any Lipschitz function, we sharpen prior work, proving that accurate recovery in the Wasserstein distance is possible with more noise than previously known. Our result immediately yields a number of applications:
1) We give a simple "linear query" algorithm for constructing a differentially private synthetic data distribution with Wasserstein-$1$ error $\tilde{O}(1/n)$ based on a dataset of $n$ points in $[-1,1]$. This bound is optimal up to log factors, and matches a recent result of Boedihardjo, Strohmer, and Vershynin [Probab. Theory. Rel., 2024], which uses a more complex "superregular random walk" method.
2) We give an $\tilde{O}(n^2/\epsilon)$ time algorithm for the linear algebraic problem of estimating the spectral density of an $n\times n$ symmetric matrix up to $\epsilon$ error in the Wasserstein distance. Our result accelerates prior methods from Chen et al. [ICML 2021] and Braverman et al. [STOC 2022].
3) We tighten an analysis of Vinayak, Kong, Valiant, and Kakade [ICML 2019] on the maximum likelihood estimator for the statistical problem of "Learning Populations of Parameters'', extending the parameter regime in which sample optimal results can be obtained.
Beyond these main results, we provide an extension of our bound to estimating distributions in $d > 1$ dimensions. We hope that these bounds will find applications more broadly to problems involving distribution recovery from noisy moment information.
arXiv:2605.19726v1 Announce Type: new
Abstract: Diffusion Language Models (DLMs) enable globally coherent, bidirectional, and controllable text generation, offering advantages over traditional autoregressive LLMs, while scaling to ultra-long sequences remains costly. Many existing block-sparse attention methods select blocks by fixed sampling patterns over the high-resolution attention space, such as tail regions or anti-diagonal stripes. Such prior-driven sampling can miss salient tokens and introduce instability under distribution shifts. In this paper, we propose the Block Approximate Sparse Attention framework (BA-Att) with block-wise pre-downsampled operation, which identifies informative regions within a compact downsampled space, avoiding reliance on brittle positional priors. To analyze its theoretical behavior, we define an oracle post-downsample attention map and formalize the approximation error between pre- and post-downsample schemes. Based on this insight, we introduce a lightweight norm-sorting module and a covariance-compensated correction that approximates full covariance using diagonal QK variances, reducing computational complexity. Extensive experiments show that our operator achieves up to 6.95x acceleration over FlashAttention in attention computation, and maintains near full-attention performance at 50% sparsity across language models, multimodal language models, and video generation models, demonstrating strong efficiency and generalization.
arXiv:2407.17200v3 Announce Type: replace-cross
Abstract: Many real-world decision problems require solving, again and again, combinatorial optimization instances drawn from a common distribution. A recent line of structured learning methods exploits this regularity by learning policies that pair a statistical model with a tractable combinatorial oracle, instead of solving each instance independently. Training such policies is notoriously difficult, however: the resulting empirical risk is piecewise constant in the model parameters, which hinders gradient-based optimization, and only a few theoretical guarantees have been provided so far. We address this issue by analyzing smoothed (perturbed) policies: adding controlled random perturbations to the direction used by the linear oracle yields a differentiable surrogate risk and improves generalization. Our main contribution is a generalization bound that decomposes the excess risk into $(\mathit{i})$ perturbation bias, $(\mathit{ii})$ statistical estimation error, and $(\mathit{iii})$ optimization error. The perturbation bias is controlled by the \emph{fan-crossing probability}, a new geometric quantity measuring the likelihood that a perturbation changes the oracle solution. We introduce two complementary conditions to bound it--the \emph{Uniformly Bounded Density} (UBD) property, yielding a sharp ${O}(\lambda)$ bias, and the weaker \emph{Uniform Weak moment} (UW) property, yielding a sub-linear bound--both capturing the geometric interaction between the statistical model and the normal fan of the feasible polytope. The statistical estimation error is controlled via a uniform deviation bound over the policy class, with rate ${O}(1/(\lambda\sqrt{n}))$ that scales inversely in the smoothing parameter. Concerning the optimization error, we exploit kernel Sum-of-Squares methods to mitigate the curse of dimensionality of global optimization.
arXiv:2603.05933v2 Announce Type: replace
Abstract: Applying Small Language Models (SLMs) to Chinese character-driven generation remains challenging due to data scarcity and the difficulty of disentangling character style. Standard Supervised Fine-Tuning (SFT) often captures surface-level semantics but produces frequent Out-Of-Character (OOC) outputs. We frame this as a controlled sentence-level style rewriting task, which isolates stylistic quality from dialogue context management. We propose a Structured Style-Rewrite Framework that decomposes character style into interpretable format signature, syntactic, and pragmatic dimensions, combined with Chain-of-Thought (CoT) supervision for explicit style planning. A CoT-Shared Direct Preference Optimization (DPO) stage further aligns style planning with surface realization by ensuring preference learning targets output-level style execution rather than reasoning trace differences. Experiments across eight characters from four diverse source domains demonstrate that our method enables a Qwen3-1.7B model to achieve a Valid Style Score of $0.632$ while maintaining strong semantic fidelity (0.878), placing on the Pareto frontier among the evaluated systems and outperforming significantly larger baselines (e.g., GLM-4.7) on consumer hardware.
arXiv:2605.18831v1 Announce Type: cross
Abstract: Cold-chain storage limits access to insulin for hundreds of millions of people; a thermally protective patch polymer could help, but the design space is too large for exhaustive experiment. Starting from that problem, we narrow to an agentic workflow: a large language model (LLM) calls physics-based tools through the Model Context Protocol (MCP), searching the discrete PSMILES space under a budget of OpenMM Packmol-matrix evaluations. The LLM acts as an implicit acquisition function conditioned on a persistent "discovery world": hypotheses, literature claims, and simulation outcomes updated each iteration. Under matched oracle budgets, the best autonomous campaign reaches an insulin-polymer interaction energy of -2263 kJ/mol, outperforming reinforcement-learning baselines by 68% and Bayesian optimization by 19%. Three independent campaigns converge on one structural motif (dense hydrogen-bond donors and acceptors per repeat unit) while physics checks reject infeasible packings and name-structure mismatches before they steer the next step. The science stage is CPU-bound and runs on commodity hardware. More broadly, the same architecture and workflow designed here applies to other protein-stabilization tasks whenever a tractable screening oracle is available.