arXiv:2506.23287v3 Announce Type: replace
Abstract: In single-cell research, tracing and analyzing high-throughput single-cell differentiation trajectories is crucial for understanding biological processes. Key to this is the robust modeling of hierarchical structures that govern cellular development. Traditional methods face limitations in computational cost, performance, and stability. VAE-based approaches have made strides but still require branch-specific network modules, limiting their scalability and stability, while often suffering from posterior collapse. To overcome these challenges, we introduce HDTree, a generative modeling framework designed for robust lineage inference. HDTree captures tree relationships within a hierarchical latent space using a unified hierarchical codebook and employs a quantized diffusion process to model continuous cell state transitions. By aligning the generative process with the Waddington landscape, this method not only improves stability and scalability but also enhances the biological plausibility of inferred lineages. HDTree's effectiveness is demonstrated through comparisons on both general-purpose and single-cell datasets, where it outperforms existing methods in lineage inference accuracy, reconstruction quality, and hierarchical consistency. These contributions enable accurate and efficient modeling of cellular differentiation paths, offering reliable insights for biological discovery.\footnote{Code is available at https://github.com/zangzelin/code\_HDTree\_icml.
Science Journals
arXiv:2605.17614v1 Announce Type: new
Abstract: Brownian rods disperse in pressure-driven flow through a coupling between axial shear, anisotropic translational diffusion and Jeffery--Brownian rotation. Classical tube Taylor--Aris theory treats transverse mixing as a scalar process, and existing passive-rod reductions have mainly addressed planar geometries. A circular tube adds two ingredients: the shear strength varies with radius and freely rotating rods sample a three-dimensional orientation space. We formulate a tensorial Taylor--Aris theory for dilute axisymmetric rods in Poiseuille flow by solving the local steady orientation Fokker--Planck problem and using its second moments to close a conservative axisymmetric transport equation. The long-wave reduction shows how each part of the diffusion tensor enters the one-dimensional limit. The radial diffusivity sets the invariant cross-sectional measure and the cell problem for the leading Taylor coefficient; the radial--axial component produces an inverse-P{\'e}clet correction to the migration speed; the axial component gives the direct diffusivity. The central mechanism is the streamwise alignment generated in high-shear annular layers. Alignment reduces radial diffusivity there, shifts the long-time sampling of the velocity profile toward slower streamlines, and amplifies the radial cell response. In strong shear this raises the Taylor coefficient by about \(23\%\) for aspect ratio \(p=1000\) and by about \(30\%\) in the infinitely slender limit, approaching the fully aligned bound. Direct simulations of the full tensorial equation validate the asymptotic coefficients. The same radial mixing operator also gives a Sturm--Liouville spectral model that tracks finite-time relaxation from different radial injections to the long-time Taylor regime.
arXiv:2605.17924v1 Announce Type: new
Abstract: Electronic waste (e-waste) is one of the fastest-growing waste streams worldwide due to rapid technological advancements and shorter device lifespans. Improper disposal releases hazardous substances that harm the environment and human health, while valuable materials such as gold, copper, and aluminum are lost if not recycled. In 2022, approximately 62 million metric tonnes of e-waste were generated globally, but only about 22% was formally recycled. India generated around 1.751 million metric tonnes in 2023-24, with only 43% processed through authorized channels. Green Grid is a full-stack web-based platform designed to simplify and encourage e-waste recycling through an E-Dumper Locator, Green Rewards System, Insights and Awareness Hub, Scheduled Pickup Service, Recycling Impact Calculator, Eco AI Assistant, and Eco-Marketplace. Developed using React.js, Node.js, Express.js, SQL, Google Maps API, and JWT authentication, the platform transforms e-waste recycling into a transparent, educational, and rewarding process. By combining technology, awareness, and incentives, Green Grid promotes responsible disposal and supports circular economy practices for a more sustainable future.
arXiv:2605.17319v1 Announce Type: new
Abstract: Inspired by cephalopod jet propulsion through compliant funnels, this study investigates elastic wave propagation and energy exchange in passively deforming cylindrical nozzles through three-dimensional, two-way fluid-structure interaction simulations. Flexible nozzles with varying stiffness ($Eh = 75 - 500~\mathrm{N\,m^{-1}}$, where $E$ and $h$ are Young's modulus and nozzle thickness, respectively) are subjected to a pulsatile jet inflow at $Re \sim 4000$. Increasing nozzle flexibility reduces the deformation-wave speed in accordance with Moens-Korteweg scaling, thereby prolonging the nozzle expansion phase. This delayed expansion enhances jet entrainment and elastic energy storage while suppressing early shear-layer roll-up and vortex formation. During contraction, the stored elastic energy is released, thereby enhancing jet acceleration and vortex formation. For the most flexible nozzle, the primary vortex-ring circulation increases by 52.13%, the vortex convection distance by 9.00%, and the peak outlet kinetic energy flux by a factor of 4.62 compared with a rigid nozzle. These effects collectively yield a 61.92% increase in total hydrodynamic impulse. These findings identify passive wave-speed tuning via nozzle compliance as a mechanism to enhance pulsed-jet thrust for bio-inspired underwater propulsion.
arXiv:2605.17938v1 Announce Type: new
Abstract: Training data attribution (TDA) should enable generative model interpretability and foster a variety of related downstream tasks. Nonetheless, current TDA approaches lack reliability and robustness, preventing their adoption in real-world setups. In this paper, we take a decisive step towards more reliable and robust TDA for diffusion models. We propose to perform TDA with mirrored unlearning and noise-consistent skew (MUCS). The idea is to fine-tune a second model with bounded mirrored gradient ascent, and to measure the normalized skew of this model with respect to the original one using consistent noise samples. We show that, while being conceptually simple and generic, MUCS systematically outperforms existing methods on three different datasets by a large margin. We additionally study the effect that core design choices have on final performance, and analyze novel aspects regarding the overlap of influential instances across generated items and the potential of ensembling TDA approaches. We believe that our findings may have broader implications for more general unlearning setups, as well as for tasks requiring the comparison of diffusion losses.
arXiv:2605.18026v1 Announce Type: new
Abstract: Roundabouts, characterized by frequent merging and yielding interactions, remain a safety-critical corner case for the development and testing of intelligent driving functions. However, extracting sufficient near-critical scenarios from naturalistic data is inefficient. Most existing scenario generation methods provide limited controllability over interaction intensity and criticality, making systematic safety testing and detailed analysis difficult. This paper presents an interaction-aware roundabout scenario generator with continuously adjustable interaction intensity. Geometric routes and temporal progress profiles are first decoupled and mapped to latent codes using pretrained autoencoders. Conditional latent generation is then performed with Wasserstein Generative Adversarial Networks (WGAN) to generate scenarios. Yielding is modeled as a controllable timing intervention via a compact yield code during the approach-to-entry segment, where interaction intensity is modulated by scaling the code with a factor $\lambda$. Results demonstrate enhanced timing-latent fidelity and plausible interaction responses compared to a baseline model. Under criticality-calibrated scaling, increasing $\lambda$ expands the safety margin, providing a scalable and controlled testing mechanism.
arXiv:2605.18028v1 Announce Type: new
Abstract: Federated fine-tuning of Large Language Models faces severe statistical heterogeneity. However, existing model-level defenses often overlook the root cause: intrinsic data distribution mismatches. In this work, we first establish Federated Self-Distillation (FedSD) as a fundamental and potent strategy. By projecting client representations into a smoothed ``model-understanding space,'' FedSD alone serves as a universal booster, demonstrating superior performance over conventional algorithms. Despite its success, we identify a subtle trade-off termed the Rewrite Paradox -- unconstrained self-distillation can inadvertently increase hallucinations and redundancy. To refine this paradigm, we further propose FedSDR (Federated Self-Distillation with Rectification), the ultimate reinforced framework. It augments FedSD with a dual-stream mechanism: a local LoRA-S (Smoothing) branch to implicitly absorb heterogeneity via distilled data, and a parallel global LoRA-R (Rectification) branch anchored to raw data to enforce factual correctness. By selectively aggregating only LoRA-R, FedSDR yields a globally aligned and faithful model. Extensive experiments verify its superior performance.
arXiv:2605.17607v1 Announce Type: new
Abstract: Autonomous pricing agents are widely deployed in online marketplaces, making algorithmic pricing a prominent application of multi-agent learning. Experimental studies often report collusive outcomes, but these findings typically rely on Q-learning in complete-information environments and lack rigorous convergence guarantees. In this paper, we study the stochastic learning dynamics of Regularized Robbins-Monro (RRM) algorithms in a Bayesian Bertrand competition with private costs. We show that this setting violates standard stability conditions, including monotonicity and the Minty variational inequality, rendering classical convergence results for gradient-based learning inapplicable. Despite this, we prove that Euclidean RRM algorithms converge almost surely to the unique, efficient Bayes-Nash equilibrium within a finite-dimensional approximation of the strategy space. By analyzing symmetric piecewise-linear pricing strategies in a duopoly, we explicitly construct a global Lyapunov function for the projected primal dynamics and establish global asymptotic stability of the equilibrium. Our analysis yields rigorous convergence guarantees for stochastic first-order learning algorithms in Bayesian Bertrand competition and provides a principled counterpoint to widespread claims of algorithmic collusion.
arXiv:2605.17812v1 Announce Type: new
Abstract: Vertical AI firms in accounting, law, healthcare, procurement, and similar domains historically bundled workflow, domain logic, and accountability into a single application. General-purpose AI agents are now unbundling that package, prompting founders and investors to advocate "going headless": cede the workflow and interface to agents and expose domain expertise as callable services. This article argues that going headless is correct for some firms and destructive for others, and that the latter often cede their value capture inadvertently through architectural choices that look like interface decisions. This is a boundary question, and the answer turns on distinguishing the interface boundary, which can often move, from the accountability boundary, which often must not. Drawing on Coase's theory of the firm, Eisenmann, Parker, and Van Alstyne's platform envelopment framework, and Teece's analysis of complementary assets and appropriability, the article shows that orchestrators operating through open protocols acquire envelopment power even as technical interoperability improves, and that durable value capture concentrates in cospecialized accountability assets: professional signoff, regulated workflows, evidence trails, and trusted systems of record. The article proposes a three-position taxonomy (component, integrated software platform, dual-track) determined not by sector but by task-accountability regime, and formalizes the construct of rule debt: the future governance, maintenance, and accountability burden that accrues to customer organizations when business rules and professional standards migrate from governed systems into prompts and agent instructions. Four principles follow: decompose by accountability not interface, invert the edges while retaining the core, position rule debt as the customer cost the integrated platform prevents, and avoid single-orchestrator dependence.
arXiv:2605.17930v1 Announce Type: new
Abstract: While the approximation properties of single-layer Transformer architectures have been studied in recent works, a rigorous theoretical understanding of the multi-layer setting remains limited. In this work, we establish that multi-layer Transformers possess fundamentally different approximation capabilities from single-layer ones: for certain retrieval tasks, any single-layer Transformer requires least $\Omega (\varepsilon^{-k})$ parameters to achieve precision $\varepsilon$, where $k$ grows linearly with sequence length $T$, whereas a two-layer Transformer with a single head per layer achieves the same approximation precision with at most $O (\varepsilon^{-1})$ parameters. To understand this separation, we identify two structural mechanisms underlying multi-layer approximation. Specifically, softmax attention can only efficiently retrieve the token attaining the maximum attention score, incurring exponential-in-length parameter cost for $k$-th largest retrieval with $k \geq 2$. Moreover, the parameter cost of decoding coupled information scales with the size of the retrieved token set. Motivated by these findings, we propose InfoFlow, a framework for multi-layer Transformers. The framework tracks an information set of accessible input positions at each token and layer, assigning an explicit approximation rate to each mode of information propagation. This abstraction recovers known approximation bounds, remains consistent with experimental observations on trained networks, and yields concrete predictions in settings where direct theoretical analysis is currently intractable. Our results provide a principled framework for reasoning about the approximation efficiency of multi-layer Transformers.
arXiv:2605.18570v1 Announce Type: new
Abstract: Cross-domain knowledge alignment is essential for integrating heterogeneous medical systems, yet existing approaches typically treat entity alignment as a static matching problem, ignoring query context and cross-system asymmetry. This limitation is particularly critical in integrative medical settings, where correspondence between concepts is inherently context-dependent, non-bijective, and direction-sensitive.
In this paper, we propose Query-Conditioned Entity Alignment (QCEA), which reformulates entity alignment as a query-conditioned correspondence problem. Instead of learning a fixed mapping between entity representations, QCEA treats the textual description of a source entity as a query and ranks candidate entities in the target graph, enabling context-dependent alignment. The framework integrates semantic encoding, graph-based representation learning, and a direction-aware transformation module to capture asymmetric and many-to-many correspondence across heterogeneous knowledge systems.
We evaluate QCEA on TCM--WM knowledge graphs derived from SymMap, covering both symptom alignment and herb--molecule alignment tasks. Experimental results show consistent improvements over representative baselines, particularly on rank-sensitive metrics such as Hit@K and MRR. Furthermore, downstream retrieval-augmented generation (RAG) experiments demonstrate that improved alignment leads to better evidence retrieval, stronger grounding, and higher answer accuracy. These findings highlight that alignment is not merely a data integration step, but a key factor that shapes knowledge accessibility and reliability in cross-system medical reasoning.
arXiv:2605.17936v1 Announce Type: new
Abstract: Recent works have illustrated that modern NLP models trained for diverse tasks ranging from sentiment analysis to language generation succumb to universal adversarial attacks, a class of input-agnostic attacks where a common trigger sequence is used to attack the model. Although these attacks are successful, the triggers generated by such attacks are ungrammatical and unnatural. Our work proposes a novel technique combining parts-of-speech filtering and perplexity based loss function to generate sensible triggers that are closer to natural phrases. For the task of sentiment analysis on the SST dataset, the method produces sensible triggers that achieve accuracies as low as 0.04 and 0.12 for flipping positive to negative predictions and vice-versa. To build robust models, we also perform adversarial training using the generated triggers that increases the accuracy of the model from 0.12 to 0.48. We aim to illustrate that adversarial attacks can be made difficult to detect by generating sensible triggers, and to facilitate robust model development through relevant defenses.
arXiv:2605.18606v1 Announce Type: new
Abstract: Neural operators approximate PDE solution maps, but they need not respect the symmetries of the governing equation. In out-of-distribution (OOD) regimes, a standard neural operator must often learn coordinate alignment and physical evolution within a single map, which can hurt generalization. We use known continuous symmetries of evolution equations on periodic domains to separate these two roles. We propose the Physics-Aligned Canonical Equivariant Fourier Neural Operator (PACE-FNO), which estimates the input frame with a Lie-algebra coordinate estimator, maps the field to a reference frame, applies a standard Fourier Neural Operator (FNO), and restores the prediction to the target frame. We train alignment and operator prediction jointly using bounded symmetry perturbations, with an optional low-dimensional refinement step that updates the estimated frame at inference. Equivariance is enforced by the input and output transformations, while the FNO architecture remains unchanged. Across 1-D and 2-D Burgers, shallow-water, and Navier-Stokes equations on periodic domains, PACE-FNO matches the in-distribution (ID) accuracy of standard neural operators and reduces out-of-distribution (OOD) relative error by up to 12x over FNO with symmetry augmentation (FNO+Aug) under translations and Galilean shifts, with smaller gains for coupled rotation-translation shifts. Ablations show that aligning the input and restoring the output frame account for most OOD gains; inference-time refinement provides a smaller correction.
arXiv:2506.01523v2 Announce Type: replace
Abstract: Alignment via reinforcement learning from human feedback (RLHF) has become the dominant paradigm for controlling the quality of outputs from large language models (LLMs). However, existing theories do not provide strong justification for the RLHF objective itself and do not allow comparisons of the guarantees between various methods because different methods are often analyzed under different frameworks. Toward a unified framework for alignment, we ask under what assumptions can we derive existing or new training objectives and obtain theoretical guarantees. To this end, we reframe alignment as distribution learning from pairwise preferences, which makes a probabilistic assumption describing how preferences reveal information about the target LM. This leads us to propose three principled alignment objectives: preference maximum likelihood estimation, preference distillation, and reverse KL minimization. We prove that they all enjoy strong non-asymptotic $O(1/n)$ convergence to the target LM, naturally avoiding degeneracy. In particular, reverse KL highly resembles the RLHF objective, providing strong justification for RLHF. Furthermore, our theory explains, for the first time, the empirical finding that on-policy objectives (e.g., RLHF) typically outperform likelihood-style objectives (e.g., DPO). Finally, empirical results indicate that the proposed objectives are competitive with strong baselines across several tasks and models.
arXiv:2605.17957v1 Announce Type: new
Abstract: As code generation becomes increasingly central to improving software development efficiency, modern code models are largely trained and evaluated on code with natural-language descriptions. In real projects, developers often implement missing functions under limited project-specific artifacts, while the local call-site context is already available in the surrounding code. This usage context provides actionable cues about expected behavior, but existing models are not explicitly optimized to leverage it reliably, leading to implementations that may not integrate smoothly with surrounding usage in repository settings.
In this work, we propose contextualized code pretraining, an invocation-aware framework that integrates calling context into both the training and evaluation of code models. Using static analysis, we automatically extract large-scale caller-callee pairs from real repositories to construct pretraining tasks and benchmarks that condition generation on the calling context. We train CallerGen, the first code models pretrained with invocation-aware objectives spanning multiple sizes, and evaluate them on CallerEval, a new benchmark featuring realistic scenarios. Experiments show that CallerGen outperforms comparable-scale models and remains competitive with larger ones across two benchmarks. Our 220M and 0.5B models achieve 16.58% and 22.81@% pass1, surpassing baselines on CallerEval. These results highlight the importance of calling context in realistic code generation.
arXiv:2605.17958v1 Announce Type: new
Abstract: Code reasoning refers to the task of predicting the output of a program given its source code and specific inputs. It can measure the reasoning capability of large language models (LLMs) and also benefit downstream tasks such as code generation and mathematical reasoning. Existing work has verified the effectiveness of reinforcement learning on the task. However, these methods design rewards solely based on final outputs or coarse-grained signals, and neglect the inherent consistency of the stepwise reasoning process in the task. Therefore, these methods often result in sparse reward or reward hacking, which limits the full play of enhanced learning capabilities. To alleviate these issues, we propose CodeThinker, a consistency-driven reinforcement learning framework for code reasoning. Specifically, CodeThinker has three key components: (1) a stepwise reasoning-aware model training module, which utilizes a consistency tracing paradigm as a template to synthesize training data that captures the stepwise reasoning process; (2) a dynamic beam sampling strategy, which aims to improve the quality of sampled outputs under a fixed sampling budget; and (3) a consistency reward mechanism that can effectively alleviate reward hacking. Experiments on three popular benchmarks show that CodeThinker achieves state-of-the-art performance across multiple LLMs. For instance, it outperforms the strongest baseline by 4.3% in accuracy when deployed on Qwen2.5-Coder-7B-Instruct. We also validate the effectiveness of CodeThinker on downstream tasks. Results show that, without additional training, CodeThinker obtains average accuracy gains of 5.33 and 3.11 percentage points on mathematical reasoning and code reasoning tasks covering 17 programming languages, respectively.
arXiv:2605.17959v1 Announce Type: new
Abstract: Text-video retrieval aims to find the most semantically similar videos with given text queries. However, since videos contain more diverse content than texts, the main semantics expressed by each text-video pair is often partially relevant. The primary methods involve the utilization of language-video attention module to align texts and videos. Though effective, this paradigm inevitably introduces prohibitive computational overhead, resulting in inefficient retrieval. In this paper, we propose a simple yet effective method called Global-Local Contrastive Consistency Learning (GLCCL) to achieve texts and videos semantics alignment. Specifically, we design a parameter-free Global-Local Interaction Module (GLIM) to generate semantic-related frame and video features in a text-guided manner. Furthermore, a Contrastive Score Consistency (CSC) loss is developed to promote consistency learning among different scores on positive pairs and suppress consistency learning on negative pairs. Empirical evidence suggests that CSC loss provides the model with robust discriminative power between positives and hard negatives. Extensive experiments on three benchmark datasets, including MSR-VTT, DiDeMo and VATEX, demonstrate the effectiveness and superiority of our approach.
arXiv:2605.17960v1 Announce Type: new
Abstract: Machine-learning-based Intrusion Detection Systems (IDS) have achieved impressive accuracy in classifying network attacks, yet they consistently fall short on the question that matters most to a security analyst: what should I do next? This paper presents a unified, end-to-end framework that closes the gap between threat detection and actionable response. The system operates in two tightly coupled stages. First, an ensemble of three independently trained binary Deep Neural Networks (DNNs) classifies network traffic flows as Benign, Denial of Service (DoS), or Distributed Denial of Service (DDoS), achieving 99.84% accuracy on the CICIDS2018 dataset and 95.30% on the UNSW-NB15 dataset. Second, a Retrieval-Augmented Generation (RAG) pipeline constructs explanation-aware prompts from the top-5 anomalous features, retrieves the most semantically and lexically relevant guidance from a knowledge base derived from authorized sources and di- rects a locally deployed language model to synthesise structured, citation-grounded mitigation reports. The RAG-enhanced reports outperform vanilla LLM outputs across all automated evaluation metrics.
arXiv:2605.17378v1 Announce Type: cross
Abstract: We introduce UPSim (UxNB Propagation Simulator), a ray tracing-calibrated, semi-deterministic solution for spatially consistent FR3 air-to-ground propagation modeling in uncrewed aerial vehicle (UAV) networks. Instead of launching rays for every receiver position, UPSim derives deterministic visibility regions from 3D building geometry via shadow projection. It then augments these regions with line-of-sight (LOS) state-specific and altitude-aware path loss, correlated large-scale fading, and small-scale fading. Calibration and validation against FR3 ray tracing data using the global 3D-GloBFP building dataset demonstrate that UPSim accurately reproduces empirical channel distributions. Furthermore, the resulting maps support route-based analysis of channel evolution over complex urban layouts, exposing critical trajectory-level statistics such as outage distances. Consequently, UPSim offers a highly scalable, practical middle ground between computationally expensive full ray tracing and purely stochastic channel generation for mobility-aware planning and radio-map construction in aerial access scenarios.
arXiv:2605.18656v1 Announce Type: cross
Abstract: Federated Learning is a leading framework for training ML and AI models collaboratively across numerous user devices or databases. We study the trade-offs among estimation accuracy, privacy constraints, and communication cost for differentially private (DP) federated M estimation. The two standard methods in the literature are FedAvg, which may suffer from high federation bias, and FedSGD, which can incur high communication cost. Aimed at improving accuracy at a reduced communication cost, we propose FedHybrid, which uses FedSGD starting with an improved initialization by the FedAvg estimator. We propose FedNewton, which averages local Newton iterations to reduce bias in FedAvg, achieving an estimation accuracy comparable to FedSGD with much fewer communication rounds when the number of clients grows sufficiently slowly. We establish finite sample upper bounds on the mean-squared error rates of the DP versions of these estimators as functions of the number of clients, local sample sizes, privacy budget, and number of iterations. We further derive a minimax lower bound on the MSE of any iterative private federated procedure that provides a benchmark to assess the optimality gap of these methods. We numerically evaluate our methods for training a logistic regression and a neural network on the computer vision datasets MNIST and CIFAR-10.
arXiv:2501.06933v3 Announce Type: replace
Abstract: Nonlinear conservation laws govern a broad class of important physical systems in science and industry and are central to scientific machine learning (SciML). Large general-purpose models offer speed, but replacing the numerical and physical structure of solvers often compromises stability, accuracy, and physical faithfulness. Here, we aim to balance the general inductive bias of conservation with the flexibility and speed of neural networks through a conservation-aware SciML backbone, which we call Neural Discrete Equilibrium (NeurDE). NeurDE places machine learning inside a kinetic solver by learning the local equilibrium closure of a Boltzmann formulation. The kinetic solver still performs transport, relaxation, moment recovery, and conservation; the neural network provides only the nonlinear equilibrium target. We test NeurDE on $6$ conserved systems, including three very challenging subsonic, transonic, and supersonic shock systems. NeurDE outperforms state-of-the-art SciML methods, including neural operators and pretrained SciML foundation models that are $10^4$ and $10^6$ times larger, respectively. Most notably, NeurDE improves upon the numerical method from which it is derived. NeurDE therefore provides a compact target for scientific machine learning in conservative simulation: learn the equilibrium law toward which the system relaxes, not the evolution law itself.
arXiv:2605.17577v1 Announce Type: new
Abstract: Large-scale pre-trained Vision-Language models (VLMs), such as CLIP, exhibit strong zero-shot generalization, yet remain highly vulnerable to imperceptible adversarial perturbations, raising serious safety concerns for open-world deployment. To enhance robustness without requiring downstream task-specific retraining, we propose TAME, a novel test-time defense. Building upon our prior Test-Time Adversarial Prompt Tuning (TAPT), TAME introduces an architectural reformulation by replacing TAPT's single adaptive prompt with an input-conditioned Mixture-of-Experts (MoE) framework, enabling more expressive and adaptive defense. Specifically, TAME maintains a bank of learnable expert prompts and employs an input-dependent routing mechanism to aggregate a customized prompt mixture for each unlabeled test sample at inference time. This test-time defense mechanism is driven by three unsupervised objectives: (1) multi-view prediction entropy minimization, (2) layer-wise alignment of visual token statistics to precomputed clean and adversarial reference distributions, and (3) MoE regularization for balanced expert utilization and prompt diversity. We evaluated TAME on 11 benchmark datasets, including ImageNet and 10 additional zero-shot datasets. The results show that TAME improves the zero-shot adversarial robustness of the original CLIP by at least 49.1% under AutoAttack while largely preserving generalization on clean samples. TAME also consistently outperforms existing adversarial prompt tuning methods across multiple prompt designs, yielding an average robustness gain of at least 30.2%.
arXiv:2602.16367v3 Announce Type: replace
Abstract: This paper addresses the challenge of efficient rendezvous in multihop cognitive radio networks, where existing channel-hopping algorithms designed for single-hop scenarios incur increased delay and coordination inefficiencies in multinode topologies. To overcome these limitations, we propose a Multihop Dual Modular Clock Algorithm (M-DMCA), which systematically extends modular clock-based rendezvous to multihop environments while preserving efficient channel coordination. The proposed scheme enables dual-channel selection per timeslot and incorporates a lightweight three-way handshake mechanism to improve coordination among intermediate nodes. Simulation results under worst-case conditions, including high primary user activity, asymmetric channel availability, and dense network settings, demonstrate that M-DMCA significantly reduces rendezvous time compared to existing approaches, achieving up to 24% improvement. These results demonstrate the suitability of M-DMCA for timely node discovery in dynamic emergency response scenarios.
arXiv:2605.17575v1 Announce Type: new
Abstract: Network traffic classification (NTC) models often suffer severe performance degradation when deployed in real-world environments due to distribution shifts caused by changing network conditions. Existing robustness-enhancing approaches are commonly coupled to specific model architectures or data settings, fail to generalize to state-of-the-art raw-byte-based NTC models, or incur significant training overhead. In this paper, we propose UniAlign, a novel model-agnostic framework that improves the robustness of deep learning-based NTC models under distribution shifts. UniAlign combines \emph{domain alignment fine-tuning}, which encourages the learning of domain-invariant traffic representations across heterogeneous network conditions, with \emph{stable model ensembling}, which enhances inference robustness by aggregating checkpoints within a flat loss region. The framework can be seamlessly integrated into existing supervised NTC models without requiring specific feature modalities or introducing non-constant additional training costs. We evaluate UniAlign on three public datasets covering diverse distribution shifts, including encryption schemes, data collection devices, and attack behaviors. Experimental results on two representative NTC models demonstrate that, compared with standard training, UniAlign improves average classification accuracy by 2.51\% and average F1 score by 2.71\%, outperforming the strongest baseline by 1.45\% in accuracy and 1.69\% in F1 score, while requiring only 12.4\%--53.9\% of the training time of all NTC-specific baselines.
arXiv:2512.23752v5 Announce Type: replace
Abstract: Recent work has shown that small transformers trained in controlled "wind-tunnel'' settings can implement exact Bayesian inference, and that their training dynamics produce a geometric substrate -- low-dimensional value manifolds and progressively orthogonal keys -- that encodes posterior structure. We investigate whether this geometric signature persists in production-grade language models. Across Pythia, Phi-2, Llama-3, and Mistral families, we find that last-layer value representations organize along a single dominant axis whose position strongly correlates with predictive entropy, and that domain-restricted prompts collapse this structure into the same low-dimensional manifolds observed in synthetic settings.
To probe the role of this geometry, we perform targeted interventions on the entropy-aligned axis of Pythia-410M during in-context learning. Removing or perturbing this axis selectively disrupts the local uncertainty geometry, whereas matched random-axis interventions leave it intact. However, these single-layer manipulations do not produce proportionally specific degradation in Bayesian-like behavior, indicating that the geometry is a privileged readout of uncertainty rather than a singular computational bottleneck. Taken together, our results show that modern language models preserve the geometric substrate that enables Bayesian inference in wind tunnels, and organize their approximate Bayesian updates along this substrate.