arXiv:2605.15240v1 Announce Type: cross
Abstract: This paper investigates the critical role of eigenalignments between the kernel matrix and learning targets in achieving robust generalization in learning problems. We establish a direct connection between generalization performance in kernel methods and the estimation of eigenvectors and eigenvalues of matrices, offering a more intuitive understanding compared to prior work with minimal assumptions. We also show that, since the prediction task in KRR is essentially the weighted sum of eigenvectors/singular vectors, by analyzing how much error can be caused by perturbations to the kernel matrix, we can then derive a bound on this generalization error using the estimation stability of matrix eigenvalues and eigenvectors. Compared with previous work, our analysis concentrates on finite-sample settings and on the generalization error arising from having a suboptimal finite training set. Our findings reveal that in kernel methods, as long as the kernel is of high rank, the near-zero reconstruction error can be trivially obtained, implying that the reconstruction error will have limited predictive power for generalization. Finally, we establish a generalization bound from an eigenvalues/eigenvectors estimation perspective, showing that strong generalization requires increasing eigenvector alignment, eigenvalue magnitude, or gaps between consecutive eigenvalues.
Science Journals
arXiv:2507.14200v2 Announce Type: replace
Abstract: Existing multi-LLM collaboration systems often encounter scalability challenges when integrating new LLMs and tasks, leading to suboptimal performance. To address this, we propose SMCS, a Scalable Multi-LLM Collaboration System designed to effectively coordinate multiple open-source LLMs. The system consists of two core components: a Retrieval-based Prior Selection (RPS) module, which dynamically selects the most suitable LLMs for each input, and an Exploration-Exploitation-Driven Posterior Enhancement (EPE) module, which fosters response diversity and selects high-quality outputs through a hybrid scoring mechanism. Experiments on eight mainstream benchmarks validate the effectiveness of our system: by integrating fifteen open-source LLMs, SMCS outperforms prevailing closed-source LLMs, e.g., GPT-4.1(+5.36%) and GPT-o3-mini(+5.28%) across multiple tasks. Remarkably, it even exceeds the average of best results on different datasets with open-source LLMs (+2.86%), significantly advancing the empirical performance frontier of open-source collaboration. The code is released at https://github.com/magent4aci/SMCS.
arXiv:2605.16157v1 Announce Type: new
Abstract: This paper explores epistemic realizability, a form of realizability in which the property that a piece of data constitutes evidence for a logical proposition is semi-decidable. In this framework, each proposition A is assigned a verifier} program that checks whether a datum X is a realizer for A, and a dual generator program that behaves as a generic realizer for X. We propose epistemic realizability interpretations for minimal logic, second-order intuitionistic logic, and higher-order intuitionistic logic, proving that each system is sound and complete under the proposed semantics.
arXiv:2212.13347v3 Announce Type: replace
Abstract: We report a diffuse Maxwellian illumination scheme for wide-field retinal laser Doppler holography. Inserting an engineered diffuser in the illumination arm transforms a spatially concentrated near-infrared laser focus into an angularly diversified illumination pattern, thereby reducing local irradiance near the anterior segment while preserving coherent interferometric detection. This configuration allows the eyepiece to be positioned closer to the cornea, increasing the digitally reconstructed retinal field of view without producing a localized corneal hot spot. We compare three illumination geometries: focused non-diffuse illumination, diffuse illumination at the same cornea--eyepiece distance, and diffuse Maxwellian illumination. Diffuse Maxwellian illumination expands the retinal field of view while preserving Doppler contrast in broad and high-frequency fluctuation bands. Light-hazard assessment is limited to the current ophthalmic standards ISO 15004-2:2024 and ANSI Z80.36-2021. Based on measured beam profiles, the recommended operating power at 852 nm is set by the most restrictive relevant exposure condition among the assessed anterior-segment, iris, and retinal limits. These results support diffuse illumination as a practical route toward safer, non-mydriatic, wide-field Doppler holography of the human retina.
arXiv:2509.23352v3 Announce Type: replace
Abstract: The integration of Reinforcement Learning (RL) into flow matching models for text-to-image (T2I) generation has driven substantial advances in generation quality. However, these gains often come at the cost of exhaustive exploration and inefficient sampling strategies due to slight variation in the sampling group. Building on this insight, we propose Dynamic-TreeRPO, which implements the sliding-window sampling strategy as a tree-structured search with dynamic noise intensities along depth. We perform GRPO-guided optimization and constrained Stochastic Differential Equation (SDE) sampling within this tree structure. By sharing prefix paths of the tree, our design effectively amortizes the computational overhead of trajectory search. With well-designed noise intensities for each tree layer, Dynamic-TreeRPO can enhance the variation of exploration without any extra computational cost. Furthermore, we seamlessly integrate Supervised Fine-Tuning (SFT) and RL paradigm within Dynamic-TreeRPO to construct our proposed LayerTuning-RL, reformulating the loss function of SFT as a dynamically weighted Progress Reward Model (PRM) rather than a separate pretraining method. By associating this weighted PRM with dynamic-adaptive clipping bounds, the disruption of exploration process in Dynamic-TreeRPO is avoided. Benefiting from the tree-structured sampling and the LayerTuning-RL paradigm, our model dynamically explores a diverse search space along effective directions. Compared to existing baselines, our approach demonstrates significant superiority in terms of semantic consistency, visual fidelity, and human preference alignment on established benchmarks, including HPS-v2.1, PickScore, and ImageReward. In particular, our model outperforms SoTA by $4.9\%$, $5.91\%$, and $8.66\%$ on those benchmarks, respectively, while improving the training efficiency by nearly $50\%$.
arXiv:2605.15695v1 Announce Type: new
Abstract: Fueled by the ability to mine real-world graph data, GNN applications have experienced phenomenal growth. Sparse Matrix-Matrix Multiplication (SpMM) is a critical operator in GNNs. However, existing SpMM designs for GNNs struggle to adapt to diverse input characteristics. In this paper, we first conduct a comprehensive analysis of existing SpMM optimizations, revealing their limitations through statistical and empirical evidence. Based on this analysis, we introduce ParamSpMM, a parametric approach for highly adaptive and efficient SpMM computation in GNNs. It incorporates a new data structure, the Parameterized Compressed Sparse Row (PCSR), to flexibly integrate existing optimization techniques. ParamSpMM enables the configuration of these optimization techniques according to various input characteristics. Furthermore, we complement ParamSpMM with an ML-based SpMM-decider that predicts optimal configurations based on carefully crafted input features. Our evaluations demonstrate that ParamSpMM outperforms Nvidia cuSPARSE with an average speedup of 1.92x, significantly enhancing GNN training efficiency.
arXiv:2605.15372v1 Announce Type: cross
Abstract: We derive an explicit formula for the intrinsic MacWilliams transform for permutation-invariant qudit codes. Such codes naturally live in symmetric power representations, where the relevant error sectors are determined by the irreducible decomposition of the conjugation action on the associated operator space. Using the multiplicity-free structure of this decomposition and the corresponding intertwiner algebra, we identify the intrinsic MacWilliams matrix with a finite Racah transform. The entries are given by a terminating hypergeometric series, and the rows of the matrix are Racah orthogonal polynomials with parameters determined explicitly by the block length and local dimension. Computing the spectrum of the degree-one twirl reveals that this spectrum lies on an affine quadratic lattice. Then we derive a tridiagonal multiplication rule from the representation theory of the adjoint sector. As consequences, we obtain closed-form orthogonality, detailed-balance, and involutivity identities for the transform. The resulting formula supplies an explicit MacWilliams matrix for computing linear programming bounds on permutation-invariant qudit codes.
arXiv:2302.09758v5 Announce Type: replace
Abstract: A recently proposed superconducting linear collider with energy recovery (ERLC) and multiple beam reuse employs twin RF structures to eliminate parasitic collisions in the linacs. Such a collider can operate in either pulsed or continuous-wave (CW) mode, achieving a luminosity of ${\cal O}(10^{36})$ cm$^{-2}$s$^{-1}$ at $2E_0$ = 250--500 GeV. This paper demonstrates that in pulsed mode, the ERLC luminosity is independent of the accelerating gradient for a fixed total power, enabling operation at the highest available gradients. A similar independence holds for the CW mode when the available power significantly exceeds the operational threshold. The luminosity scales with the cavity quality factor as $L\propto Q_0^{1/2}$. We also present, for the first time, a study of a twin $e^-e^-$ ERLC and estimate its performance. This configuration is simpler than the $e^+e^-$ version as it eliminates the need for beam recirculation; electrons can be generated anew for each cycle. In this case, the luminosity scales as $L\propto Q_0^{1/4}$. Furthermore, the use of traveling-wave (TW) RF structures allows for higher gradients and reduced thermal loading. We show that an ERLC with $G$ = 40 MeV/m can operate in CW mode, reaching luminosities of $L_{e^+e^-}$= (1-2.5)$\times 10^{36}$ and $L_{e^-e^-}$= (3-7)$\times 10^{36}$ cm$^{-2}$s$^{-1}$ at $2E_0$ = 250 and 500 GeV, respectively, with a total power consumption of 150-300 MW. These results position the ERLC as a highly promising candidate for a future Higgs factory.
arXiv:2605.15990v1 Announce Type: new
Abstract: Tremendous efforts have been put into evaluating the inclusivity and effectiveness of AI systems across cultures. However, the cultural capabilities considered in much of the literature remain vaguely defined, are referred to using interchangeable terminology, and are typically limited to recalling accurate information about various demographics, regions, and nationalities. To address this construct ambiguity, we draw from Intercultural Communication scholarship and propose a three-level taxonomy of AI-relevant cultural capabilities: Cultural Awareness answers "Does the model know?", Cultural Sensitivity answers "How does it frame its knowledge?", and Cultural Competence answers "Can it adapt as the interaction evolves?". Beyond conceptual clarification, we position this taxonomy as a practical tool for improving the validity and interpretability of AI evaluation in real-world, multicultural settings. Without such construct clarity, evaluation results risk overstating model capabilities and may lead to inappropriate deployment decisions in culturally sensitive contexts.
arXiv:2605.16143v1 Announce Type: new
Abstract: Large language model based agents often fail in unfamiliar environments due to premature exploitation: a tendency to act on prior knowledge before acquiring sufficient environment-specific information. We identify autonomous exploration as a critical yet underexplored capability for building adaptive agents. To formalize and quantify this capability, we introduce Exploration Checkpoint Coverage, a verifiable metric that measures how broadly an agent discovers key states, objects, and affordances. Our systematic evaluation reveals that agents trained with standard task-oriented reinforcement learning consistently exhibit narrow and repetitive behaviors that impede downstream performance. To address this limitation, we develop a training strategy that interleaves task-execution rollouts and exploration rollouts, with each type of rollout optimized by its corresponding verifiable reward. Building on this training strategy, we propose the Explore-then-Act paradigm, which decouples information-gathering from task execution: agents first utilize an interaction budget to acquire grounded environmental knowledge, then leverage it for task resolution. Our results demonstrate that learning to systematically explore is imperative for building generalizable and real-world-ready agents.
arXiv:2605.15411v1 Announce Type: cross
Abstract: We study contextual dynamic pricing in a semiparametric scalar-index valuation model where the latent value is $v_t=\mu_\ast(\mathsf c_t)+\xi_t$, with an unknown utility map $\mu_\ast$ and an unknown additive noise distribution. The key decision object is the one-dimensional oracle price map $u\mapsto p^\ast(u)$ induced by the scalar index $u=\mu_\ast(\mathsf c)$ and the noise tail. Under the $\beta$-H\"older smoothness of the tail function for $\beta\geq 2$ and a revenue-geometry condition that gives a unique, stable, interior maximizer, this oracle map is itself $(\beta-1)$-smooth. We exploit such structure through $\mathsf{ORBIT}$, a modular coarse-to-fine policy that takes a scalar pilot index as input, localizes a benchmark price in each active bin, and learns a local polynomial approximation of the oracle map inside a trust region via bandit convex optimization. For the baseline linear utility model $\mu_\ast(\mathsf c)=\mathsf c^\top\theta_\ast$, an adaptive elliptical exploration scheme constructs the required scalar pilot online without distributional assumptions on the contexts. The resulting policy achieves regret $\widetilde{O}\big(T^{\frac{2\beta-1}{4\beta-3}}+\sqrt{dT}\big)$. For fixed $d$, we establish a matching lower bound in the horizon dependence, unveiling that the nonparametric oracle-map learning term is minimax sharp. The same scalar-pilot interface also yields extensions to sparse high-dimensional linear utility and nonparametric H\"older utility.
arXiv:2505.18511v3 Announce Type: replace
Abstract: Stochastic Partial Differential Equations (SPDEs) driven by random noise play a central role in modeling physical processes with rough spatio-temporal dynamics, such as turbulence flows, superconductors, and quantum dynamics. Although machine learning (ML)-based surrogate models have shown promise for efficiently approximating such dynamics, progress remains limited by the lack of a unified benchmark with controlled data generation and comprehensive evaluation. This gap is particularly significant for singular SPDEs, for which benchmark datasets are largely unavailable and reliable simulation requires numerically delicate schemes based on renormalization. Moreover, subtle differences in data-generation procedures, such as noise approximation, basis choice, and the inclusion of renormalization, can significantly affect the resulting datasets and, consequently, model evaluation. We introduce SPDEBench, the first unified benchmark for ML-based SPDE learning. SPDEBench provides ready-to-use datasets for physically and mathematically significant SPDEs on 1-3D domains with periodic or Dirichlet boundary condition. Both regular and singular SPDEs are taken into consideration. SPDEBench also incorporates representative ML baselines in operator learning, together with 7 evaluation metrics, including Sobolev and distributional metrics beyond the standard $L^2$-error. Supported by SPDEBench, we conduct systematic evaluations of model accuracy, robustness, and out-of-distribution generalization under controlled data variations. Our numerical results show that SPDE-aware architectures generally achieve stronger performance than generic operator-learning baselines. These findings establish SPDEBench as a reproducible and extensible resource, paving pathway for principled benchmarking and architecture design for stochastic spatio-temporal dynamics.
arXiv:2605.16149v1 Announce Type: new
Abstract: Space plasmas are generally characterized by non-Maxwellian distributions with suprathermal populations, as routinely revealed by in situ observations. Such departures from standard Maxwellian distributions can be understood as signatures of quasiequilibrium states, in which the distribution of the medium can be expressed as a continuous superposition of Maxwellian distributions, namely through superstatistics. Here, we construct macroscopic relations linking fluxes to their associated driving forces in such plasmas, where superstatistical effects enter the picture through the transport coefficients. After comparing the resulting superstatistical distributions with observed electron distributions in the solar wind, we turn to the kinetic response of quasiequilibrium plasmas and derive the corresponding transport coefficients, including the electric and thermal conductivities, the mobility, and the diffusion coefficient. We further extend the analysis to viscous plasmas and compute the shear and bulk viscosity coefficients. Overall, quasiequilibrium effects are found to systematically enhance the transport coefficients relative to their Maxwellian values. We quantify this enhancement for the three main universality classes of superstatistics, which are the most commonly encountered in experimental and observational situations, and interpret it as a consequence of the increased population of energetic particles in the non-Maxwellian tails.
arXiv:2503.02597v3 Announce Type: replace
Abstract: Recent Multimodal Large Language Models (MLLMs) have demonstrated significant progress in perceiving and reasoning over multimodal inquiries, ushering in a new research era for foundation models. However, vision-language misalignment in MLLMs has emerged as a critical challenge, where the textual responses generated by these models are not factually aligned with the given text-image inputs. Existing efforts to address vision-language misalignment have focused on developing specialized vision-language connectors or leveraging visual instruction tuning from diverse domains. In this paper, we tackle this issue from a fundamental yet unexplored perspective by revisiting the core architecture of MLLMs. Most MLLMs are typically built on decoder-only LLMs consisting of a causal attention mechanism, which limits the ability of the earlier modalities (e.g., images) to incorporate information from the latter modalities (e.g., text). To address this problem a MLLM that unlocks causal attention into our proposed modality-mutual attention (MMA) to enable image tokens to attend to text tokens. This simple yet effective design allows MMA to achieve state-of-the-art performance in 12 multimodal understanding benchmarks (+6.2% on average across 3 LLMs backbones) without introducing additional parameters. Our MMA design is intended to be generic, allowing for applications across various modalities, and scalable to accommodate diverse multimodal scenarios.
arXiv:2605.16154v1 Announce Type: new
Abstract: Reinforcement learning (RL) allows vision-language-action (VLA) policies to generalize beyond their training distribution by optimizing directly for task success, but post-training is computationally expensive. A natural response has been to speed rollout collection through faster simulators and world models. In GRPO-based VLA RL, we find that the dominant cost lies elsewhere: gradient computation accounts for approximately 78% of wall-clock time per step in our runs, while rollout collection accounts for only 21%. Gradient cost dominates because much of this computation is spent on phases that contribute little to learning. GRPO's learning signal is driven by advantage variance: only phases where successful and failed rollouts diverge produce learning signal. However, GRPO assigns the same advantage to every chunk in a rollout. As a result, actor-update compute is spent uniformly across the trajectory, including phases the policy already handles after pre-training and supervised fine-tuning. This paper presents Probabilistic Chunk Masking (PCM), a drop-in modification to GRPO that allocates gradient computation to a small, probabilistically selected subset of chunks per trajectory. PCM scores semantic phases using success-failure action variance, a rollout-derived proxy for per-phase gradient variance, and samples a fixed chunk budget with online-updated phase-level keep probabilities. We formalize per-phase gradient variance as the quantity determines where gradient computation is useful and show that success-failure action variance provides a measurable proxy for it. PCM requires no reward model or learned critic. On three LIBERO benchmarks, PCM matches the final success rate of standard GRPO while achieving 2.38 times wall-clock speedup, 4.8 times faster gradient updates, and 60% lower peak activation memory, while backpropagating through fewer than 20% of trajectory chunks.
arXiv:2504.09544v3 Announce Type: replace
Abstract: Recent advances in self-supervised deep learning have improved our ability to quantify cellular morphological changes in high-throughput microscopy screens, a process known as morphological profiling. However, most current methods only learn from images, despite many screens being inherently multimodal, as they involve both a chemical or genetic perturbation as well as an image-based readout. We hypothesized that incorporating chemical compound structures during self-supervised pre-training could improve learned representations of images from high-throughput microscopy screens. We introduce a representation learning framework, MICON (Molecular-Image Contrastive Learning), that models chemical compounds as treatments that induce transformations of cell phenotypes. MICON significantly outperforms classical hand-crafted features such as CellProfiler and existing deep-learning-based representation learning methods in challenging evaluation settings where models must identify reproducible effects of drugs across independent replicates and data-generating centers. We demonstrate that incorporating chemical compound information into the learning process provides small, but consistent improvements in performance and that modeling compounds specifically as treatments outperforms approaches that directly align images and compounds in a single representation space. Our findings point to a new direction for representation learning in morphological profiling, suggesting that methods should explicitly account for the multimodal nature of microscopy screening data.
arXiv:2603.29617v2 Announce Type: replace-cross
Abstract: Understanding how the brain processes linguistic constructions is a central challenge in cognitive neuroscience and linguistics. Recent computational studies show that artificial neural language models spontaneously develop differentiated representations of Argument Structure Constructions (ASCs), generating predictions about when and how construction-level information emerges during processing. The present study tests these predictions in human neural activity using electroencephalography (EEG). Ten native English speakers listened to 200 synthetically generated sentences across four construction types (transitive, ditransitive, caused-motion, resultative) while neural responses were recorded. Analyses using time-frequency methods, feature extraction, and machine learning classification revealed construction-specific neural signatures emerging primarily at sentence-final positions, where argument structure becomes fully disambiguated, and most prominently in the alpha band. Pairwise classification showed reliable differentiation, especially between ditransitive and resultative constructions, while other pairs overlapped. Crucially, the temporal emergence and similarity structure of these effects mirror patterns in recurrent and transformer-based language models, where constructional representations arise during integrative processing stages. These findings support the view that linguistic constructions are neurally encoded as distinct form-meaning mappings, in line with Construction Grammar, and suggest convergence between biological and artificial systems on similar representational solutions. More broadly, this convergence is consistent with the idea that learning systems discover stable regions within an underlying representational landscape - recently termed a Platonic representational space - that constrains the emergence of efficient linguistic abstractions.
arXiv:2605.10867v2 Announce Type: replace
Abstract: Continuous authentication in high-stakes digital environments requires datasets with fine-grained behavioral signals under realistic cognitive and motor demands. But current benchmarks are often limited by small scale, unimodal sensing or lack of synchronised environmental context. To address this gap, this paper introduces BEACON (Behavioral Engine for Authentication & Continuous Monitoring), a large-scale multimodal dataset that captures diverse skill tiers in competitive Valorant gameplay. BEACON contains approximately 430 GB of synchronised modality data (461 GB total on-disk including auxiliary Valorant configuration captures) from 79 sessions across 28 distinct players, estimated at 102.51 hours of active gameplay, including high-frequency mouse dynamics, keystroke events, network packet captures, screen recordings, hardware metadata, and in-game configuration context. BEACON leverages the high precision motor skills and high cognitive load that are inherent to tactical shooters, making it a rigorous stress test for the robustness of behavioral biometrics. The dataset allows for the study of continuous authentication, behavioral profiling, user drift and multimodal representation learning in a high-fidelity esports setting. The authors release the dataset and code on Hugging Face and GitHub to create a reproducible benchmark for evaluating next-generation behavioral fingerprinting and security models.
arXiv:2605.15426v1 Announce Type: cross
Abstract: Entanglement in continuous-variable Gaussian systems is a key resource, and common reservoirs can both suppress and generate correlations. Existing work focused on pre-entangled states or Markovian baths, leaving open whether separable squeezed inputs entangle in structured environments or under modulation. We study two bosonic modes coupled to a common reservoir, each initialized in a separable squeezed vacuum. Dynamics are analyzed utilizing Gaussian covariance methods, evolved under approximate Non-Markovian quantum state diffusion (QSD), finite-temperature pseudomode embeddings, and Bures-based non-Markovian diagnostics. We identify three mechanisms absent in Markovian dynamics: (1) A detuning condition that freezes entanglement trajectories across reservoir correlation times; (2) birth, death, and revival of entanglement from orthogonal inputs; and (3) integer-locked beating with square-wave oscillations produced by periodic detuning. All mechanisms persist at finite temperature, with deviations bounded within 5% in cryogenic regimes and 20% at moderate occupations. These deviation bounds align with cryogenic cavity, phononic, and optomechanical platforms, where structured spectral densities and detuning modulation are already accessible. Structured reservoirs are shown to emerge as tunable entanglement resources for continuous-variable quantum technologies.
arXiv:2604.09173v2 Announce Type: replace
Abstract: Managing large-scale vector datasets with disk-resident graph approximate nearest neighbor search (ANNS) systems incurs substantial storage overhead due to the co-location of vector data and auxiliary index metadata, which prevents the storage layer from exploiting their distinct compressibility. We present COMPASS, a component-aware compressed storage framework for disk-resident graph vector search. Leveraging data-index decoupling as a foundation, COMPASS losslessly compresses each component according to its distinct compressibility characteristics, thereby significantly reducing storage space. It further adapts the search and update paths to preserve their performance under compressed storage layouts. Evaluation on real-world public and proprietary billion-scale datasets shows that COMPASS reduces storage space by up to 58.7%, while delivering improved or competitive search and update performance compared to state-of-the-art disk-resident graph ANNS systems.
arXiv:2605.15438v1 Announce Type: cross
Abstract: The fluidic pinball presents a significant benchmark for nonlinear flow control, managing the complex interactions of three cylinder wakes. This study addresses the stabilization of the fluidic pinball to its unstable steady-state solution using a model-based nonlinear feedback strategy. We propose a framework that combines interpolatory model order reduction (IMOR) with the quadratic-quadratic regulator (QQR), a feedback control methodology that is specifically suited to the quadratic nonlinearity of the Navier-Stokes equations. A finite element model (FEM) of the problem coupled with IMOR is used to produce a reduced-order model (ROM) that accurately represents the input-output dynamics of the actuated wake. The performance of the QQR control is evaluated against the traditional linear feedback control for two different Reynolds numbers, $Re_D = 30$ and $Re_D = 50$. At $Re_D = 30$, the QQR controller is able to stabilize the wake and reaches the desired performance criteria 40.1\% faster than using a linear feedback controller. More significantly, at $Re_D = 50$, the QQR controller successfully stabilizes the wake, whereas the linear controller fails to overcome the nonlinearity of the flow. The QQR control effectively suppresses vortex shedding, resulting in the elimination of lift oscillations and a reduction in the drag coefficient. These results demonstrate that the IMOR-QQR framework provides an effective model-based control strategy that can manage nonlinear hydrodynamic instabilities in such complex wake flows.
arXiv:2507.01679v3 Announce Type: replace
Abstract: Existing LLMs-post-training techniques are broadly categorized into supervised fine-tuning (SFT) and reinforcement fine-tuning (RFT). Each paradigm presents a distinct trade-off: (1) SFT excels at mimicking demonstration data, but can lead to problematic generalization as a form of behavior cloning. (2) Conversely, RFT can significantly enhance a model's performance but is prone to learning unexpected behaviors, and its performance is sensitive to the initial policy. In this paper, we propose a unified view of these methods and introduce Prefix-RFT, a hybrid approach that synergizes learning from both demonstration and exploration. Using mathematical reasoning problems as a test bed, we empirically demonstrate that Prefix-RFT is simple yet effective. Not only does it surpass the performance of standalone SFT and RFT, but it also outperforms parallel mixed-policy RFT methods. Our analysis highlights the complementary nature of SFT and RFT, validating that Prefix-RFT effectively harmonizes them. Further ablation studies confirm the method's robustness to variations in the quality and quantity of demonstration data.
arXiv:2603.04299v4 Announce Type: replace
Abstract: Large Language Models (LLMs) often exhibit highly agreeable and reinforcing conversational styles, also known as AI-sycophancy. Although this pattern arises from training objectives that reward user satisfaction over accuracy, it may become problematic when interacting with user prompts that reflect negative social tendencies. Such responses risk amplifying harmful behavior rather than mitigating it. In this study, we examine how LLMs respond to user prompts expressing varying degrees of Dark Triad traits (Machiavellianism, Narcissism, and Psychopathy) using a curated dataset. Our analysis reveals differences across models, whereby all models predominantly exhibit corrective behavior, while showing reinforcing output in certain cases. Model behavior also depends on the severity level and differs in the sentiment of the response. Our findings raise implications for designing safer conversational systems that can detect and respond appropriately when users escalate from benign to harmful requests.
arXiv:2605.16122v1 Announce Type: new
Abstract: Diffusion-based image synthesis has made AI-generated images (AIGI) increasingly photorealistic, raising urgent concerns about authenticity in applications such as misinformation detection, digital forensics, and content moderation. Despite the substantial advances in AIGI detection, how to correct detected AI-generated images with visible artifacts and restore realistic appearance remains largely underexplored. Moreover, few existing work has established the connection between AIGI detection and artifact correction. To fill this gap, we propose GenShield, a unified autoregressive framework that jointly performs explainable AIGI detection and controllable artifact correction in a closed loop from diagnosis to restoration, revealing a mutually reinforcing relationship between these two tasks. We further introduce a Visual Chain-of-Thought based curriculum learning strategy that enables self-explained, multi-step ``diagnose-then-repair'' correction with an explicit stopping criterion. A high-quality dataset with large-scale ``artifact-restored'' pairs is also constructed alongside a unified evaluation pipeline. Extensive experiments on our correction benchmark and mainstream AIGI detection benchmarks demonstrate state-of-the-art performance and strong generalization of our method. The code is available at https://github.com/zhipeixu/GenShield.
arXiv:2510.10454v2 Announce Type: replace
Abstract: Large language models (LLMs) offer a generalizable approach for modeling patient trajectories, but suffer from the long and noisy nature of electronic health records (EHR) data in temporal reasoning. To address these challenges, we introduce Traj-CoA, a multi-agent system involving chain-of-agents for patient trajectory modeling. Traj-CoA employs a chain of worker agents to process EHR data in manageable chunks sequentially, distilling critical events into a shared long-term memory module, EHRMem, to reduce noise and preserve a comprehensive timeline. A final manager agent synthesizes the worker agents' summary and the extracted timeline in EHRMem to make predictions. In a zero-shot one-year lung cancer risk prediction task based on five-year EHR data, Traj-CoA outperforms baselines of four categories. Analysis reveals that Traj-CoA exhibits clinically aligned temporal reasoning, establishing it as a promisingly robust and generalizable approach for modeling complex patient trajectories. Implementation of Traj-CoA is available on https://github.com/zengsihang/Traj-CoA.