arXiv:2606.06497v2 Announce Type: replace Abstract: Generative video models have achieved remarkable visual fidelity, yet their prompt-only interface offers thin creative agency and obscures the model's material process from the artists working with it. We present Real-Time AttentionBender, a tool that extends the practice of network bending across the full depth of the video diffusion transformer (DiT) and brings it into live, interactive generation. Built as a plugin within the DayDream Scope ecosystem and wrapping open-source real-time Wan pipelines, the tool exposes self-attention, cross-attention, and the feed-forward network as independently manipulable surfaces, with targeting down to individual diffusion steps, DiT layers, prompt tokens, and hidden neurons. The immediacy of live manipulation affords what we call "material intimacy" with the model: a responsive, near-mechanistic feel for how specific layers and neurons shape generated video. We position the tool as simultaneously an XAIxArts probe into transformer internals and an expressive instrument for discovering aesthetics outside the model's default representational space.
Science Journals
arXiv:2507.00260v3 Announce Type: replace-cross Abstract: When predictors are statistically dependent, the appropriate definition of feature importance depends on the operational goal. Conditional-incremental measures are well-suited for feature selection, acquisition, and compression, where shared predictive information is treated as redundancy. For post-hoc interpretation, however, the goal is often to attribute predictive signals across correlated measurement channels. We introduce Disentangled Feature Importance (DFI), a population-level attribution framework for this setting. DFI maps covariates to an independent latent representation under a specified entropic optimal transport geometry, computes latent importance, and attributes it back to the original covariates through barycentric sensitivities. We show that broad conditional-incremental FI functionals target conditional incremental predictive value under squared-error loss, and therefore answer a different question from attribution of shared predictive signal under dependence. Under fixed transport cost, reference law, and regularization level, DFI defines a well-specified family of estimands. Latent scores admit a functional ANOVA interpretation, and in the Gaussian linear case, the attributed DFI recovers the classical $R^2$ decomposition for correlated regressors. We derive influence-function-based inference under nuisance-rate and smoothness conditions, and show in simulations and an HIV-1 neutralization-resistance analysis that DFI yields stable, interpretable, uncertainty-quantified attributions of shared predictive signal.
arXiv:2606.09517v1 Announce Type: new Abstract: As renewable energy integration increases market volatility, probabilistic electricity price forecasting has become essential for effective risk management. However, current-proper-scoring rules often prioritize forecast sharpness at the expense of calibration, leading to overconfident and statistically unreliable uncertainty estimates. This work highlights the critical gap between theoretical scoring and practical calibration, demonstrating that models can become mere proxies for deterministic forecasts when reliability is neglected. We conclude that future research must shift toward calibration-aware objectives and architectures to ensure the distributional integrity of energy market forecasts.
arXiv:2606.09523v1 Announce Type: new Abstract: We introduce a unified theoretical framework for single-plasmon transport in one-dimensional nanowires, bridging the quantized electromagnetic Green's tensor formalism with effective non-Hermitian Hamiltonian models. This approach naturally incorporates propagating surface plasmon polaritons, high-order modes dissipative channels, and intrinsic losses. We investigate both the stationary regime and the spatio-temporal dynamics of a single-plasmon pulse travelling through an atomic chain coupled to a dispersive nanowire. We analyze modal contributions to reflection and transmission spectra for quantum emitter coupled to a silver nanowire, a configuration proposed as a single-plasmon transistor, and we demonstrate that optimized multi-emitter systems offer significant advantages. In case of one quantum emitter coupled to a silver nanowire at telecom wavelengths, we predict a single-plasmon transmittivity down to 7\% under realistic conditions, and an atomic qubit population of 12\%. Extension to multi-emitter systems using a L\"owdin orthogonalization procedure enables a consistent treatment of collective interactions. We show that optimized positioning with just five emitters enhances plasmon modulation, achieving a transmittivity of 2\% but also reduces coupling losses to one-third compared to the single-emitter case. Our results establish a robust foundation for analyzing and designing plasmonic waveguide quantum electrodynamics systems.
arXiv:2606.09525v1 Announce Type: new Abstract: During instruction fine-tuning (IFT), large language models (LLMs) learn to follow instructions by using the provided context to answer a query. While prior work has studied how context characteristics correlate with context usage by the LLM, this analysis has been limited to inference time, leaving open how these relationships are acquired in the first place. Here, we measure how models' sensitivity to such characteristics shifts across successive IFT stages: supervised fine-tuning (SFT), direct preference optimization (DPO), and reinforcement learning with verifiable rewards (RLVR). Experiments across four models and three datasets show that SFT makes models more likely to use contexts that are easy to understand, such as containing high length, context-query similarity, and fluency. Post-SFT dynamics may either reinforce or resolve these preferences depending on the training dataset. Our findings reveal that context usage is actively reshaped at each IFT stage, and designing a balanced IFT dataset is important in ensuring robust context utilization of instruction-tuned models.
arXiv:2606.09528v1 Announce Type: new Abstract: Dynamic analysis of Android's application layer typically relies on physical devices, limiting scalability and reproducibility. To compensate, we introduce a systematic re-hosting method that relocates the Android framework and pre-installed software from real device firmware into a fully emulated environment. Our approach integrates vendor-specific components into the Android Open Source Project (AOSP) build system using tailored extraction and injection strategies, producing vendor-flavoured emulator images that preserve system integrity and runtime compatibility. This enables dynamic execution of real-world framework and application-layer components, including proprietary binaries and pre-installed apps, across multiple SDK versions. We evaluate our method on 184 firmware samples from SDK 31-33. It achieves high build and boot success rates, with residual failures primarily occurring during core-service initialization due to baseline strategy limitations, missing dependencies, device-protection checks, or emulator constraints. However, the modular design allows injection strategies to be extended for specific firmware, supporting broader compatibility and future research on automated, adaptive re-hosting. Though we identified potential for optimization through engineering vendor-specific solutions, our research demonstrates the feasibility of vendor-flavoured emulators for scalable, reproducible dynamic analysis.
arXiv:2606.09529v1 Announce Type: new Abstract: The Traveling Salesman Problem (TSP) is a classical NP-hard combinatorial optimization problem that aims to find the shortest Hamiltonian cycle visiting each city exactly once and returning to the starting point. This paper proposes a hybrid metaheuristic for the TSP by combining the Dragonfly Algorithm (DA), a swarm-intelligence-based global search method, with Tabu Search (TS), a memory-based local search technique. The proposed method follows a High-Level Relay Hybridization (HRH) scheme, in which DA is first used to explore the solution space and generate a promising initial tour, while TS subsequently refines this solution through neighbourhood-based improvement and tabu memory. The hybrid approach is evaluated on standard TSPLIB benchmark instances, including burma14, att48, and ch150, and compared with standalone DA, standalone TS, and several classical metaheuristics such as Genetic Algorithm, Ant Colony Optimization, Particle Swarm Optimization, and Random Search. A systematic grid-search procedure is also conducted to study the influence of the main hyperparameters on solution quality and execution time. The experimental results indicate that the proposed hybrid can improve tour quality compared with the standalone DA and TS on the tested instances, highlighting the benefit of combining global exploration with local exploitation. However, the results also suggest that performance remains sensitive to parameter settings and problem size, motivating further validation on larger benchmarks and stronger TSP-specific baselines.
arXiv:2606.09530v1 Announce Type: new Abstract: This contribution presents an alternative stabilization technique for the virtual element method (VEM) based on reduced integration combined with a scaled boundary parametrization. To this end, a Taylor series expansion of the constitutive quantities with respect to the sectional center is carried out, enabling analytical integration of the weak form and reducing the need for integration points to only one per section. The accuracy of the proposed formulation is shown by several numerical examples, including a non-linear patch test. Different loading, e.g. compression under large deformations, and material conditions, such as hyperelastic anisotropy and elasto-plasticity, are considered. The biquadratic serendipity finite element formulation (Q2) and the low-order finite element formulation with hourglass stabilization (Q1STc+) are used for comparison. While the patch test was not fulfilled using higher order shape functions, the formulation led to good results and was able to capture the structure's response accurately. Furthermore, the formulation performed better when the physical element resembled the pre-assigned parent elements. The example of the asymmetrically notched specimen under elasto-plastic material behavior showed that the proposed formulation is able to capture inelasticities.
arXiv:2606.09532v1 Announce Type: new Abstract: Crises alter both how people move and how they communicate. During emergencies such as wildfires and pandemics, changes in mobility patterns and online emotional discourse evolve jointly, yet they are typically studied in isolation. This paper presents a unified and interpretable pipeline that integrates mobility and social media data to identify cross-domain behavioral patterns in crisis settings. The framework is evaluated through two case studies: a short-horizon analysis of the January 2025 Los Angeles wildfires (prototype case) and a longitudinal analysis of UAE COVID-19 behavior from March 2020 to December 2021 (primary case, 671 days). The pipeline aligns heterogeneous daily signals, transforms them into binary behavioral states, applies Formal Concept Analysis (FCA) to extract co-occurrence structure, mines association rules, and validates rule stability through chronological holdout testing. A structured policy-translation layer renders robust rules as operational briefs specifying triggers, lead times, and action playbooks. Results reveal clear cross-domain behavioral structure in both crises. In the wildfire case, traffic stress, fear/anger sentiment, and governance discourse are tightly coupled within a 33-day window, with key rules reaching 100\% confidence and lift scores up to 2.5. In the COVID case, repeated mobility adaptation and sentiment volatility yield 8 stable same-day rules (88\% holdout pass rate) and 40 clean predictive rules with 2--7 day lead horizons. The work demonstrates that interpretable multimodal fusion can produce both scientifically credible and policy-actionable crisis intelligence.
arXiv:2606.09724v1 Announce Type: new Abstract: Retrieval-Augmented Generation (RAG) has become a standard architectural response to unreliability in legal AI, yet high-profile failures, including fabricated citations submitted to courts and anachronistic legal content presented as current, continue to appear across jurisdictions. We argue that these failures are not residual confabulations to be eliminated by scaling language models, but symptoms of an architectural mismatch between probabilistic retrieval and the hierarchical, temporal, and institutional structure of legal knowledge. We develop the argument in three moves. First, we articulate the ontological commitment of legal knowledge as a triad of properties derivable from classical legal theory: hierarchical and mereological structure, diachronic dynamism under operational closure, and causal traceability of institutional provenance grounded in the duty of justification. Second, we identify three corresponding pathologies of retrieval (mereological blindness, diachronic blindness, and causal opacity), each developed with an operational definition, a failure mechanism, a canonical example, and detection criteria for diagnostic use. Third, we review the state of the art through this lens, showing that existing approaches address these requirements unevenly and do not yet compose into a paradigm that treats them as co-constitutive. From this analysis we derive four architectural commitments that characterize the deterministic-by-design direction for legal retrieval: ontological primacy, event reification, bitemporal correctness, and deterministic interaction protocols. The framework concerns quaestio juris (which norms apply and in what state) rather than the downstream tasks that act on identified norms, and addresses legislative and constitutional retrieval primarily, with interpretive time as an explicit extension.
arXiv:2606.08659v1 Announce Type: new Abstract: Since its classical origin, Hamilton's principle has been formulated under an exact closure condition: admissible variations vanish at the boundaries of the variational domain. This condition removes the boundary term in the first variation of the action and yields the Euler--Lagrange equation. Although natural for isolated deterministic systems, fixed boundary admissibility is usually treated as a technical condition rather than as a physical closure hypothesis. Here we ask what follows when this hypothesis is made explicit and relaxed. We introduce \emph{variational openness} as the retention of the boundary contribution in the variational balance. The retained term defines a boundary-openness density, which must be projected onto admissible variations before it becomes a dynamical source. In this formulation, the classical Euler--Lagrange equation is recovered as the exact-closure limit of an open variational balance; the source term is therefore identified with incomplete variational closure rather than with an externally imposed force. The framework is illustrated through three elementary examples: an open harmonic oscillator, a finite-compliance boundary, and a delayed oscillator with memory. These examples show how boundary openness can produce forcing, partial closure, history dependence, and non-Markovian structure while preserving standard mechanics in the closed limit. The resulting perspective suggests that Hamiltonian mechanics may be understood as the mechanics of variationally closed systems and motivates an open Hamilton--Jacobi theory in which admissibility itself becomes dynamical.
arXiv:2601.01279v3 Announce Type: replace-cross Abstract: When competing sellers delegate pricing to a shared AI model, such as a large language model, correlated recommendations combined with performance-driven updates aggregating seller feedback raise a key question: can standard AI deployment practices inadvertently produce supracompetitive pricing? We develop a stylized duopoly model in which two sellers receive pricing recommendations from a shared AI characterized by two parameters: a propensity parameter capturing the model's tendency to set high prices and an output-fidelity parameter measuring alignment between this tendency and actual outputs, with propensity updated via periodic retraining on observed outcomes. We find that configuring AI models for robustness and reproducibility can lead to supracompetitive pricing via a phase transition. Below a critical output-fidelity threshold, competitive pricing is the unique stable outcome. Above it, the model exhibits bistability: both competitive and supracompetitive pricing are locally stable, with the realized outcome determined by the model's initial propensity. Supracompetitive pricing raises average prices, but occasional low-price recommendations complicate detection. With perfect output fidelity, full price coordination emerges from any interior initial propensity. For finite training batches of size $b$, when the initial propensity lies in the supracompetitive basin, the probability of supracompetitive pricing approaches 1 as $b$ increases, with the region of indeterminate outcomes shrinking at rate $O(1/\sqrt{b})$. Any factor reducing alignment between the model's propensity and sellers' actual pricing, whether through diversifying AI providers, introducing recommendation noise, or reducing seller adherence, pushes the market toward competitive outcomes.
arXiv:2606.09754v1 Announce Type: new Abstract: AI-mediated information manipulation increasingly takes the form of social cyber attacks that target trust, attention, credibility, reputation, and decision-making rather than only technical infrastructures or isolated false contents. Existing defensive approaches often oscillate between incident-level analysis, which fragments campaigns into weak signals, and attribution-first analysis, which may delay mitigation until responsibility is established. This paper proposes a SOCMINT framework based on Information Manipulation Sets (IMS) as an intermediate operational unit between individual incidents and strategic attribution. Building on the VIGINUM/EEAS use of IMS in counter-FIMI analysis, the framework treats manipulation as a coherent process involving narratives, accounts, infrastructures, temporal patterns, cross-platform migration, synthetic amplification, and cognitive targeting. The proposed pipeline moves from signal detection and diagnostic triage to IMS hypothesis construction, confidence/severity assessment, mitigation selection, and iterative update. A compact scenario illustrates how IMS-based analysis captures what content-level and attribution-first approaches miss. The paper also proposes a tabletop evaluation protocol to assess decision quality, confidence calibration, and mitigation proportionality. The main implication is that human-centred risk mitigation requires not only better detection, but also structured reasoning under uncertainty, auditable decision-making, and safeguards against over-securitising legitimate dissent.
arXiv:2606.09697v1 Announce Type: new Abstract: Large language models (LLMs) routinely face requests that should be refused, creating a trade-off between helpfulness and harm prevention. However, refusals themselves can be helpful. In high-risk interactions involving crisis, coercion, or escalating intent, blunt non-compliance may prevent direct harm while still failing to support the needs of the person behind the request. We present PsychoSafe, a psychologically-informed refusal framework that reframes refusal as structured supportive communication grounded in evidence-based intervention strategies. To develop PsychoSafe, we construct a corpus of 8019 prompt-response pairs spanning five psychologically salient risk domains and apply prompting and parameter-efficient fine-tuning to Qwen 3.5 27B. On a balanced validation set of 500 prompts, evaluated with an LLM judge and validated through human ratings, PsychoSafe prompting improves overall refusal quality by 28.1% over a generic baseline, with particularly strong gains in external resource referral (+46.8%) and psychological grounding (+34.8%), while preserving downstream performance on non-refusal tasks. Fine-tuning achieves near-perfect refusal and resource-referral rates but reduces response relevance. Additional evaluations on SORRY-Bench and XSTest show strong in-domain robustness but limited out-of-domain generalization, suggesting that future work should diversify fine-tuning data to help models apply interventions selectively rather than schematically.
arXiv:2606.09698v1 Announce Type: new Abstract: We study the problem of optimally sending a real-valued source through multiple uses of a channel with feedback. First, we state a set of conditions that are sufficient for an encoder to achieve maximal mutual information between the source and all the channel outputs. This set of conditions are also necessary when the channel is input-identifiable, a condition widely satisfied by common channel models. More notably, we further study the information maximization-distortion minimization problem, where the mutual information between the source and all channel outputs still needs to be maximized, while at each step, the MMSE of estimating the source from the channel outputs so far also needs to be minimized. We derive a solution to this problem for discrete channels with certain symmetries, e.g. $k$-ary symmetric or $k$-ary erasure channels. We show that for such channels the famous posterior matching scheme, while not necessary for information maximization alone, is sufficient and essentially necessary for achieving both information maximization and distortion minimization. This work also provides a new perspective of regularizing distortion-minimizing feedback communication through information maximization, which enables us to find the optimal solution that otherwise would be intractable.
arXiv:2606.09699v1 Announce Type: new Abstract: The state-of-the-art generative models, such as CycleGAN, Pix2Pix, and diffusion models have demonstrated remarkable performance in the face generation task. However, they fail to effectively capture cross-modality semantic information in craniofacial reconstruction when translating from the skull (x-ray) to the face (optical) domain, due to a mismatch in the alignment of structural identity across modalities. To address this issue, we propose Cranio-Diff, a diffusion-based framework for cross-domain cranio-facial reconstruction from 2D X-ray skull images. The proposed approach integrates skull-conditioned structural guidance through ControlNet with biometric text conditioning to generate a face which is more semantically and structurally aligned with the given skull. The proposed Cranio-diff method is evaluated on skull-face dataset obtained from X-ray scans of 120 subjects in lateral and frontal views. To enable controlled evaluation, each face image is synthesised across three age groups (25, 45, 65) and three BMI variations of -10%, baseline and +10%, yielding 4320 paired samples. To the best of our knowledge, this is the only X-ray-face dataset with this magnitude. Extensive experiments showed that the proposed method outperforms recent existing approaches in both generated image quality and retrieval task. Finally, to evaluate the performance of our proposed method, we have evaluated the quality of the generated image using FID, IS, SSIM, LPIPS, PSNR and ArcFace score. Additionally, retrieval performance is evaluated using recall@k, mAP@k and MRR@k. Obtained experimental results demonstrate that the proposed method can be used as an alternate tool in providing aid in forensic investigations.
arXiv:2606.09536v1 Announce Type: new Abstract: Conventional one-hot encodings often yield poorly calibrated models, being overconfident under attack, and letting entropy-based detection algorithms fail. Previous image classification works have demonstrated that Hadamard-coded output representations can improve adversarial robustness. However, attempts to integrate Hadamard codes into semantic segmentation fall far behind state-of-the-art models in mean intersection-over-union performance. Regarding object detection, such output encodings have not yet been investigated at all. Further, no prior art addressed intrinsic codeword inconsistencies or actually exploited intrinsic codeword redundancy. Accordingly, we first derive a novel decoding procedure for Hadamard codewords towards optimal class-wise probabilities, solving the underlying optimization problem by using the projection onto the probability simplex. Second, our optimization delivers a measure of prediction inconsistency. Third, we are the first to show how to exploit these inconsistencies for adversarial attack and disturbance detection. Fourth, we introduce HadamardNet, a framework employing Hadamard codes as output representations for semantic segmentation and object detection models and tasks. We conduct a comprehensive evaluation both on disturbances and adversarial attacks, achieving state-of-the-art perturbation detection performance for both tasks in only a single detection pass, while delivering equivalent or close-by reference performance on clean data.
arXiv:2606.09758v1 Announce Type: new Abstract: Parametric imitation learning via behavior cloning can suffer from poor generalization to out-of-distribution states due to compounding errors during deployment. We show that reusing the training data during inference via a semi-parametric retrieval-based imitation learning approach can alleviate this challenge. We present Difference-Aware Retrieval Policies for Imitation Learning (DARP), a semi-parametric retrieval-based imitation learning approach that addresses this limitation by reparameterizing the imitation learning problem in terms of local neighborhood structure rather than direct state-to-action mappings. Instead of learning a global policy, DARP trains a model to predict actions based on $k$-nearest neighbors from expert demonstrations, their corresponding actions, and the relative distance vectors between neighbor states and query states. DARP requires no additional assumptions beyond those made for standard behavior cloning -- it does not require additional data collection, online expert feedback, or task-specific knowledge. We demonstrate consistent performance improvements of 15-46% over standard behavior cloning across diverse domains, including continuous control and robotic manipulation, and across different representations, including high-dimensional visual features. Code and demos are available at https://weirdlabuw.github.io/darp-site/.
arXiv:2606.08682v1 Announce Type: new Abstract: Activation steering has emerged as a popular inference-time technique for modulating the behavior of large language models (LLMs). By constructing a steering vector from examples of a target behavior and injecting it into intermediate activations during inference, activation steering enables flexible behavioral control while avoiding the permanent parameter updates required by finetuning. Meanwhile, recent work has identified emergent misalignment (EM) as a significant safety concern, wherein models finetuned on unsafe examples from a narrow task may unexpectedly generalize to broadly unsafe behavior on unrelated tasks. Although finetuning-induced EM has been extensively studied, whether activation steering can induce EM remains comparatively under-explored, despite its increasing use as a model-control technique. In this paper, we present a comprehensive study of activation-steering-induced emergent misalignment, substantially expanding the evaluation scope beyond existing pioneering work. First, we show that activation steering can induce broad misalignment, even in the recent Qwen-3.5 series. Moreover, activation-steered models produce harmful responses with stronger semantic relevance and higher coherence than their finetuned counterparts, making the resulting misalignment potentially more harmful. Second, we characterize properties of AS-induced EM by analyzing key steering-specific factors, including steering magnitude, the low-rank structure of the steering subspace, and the number of epochs during steering-vector construction. Third, we evaluate the robustness and sensitivity of AS-induced EM across diverse model families, model scales, target tasks, and intervention layers. Our findings reveal activation steering as a significant yet under-examined source of emergent misalignment and provide an activation-space perspective for understanding the mechanisms and safety risks of EM.
arXiv:2606.09541v1 Announce Type: new Abstract: Single-Molecule Force Spectroscopy (SMFS) provides unprecedented insights into biomolecular mechanics, yet the high-throughput generation of force-extension trajectories creates a severe data curation bottleneck. Identifying rare molecular unbinding events within thousands of noise-dominated curves traditionally relies on tedious, non-scalable manual auditing. Here, we present a system-agnostic, interpretable deep learning framework tailored to overcome extreme class imbalance in automated SMFS triage. Utilizing 1D-to-2D rasterized geometric matrices, we deployed a modified ResNet18 architecture governed by an asymmetric Focal Loss objective function. We evaluated this framework on the complex mechanical unfolding pathways of the R. champanellensis cellulosome. Under hyper-imbalanced test conditions where the target interaction constituted only 1.34% of the dataset (13 true events out of 970 traces), the model achieved an overall accuracy of 0.9196 and a remarkable True Positive Rate (Recall) of 0.9231. By implementing an empirically calibrated dual-threshold triage system, the pipeline automatically discarded 880 unambiguous background noise traces , reducing the manual curation workload by over 90% while safely preserving high-value rare data. Finally, Gradient-weighted Class Activation Mapping (Grad-CAM) visually validated that the network's decisions are firmly anchored in the relevant geometric features of the force curves, specifically localizing on the structural unbinding regions, effectively mitigating 'black-box' skepticism. Built for free cloud-based execution, this open-source tool democratizes scalable, highly precise molecular discovery across the biophysics community.
arXiv:2606.09550v1 Announce Type: new Abstract: While LLM-based agents are increasingly used in scientific workflows, it remains unclear whether they are truly qualified for the dynamic and uncertain process of discovery. Existing static evaluations often conflate genuine reasoning with rote memorization. We introduce InquiTree, a diagnostic environment that formalizes scientific inquiry as interactive Research Trees: directed acyclic graphs capturing the logical dependencies among hypothesis formulation, study design, result interpretation, and belief updating. Evaluating agents on a 30-paper test pool and releasing the open-access InquidTree-18(IT-18) subset, we identify two key limitations. First, agents exhibit an "Erosion of Marginal Capabilities": during long-horizon interactions, they develop "cognitive tunneling," where critical judgment and anomaly detection degrade relative to their intrinsic baselines. Second, performance drops on papers published after model training cutoffs, revealing a boundary between interpolation and extrapolation and suggesting that apparent competence is partly driven by parametric memory. These findings indicate that scaling context alone is insufficient for reliable AI scientists; stronger architectures or human oversight may be required to preserve critical evaluation and generalization.
arXiv:2606.04752v2 Announce Type: replace Abstract: Transformers consuming multi-channel scalar signals must embed $C$ simultaneous values into one $d_{\text{model}}$-dimensional vector per time step. We audit eight input encoders -- a shared-scalar baseline, per-channel linear projections, an orthogonality regulariser, a nonlinear MLP, block-partitioned concatenation, channel-independent and channel-as-token architectures, and a projected positional encoding -- on a synthetic benchmark where channel identity is informative and on ETTh1, scored by next-step negative log-likelihood. The headline is practical near-equivalence within a wide "top tier": the standard per-channel linear projection matches every alternative up to small, statistically real but practically modest differences. A direct geometric probe attributes this to a spontaneous orthogonalisation of the per-channel projections: they end up near-orthogonal with no explicit regulariser, letting the standard linear recover channel identity from the summed embedding. Two encoders lose decisively: the shared-scalar baseline collapses for information-theoretic reasons we make explicit, and the channel-independent PatchTST-spirit baseline overfits universally on the synthetic benchmark and underperforms on both. Paired tests resolve two small gaps: projecting the sinusoidal positional encoding through a learned linear layer edges the rest at small $C$ by extending this orthogonality to the positional subspace; a nonlinear MLP stem edges them at the largest $C$, with the gap shrinking under more training data. The practical recommendation: use the standard per-channel linear projection by default; reach for something more elaborate only when the task calls for it.
arXiv:2205.00028v2 Announce Type: replace-cross Abstract: Dielectric elastomers are promising for soft robotics and wearable electronics and sensors, but their use is hindered by the high electric fields required. Maximizing electromechanical coupling through molecular mechanisms is essential. However, progress on the role of dipole-dipole interactions between monomers has been limited, in part because the resulting energy landscapes, characterized by multiple symmetric wells separated by high barriers, are difficult to sample with standard methods. This work develops a symmetry-adapted Markov chain Monte Carlo method that exploits the invariance of dipole-dipole interaction energies under simultaneous reflection of neighboring monomer orientations about the plane orthogonal to the applied electric field. Variable sized clusters of neighboring monomers are constructed and reflected, enabling rapid transitions between symmetric energy wells and rendering feasible simulations that are otherwise intractable. Freely jointed chains with anisotropic monomer polarizability reveals qualitatively distinct phenomena depending on the orientation of the monomer dipole relative to its backbone axis. Field-aligning chains exhibit electrically induced tautness, and an apparent compressive stiffness, while field-disaligning chains exhibit local folding and electrically induced collapse at large enough fields. Monomer orientational order is quantified across the applied field and susceptibility parameter space, revealing sharp transitions suggestive of underlying phase transitions. For field-disaligning chains, these orientational transitions correspond directly with sharp changes in chain polarization, linking microstructural rearrangement to the macroscopic dielectric response. The symmetry-adapted approach generalizes naturally to other multifunctional polymer systems whose energy landscapes possess analogous discrete symmetries.
arXiv:2311.17210v5 Announce Type: replace-cross Abstract: We determine sufficient conditions under which certain recursively defined functions are well defined for all real inputs. Given a function $f:\mathbb R\to\mathbb R$, call a decreasing sequence $x_1>x_2>x_3>\cdots$ "$f$-bad" if $f(x_1)>f(x_2)>f(x_3)>\cdots$, and call the function $f$ "ordinal decreasing" if there exist no infinite $f$-bad sequences. We prove the following result: Given ordinal decreasing functions $f,g_1,\ldots,g_k,s$ that are everywhere larger than $0$, define the recursive algorithm "$M(x)$: if $x<0$ return $f(x)$, else return $g_1(-M(x-g_2(-M(x-\cdots-g_k(-M(x-s(x)))\cdots))))$". Then $M(x)$ halts and is ordinal decreasing for all $x \in \mathbb{R}$. The recursive algorithms $M$ and $M_n$ previously studied in the context of fusible numbers by Ericskon et al. (2022) and Bufetov et al. (2024), respectively, are special cases of this scheme. Moreover, given an ordinal decreasing function $f$, denote by $o(f)$ the ordinal height of the root of the tree of $f$-bad sequences. Then we prove that, for $k\ge 2$, the function $M(x)$ defined by the above algorithm satisfies $o(M)\le\varphi_{k-1}(\gamma+o(s)+1)$, where $\gamma$ is the smallest ordinal such that $\max\{o(s),o(f),o(g_1), \ldots, o(g_k)\} <\varphi_{k-1}(\gamma)$.
arXiv:2606.09700v1 Announce Type: new Abstract: Large language model (LLM)-powered content moderation systems have become a critical defense against harmful online content. However, these systems primarily operate on tokenized text and largely ignore the visual cues that humans naturally rely on when interpreting content. We show that this discrepancy creates a fundamental perceptual mismatch: content that is readily recognized as harmful by humans can become effectively invisible to automated moderation systems. To study this vulnerability, we introduce a class of Human-Perceptible Adversarial Attacks (HPAA), in which harmful expressions are embedded into otherwise benign text through visually salient typographic manipulations. Our key insight is that typographic features, including spacing, visual emphasis, and spatial arrangement, can be strategically combined to preserve human recognition of harmful content while substantially reducing machine detectability. Operating in black-box settings with only a small query budget, our attack automatically generates evasive content without requiring model access or gradient information. We evaluate the attack across multiple datasets and ten deployed moderation systems, including commercial APIs and state-of-the-art open-source guardrails. Results reveal a striking gap between human and machine perception: with only three detector queries, generated attacks achieve over 86\% human recognition while maintaining detection rates below 1\% across the evaluated systems. We further conduct ablation studies to identify the typographic factors driving successful evasion, analyze why current moderation architectures fail to capture these signals, and discuss practical defenses. Our findings expose a fundamental blind spot in today's LLM-based moderation ecosystem and highlight need for moderation systems that reason about content in a manner more consistent with human perceptual understanding.