arXiv:2605.16486v1 Announce Type: cross
Abstract: Diffusion and flow-based models are ubiquitously used for generative modelling and density estimation. They admit a deterministic probability flow ordinary differential equation (PF-ODE), analogous to continuous normalizing flows (CNFs), which describes the transport of the probability mass. Obtaining the likelihood from these models is of interest to many workflows, especially Bayesian analysis, and requires solving the trace of the Jacobian to compute the divergence of the learned PF-ODE, which is either $\mathcal{O}(D^2)$ to compute exactly or $\mathcal{O}(D)$ with a noisy estimate. We introduce StAD, a new distillation method to predict and learn the divergence of the PF-ODE using the Langevin-Stein operator without ever computing the Jacobian. We show that our method is competitive with the Hutchinson and Hutch++ on CIFAR-10, ImageNet and other density estimation tasks, consistently improving the variance and speed of the likelihood predictions compared to the Hutchinson. We additionally show our method will generalize to a varied class of generative models, and show that under some regularity conditions these learned vector fields can be made to satisfy the Stein class.
Science Journals
arXiv:2504.07347v3 Announce Type: replace-cross
Abstract: As demand for Large Language Models (LLMs) and AI agents grows rapidly, optimizing systems for efficient LLM inference becomes critical. While significant efforts have targeted system-level engineering, little has been explored from a mathematical modeling and queueing perspective. In this paper, we develop the queueing fundamentals for LLM inference. In particular, we study the throughput aspect of LLM inference systems. We prove that a large class of `work-conserving' scheduling algorithms achieve maximum throughput for both individual requests and AI-agent workloads with directed acyclic graph (DAG) and fork-join routing topologies, establishing `work-conserving' as a key design principle for practitioners. Technically, we develop a fluid-limit framework for multi-class batched processing networks under $K$-FCFS scheduling, which may be of independent interest. Evaluations of real-world systems confirm that Orca and Sarathi-Serve are throughput-optimal, reassuring practitioners, while FasterTransformer and vanilla vLLM are not maximally stable and should be used with caution. Our analysis also reveals how constraints such as batch size limits and cyclic routing topologies complicate the throughput picture, pointing to rich open questions at the intersection of queueing theory and LLM system design.
arXiv:2602.05172v2 Announce Type: replace-cross
Abstract: We derive finite-particle rates for the regularized Stein variational gradient descent (R-SVGD) algorithm introduced by He et al. (2024) that corrects the constant-order bias of the SVGD by applying a resolvent-type preconditioner to the kernelized Wasserstein gradient. For the resulting interacting $N$-particle system, we establish explicit non-asymptotic bounds for time-averaged (annealed) empirical measures, illustrating convergence in the \emph{true} (non-kernelized) Fisher information and, under a $\mathrm{W}_1\mathrm{I}$ condition on the target, corresponding $\mathrm{W}_1$ convergence for a large class of smooth kernels. Our analysis covers both continuous- and discrete-time dynamics and yields principled tuning rules for the regularization parameter, step size, and averaging horizon that quantify the trade-off between approximating the Wasserstein gradient flow and controlling finite-particle estimation error.
arXiv:2605.16689v1 Announce Type: cross
Abstract: AI-native 6G visions increasingly invoke wireless foundation models, large multimodal models, and wireless world models as the natural endpoint of AI-native networking, drawing an analogy to recent developments in large language models (LLMs). We argue that this analogy is structurally incomplete. The success of LLMs is based on a broad, reusable, and largely self-contained tokenized data substrate, whereas the wireless domain lacks an equivalent data foundation. Unlike text, code, or images, wireless data such as CSI tensors, IQ samples, or scheduler logs are not self-contained: their meaning is configuration-dependent, simulator-conditioned, task-disaggregated, and weakly grounded in operational feedback, all structural bottlenecks that undermine current pre- and post-training recipes. We therefore argue that monolithic models, including mixture-of-experts (MoE) and wireless world models, are not the most realistic near-term path toward deployable AI-native networks. Instead, emerging evidence points toward composable and agentic network architectures, where general reasoning models orchestrate specialized signal processing models, classical algorithms, digital twins, standards-aware retrieval, and safety checks through explicit programmable interfaces.
arXiv:2605.18463v1 Announce Type: new
Abstract: Advanced regulatory control (ARC), also known as advanced PID architectures, is a simple and robust way of controlling processes with changing and possibly conflicting constraints, where it previously was believed - at least in academia - that model-based solutions, such as MPC, were the only effective solution. To illustrate this, ARC is applied in two case studies. The first is a gas-liquid separation process, in which selectors and split-parallel control are combined to achieve bidirectional inventory control in which the throughput manipulator moves automatically to the most optimal position. The second case study is on keeping acceptable air quality (CO2-level) and temperature in a room (in this case, a barn for cows). The CO2 and temperature constraints can be conflicting, leading to a hierarchical switching network of PID controllers.
Note: this is an extended version (with simulations) of paper at IFAC World Congress, August 2026, Korea.
arXiv:2605.16693v1 Announce Type: cross
Abstract: Many biological processes, from cell division to viral lysis, are triggered when an internal stochastic variable reaches a threshold. Here we introduce Branching under First-Passage Resetting, a general framework in which replication events arise endogenously from first-passage dynamics rather than from externally imposed lifetime clocks. We show that the resulting population dynamics obey an exact renewal equation linking single-trajectory first-passage statistics to the population growth rate. This mapping shows that, for fixed offspring number and fixed mean replication time, stochastic timing fluctuations necessarily enhance growth relative to a deterministic clock. When offspring yield depends on the first-passage time, however, fluctuations have non-trivial effects and expose a fundamental yield-delay trade-off: waiting longer can increase the number of descendants, but delays all future lineages. Our framework allows us to address this optimization problem analytically, and upon application to bacteriophage lysis, gives an optimal lysis time and growth rate consistent with empirical data.
arXiv:2605.17510v1 Announce Type: cross
Abstract: We study group decision-making in artificial societies where the rules of play are themselves subject to collective amendment. Using the self-amending game Nomic, we compare multiple scales across two LLM families and find that collective adaptation does not improve monotonically with model size. Instead, both families exhibit a narrow mid-scale regime that supports sustained rule adoption, diverse amendments, and balanced consensus. Smaller models tend to remain rule-inert, whereas larger models often converge on restrictive voting patterns, and heterogeneous mixed-size groups collapse into veto-driven gridlock. These cross-scale contrasts persist under temperature perturbations and under a shift from unanimity to majority voting, although latent-state structure varies by family and scale. Hidden-state divergence alone does not explain collective performance: high representational divergence can coincide with poor behavioural outcomes. Linear probes reveal regime-selective coupling between latent vote-predictive signals and collective behaviour, but decodability is necessary rather than sufficient for adaptive play. Overall, the recurring regularity is non-monotonicity, not the particular scale at which the optimum appears. Self-amending games therefore provide a controlled testbed for studying collective adaptation in artificial societies beyond raw model scale.
arXiv:2605.18162v1 Announce Type: new
Abstract: Vision-Language Models (VLMs) have made striking progress, yet their spatial reasoning remains fragile: models that answer an original input correctly can still fail under paired transformations with predictable answer mappings, revealing a gap between instance-level correctness and robust spatial reasoning. To address this, we propose Spatial Alignment via Geometric Evolution (SAGE), a self-evolving framework that enforces logical consistency in VLMs through geometric and linguistic duality operations. SAGE incorporates duality consistency as an auxiliary reward within GRPO training, encouraging models to produce logically coherent answers across original and transformed inputs. A dynamic operation pool continuously probes for inconsistencies, promoting challenging operations and retiring mastered ones, so that training focuses on the most informative signals. SAGE is model-agnostic, data-efficient compared to prior GRPO methods, and can be applied as a lightweight post-training stage to any existing VLM. Experiments on video and spatial reasoning benchmarks demonstrate consistent improvements over strong baselines and enhanced generalization to unseen data.
arXiv:2605.17546v1 Announce Type: cross
Abstract: Understanding galaxy morphology evolution across cosmic time requires models that can generate realistic galaxy populations conditioned on redshift. In this work, we study efficient redshift-conditioned generative modeling for astrophysical image synthesis using diffusion models and pixel-MeanFlow. We first review the connections between score-based diffusion models, Flow Matching, one-step generative models, and modern diffusion samplers. We then evaluate DDPM, DDIM, DEIS-AB2, DPM++2M, and one-step pixel-MeanFlow on the GalaxiesML-64 dataset using morphology-based metrics, including ellipticity, semi-major axis, S\'ersic index, and isophotal area. Our results show a clear accuracy-efficiency trade-off: standard DDPM sampling achieves the best distributional fidelity but requires high computational cost, while second-order samplers substantially improve efficiency over DDIM. Pixel-MeanFlow enables single-step generation and achieves competitive performance on several morphology statistics, though it remains weaker than many-step DDPM for fine-grained structure. Our results demonstrate that one-step generative models can recover key galaxy morphology statistics at orders-of-magnitude lower computational cost, opening a path toward efficient conditional simulators for large cosmological surveys and simulation-based scientific inference.
arXiv:2605.17705v1 Announce Type: cross
Abstract: Panel data, in which multiple units are repeatedly observed over time, arise throughout science and engineering. Quantifying predictive uncertainty in such settings is challenging because conformal prediction, while distribution-free and model-agnostic, classically relies on exchangeability assumptions that fail under temporal dependence and unit heterogeneity. We propose a simple online conformal framework for non-exchangeable panel data. The method exploits a key feature of online panel prediction: when a forecast is required for one unit, contemporaneous outcomes from related units may already be observed and can serve as a calibration panel. At each round, prediction sets are formed using currently observed calibration units together with two adaptive quantities: history-based similarity weights that emphasize calibration units resembling the target, and an adaptive miscoverage level that is updated whenever target feedback is revealed. This two-state design yields a stepwise coverage bound and a long-run coverage guarantee. Empirically, across synthetic and real panel data sets, the method improves coverage on the worst-covered target units through adaptive interval-width allocation rather than uniform inflation. The two states are complementary: similarity weights protect coverage when target feedback is sparse, while the adaptive level further improves coverage as feedback accumulates.
arXiv:2605.18177v1 Announce Type: new
Abstract: Query-based Vision Transformer segmentation models typically reconstruct dense spatial feature maps to predict masks, inheriting design patterns from convolutional architectures. We show that this explicit image-space reconstruction is not required. We introduce TokenMask, a token-space mask head that computes mask logits directly from query-token affinities and performs interpolation in logit space rather than feature space. This reformulation preserves the original linear scoring mechanism while simplifying the computational structure. Across diverse ViT backbones, datasets and segmentation tasks, TokenMask consistently improves efficiency over prior approaches by reducing computational and memory requirements while maintaining competitive accuracy, leading to tangible speedups on NVIDIA Jetson AGX Orin using TensorRT FP16 inference. Overall, TokenMask yields a simpler and more deployment-friendly design for embedded vision systems.
arXiv:2601.08013v2 Announce Type: replace
Abstract: Accurate forecasts of segment-level sailing durations are fundamental to enhancing maritime schedule reliability and optimizing long-term port operations. However, conventional estimated time of arrival (ETA) models are primarily designed for the immediate next port of call and rely heavily on real-time automatic identification system (AIS) data, which is inherently unavailable for future voyage segments. To address this gap, the study reformulates future-port ETA prediction as a segment-level time-series forecasting problem. We develop a transformer-based architecture that integrates historical sailing durations, destination port congestion proxies, and static vessel descriptors. The proposed framework employs a causally masked attention mechanism to capture long-range temporal dependencies and a multi-task learning head to jointly predict segment sailing durations and port congestion states, leveraging shared latent signals to mitigate high uncertainty. Evaluation on a real-world global dataset from 2021 demonstrates the proposed model consistently outperforms a comprehensive suite of competitive baselines. The result shows a relative reduction of 4.70% in mean absolute error (MAE), 4.95% in mean absolute percentage error (MAPE) and 2.59% in root mean squared error (RMSE) compared with sequential deep learning models. The relative reductions compared with gradient boosting machines are 7.03% in MAE, 39.49% in MAPE and 4.37% in RMSE. The case study conducted on one major destination port further illustrates the model's superior accuracy.
arXiv:2605.17944v1 Announce Type: cross
Abstract: Rapid advancements in cloud based platforms providing access to quantum computing capabilities have opened up several challenges for efficient usage of these highly delicate and costly devices. Although most of the current systems use a priority based access protocol, they are unable to fully support reliable, efficient, and scalable execution of larger-scale applications. To overcome this limitation, we propose a comprehensive solution for efficient allocation of quantum programs to appropriate quantum devices, considering all the relevant cost metrics into account including, fidelity, execution time and communication overhead. We also formulate use-cases for distributed quantum workflow and propose modified graph based algorithms to solve for allocation of such use-cases, assuming a hybrid classical-quantum network. Since hardware advancements in large standalone devices is an ongoing process, it is critical to investigate such distributed workflows to maximize the best utilization of current NISQ devices. Our empirical study shows that the proposed techniques perform better than state-of-the-art methods for almost all evaluation parameters, with average improvements of approximately $5\%$ in execution time, $30\%$ in communication overhead, $40\%$ in wait time and $2\%$ in fidelity, providing better solutions to efficient allocation strategies.
arXiv:2605.18204v1 Announce Type: cross
Abstract: Discrete diffusion models are a powerful class of generative models with strong performance across many domains. For efficiency, however, discrete diffusion typically parameterizes the generative (reverse) process with factorized distributions, which makes it difficult for the model to learn the target process in a small number of steps and necessitates a long, computationally expensive sampling procedure. To reduce the gap between the target and model distributions and enable few-step generation, we propose Forward-Learned Discrete Diffusion (FLDD), which introduces discrete diffusion with a learnable forward (noising) process. Rather than fixing a Markovian forward chain, we adopt a non-Markovian formulation with learnable marginal and posterior distributions. This allows the generative process to remain factorized while matching the target defined by the noising process. We train all parameters end-to-end under the standard variational objective. Experiments on various benchmarks show that, for a given number of sampling steps, our approach produces a higher quality samples than conventional discrete diffusion models using the same reverse parameterization.
arXiv:2605.18625v1 Announce Type: cross
Abstract: We study the near-kernel sector of the Hamiltonian constraint operator in the one-vertex model of quantum reduced loop gravity using variational Monte Carlo methods with neural quantum states. The analysis is based on the symmetric Hamiltonian containing both Euclidean and Lorentzian contributions, and on the variational minimization of the positive quadratic operator $\hat{\mathcal Q}=\hat C \hat C^\dagger$ in truncated Hilbert spaces with spin cutoff up to $j_{\mathrm{max}}=1001$. The resulting near-kernel states are found to organize into three qualitatively distinct classes. At low cutoffs, we find solutions that do not factorize across the three edge degrees of freedom. At larger cutoffs, we find two different factorized branches, both described to very high accuracy by products of one-edge wavefunctions but localized in different spin regimes. One of these branches is matched with near-unit fidelity by reduced Thiemann coherent states, providing evidence for an emergent semiclassical organization of the near-kernel sector. The other is likewise strongly factorized, but its one-edge factors are not well described by the same coherent-state family.
arXiv:2402.12705v5 Announce Type: replace
Abstract: Reconfiguration problems ask whether one feasible solution can be transformed into another by a sequence of local moves while maintaining feasibility throughout. For integers $d \geq 1$ and $k \geq d+1$, the Distance Coloring problem asks if a given graph $G$ has a $(d, k)$-coloring, i.e., a coloring of the vertices of $G$ by $k$ colors such that any two vertices within distance $d$ from each other have different colors.
For ordinary proper colorings ($d=1$), the $k$-Coloring Reconfiguration problem is polynomial-time solvable for $k\le 3$ [Cereceda, van den Heuvel, and Johnson, J. Graph Theory 67(1):69--82, 2011] but is $\mathsf{PSPACE}$-complete for every fixed $k\ge 4$, even on bipartite graphs [Bonsma and Cereceda, Theor. Comput. Sci. 410(50):5215--5226, 2009]. In this work, we initiate a study of the distance-$d$ analogue, for $d \geq 2$. We show that even for planar, bipartite, and $2$-degenerate graphs, $(d, k)$-Coloring Reconfiguration remains $\mathsf{PSPACE}$-complete for every $d \geq 3$ via a reduction from the well-known Sliding Tokens problem. Our construction uses $k = k_0 + 2 + n(\lceil d/2\rceil-1)$ colors on instances of size $n$, where $k_0\in\{3d+3,3d+6\}$ (depending on the parity of $d$). For $d = 2$, the same reduction scheme can be adapted to show that the problem is $\mathsf{PSPACE}$-complete on planar and $2$-degenerate graphs with same values of $k$. Additionally, on split graphs, there is an interesting dichotomy: the problem is $\mathsf{PSPACE}$-complete when $d = 2$ and $k$ is large but can be solved efficiently when $d \geq 3$ and $k \geq d+1$. For chordal graphs, we show that the problem is $\mathsf{PSPACE}$-complete for even values of $d \geq 2$. Finally, we design a quadratic-time algorithm to solve the problem on paths for any $d \geq 2$ and $k \geq d+1$.
arXiv:2411.10636v3 Announce Type: replace
Abstract: In this study, we investigate extrinsic gender bias in Bangla pretrained language models, a largely underexplored area in low-resource languages. To assess this bias, we construct four manually annotated, task-specific benchmark datasets for sentiment analysis, toxicity detection, hate speech detection, and sarcasm detection. Each dataset is augmented using nuanced gender perturbations, where we systematically swap gendered names and terms while preserving semantic content, enabling minimal-pair evaluation of gender-driven prediction shifts. We then propose RandSymKL, a randomized debiasing strategy integrated with symmetric KL divergence and cross-entropy loss to mitigate the bias across task-specific pretrained models. RandSymKL is a refined training approach to integrate these elements in a unified way for extrinsic gender bias mitigation focused on classification tasks. Our approach was evaluated against existing bias mitigation methods, with results showing that our technique not only effectively reduces bias but also maintains competitive accuracy compared to other baseline approaches. To promote further research, we have made both our implementation and datasets publicly available: https://github.com/sajib-kumar/Mitigating-Bangla-Extrinsic-Gender-Bias
arXiv:2602.07730v2 Announce Type: replace
Abstract: Across scientific disciplines, Laplacian eigenvectors serve as a fundamental basis for simplifying complex systems, from signal processing to quantum mechanics. In reinforcement learning (RL), they similarly form a basis over the state space, enabling reward functions to be approximated by projection onto a small set of eigenvectors. This projection makes zero-shot control possible, but it also imposes a fundamental limitation: the induced policies are only as expressive as the linear span of the chosen eigenvectors. We introduce the Laplacian Keyboard (LK), a hierarchical framework that goes beyond this linear span. LK constructs a task-agnostic library of behaviors from these eigenvectors, forming a behavior basis guaranteed to contain the optimal policy for any reward within the linear span. A meta-policy learns to stitch these behaviors dynamically, enabling efficient learning of policies outside the original linear constraints. We establish theoretical bounds on zero-shot approximation error and demonstrate empirically that LK improves over the zero-shot solution while achieving better sample efficiency compared to standard RL methods.
arXiv:2605.17499v1 Announce Type: new
Abstract: Multimodal deep neural networks enhance deep comprehension by integrating diverse data modalities. Data from different modalities are typically projected into a shared latent space for similarity computation, but this process is resource intensive due to large image encoders and equal processing of test data during prediction. Early exit methods reduce computational load by utilizing intermediate layers, saving time and memory. However, developing such methods is challenging for multimodal data like image-text pairs. This study investigates the semantic content distributions present in intermediate layers of encoders such as CLIP, which can be derived from textual descriptions. We introduce Text-Guided Exit Modules (T-GEMs) and a rate-based regularizer to control encoder usage costs while maintaining cross-modal understanding performance.
arXiv:2605.18591v1 Announce Type: new
Abstract: Natural policy gradients improve optimization by accounting for the geometry of distribution space, but their practical use is limited by the cost of estimating and inverting the Fisher matrix. We present Randomized Advantage Transformation (RAT), a method for estimating Tikhonov-regularized natural policy gradients via direct backpropagation. By applying the Woodbury formula, we reformulate the regularized natural policy gradients as vanilla policy gradients with a transformed advantage. RAT computes this transformation efficiently via randomized block Kaczmarz iterations on on-policy mini-batches, avoiding explicit Fisher construction, conjugate-gradient solvers, and architecture-specific approximations. We provide convergence guarantees for RAT and demonstrate empirically that it matches or exceeds established natural-gradient methods across continuous and visual control benchmarks, while remaining simple to implement and compatible with various architectures.
arXiv:2605.17691v1 Announce Type: new
Abstract: Automating the classification of negative treatment in legal precedent is a critical yet nuanced NLP task where misclassification carries significant risk. To address the shortcomings of standard accuracy, this paper introduces a more robust evaluation framework. We benchmark modern Large Language Models on a new, expert-annotated dataset of 239 real-world legal citations and propose a novel Average Severity Error metric to better measure the practical impact of classification errors. Our experiments reveal a performance split. Google's Gemini 2.5 Flash achieved the highest accuracy on a high-level classification task (79.1%), while OpenAI's GPT-5-mini was the top performer on the more complex fine-grained schema (67.7%). This work establishes a crucial baseline, provides a new context-rich dataset, and introduces an evaluation metric tailored to the demands of this complex legal reasoning task.
arXiv:2505.09203v2 Announce Type: replace-cross
Abstract: Developing inverse design methods for functional materials with specific properties is critical to advancing fields like renewable energy, catalysis, energy storage, and carbon capture. Generative models based on diffusion principles can directly produce new materials that meet performance constraints, thereby significantly accelerating the material design process. However, existing methods for generating and predicting crystal structures often remain limited by low success rates. In this work, we propose a novel inverse material design generative framework called InvDesFlow-AL, which is based on active learning strategies. This framework can iteratively optimize the material generation process to gradually guide it towards desired performance characteristics. In terms of crystal structure prediction, the InvDesFlow-AL model achieves an RMSE of 0.0423 {\AA}, representing an 32.96% improvement in performance compared to exsisting generative models. Additionally, InvDesFlow-AL has been successfully validated in the design of low-formation-energy and low-Ehull materials. It can systematically generate materials with progressively lower formation energies while continuously expanding the exploration across diverse chemical spaces. These results fully demonstrate the effectiveness of the proposed active learning-driven generative model in accelerating material discovery and inverse design. To further prove the effectiveness of this method, we took the search for BCS superconductors under ambient pressure as an example explored by InvDesFlow-AL. As a result, we successfully identified Li\(_2\)AuH\(_6\) as a conventional BCS superconductor with an ultra-high transition temperature of 140 K. This discovery provides strong empirical support for the application of inverse design in materials science.
arXiv:2506.04105v2 Announce Type: replace-cross
Abstract: We consider a network of users connected by pairwise quantum key distribution (QKD) links. Using these pairwise secret keys and public classical communication, the users want to generate a common (conference) secret key at the maximal rate. We propose an algorithm based on spanning-tree packing (a known problem in graph theory) and prove its optimality. This algorithm enables optimal conference key generation in modern quantum networks of arbitrary topology. Additionally, we discuss how it can guide the optimal placement of new bipartite links in the network design.
arXiv:2509.13045v4 Announce Type: replace-cross
Abstract: Boron-doped diamond crystals (BDD, C$_{1-x}$B$_{x}$) exhibit exceptional mechanical strength, electronic tunability, and resistance to radiation damage. This makes them promising materials for use in gamma-ray crystal-based light sources. To better understand and quantify the structural distortions introduced by doping, which are critical for maintaining channelling efficiency, we perform atomistic-level molecular dynamics simulations on periodic C$_{1-x}$B$_{x}$ systems of various sizes. These simulations allow the influence of boron concentration on the lattice constant and the (110) and (100) inter-planar distances to be evaluated over the concentration range from pure diamond (0%) to 5% boron at room temperature (300 K). Linear relationships between both lattice constant and inter-planar distance with increasing dopant concentration are observed, with a deviation from Vegard's Law. This deviation is larger than that reported by other theoretical and computational studies; however, this may be attributed to an enhanced crystal quality over these studies, a vital aspect when considering gamma-ray crystal light source design. The methodology presented here incorporates several refinements to closely reflect the conditions of microwave plasma chemical vapour deposition (MPCVD) crystal growth. Validation of the methodology is provided through a comprehensive statistical analysis of the structure of our generated crystals. These results enable reliable atomistic modelling of doped diamond crystals and support their use in the design and fabrication of periodically bent structures for next-generation gamma-ray light source technologies.
arXiv:2602.18895v2 Announce Type: replace-cross
Abstract: Large language models (LLMs) have shown promise in translating model-based explanations into human-readable narratives. This study evaluates whether LLMs can serve as post-hoc explainability interfaces for credit risk models, focusing on their ability to preserve feature-importance rankings and generate autonomous explanations. Using a LendingClub dataset, we compare LLM outputs with SHAP and coefficient-based attributions on three major LLMs, including GPT-4-turbo, Claude-Sonnet-4.5, and Gemini-2.5-Flash. Results indicate that LLMs reliably reproduce reference rankings under controlled prompts but show limited alignment when generating explanations autonomously. These findings suggest that LLMs are best deployed as narrative interfaces rather than substitutes for formal attribution methods in credit risk governance.