Perturbations

The wide use of Convolutional Neural Networks (CNN) in numerous domains and real-world classification applications is justified by their high precision and automation speed, helping users concentrate on higher-expertise tasks. To better understand the models and avoid bias during deployment, eXplainable Artificial Intelligence (XAI) techniques can be used after training. But as the list of XAI solutions expand, comparisons between them diverge, and consensus over their evaluation cannot be reached. This paper proposes a variation of Fidelity-based XAI metrics, with a focus on real-conditions applications, where the number of classes is often low. The approach generates in-distribution, uncertainty-provoking perturbations, to ensure proper measurement of the XAI methods faithfulness. As demonstration of the evaluation framework usefulness, it is compared with human-centric object localization and segmentation metrics. Once applied to both medical and natural imaging applications, it highlights the intricate correlation between domain, data curation, and XAI solution choices in order to validate training of a new CNN model.

Mean opinion score (MOS) prediction models are widely used as proxy metrics in text-to-speech (TTS) research, yet their ability to capture quality differences beyond acoustic fidelity remains unclear. We investigate this via controlled perturbations on speech: acoustic degradation, prosodic errors, and manipulation of speaker-specific characteristics such as pitch and speaking rate. We obtained MOS predictions for these speech samples from both human listeners and the model, and analyzed the differences in their perceptual characteristics. Results show that most models track acoustic degradation well, while all are insensitive to prosodic errors despite large subjective score drops. For speaker characteristics, models exhibit a double dissociation: strong mean fundamental frequency (F0) biases absent in human ratings, yet insensitivity to speaking rate and F0 variability that humans notice. These findings highlight limitations of scalar MOS prediction beyond acoustic fidelity.

We study adversarial bandit optimization in which the loss functions may be non-convex and non-smooth. In each round, the learner selects an action and observes only the loss incurred at that action. The loss consists of an underlying convex and $β$-smooth component and an adversarial perturbation that may be chosen after observing the learner's action. The perturbations are subject to a global budget controlling their cumulative magnitude over time. This framework extends the globally budgeted, post-action perturbation model from underlying linear losses to general convex and $β$-smooth losses. For this broader class, we establish expected regret guarantees that explicitly characterize the effect of the perturbation budget. To establish these guarantees, we modify a standard bandit optimization algorithm and develop an analysis that controls the additional regret caused by the perturbations. In the absence of perturbations, our results reduce to regret guarantees for the standard bandit convex optimization setting with $β$-smooth losses.

Math and science reasoning benchmarks rely on pass@k, the fraction of sampled chains that reach gold, as the canonical per-example difficulty signal. The same signal drives RL with verifiable rewards, math data curation, synthetic curricula, and verifier training. We show this proxy has a persistent blind spot on its hardest stratum: on the eight free-form math cells we test (GSM8K and MATH across four open-weight models), 10.3-22.9% of the examples that no sampling seed solves in six tries are instead solved at matched compute by a six-chain deterministic regime. These are greedy decoding plus five cheap residual-stream perturbations applied via activation grafting, while greedy alone solves at most 6% on these math cells. Recovery scales with the additional budget, across perturbations whose mechanistic distinctness we verify across all twelve cells (cross-kind fix-set Jaccard <= 0.47 in every setup). Activation grafting is used as an intervention on internal representations, not a decoding method; we use it purely as a diagnostic and diversification tool, and our recovered items show that the pass@k= 0 % stratum is structurally identifiable in the residual stream rather than that the unmodified model reaches them under ordinary inference.

Score-based diffusion models typically use Brownian perturbations, which provide tractable reverse-time dynamics but impose memoryless noising. We introduce Volterra generative models, a continuous-time score-based framework whose forward process injects path-dependent noise through fractional kernels. To handle the non-Markovian and non-semimartingale dynamics, we construct finite-dimensional Markovian lifts using Gaussian quadrature in both regimes and a hybrid finite-difference exponential approximation in the smooth regime. We prove squared error bounds, derive an augmented linear-Gaussian forward process, and show that the learning can remain data-dimensional by considering residual states and analytic auxiliary Gaussian scores. We also identify covariance and reverse-time degeneracies caused by shared Brownian factors and signed smooth-regime weights. The degeneracy motivates stabilized conditioning and, for stiff larger lifts, a Gaussian-bridge reconstruction sampler. Experiments on MNIST and CIFAR-10 show that persistent fractional perturbations with small Markovian lifts can improve score-based generation on MNIST and provide a promising extension to natural images, while the bridge sampler provides a stability mechanism for larger lifts.

Evaluating large language models (LLMs) is important for understanding their capabilities, comparing competing systems, and supporting the deployment of reliable models in practice. For open-ended tasks, pairwise evaluation has become a popular paradigm, in which two responses to the same prompt are compared and the resulting judgments are aggregated into an overall ranking. A central challenge of this paradigm is intransitivity: the induced comparison outcomes may fail to support any coherent global ranking. For example, one may observe cyclic preferences such as $A \succ B \succ C \succ A$, or inconsistencies involving ties such as $A \equiv B\equiv C\neq A$. Such contradictions make the resulting leaderboard unstable and challenging to interpret. In this paper, we propose a prompt perturbation framework for improving the consistency of pairwise LLM evaluation. Our approach generates perturbed variants of each prompt, uses the resulting comparison graphs to identify and filter out structurally inconsistent comparison patterns, and then applies standard ranking methods to the filtered comparisons. A key feature of the proposed framework is that graph-level structural consistency is incorporated explicitly into the evaluation pipeline before ranking aggregation. This provides a simple and principled way to reduce cyclic inconsistencies and improve the reliability of LLM rankings.

Deep neural networks have achieved remarkable performance across medical imaging tasks, yet their tendency to overgeneralize under distributional shifts poses a major obstacle to safe clinical deployment. Out-of-Distribution (OOD) detection methods aim to mitigate this risk, but most existing approaches rely on opaque internal signals with poorly understood semantic meaning, limiting trust in safety-critical settings. In this work, we propose an interpretable OOD detection framework that probes the stability of model predictions under class-conditioned semantic perturbations. Leveraging sparse autoencoders (SAEs), we learn class-specific concept vectors from in-distribution data that disentangle dense intermediate representations into sparse, semantically meaningful components. At inference, we perturb deeper-layer representations using the concept vectors associated with the model's predicted class and measure the class logits stability. We hypothesize that in-distribution samples exhibit low sensitivity to such perturbations, as their representations align with class-specific semantic directions, whereas OOD samples show amplified deviations due to representational misalignment. By framing OOD detection as a concept conditioned stability analysis, our approach provides both a discriminative OOD signal and an interpretable lens into the internal mechanisms driving model uncertainty, making it particularly suitable for high stakes medical applications.

With the rapid deployment of speech generation systems in open environments, providing verifiable source attribution and copyright accountability for audio content has become critical. A gap in current research is the lack of a unified benchmark that systematically compares different watermark injection methods under realistic distribution shifts. To address this, we build VoxWatermark by applying 10 watermarking methods (4 neural and 6 traditional) with unified injection and annotation on multilingual, multi-source corpora, and introducing no-box, black-box, and white-box perturbations to simulate real recording and transmission conditions. Based on this benchmark, we propose AudioWMD as a robust baseline detector for large-scale, multi-method, cross-distribution settings. Results show that injection-method diversity and distribution shifts affect detection stability, while validating the effectiveness and scalability of AudioWMD. Dataset and code are publicly available.

Proof autoformalization aims to translate a mathematical informal proof written in natural language into a formal proof in a formal language such as Lean~4. Several works have developed LLM-based models for proof autoformalization. However, existing evaluations have typically focused on translating well-formed informal proofs from curated datasets. We argue that a robust proof autoformalizer must remain faithful even for informal proofs that diverge from these idealized ones, and we present the first study on the robustness of proof autoformalization models. We formulate two categories of perturbations and evaluate robustness under each: a global perturbation paraphrases the informal proof in a different style, under which the formalization should remain consistent; a local perturbation alters a value, symbol, or proof step, possibly in a counterfactual way, and a robust formalization should faithfully reflect the perturbation rather than reverting to the original one or inferring a different one on its own. We build a benchmark with both perturbations on miniF2F and MATH-500, and automatically measure how stable a proof autoformalization's correctness is under global perturbations and how faithfully its output reflects local perturbations. We evaluate seven recent models, all of which are sensitive to global perturbations and mostly fail to remain faithful under local perturbations. Code and data are available via https://github.com/ucr-rai/robust-proof-autoformalization.

As AI systems are deployed in high-stakes ethical contexts such as healthcare triage, autonomous vehicle control, and employment screening, formal methods for evaluating their robustness against adversarial manipulation of ethical reasoning remain underdeveloped. This paper introduces the Ethical Robustness Testing System (ERTS), a closed-pipeline framework that: (1) encodes ethical dilemmas into a 22-dimensional Ethical Consequence Space (ECS) grounded in established ethical theory; (2) applies 17 semantic perturbation functions subject to 6 validity constraint classes including a novel semantic coherence constraint; (3) measures decision deviation via a 4-component Ethical Instability Index (EII); and (4) produces domain-adaptive pre-deployment robustness assessment verdicts. We evaluate 4 structured baseline models and 2 production LLMs (Gemini 2.0 Flash and Llama 3.2) across 50 ethical scenarios spanning 8 deployment domains, generating 1,500 adversarial test cases. Results demonstrate that only 33% of models achieve assessment clearance, with the local Llama-3.2 model proving particularly vulnerable to fairness corruption and information degradation attacks (ERS = 0.737). To the best of our knowledge, no existing framework combines a bounded ethical consequence space, semantic coherence constraints, and domain-adaptive assessment in a single adversarial testing pipeline.

Spectral methods rely fundamentally on the stability of principal eigenspaces under random perturbations. Classically, this stability is quantified by the Davis-Kahan and Wedin theorems, which bound the eigenspace error using the operator norm of the noise and the relevant spectral gaps. While these worst-case bounds are sharp for arbitrary deterministic perturbations, they can be wasteful in the low-rank signal-plus-random-noise setting, as they fail to capture the fine-grained interaction between the signal geometry and the noise distribution. In this paper, we study the spectral perturbation of signal-plus-noise matrices corrupted by sparse, random noise with an arbitrary, inhomogeneous variance profile. We demonstrate that under heterogeneous noise variances, the empirical eigenvectors suffer a systematic, deterministic geometric bias that is entirely invisible to classical perturbation bounds. By leveraging the Quadratic Vector Equation (QVE) and establishing fine-grained isotropic local laws, we derive near-optimal, non-asymptotic perturbation bounds for the leading eigenspaces in the operator and $2\to\infty$ norms. The bounds separate the usual signal-to-noise contribution, stochastic fluctuations, and structured geometric bias terms determined by the alignment between the signal eigenspaces and the row-wise variance profile.

With AI increasingly deployed in safety-critical systems, providing formal robustness guarantees for the underlying models is essential. Existing verification methods either rely on overly conservative approximations or incur prohibitive computational costs. For example, the use of lp-norm perturbations in video settings encodes the belief that the adversary can inject noise in every video frame. In practice, adversarial perturbations exhibit structured spatial and temporal correlations, constrained to lower-dimensional, semantically meaningful subspaces. In this work, we study robustness verification of 3D CNNs processing video and volumetric inputs, targeting applications in action recognition (UCF-101), autonomous driving (Udacity), and medical imaging (MedMNIST) exploiting realistic assumptions on adversarial strength by modelling them as spatio-temporal constraints - where the attacker can modify either a subset of frames or patches within a set of consecutive frames. We demonstrate that modelling realistic constraints enables tighter approximations. We introduce Spatio-Temporal Bound Propagation (STBP), a verification framework that computes an exact closed-form characterization of the first convolutional layer and propagates certified bounds through subsequent layers using scalable approximations. Computing the exact closed form provides the tightest bounds for the first convolutional layer. Thus, we utilise approximation methods in the remainder of the network. To spur further progress in this field, we propose ST-Bench, a verification benchmark for autonomous driving and activity recognition, to systematically evaluate verifiable robustness. Compared to existing verification-based approaches, STBP provides stronger robustness guarantees with significantly improved scalability, achieving 1.7x higher certified robust accuracy under identical perturbation budgets.

Tissue graph counterfactuals ask how a cell's expression would change under altered spatial neighbor contexts. Such queries are central to predicting cell behavior in tissues, but lack a unified definition, with existing methods targeting specific intervention types or treating cells as i.i.d. In this work, we first formalize tissue graph counterfactuals as a class of spatial interventions that either rewire connections between cells (edge perturbation) or modify the expression of their neighbors (node perturbation). We then introduce Cellina (https://cellina.readthedocs.io) - a framework that uses supervised disentanglement to decompose a cell's intrinsic state from its spatial context, using the latter as a conditioning input for counterfactual predictions. Across benchmarks spanning over 2.5 million spatially-resolved cells in colorectal cancer and mouse brain, Cellina outperforms spatially-informed and non-spatial competitors in in-silico graph perturbations, disentanglement, and scalability. Additionally, we show that Cellina reveals biologically distinct cancer subdomains in an unsupervised manner and enables targeted neighbor perturbation simulations.

Training Data Attribution (TDA) seeks to trace a model's predictions back to its training data. The gold standard for TDA relies on causal interventions, observing how a model changes when data is added or removed, but repeated retraining is computationally challenging for Large Language Models (LLMs). Consequently, most approaches approximate this effect in the parameter space using gradients. However, tracking gradients across billions of parameters is not only prohibitively expensive but relies on local approximations. In this work, we propose a shift: rather than estimating parameter changes, we model the functional effect of training data in the activation space. We introduce STRIDE (Steering-based Training Data Influence Decomposition), a framework that formulates TDA as a sparse recovery problem in the spirit of compressive sensing. STRIDE learns lightweight "steering operators" that mimic the behavioral shift caused by training on data subsets. By measuring how these operators perturb test predictions, we recover individual training example influences via sparse linear decomposition. STRIDE achieves state-of-the-art for LLM pre-training attribution while being an order of magnitude ($13\times$) faster than previous art. We further validate its practical utility through downstream applications including data selection, data contamination, and qualitative analysis.

Muon has recently demonstrated strong empirical performance in large language model training, but the theoretical role of momentum in Muon remains unclear. Existing analyses of Muon either remove momentum to study spectral updates in isolation, or retain momentum without explaining why it improves empirical performance. Our work bridges this gap by showing momentum in Muon acts as a spectral filter. Under a structured signal-plus-perturbation gradient model, we prove that momentum suppresses perturbations while preserving the dominant signal, thereby enlarging the spectral gap between them. This enlarged gap stabilizes the singular subspaces of the matrix passed to Muon's orthogonalization step, making the resulting update more reliable. We further show that applying momentum before orthogonalization achieves provably stronger alignment with the signal component of the gradient than either reversing this order or simply removing momentum. Experiments across diverse tasks, including LLM pretraining, support our theoretical analysis. More broadly, our theory offers a starting point for understanding the benefits of momentum in other matrix-based optimizers.

Cell Painting combines multiplexed fluorescent staining, high-content imaging, and quantitative analysis to generate high-dimensional phenotypic readouts to support diverse downstream tasks such as mechanism-of-action (MoA) inference, toxicity prediction, and construction of drug-disease atlases. However, existing workflows are slow, costly and difficult to interpret. Approaches for drug screening modeling predominantly focus on molecular representation learning, while neglecting actual experimental context (e.g., cell line, dosing schedule, etc.), limiting generalization and MoA resolution. We introduce CP-Agent, an agentic multimodal large language model (MLLM) capable of generating mechanism-relevant, human-interpretable rationales for cell morphological changes under drug perturbations. At its core, CP-Agent leverages a context-aware alignment module, CP-CLIP, that jointly embeds high-content images and experimental metadata to enable robust treatment and MoA discrimination (achieving a maximum F1-score of 0.896). By integrating CP-CLIP outputs with agentic tool usage and reasoning, CP-Agent compiles rationales into a structured report to guide experimental design and hypothesis refinement. These capabilities highlight CP-Agent's potential to accelerate drug discovery by enabling more interpretable, scalable, and context-aware phenotypic screening -- streamlining iterative cycles of hypothesis generation in drug discovery.

The effectiveness of a soft wearable back-support device in enhancing postural stability was investigated under trip-like perturbations using two experimental paradigms: perturbed standing and perturbed walking. Healthy subjects completed trials under three different back-support conditions: no device, device worn with low stiffness, and device activated with high stiffness. Whole-body stability was quantified using the minimum Margin of Stability (MOS) at the point of maximal instability. Results demonstrated increased MOS during device use, indicating enhanced postural stability. In standing, MOS increased significantly with device stiffness, whereas in walking, both device conditions improved MOS relative to no device but did not differ significantly from each other. These findings highlight the potential of soft wearable back-support devices with adjustable stiffness to improve reactive balance control against external perturbations, with important implications for fall prevention. Future research should explore personalized stiffness optimization and evaluate efficacy in populations at elevated risk of falls.

Robot localization systems are critical for autonomous navigation and safety. Adversarial perturbations can mislead these systems, resulting in mislocalization, navigation errors, or unsafe interactions, especially in mission-critical scenarios. This paper investigates the vulnerability of deep learning based localization pipelines to adversarial attacks. We propose a novel framework for generating adversarial queries that specifically target Product Quantization (PQ) in visual localization systems. Our method employs a Lightweight Product Quantization Network (LPQN) to perturb query feature encodings, misleading the retrieval process by returning semantically irrelevant database entries. Adversarial queries are generated via a two-phase procedure: a forward pass that perturbs feature distributions and a backward pass that refines the perturbation through optimization. The lightweight design of LPQN allows the creation of subtle yet highly effective perturbations with minimal computational overhead. Extensive experiments in both controlled and real-world robotic environments demonstrate that our approach substantially degrades PQN performance, exposing critical vulnerabilities in practical applications.

Developing effective anticancer therapeutics remains challenging due to tumor heterogeneity and the absence of well-defined molecular targets across cancer subtypes. Generative models conditioned on cancer genotypes offer a promising avenue for personalized drug discovery, yet existing approaches lack explicit optimization for simultaneous sensitivity, synthesizability, and mechanistic binding plausibility. We present a latent-space optimization approach for a pretrained genotype-to-drug diffusion model, introducing a learnable perturbation over the molecular latent space optimized via gradient ascent to maximize a composite reward combining predicted drug sensitivity (AUC), drug-likeness (QED), and synthetic accessibility (SAS). Critically, biological realism is enforced by grounding both reward design and evaluation in experimentally-derived cancer cell line data and validated pharmacologic signals, anchoring candidate generation in real-world clinical evidence. Mechanistic consistency plausibility is further assessed by a multi-agent LLM pipeline grounded in the diffusion model's attention mechanism. Experiments across 15 cancer cell lines from three held-out evaluation sets demonstrate consistent and noticeable improvements over competing baselines in sensitivity, drug-likeness, synthesizability, and chemical validity.

Adversarial Training (AT) is a leading defense against adversarial examples but often suffers from Catastrophic Overfitting (CO) in efficient single-step variants, where robustness to multi-step attacks collapses despite high single-step performance. We address this failure mode with two contributions. First, we formalize Epsilon Overfitting (EO), a perspective in which fixed perturbation magnitudes and directions exacerbate CO, and show that introducing perturbation variability significantly improves robust generalization across different architectures and datasets. Second, we propose PertAlign (Perturbation Alignment), a theoretically grounded, computationally negligible metric that predicts CO onset by measuring gradient alignment across attack stages. Leveraging these insights, we introduce SORA, an adaptive step-size AT method that dynamically adjusts perturbations based on loss surface geometry. SORA consistently prevents CO, achieves state-of-the-art robustness and clean accuracy, and generalizes across datasets and architectures using a single fixed set of hyperparameters, which is essential for applicability in fast AT. Extensive experiments on diverse datasets and architectures show that SORA matches or surpasses the robustness of prior methods while delivering higher clean accuracy and superior efficiency. Code is available at https://github.com/SecondOrderAT/SORA.

Large perturbation models require training data encompassing chemical, cellular, and assay diversity. Current transcriptomic resources for small-molecule modeling, however, are fragmented across technologies, metadata conventions, controls, doses, and preprocessing pipelines. We introduce Chem-PerturBridge, a harmonized multi-dataset resource comprising over 37k compounds, 136 cellular contexts, and 1.25M transcriptomic samples across eight assay types, with standardized identifiers, metadata, and replicate-aware condition-level effects. We use the resource to evaluate matched-condition agreement across datasets and replicate agreement within datasets. Matched same-compound conditions generally show weak agreement in fine-grained logFC rankings and magnitudes across most dataset pairs, often falling below same-context different-compound baselines. In contrast, logFC direction agreement is substantially more stable and usually exceeds these baselines. We further evaluate Chem-PerturBridge as a pretraining resource for compound representation learning. Under a compound-held-out OP3 evaluation split, embeddings pretrained on Chem-PerturBridge improve over L1000-only embeddings, Morgan fingerprints, and the descriptor-free OP3 baseline across metrics. An extensive molecule-holdout evaluation across 11 datasets further shows that models trained on Chem-PerturBridge outperform or match those that are not. Chem-PerturBridge therefore supports both diagnostic evaluation of cross-dataset signature agreement and model-oriented reuse of heterogeneous perturbation transcriptomic data.

Watermarking embeds statistical signatures in AI-generated text for detection and attribution. We reveal a fundamental vulnerability: when users access multiple models (today's reality), watermarks trivially fail. Watermarks perturb output distributions away from the original, and in competitive markets, these perturbations are typically independent across providers. We theoretically prove that averaging output probability distributions recovers the unwatermarked distribution with up to a second-order error term. Empirically, simply averaging 3-5 models cancels out these perturbations. We introduce WASH (Watermark Attenuation via Statistical Hybridisation), which solves practical challenges in ensemble generation: vocabulary misalignment and tokenisation differences across heterogeneous models. Experiments across six watermarking schemes and three LLMs show that averaging across 3 models suppresses detection z-scores from 5-300 to below 2 (below the detection threshold of 4) and reduces TPR at 5% FPR to below 50%, while improving quality by 27.5% and running 6 times faster than the best baseline on the long sequence generation. Our results suggest that robust AI-text detection via watermarking requires either accepting this fundamental vulnerability or unprecedented coordination among model providers.

The work presents an approach for addressing the challenge of robustness in Large Language Models (LLMs) to alterations and potential errors caused by semantically similar but textually different prompts. Recent works have shown that these kinds of prompt variations can significantly impact the performance of LLMs on tasks. The central question is: can LLMs' robustness to semantically-neutral prompt alterations be acquired without expensive retraining of the entire model? We address this question both theoretically and through experiments. Our theoretical analysis reveals a crucial factor impacting model robustness - a systematic expected shift or perturbation-induced bias in neural network module outputs. Motivated by this analysis, we show that robustness can be achieved via a simple fine-tuning process: debiasing for robustness. We identify conditions when debiasing helps and when it does not, and demonstrate, through both theory and extensive experiments, that debiasing for robustness may indeed be a quick and efficient tool to enhance robustness and provide certification against random prompt perturbations.

LLM-based coding assistants are seeing rapid adoption, offering substantial gains in developer productivity. As organizations increasingly ship code these agents produce, the security of that code becomes critical. Prior work has shown that minor prompt perturbations degrade the functional correctness of LLM-generated code, but whether they also compromise code security has remained unstudied. We apply token-level mutations to prompts across three models and five programming languages, and show that mutations as small as a single-character change can flip generated code from secure to vulnerable. Probing the models' hidden states reveals that this fragility is partially encoded in prompt representations, but unevenly so. Input-handling vulnerabilities, where the model omits validation or sanitization, are more predictable (mean AUC 0.753) than secure-defaults vulnerabilities, where insecure code stems from one local choice such as a weak algorithm or unsafe parameter (mean AUC 0.674). These results show that the threat model for LLM-assisted coding extends beyond prompt injection to ordinary prompt variation, and indicate that input-handling flaws can be caught before generation while secure-defaults flaws require intervention during decoding.

Deep neural networks process data through a cascade of representations: input features, hidden activations, logits, and loss. While perturbations at the input, logit, and label levels have been systematically studied, the intermediate hidden activations, which constitute the bulk of the network's computation, have received no unified perturbation analysis. In this paper, we establish a unified framework for hidden activation perturbation, revealing that Dropout, Manifold Mixup, adversarial feature perturbation, and related methods all impose specific forms of activation perturbation but with class-agnostic or random strategies. We conjecture that expansive perturbation (increasing activation norm) acts as positive augmentation, while contractive perturbation (decreasing activation norm) acts as negative augmentation, and that the perturbation layer determines whether the effect resembles input-level augmentation (shallow layers) or logit-level manipulation (deep layers). We propose Learning to Perturb Activations (LPA), which adaptively perturbs activations at a selected hidden layer with class-level perturbations learned via PGD. We further provide theoretical analysis connecting activation perturbation to flat minima and perturbation amplification through layers. Experiments on balanced classification, long-tail classification, and domain generalization demonstrate that LPA consistently outperforms existing methods and provides complementary benefits to logit perturbation methods such as LPL.

Deep neural network training involves both forward propagation (from features through logits to loss) and backward propagation (from loss through gradients to parameter updates). While perturbations along the forward chain, including feature perturbation, logit perturbation, and label perturbation, have been extensively studied, the backward chain's gradient perturbation has received little systematic investigation. In this paper, we establish a unified framework for gradient perturbation, revealing that existing methods such as Sharpness-Aware Minimization (SAM), gradient clipping, and gradient noise injection can all be interpreted as imposing specific forms of gradient perturbation. Analogous to the recently proposed Logit Perturbation Learning (LPL), we conjecture that amplifying the gradient norm for a class acts as positive augmentation (enhancing learning), while dampening it acts as negative augmentation (suppressing overfitting). Based on these observations, we propose Learning to Perturb Gradients (LPG), which adaptively perturbs logit-level gradients at the class level to achieve category-aware training. We also establish theoretical connections between gradient perturbation bounds and generalization guarantees via PAC-Bayesian analysis. Experiments on balanced classification, long-tail classification, and noisy label learning demonstrate that LPG consistently outperforms existing methods and can be combined with them as a plug-in module.

Vision-Language-Action (VLA) models with integrated reasoning have been proposed for end-to-end autonomous driving, assuming a tight coupling between reasoning and trajectory generation. However, the robustness of such systems under realistic input perturbations remains largely unexplored. We show that these models are highly vulnerable to realistic input perturbations, achieving up to 89% attack success rate (ASR) on reasoning and up to 72% on trajectory manipulation in closed-loop simulation, leading to increased collision rates and degraded safety metrics. Using NVIDIA's recent Alpamayo models as representative industry-developed VLAs, we conduct the first systematic black-box study of reasoning-enabled VLA models under realistic textual input corruptions, evaluating their impact on reasoning and driving behavior. We introduce a reasoning-aware evaluation framework capturing both semantic and structural aspects of reasoning, along with safety-centric measures. We also introduce a benchmark for evaluating attacks and defenses on reasoning-trajectory interactions in autonomous driving. Our results highlight the need for rigorous evaluation and improved defenses to ensure the safety of reasoning-enabled VLA systems in autonomous driving.

When machine learning systems under-perform for particular subgroups, affected users typically have no way to correct these disparities without relying on platform-level fixes. Existing approaches to algorithmic fairness rely on provider-centric approaches to correct these failures, leaving users with no external lever when faced with harm. Recent work in Algorithmic Collective Action shows that coordinated users can steer an algorithmic system toward a collective goal, but the existing mechanisms require the provider to retrain on the collective's modified data which users may not have control over. We propose Test-Time Collective Action (TTCA), a framework through which a group of users who share query access to the platform, can correct disparities affecting under-served subgroup without participating in the platform's training loop. We implement this through a proxy-based mechanism where the collective pools query access to a black-box API to extract a proxy of the platform, then optimizes a per-class universal perturbation against the proxy. Each member applies this perturbation to their own inputs at submission time, requiring no cooperation from the platform. We empirically evaluate the mechanism on CIFAR-10, CIFAR-100, and FairFace, showing that modestly-sized collectives close most of the subgroup accuracy gap, transfer across architectures (a small proxy can attack a larger platform), and improve worst-group accuracy, equal-opportunity gap, and disparate impact. A query-budget analysis comparing a per-user black-box attack baseline shows that pooling is cheaper than each subgroup member attacking alone. Test-time collective action thus offers corrective intervention to users when platform-side remediation is unavailable or delayed.

Multi-agent LLM systems decompose workflows across agents, tools, shared context, memory, and decision gates. This modularity improves interpretability, but creates a propagation risk: a bounded perturbation to one component can be reused by other agents and amplified into system-level harm. We introduce HARP (Harm Amplification through Role Perturbation), a trace-first methodology for studying local-to-global harm amplification in multi-agent LLM systems. HARP compares paired clean and perturbed executions and records specialist outputs, tool calls, memory reads/writes, guard events, oracle logs, latency, token cost, and decisions. We define local harm as deviation from targeted agents or corrupted channels, global harm as deviation over the full trace, and harm amplification as (H_global/H_local). This complements attack success rate with a measure of how strongly orchestration spreads harm beyond the attack point. We instantiate HARP in a finance-oriented seven-agent system with a deterministic decision gate and configurable attack harness for specialist compromise, collusion, shared-context corruption, and temporal or memory-persistent attacks. Across five defenses, prompt-only defenses preserve benign utility but leave high success and stealth; pre-tool and step-level guards reduce some failures with utility or latency costs; and IntegrityGuard, a trace-consistency defense, achieves the lowest attack success and global harm but introduces utility/cost trade-offs. Results show that single-specialist compromise produces the strongest amplification, shared-context corruption yields the highest attack success, and temporal persistence produces the largest malicious impact. HARP argues that secure multi-agent evaluation must measure not only bypass, but propagation.

Generative surveying -- where collections of LLM-based personas provide feedback on messages -- has emerged as a cheap and scalable alternative to traditional market research. However, LLMs are sensitive to small variations in prompt design and conclusions drawn from generative surveys may depend on arbitrary phrasing choices. Controlling for this sensitivity requires including semantically equivalent perturbations in the analysis. In this paper, we show that standard hypothesis tests, including the sign test and Wilcoxon signed-rank test, are invalid under a statistical model for generative surveying that includes realistic perturbation structure. We propose a permutation test that is valid under this model and formally characterize the conditions under which standard tests fail. Applying our framework to a simple generative surveying problem, we estimate relevant parameters, characterize the power of the permutation test under realistic conditions, and provide practical guidance on budget allocation across personas, perturbations, and replicates. Finally, we show that both the magnitude and direction of the estimated effect are sensitive to the choice of model, even within the same model family.

In-context learning (ICL) is often motivated by the intuition that demonstrations help because they provide correct input-output examples. However, we reveal a counterintuitive phenomenon: correctness does not guarantee exemplar utility, and some correct demonstrations can even reduce ICL accuracy. To study this correctness-utility gap, we introduce task-preserving perturbations, where only the exemplar input is changed, while the example remains a correct instance of the same task. Concretely, each perturbed exemplar is assigned the target induced by the task mapping. This framework covers both label-updating perturbations, where task-relevant semantics change and targets are recomputed, and stricter target-preserving perturbations, where the original target remains valid. We formalize the resulting failure mode as contextual evidence shift: task-preserving perturbations can change the effective mixture of evidence used by the model for contextual inference, thereby separating exemplar correctness from exemplar utility. Across sentiment classification, logical reasoning, and math word problems, we find that task-preserving perturbed demonstrations can substantially degrade ICL performance, especially for smaller models, harder tasks, and higher perturbation ratios. Our results show that robust ICL requires evaluating not only whether demonstrations are correct, but also how they influence contextual inference. Code is available at https://github.com/Chenghao-Qiu/Task-Preserving-ICL.

We document an empirical phenomenon in chain-of-thought and ReAct agents driven by ten large language models from seven architecture families: meaning-bearing perturbations (e.g., paraphrase, synonym) alter final answers more often than presentation perturbations (e.g., formatting, reordering) of comparable severity. Across 68 cells spanning GSM8K, MATH, and HotpotQA (1,530 originals and $\sim$11,150 variants), the inconsistency gap averages +19.69 pp after severity matching (paired $t=9.58$, $p<0.0001$), with 64/68 cells positive. The gap survives four severity-proxy audits and remains significant when excluding qwen models (+11.10 pp, $p<0.0001$). Several stress tests fail honestly: cluster-bootstrap significance disappears under stricter assumptions, tractability contrasts do not replicate, cross-architecture generator swaps break per-cell rankings, and a second LLM judge yields only moderate agreement ($κ=0.50$). We then validate the headline effect on a fully held-out 11th model (qwen2.5-14B-Instruct; 1,800 trajectories) and re-test a pre-registered capability$\times$tractability partition, observing a small but positive held-out effect (3/4 cells positive; pooled Welch $t=3.81$, $p=9.6\times10^{-4}$). Using held-out trajectories, we probe four trace-level mechanism signals. Two prior mechanism claims fail to replicate and are explicitly retracted. Two new probes instead support a \emph{stealth-divergence} picture: semantic perturbations often preserve the first action but induce divergence in intermediate reasoning from later steps onward, accompanied by slightly deeper trajectories. We position this as a measurement contribution with held-out replication and a partial trace-level account of how semantic perturbations propagate through agent reasoning. Code, perturbation corpus, raw trajectories, and analysis scripts are released anonymously for review.

Code agents are currently having skillful performance on repository-level software engineering benchmarks, but it remains unclear whether success on end-to-end tasks such as issue resolution truly reflects repository context reasoning, the ability to identify the task-relevant information across multiple files and reason over the relations among them. To investigate this question, we introduce RepoMirage, a two-stage evaluation suite built on SWE-Bench Verified that adopts perturbation as a diagnostic tool to increase the demand for context reasoning by transforming how the repository is exposed. First, RepoMirage-Perturb applies three types of semantics-preserving repository-level perturbations, revealing a clear performance drop when correct solving requires broader context access. RepoMirage-Extend further turns perturbation-targeted structural bottlenecks into explicit tasks beyond issue resolution, where the average performance declines from 66.8% in the original setting to 25.3%, indicating a significant deficiency in repository context reasoning. Further trajectory analysis reveals an exploration drift, where agents access broader repository context but fail to turn it into effective structure information. Motivated by this observation, we propose RepoAnchor, a structure-first prototype workflow that separates repository exploration from downstream problem solving, and show that explicit structural scaffolding yields notable gains. These results uncover an previously overlooked gap in repository context reasoning for code agents and suggest that stronger structure-aware methods are potential to improve them.

Particle Swarm Optimization (PSO) frequently suffers from premature convergence. This paper introduces a family of problem-informed diversity-enhancing strategies that manipulate the swarm's social and cognitive components. These include opposing-best strategies that repel particles from optimal regions, negative learning strategies that guide exploration toward poor solutions, and reverse learning strategies that push particles away from inferior regions. These socio-cognitive mechanisms are evaluated against an analogous suite of problem-unaware, explicit randomization strategies that inject randomness either into velocity update components or directly into position updates. The results reveal that the effectiveness of diversity enhancement is determined primarily by how it is embedded within the swarm dynamics, rather than by the mere presence of extraneous problem-informed guidance. Particularly, random perturbations introduced at the velocity-update level consistently outperform those applied directly to particle positions.

Existing scaling laws for Large Language Models (LLMs), predominantly monotonic power laws, fail to explain emerging non-monotonic phenomena such as catastrophic overtraining and quantization-induced degradation, where performance deteriorates despite increased compute. We propose the Shannon Scaling Law, a unified theoretical framework that models LLM training as information transmission over a noisy channel, grounded in the Shannon-Hartley theorem. By mapping model parameters to channel bandwidth and training tokens to signal power, our formulation explicitly captures the interaction between learning signal and intrinsic noise. This perspective reveals a fundamental Shannon capacity for LLMs: scaling model size or data without preserving a sufficient signal-to-noise ratio (SNR) inevitably amplifies noise, inducing a transition from monotonic improvement to U-shaped performance degradation. We validate our theory through experiments on Pythia and OLMo2 under perturbations, including Gaussian noise, quantization and supervised fine-tuning on math, QA and code tasks. The Shannon Scaling Law consistently outperforms classical scaling laws and recent perturbation-aware laws, achieving strong $R^2$ scores and accurately capturing loss basins missed by prior approaches. It also extrapolates: fitted on $\leq$6.9B Pythia models with $\leq$180B tokens, it predicts the unseen 12B model up to 307B tokens at pooled $R^2{=}0.847$, while monotonic baselines collapse.

Gradient-flow sampling interprets a Gibbs distribution as the minimizer of an energy functional over probability measures and generates dynamics converging to this target. Under spherical Hellinger-Kantorovich (SHK) geometry, the flow couples transport and reaction and coincides with birth-death Langevin dynamics. In this work, we develop a perturbation theory for SHK gradient flows. For two potentials $V$ and $V^{\prime}$, we compare the associated flows from a common initialization and quantify how potential discrepancies propagate over time. A uniform perturbation bound yields dimension-free, pointwise control of the log-likelihood ratio and Rényi divergence, while additional structure allows us to derive bounds for the KL divergence as well. We apply these results to approximate sampling for the exponential mechanism in differential privacy. The likelihood-ratio control provides explicit time-dependent Pure-DP guarantees for SHK-based samplers, while the KL bound yields Approximate-DP certificates via hockey-stick divergence. We also derive a utility bound separating intrinsic exponential-mechanism suboptimality from finite-time sampling error.

Interpretable autonomous driving planners depend not only on generating explanations, but also on those explanations remaining reliable under real-world sensor degradation. In this paper we present a controlled perturbation study of Vision-Language-Action (VLA) robustness in autonomous driving, evaluating Alpamayo R1 (10B parameters) across 1,996 scenarios under eight sensor perturbations (Gaussian noise at four intensities, two lighting extremes, and two fog levels; ${\sim}18{,}000$ inference trials). We find that reasoning consistency is a high-fidelity indicator of trajectory reliability: when Chain-of-Causation (CoC) explanations change after perturbation, trajectory deviation spikes $5.3{\times}$ (21.8m vs 4.1m), with $r\!=\!0.99$ across attack types and $r_{pb}\!=\!0.53$ per-sample (Cohen's $d\!=\!1.12$). A controlled ablation provides evidence that enabling CoC generation is associated with improved trajectory accuracy (11.8% on average across conditions; $p < 0.0001$) under matched inference settings. Over the tested noise range ($σ\in \{10, 30, 50, 70\}$), degradation is approximately linear ($R^2\!=\!0.957$), while standard input preprocessing defenses provide only marginal relief. Together, these results establish CoC consistency as a quantitative proxy for planning safety and motivate reasoning-based runtime monitoring for safer VLA deployment.

Single-cell perturbation modeling is fundamental for understanding and predicting cellular responses to genetic perturbations. However, existing approaches, from causal representation learning to foundation models, often struggle with an overlooked challenge: gene expression is dominated by perturbation-invariant information, while perturbation-specific signals are intrinsically sparse. As a result, learned representations either entangle invariant and perturbation-specific information, leading to spurious and non-generalizable predictors, or suppress perturbation-specific signals altogether, rendering them ineffective for prediction. To address this, we propose PerturbedVAE, a general framework designed to resolve this signal imbalance. The framework explicitly separates perturbation-specific information from dominant invariant structure and recovers causal representations to effectively utilize such information for prediction. We further provide an identifiability analysis that characterizes the conditions under which sparse perturbation effects can be reliably recovered, thereby clarifying how the framework can be concretely specified under such conditions. Empirically, PerturbedVAE achieves state-of-the-art performance on a widely used benchmark across multiple evaluation settings, yielding significant gains on out-of-distribution combinatorial predictions and uncovering interpretable perturbation-response programs.

Robustness is a critical requirement for deploying autonomous driving systems in the real world. Existing robustness benchmarks for autonomous driving have made important progress in studying the effects of image-level corruptions, such as adverse weather or camera degradation, on perception modules and open-loop planning outputs. However, deployment can also involve system-level imperfections, such as inference latency and ego-state estimation errors, which remain less studied in closed-loop E2E-AD evaluation. These imperfections can accumulate through the feedback loop and destabilize control. In this work, we present Bench2Drive-Robust, to our knowledge the first device-centric robustness benchmark for closed-loop end-to-end autonomous driving under realistic deployment perturbations. We systematically evaluate deployment-oriented perturbations arising from three major sources: camera-stream failures (frame drop, partial observation), ego-state estimation errors (GPS noise, and speed or odometry errors), and compute-induced control delay (model inference delay). We evaluate representative end-to-end driving methods and analyze their robustness under different perturbation severities. Our results show that these deployment-related perturbations can substantially degrade closed-loop driving performance, revealing robustness challenges that are not fully captured by conventional image-level corruption evaluations. By establishing a closed-loop evaluation protocol and demonstrating the substantial impact of these deployment-oriented perturbations, Bench2Drive-Robust defines practical robustness problems for end-to-end autonomous driving and encourages further research on deployment-aware robust driving systems.

We introduce a dynamics-level approach to watermarking generative models. Rather than embedding signals into model weights or outputs, we embed the watermark directly into the learned continuous dynamics -- the velocity field of a flow matching model. We formulate this as random coding over a continuous channel: a key-dependent perturbation is added during training, and the message is recovered at detection time from black-box queries. The perturbation is designed to leave the generated distribution unchanged. Experiments on MNIST and CIFAR-10 across different architectures confirm reliable message recovery, preserved generation quality, and chance-level decoding accuracy without the secret key.

Large vision-language models (LVLMs) have emerged as a powerful paradigm for multimodal intelligence, but their growing deployment also expands the attack surface of prompt injection. Despite this growing concern, existing attacks still suffer from a critical limitation: the injected prompt for one modality only steers the model's interpretation of that singular input. Alternatively, these attacks remain multimodal but fail to achieve cross-modal prompt perturbation. To bridge this gap, we introduce a novel cross-modal prompt injection attack CrossMPI, which can steer the model's interpretation of both textual and visual inputs via image-only prompt injection. Our design is underpinned by the following key breakthroughs. First, we turn the focus of the injected prompt perturbation optimization from the visual embedding space (typically with only $10^5$ parameters) to the model hidden state space (for multimodal information integration and with $10^7$ parameters). Then, two strategies are adopted to mitigate the optimization challenges posed by the larger parameter space. To constrain the optimized model parameter space, we introduce a layer selection strategy that identifies the layers most critical to multimodal integration. Interestingly, deviating from the past experience, our analysis reveals that the optimal layers for LVLM prompt perturbation reside in the middle of the model rather than the last. To constrain the image perturbation space, we propose a new distance-decremental perturbation budget assignment strategy that allocates budgets decrementally as the pixel distance to semantic-critical regions increases. Extensive experiments across multiple LVLMs and datasets show that our method significantly outperforms baseline approaches.

Recent advancements have established Reinforcement Learning (RL) as a pivotal paradigm for aligning generative models with human intent. However, group-based optimization frameworks (e.g., GRPO) face a critical limitation: the rapid decay of intra-group variance. As the distinctiveness among samples within a group diminishes, the variance approaches zero. This eliminates the very learning signal required for optimization, rendering the process unstable and forcing the policy into premature stagnation or reward hacking. Existing strategies, such as varying the initial noise or increasing group sizes, often fail to address this fundamental issue, resulting in training instability or diminishing returns. To overcome these challenges, we propose $\textbf{Embedding-perturbed Exploration Preference Optimization (}E^2\textbf{PO)}$, a novel framework that sustains optimization through embedding-level perturbation. Our method introduces structured, embedding-level perturbations within sample groups, guaranteeing a robust variance that preserves the discriminative signal throughout the training process. Extensive experiments demonstrate that our approach significantly outperforms state-of-the-art baselines, achieving a more faithful alignment with human preference.

Evaluation leaderboards such as LMArena play a central role in benchmarking large language models by aggregating pairwise human preferences into model rankings, yet the robustness of these rankings remains poorly understood. We present a unified perturbation framework for analyzing Bradley-Terry leaderboards under structured data modifications using influence-based approximations. Our framework studies three match-level perturbations -- Drop, Add, and Flip -- together with player removal, and evaluates their effects on top-k membership, global ranking consistency via Kendall's tau, and confidence-interval-based uncertainty. Across Chatbot Arena and six additional pairwise-comparison datasets, we show that modern leaderboards are non-robust across all three objectives: sub-1% targeted perturbations can change the top-ranked model, degrade Kendall's tau, and alter confidence intervals. Beyond robustness auditing, we show that the same influence scores enable efficient targeted perturbations, promoting or demoting specific models and reducing target-model uncertainty with fewer actions than previous manipulation and active-sampling baselines. By summarizing these effects with normalized dataset-level robustness scores, our framework provides a practical and helpful tool for auditing leaderboard stability and motivating more robust evaluation protocols.

Deep neural networks are highly vulnerable to adversarial examples, i.e.,small perturbations that can significantly degrade model performance. While adversarial training has become the primary defense strategy, most studies focus on balanced datasets, overlooking the challenges posed by real-world long-tail data. Motivated by the fact that perturbations in adversarial examples inherently alter the training distribution, we theoretically investigate their impact. We first revisit adversarial training for long-tail data and identify two key limitations: (i) a skewed training objective caused by class imbalance, and (ii) unstable evolution of adversarial distributions. Furthermore, we show that perturbations can simultaneously address both adversarial vulnerability and class imbalance. Based on these insights, we propose RobustLT, a plug-and-play framework that adaptively adjusts perturbations during adversarial training. Extensive experiments demonstrate that RobustLT consistently enhances adversarial robustness and class-balance on long-tailed datasets. The code is available at \href{https://github.com/zhang-lilin/RobustLT}{https://github.com/zhang-lilin/RobustLT}.

Recent advancements in large language models demonstrate that injecting perturbations can substantially enhance extrapolation performance. However, current approaches often rely on discrete perturbations with fixed designs, which limits their flexibility. In this work, we propose a framework where token prefixes are perturbed by a learnable transformation of a continuous latent vector within an embedding space. To overcome the challenge of an intractable marginal likelihood, we derive unbiased estimating equations for model parameters and optimize them via stochastic gradient descent. We establish the statistical properties of the resulting estimator in over-parameterized regimes. Empirical evaluations on both synthetic and real-world datasets demonstrate that our proposal yields significant gains in out-of-domain settings over a range of state-of-the-art baseline methods.

Randomized smoothing provides strong, model-agnostic robustness certificates, but existing guarantees are limited to single modalities, treating continuous and discrete inputs in isolation. This limitation becomes critical in multimodal models, where decisions depend on cross-modal semantics and adversaries can jointly perturb heterogeneous inputs, rendering unimodal certificates insufficient. We introduce a unified randomized smoothing framework for mixed discrete--continuous inputs based on an analytically tractable Neyman--Pearson formulation of the joint worst-case problem. By analyzing the joint likelihood ordering induced by factorized discrete and continuous noise, our approach yields a closed-form, one-dimensional certificate that strictly generalizes both Gaussian (image-only) and discrete (text-only) randomized smoothing. We validate the framework on multimodal safety filtering, providing, to our knowledge, the first model-agnostic Neyman--Pearson certificate for joint discrete-token and continuous-image perturbations in interaction-dependent text--image safety filtering.

Accurately and consistently indexing biomedical literature by publication type and study design is essential for supporting evidence synthesis and knowledge discovery. Prior work on automated publication type and study design indexing has primarily focused on expanding label coverage, enriching feature representations, and improving in-domain accuracy, with evaluation typically conducted on data drawn from the same distribution as training. Although pretrained biomedical language models achieve strong performance under these settings, models optimized for in-domain accuracy may rely on superficial lexical or dataset-specific cues, resulting in reduced robustness under distributional shift. In this study, we introduce an evaluation framework based on controlled semantic perturbations to assess the robustness of a publication type classifier and investigate robustness-oriented training strategies that combine entity masking and domain-adversarial training to mitigate reliance on spurious topical correlations. Our results show that the commonly observed trade-off between robustness and in-domain accuracy can be mitigated when robustness objectives are designed to selectively suppress non-task-defining features while preserving salient methodological signals. We find that these improvements arise from two complementary mechanisms: (1) increased reliance on explicit methodological cues when such cues are present in the input, and (2) reduced reliance on spurious domain-specific topical features. These findings highlight the importance of feature-level robustness analysis for publication type and study design classification and suggest that refining masking and adversarial objectives to more selectively suppress topical information may further improve robustness. Data, code, and models are available at: https://github.com/ScienceNLP-Lab/MultiTagger-v2/tree/main/ICHI

Distilled one-step (T=1) or few-step (T$\leq$4) diffusion models enable real-time image generation but often exhibit reduced sample diversity compared to their multi-step counterparts. In multi-step diffusion, diversity can be introduced through schedules, trajectories, or iterative optimization; however, these mechanisms are unavailable in the few-step or single-step setting, limiting the effectiveness of existing diversity-enhancing methods. A natural alternative is to perturb intermediate features, but naive feature perturbation is often ineffective, either yielding limited diversity gains or degrading generation quality. We argue that effective diversity injection in few-step models requires perturbations that respect the model's learned feature geometry. Based on this insight, we propose STRIDE, a training-free and optimization-free method that operates in a single forward pass. STRIDE injects spatially coherent (pink) noise into intermediate transformer features, projected onto the principal components of the model's own activations, ensuring that perturbations lie on the learned feature manifold. This design enables controlled variation along meaningful directions in the representation space. Extensive experiments on FLUX.1-schnell and SD3.5 Turbo across COCO, DrawBench, PartiPrompts, and GenEval show that STRIDE consistently improves diversity while maintaining strong text alignment. In particular, STRIDE reduces intra-batch similarity with minimal impact on CLIP score, and Pareto-dominates existing training-free baselines on the diversity-fidelity frontier. These results highlight that, in the absence of iterative refinement, improving diversity in few-step and one-step diffusion depends not on increasing perturbation strength, but on aligning perturbations with the model's internal representation structure.

Hardware-aware training (HAT) is widely used to improve the robustness of neural networks on non-ideal AI accelerators, such as analog in-memory computing (IMC) systems. However, not all hardware-induced distortions are equally compensable by training. This paper presents a diagnostic framework that models hardware non-idealities as structured perturbations of the forward operator and evaluates their compatibility with gradient-based optimization. We analyze six representative perturbation classes--read noise, variability, drift, stuck-at faults, IR-drop, and ADC discretization--and identify three key diagnostics: gradient expectation consistency, bounded gradient variance, and non-degenerate sensitivity. Our results show a clear separation between perturbations that can be compensated by HAT and those that consistently break optimization. This provides practical guidance for hardware-software co-design, clarifying which non-idealities can be addressed at the training level and which require circuit-, architecture-, or calibration-level mitigation. This study should be interpreted as a controlled empirical analysis under vanilla forward-perturbation HAT, rather than as a universal theory of hardware-aware training.

Log anomaly detection is a critical task for system operations and security assurance. However, in networked systems at scale, log data are generated at massive scale while instance-level annotations are prohibitively expensive, posing great difficulties to fine-grained anomaly localization. To address this challenge, we propose LogMILP (Log anomaly localization based on Multi-Instance Learning enhanced by prototypes and Perturbation), a weakly supervised framework that enables both bag-level anomaly detection and instance-level anomaly localization using only bag-level labels. Our method guides the model to pinpoint the critical log entries using prototype-guided structural modeling with counterfactual perturbation consistency regularization, thereby improving localization reliability and interpretability under coarse-grained supervision. Experimental results on three public datasets demonstrate that LogMILP achieves competitive detection performance while yielding significantly more reliable instance-level localization. Our code is open-sourced at https://github.com/YUK1207/LogMILP.

These notes introduce the theory of susceptibilities as developed in [arXiv:2504.18274, arXiv:2601.12703] for interpreting neural networks. The susceptibility of an observable $φ$ to a data perturbation is defined as a derivative of a posterior expectation, which by the fluctuation--dissipation theorem equals a posterior covariance. Different choices of $φ$ yield different objects: per-sample losses give the influence matrix (the Bayesian influence function of [arXiv:2509.26544]), while component-localized observables give the structural susceptibility matrix that pairs model components with data patterns. The susceptibility matrix is (up to a factor of $nβ$) the Jacobian of the map from data distributions to structural coordinates; its pseudo-inverse provides a linearized solution to the patterning problem of [arXiv:2601.13548]: finding data perturbations that produce a desired structural change. We motivate the theory from its statistical-mechanical foundations, then give a detailed exposition of susceptibilities, their empirical estimators, and their connection to the geometry of the loss landscape.

The Danish Technological Institute (DTI) focuses on transferring advanced technologies (including robots) to the industry and the public sector. One key application is laptop refurbishment using specialized robots, aimed at promoting reuse, reducing electronic waste, and supporting the European Circular Economy Action Plan. The software of such robots often includes features that use object detection models to detect objects for various purposes, such as identifying screws for laptop disassembly or detecting stickers to remove them. Ensuring the robustness of such models to small input variations remains a critical challenge, and addressing it is important to avoid potential damage to laptops during refurbishment. In this paper, we propose PROBE, a search-based robustness testing approach that leverages multi-objective optimization to identify minimal, localized perturbations that expose failures in object detection models used in the software of laptop refurbishing robots. PROBE employs NSGA-II to systematically explore the perturbation space, optimizing for failure induction considering both localization and confidence, and perturbation magnitude, while enabling the discovery of diverse failure cases. Results show that PROBE is 3$\times$ to 7$\times$ more effective than random search in generating failure-inducing perturbations, while requiring smaller perturbation magnitudes, and that the generated perturbations transfer across models. We further show that metamorphic relations provide additional insights into model robustness, enabling the assessment of stability even in non-failing cases.

Benchmarks like GSM8K are popular measures of mathematical reasoning, but leaderboard gains can overstate true capability due to memorization of fixed test sets. Most robustness variants apply surface-level perturbations (paraphrases, renamings, number swaps, distractors) that largely preserve the underlying facts, and static releases can themselves become memorization targets over time. We introduce GSM-SEM, a reusable and stochastic framework for generating semantically diverse benchmark variants with substantially higher semantic variance than prior approaches. GSM-SEM perturbs problem statements by modifying entities, attributes, and/or relationships, frequently altering underlying facts and requiring models to recompute solutions under new conditions, while constraining generation to preserve the original calculations/answer and approximate problem difficulty. GSM-SEM generates fresh variants on each run without requiring re-annotation, reducing reliance on static public benchmarks for evaluation and thereby lowering the bias of memorization. We apply GSM-SEM on GSM8K and two existing variation suites (GSM-Symbolic and GSM-Plus), producing GSM8K-SEM, GSM-Symbolic-SEM, and GSM-Plus-SEM. Evaluating 14 SOTA LLMs, we observe consistent performance drops with larger decline when semantic perturbations are coupled with symbolic/plus variations (average drop rate 28% in maximum strictness configuration of GSM-SEM). We publicly release the three SEM variants as fully human-validated datasets. Finally, to demonstrate applicability beyond GSM-style math problems, we apply GSM-SEM to additional benchmarks including BigBenchHard, LogicBench, and NLR-BIRD.

Objective perturbation is a standard mechanism in differentially private empirical risk minimization. In particular, Linear Objective Perturbation (LOP) enforces privacy by adding a random linear term, while strong convexity and stability are ensured by an additional deterministic quadratic term. However, this approach requires the strong assumption of bounded gradients of the loss function, which excludes many modern machine learning models. In this work, we introduce Quadratic Objective Perturbation (QOP), which perturbs the objective with a random quadratic form. This perturbation induces strong convexity and enforces stability of the problem through curvature, thereby enabling privacy and allowing sensitivity to be controlled through spectral properties of the perturbation rather than assumptions on the gradients. As a result, we obtain $(\varepsilon, δ)$-differential privacy under weaker assumptions, in the interpolation regime. Furthermore, we extend the analysis to account for approximate solutions, showing that privacy guarantees are preserved under inexact solves. Additionally, we derive utility guarantees in terms of empirical excess risk, and provide a theoretical and numerical comparison to LOP, highlighting the advantages of curvature-based perturbations. Finally, we discuss algorithmic aspects and show that the resulting problems can be solved efficiently using modern splitting schemes.

In grokking, a model first fits the training data while test accuracy remains low, and only later begins to generalize. We ask whether this transition can be localized from observed training trajectories before the test accuracy rises, and formulate grokking transition localization as a diagnostic problem with an explicit threshold/FPR/lead-time trade-off. Task-dependent observables are summarized as empirical distributions, mapped to Wasserstein/quantile coordinates, and analyzed by Hankel dynamic mode decomposition (DMD); the resulting reconstruction residual, together with spectrum and effective rank, forms the diagnostic output. On held-out modular-addition Transformer runs, the residual achieves AUROC $\approx$ 0.93 for grokking-vs-non-grokking discrimination at the run level; under a fixed sustained-threshold operating rule, true-positive alarms can precede onset, with lead time reported jointly with false-alarm rate and uncertainty intervals. Perturbation experiments show that, in the tested $wd=1$ pool, high-residual windows exhibit about $3\times$ larger short-horizon perturbation deviation than low-residual windows. In a same-data norm-window control, perturbation sensitivity aligns with the residual ordering rather than total-parameter-norm ordering, suggesting that the residual is not merely a total-norm proxy at the window level in the studied $wd=1$ dynamics. Norm signals remain strong run-level regime indicators, and log-probability performs best among the observables tested under the current protocol. We position the residual as a window-level monitoring and localization signal in the studied modular-arithmetic Transformer settings, not a universal early-warning predictor or an intervention rule.

Reinforcement learning with verifiable rewards, particularly Group Relative Policy Optimization (GRPO), has significantly advanced the reasoning capabilities of Large Language Models (LLMs). However, in complex tasks, GRPO frequently suffers from the ``zero-advantage problem'': when all sampled rollouts for a query fail, the relative advantage collapses to zero. Consequently, the model loses effective training signals for these questions, wasting the training data and computational budget. While simply increasing the sampling budget for these questions is a common remedy, the static sampling policy inherently constrains reasoning exploration, limiting the success rate. In this paper, we propose Lorem Perturbation for Exploration (LoPE), a simple yet effective training framework to break this exploration bottleneck. We posit that task-irrelevant prompt-space perturbations can shift the model's output distribution enough to unlock orthogonal reasoning pathways for hard questions. Specifically, LoPE prepends sequences stochastically assembled from Lorem Ipsum vocabulary (a pseudo-Latin placeholder text) to the prompts before resampling. Experiments across 1.7B, 4B, and 7B models demonstrate that LoPE significantly outperforms resampling with the original prompts. Further analysis reveals that other Latin-based random sequences with low perplexity are also effective perturbations. Our results establish LoPE as a strong baseline for broadening exploration in LLM reinforcement learning.

We introduce a simple yet powerful framework for training large language models. In contrast to the standard autoregressive next-token prediction based on an exact prefix, we propose a perturbation-based procedure that first transforms the prefix into a semantic neighbor and then conditions on this perturbed variant for next-token prediction. This yields a hierarchical model with a pre-post-additive noise structure. Within this framework, we develop a rigorous theory of extrapolability, namely, the capacity of a model class to make reliable predictions for token sequences that lie outside the empirical support of the training corpus. We evaluate the finite-sample performance of the proposed procedure using both synthetic and real-world language data. Results show that the proposed method consistently improves out-of-support prediction while maintaining competitive in-support performance, demonstrating that perturbation offers a practical route to language modeling.

Multi-constraint instruction following requires verifying whether a response satisfies multiple individual requirements, yet LLM judges are often assessed only through overall-response judgments. We introduce MCJudgeBench, a benchmark for constraint-level judge evaluation in multi-constraint instruction following. Each instance includes an instruction, a candidate response, an explicit constraint list, per-constraint gold labels in {yes, partial, no}, and controlled response-side perturbations. The evaluation protocol further includes evaluation prompt variants to test judge stability. We evaluate proprietary and open-source LLM judges using both correctness and inconsistency metrics, distinguishing intrinsic inconsistency under stochastic decoding from procedural inconsistency under prompt and response perturbations. Our results show that judge reliability has multiple dimensions: strong overall performance does not guarantee equally reliable detection across label categories, especially for rarer partial and no cases. Judges with higher correctness do not always have lower inconsistency. Evaluation with reasoning improves correctness but does not uniformly improve stability. These findings motivate evaluating LLM judges at the constraint level to study these failure modes.

Privacy concerns in distributed learning often lead clients to return intentionally altered gradient information. We consider the problem of learning convex and $L$-smooth functions under adversarial gradient perturbation, where a client's gradient reply to a server query can deviate arbitrarily from the true gradient subject to a distance bound. Our study focuses on two fundamental questions: (i) what is the smallest achievable sub-optimality gap (i.e., excess error in optimization) under such responses, and (ii) how many queries are sufficient to guarantee a given sub-optimality gap? We establish tight feasibility thresholds on the sub-optimality gap and provide algorithms that achieve these thresholds with provable query complexity guarantees.

We show that a threshold in decision capacity determines whether self-play reinforcement learning agents collapse under asymmetric rule perturbations. Across poker variants, matrix games, a dice game, and multiple learning algorithms, eliminating all positive-reach contingent decisions causes rapid convergence to a deterministic exploitation attractor, a fixed point at near-maximal loss. Preserving even a single positive-reach contingent decision point prevents this collapse. A frozen baseline and fixed-opponent control confirm that the mechanism is co-adaptation under constraint, not the perturbation itself. The phenomenon is timing-invariant, fully reversible upon action restoration, and intensifies under function approximation. These results establish a sharp threshold at zero reach-weighted contingent action capacity, with severity scaling continuously via reach-weighted capacity in the tested domains.