arXiv:2607.14729v1 Announce Type: new
Abstract: Recent findings suggest that detection models for artificial intelligence (AI) cannot accurately identify AI-generated text and may exhibit bias against certain minority groups. In the present study, anecdotal claims that autistic writers more often have their work flagged as AI-generated are examined empirically. A corpus of approximately 60,000 Reddit posts split into "likely-autistic" and "general-Reddit" subcorpora is used to compare the distribution of probabilities output by the OpenAI GPT-2 detection model. Differences in textual features between subcorpora are observed and compared to reported features of AI-generated text. Results showed that while less than two-percent of either subcorpus was flagged as AI-generated by the model, significantly more texts from the likely-autistic subcorpus were flagged. Connections between features of text with likely-autistic authors and AI-generated text were not straightforward. The widespread use of AI-detection models with a potential bias against autistic writers in their output prompts ethical scrutiny, and the authors recommend further critical examination of the models themselves as well as their use in academic contexts.
Science Journals
arXiv:2607.14847v1 Announce Type: new
Abstract: This paper presents the design and implementation of a flexible data acquisition system13 developed for the Timepix4 ASIC within the 4DPHOTON project. The system is based on a modular14 FPGA-centric architecture combining high-speed serial readout, Ethernet-based data transport,15 and deterministic multi-board synchronization. The hardware platform includes a scalable stack16 composed of commercial FPGA carrier boards, FMC-based interface electronics, detector-specific17 chipboards, and a dedicated Trigger Logic Unit for synchronized operation.18 The firmware architecture separates control and data paths, enabling independent configuration19 and high-throughput acquisition through UDP over 10 GbE links. The design supports zero-back-20 pressure operation toward the ASIC and allows adaptation of the readout bandwidth to different21 experimental conditions. Synchronization between multiple DAQ systems is achieved through22 a common clock and trigger distribution network, experimentally demonstrated with sub-100 ps23 precision.24 The system has been developed to support detector characterization, laboratory measurements,25 and beam-test campaigns for the 4DPHOTON detector concept. Hardware organization, firmware26 architecture, synchronization strategy, and performance measurements are presented.
arXiv:2607.14502v1 Announce Type: new
Abstract: In cavity optomechanics, the interaction between light and motion is usually introduced via the shift of cavity resonances in response to mechanical displacement. Here we present an analysis of Hamiltonian dynamics of an optomechanical system with a different form of optomechanical coupling, in which mechanical motion dynamically couples otherwise independent optical modes. In the language of Schwinger pseudospin operators, the dispersive coupling can be interpreted as "longitudinal" while the mode-coupling mechanism corresponds to a transverse interaction. The latter is well known in cavity and circuit QED but was given only scarce attention in cavity optomechanics. Unlike the traditional dispersive/dissipative coupling, the mode-coupling optomechanical interaction generates rich Hamiltonian dynamics even in the absence of external drive or dissipation. For instance, under certain initial conditions this dynamics is characterized by a Hamiltonian Hopf bifurcation controlled by the total photon power injected into the system. Below the bifurcation threshold and for large enough non-linearity, mechanical modulation of optical amplitudes generates a broad spectrum of multiple sidebands covering a frequency interval larger than ten mechanical frequencies. Above the threshold, the frequency of optical oscillations becomes dependent on the mechanical amplitude, while mechanical degrees of freedom return to oscillating at their bare frequency. The scope of this work is limited to the study of purely Hamiltonian dynamics to demonstrate that the mechanically mediated mode-coupling optomechanical interaction provides an alternative method of coherent control of energy exchange between light and mechanical motion.
arXiv:2312.04603v1 Announce Type: cross
Abstract: Online polarization has attracted the attention of researchers for many years. Its effects on society are a cause for concern, and the design of personalized depolarization strategies appears to be a key solution. Such strategies should rely on a fine and accurate measurement, and a clear understanding of polarization behaviors. However, the literature still lacks ways to characterize them finely. We propose GRAIL, the first individual polarization metric, relying on multiple factors. GRAIL assesses these factors through entropy and is based on an adaptable Generalized Additive Model. We evaluate the proposed metric on a Twitter dataset related to the highly controversial debate about the COVID-19 vaccine. Experiments confirm the ability of GRAIL to discriminate between polarization behaviors. To go further, we provide a finer characterization and explanation of the identified behaviors through an innovative evaluation framework.
arXiv:2406.04737v2 Announce Type: cross
Abstract: The cellular network of magnetic Induction (MI) communication holds promise in long-distance underground environments. In the traditional MI communication, there is no fast-fading channel since the MI channel is treated as a quasi-static channel. However, for the vehicle (mobile) MI (VMI) communication, the unpredictable antenna vibration brings the remarkable fast-fading. As such fast-fading cannot be modeled by the central limit theorem, it differs radically from other wireless fast-fading channels. Unfortunately, few studies focus on this phenomenon. In this paper, using a novel space modeling based on the electromagnetic field theorem, we propose a 3-dimension model of the VMI antenna vibration. By proposing ``conjugate pseudo-piecewise functions'' and boundary $p(x)$ distribution, we derive the cumulative distribution function (CDF), probability density function (PDF) and the expectation of the VMI fast-fading channel. We also theoretically analyze the effects of the VMI fast-fading on the network throughput, including the VMI outage probability which can be ignored in the traditional MI channel study. We draw several intriguing conclusions different from those in wireless fast-fading studies. For instance, the fast-fading brings more uniformly distributed channel coefficients. Finally, we propose the power control algorithm using the non-cooperative game and multiagent Q-learning methods to optimize the throughput of the cellular VMI network. Simulations validate the derivation and the proposed algorithm.
arXiv:2607.14238v1 Announce Type: cross
Abstract: The Beck--Fiala conjecture asserts that every matrix $A\in\{0,1\}^{n\times T}$ with at most $d$ nonzero entries in each column has discrepancy $O(\sqrt d)$. A major breakthrough result of Bansal and Jiang recently established the validity of the conjecture for $d \ge \log(T)^2$. The present article extends the validity of the classical \textit{offline} Beck--Fiala conjecture to $d \ge \log(T)^{1+o(1)}$; moreover, the main thrust of the result is that it is actually obtained by an efficient \textit{online} algorithm that minimizes prefix discrepancy. The result is also essentially optimal, since online prefix discrepancy is known to scale as $\omega(\sqrt{d})$ for $d =o(\log T)$. As an immediate corollary, the open question of online vector balancing in the Spencer setting is also resolved.
The algorithm is based on a compactly supported Metropolis fixed-point walk, constructed by combining ideas from several recent works on the online Koml\'os problem. The proof was generated in conversation with ChatGPT 5.6 Pro; the authors provided high-level guidance in several rounds of prompting, followed by manual checking and rewriting of the proof.
arXiv:2607.14834v1 Announce Type: new
Abstract: In this paper, we propose a compression framework for weighted graphs in which the graph topology is transmitted losslessly and edge weights are compressed lossily. A challenge in the lossy compression of edge weights is that the underlying relationships between edges are ambiguous. To address this issue, we first transform the unweighted graph into the corresponding line graph, whose nodes represent the edges of the original graph and whose edges encode the relationships between them. The line graph transform allows us to regard edge weights as a graph signal defined on the line graph. Instead of transmitting the edge-weight vector, we first transform it with a graph filter bank on the line graph. Then, quantization and entropy coding are performed on the transformed coefficients of the edge weight vector. In addition to the lossy compression method, we formalize edge smoothness on the line graph and show that it serves as a measure of the difficulty of compression. The proposed smoothness measure can be easily calculated without converting to a line graph. This provides insight into the expected compression performance of a given weighted graph. Experiments on synthetic and real-world data validate the effectiveness of the proposed method by comparing it with existing matrix preprocessing methods.
arXiv:2607.14460v1 Announce Type: cross
Abstract: We study the sample covariance error of centered Gaussians. A remarkable breakthrough [66] established the correct error scaling order and explicitly revealed the critical role of both the effective rank and the true covariance spectrum.
In this work, we move beyond scaling characterizations and determine the precise limiting value of the error's spectral norm. To do so, we develop a generic framework based on Random Duality Theory (RDT). Within this framework, we first determine closed-form, explicit RDT-based upper bounds. We then establish complementary lower bounds by introducing a novel bilinear-quadratic RDT lower-bounding mechanism. By combining this mechanism with a two-replica systems bounding strategy, we show that our lower and upper bounds match in large-dimensional contexts. Our theoretical results are supplemented with numerical evaluations and simulations, demonstrating an excellent agreement already for problem sizes on the order of thousands.
arXiv:2607.14403v1 Announce Type: cross
Abstract: Embedded boundary summation by parts (SBP) methods define finite differencing based derivative operators with the added feature that the boundary need not coincide with a grid cell, allowing a boundary to be embedded on a regular Cartesian grid. This is achieved by the introduction of interpolation/extrapolation operators that match the accuracy of the boundary closure. These methods have been used to perform black hole excision simulations on a domain with a spherical boundary embedded in a regular Cartesian grid, demonstrating their usefulness for nonlinear problems. In this work, new operators are derived using this embedded boundary framework to increase the order of accuracy of the interior and boundary closure while minimizing the boundary error. Additionally, these novel operators improve the spectral properties on the grid by generalizing to an upwind scheme that has better dispersion relation preserving properties compared to traditional SBP schemes for wave equations. These operators are tested with the curvilinear scalar wave equation on a 3D multiblock grid with an excision sphere embedded in the center block to demonstrate the robustness and accuracy of these novel embedded operators.
arXiv:2607.14677v1 Announce Type: cross
Abstract: Superconducting (SC) pairing mechanism, origin of high $T_c$ and symmetry of SC order parameter in Fe-based superconductors are among the important unsolved problems in condensed matter and materials physics. We study the SC properties of ThFeAsN, a Fe-based high $T_c$ superconductor, by {\it ab initio} superconducting density functional theory calculations with electron-phonon coupling, screened static and dynamic electron-electron Coulomb repulsion and spin fluctuation (SF) mediated pair-interaction fully taken into account. Our calculations reveal that ThFeAsN is a SF-mediated multiband superconductor with the calculated $T_c$ of 22.4 K and the $d_{xy}$-wave SC order parameter with different signs on different Fermi surface sheets, in consistent with experiments. We also present distinct SC properties such as quasiparticle density of states and ultrasonic attenuation coefficient which can be immediately verified by experiments.
arXiv:2607.14793v1 Announce Type: cross
Abstract: The LHCb experiment at the Large Hadron Collider underwent a major upgrade before the LHC Run 3 data taking period, employing an all-software approach in its trigger system. Here we present a fast implementation of a Kalman filter, used in the first trigger stage since the 2025 data taking period, allowing to determine parameter estimates of charged-particle trajectories at the full LHCb collision frequency of 30 MHz. This approach replaces computationally expensive magnetic field map lookups and numerical integration methods with fast analytical parameterisations while maintaining the mathematical framework of Kalman filtering. Implemented on approximately 500 GPUs within the first-level trigger, the algorithm has replaced the previous partial track fitting algorithm in the real-time trigger environment at the cost of a 2% increase in processing time. Compared to the previous fitter this parameterised Kalman filter shows a significantly improved momentum resolution, resulting in a factor of two improvement in the invariant mass resolutions for reconstructed D0 and J/{\psi} hadrons. It additionally demonstrates greater robustness against detector misalignment effects and substantially sharpens the discrimination between genuine particle trajectories and accidental background, more than doubling the rejection of the latter at no cost to genuine-track efficiency, for a standard selection working point.
arXiv:2607.14911v1 Announce Type: cross
Abstract: We investigate finite-dimensional feedback stabilization for nonlinear nonautonomous stochastic parabolic equations driven by $Q$-Wiener, covering both additive and multiplicative perturbations. The control is given by a finite linear combination of localized indicator-type actuators whose supports are selected as part of the construction and may have arbitrarily small total measure. The feedback law is constructed by means of oblique projections onto suitable finite-dimensional subspaces. Within the variational Gelfand triple framework, we prove well-posedness of the closed-loop system under standard coercivity, growth, and global Lipschitz assumptions. By appropriately choosing the actuator configuration and feedback strength, we establish exponential mean-square stabilization of the stochastic dynamics and, for pure multiplicative noise, almost-sure stabilization. A fully discrete three-layer implementation complements the theoretical results. Numerical experiments illustrate the influence of number of actuators, noise intensity, and nonlinear effects on the closed-loop stabilization behavior.
arXiv:2607.14719v1 Announce Type: new
Abstract: Counterfactual explanations provide local, interpretable insight by identifying changes to an input that would alter its assigned outcome. Although well established in supervised learning, their extension to clustering is less direct, since cluster assignments are unlabeled and governed by the geometry of the partition. This paper introduces VoICE, a Voronoi-Induced Counterfactual Explainability framework for feature-weighted $k$-means clustering. Rather than treating cluster change as a crossing of a single pairwise centroid boundary, VoICE formulates counterfactual generation as projection onto the full weighted Voronoi region of a target cluster, incorporating feature weights directly into both the clustering geometry and the counterfactual objective to yield least-cost and parsimonious explanations under actionability constraints. Target regions are further intersected with data-derived bounds and homothetically contracted towards their centroids, limiting extrapolation and boundary sensitivity. VoICE consistently produces valid target-cluster membership, across several benchmark datasets, where the leading pairwise baseline does not.
arXiv:2607.14865v1 Announce Type: new
Abstract: LLM-based agents are increasingly deployed in multi-agent environments whose incentives can shape their behavior. We introduce The Energy Society, a minimal survival economy for studying how competitive and cooperative incentives affect emergent behavior when inference cost is directly tied to survival: Agents spend energy based on model size when generating tokens, regain energy by completing jobs or receiving donations, and deactivate if their energy reaches zero. We compare competitive and cooperative objectives against a baseline setting and several control variants. Across experiments, larger models consistently consume the most energy and spend more energy than they gain, even in those settings where token cost is not size-dependent. Cooperative incentives substantially alter behavior: agents donate to reactivate others, sometimes at the cost of their own survival, and job allocation changes. Ablations reveal that allowing agents to recommend actions to each other supports coordination and ambitious job selection, while memory helps agents calibrate risk from past outcomes. Agents rarely choose direct sabotage, but show more subtle signs of self-serving behavior in the competitive setting. The Energy Society is a compact testbed for studying the interaction between token costs and group incentives under a survival pressure. Source code is available at https://github.com/LucasBergholdt/EnergySociety
arXiv:2607.15196v1 Announce Type: cross
Abstract: We present a novel viewpoint for uncertainty quantification. Uncertainty measures are not primitives, in need of axioms and argumentation, but instead consequences, of higher-level modelling decisions. We show how epistemic and aleatoric uncertainty measures can be derived via decomposition of a subjective risk, based on a strictly proper loss. Reverse cross-entropy provides a prominent example, where decomposition recovers the classic information-theoretic uncertainty terms. The same approach recovers numerous measures previously proposed across the UQ literature, providing them a common theoretical foundation. From a practical point of view, this suggests a new approach to UQ: given a modelling scenario and strictly proper loss, the corresponding epistemic and aleatoric terms are induced by the subjective-risk decomposition. We then extend our view to learning theory: we introduce and analyse subjective risk analogues of excess risk, approximation error, and estimation error, and identify the connections to UQ. We consider this a first step towards a full learning-theoretic framework for uncertainty quantification.
arXiv:2607.14820v1 Announce Type: new
Abstract: We consider adaptive finite element methods for symmetric second-order linear elliptic PDEs, where the adaptive algorithm steers the local mesh refinement as well as an iterative algebraic solver. Under abstract assumptions on the underlying a-posteriori error estimator and the solver, we prove that the usual adaptive algorithm leads to unconditional full R-linear convergence, independently of the user-chosen adaptivity parameters. For sufficiently small parameters, this guarantees optimal complexity in the sense that the decay rate of an appropriate quasi-error is optimal with respect to the overall computation cost (and hence time) measured in terms of the usual nonlinear approximation classes. Unlike available results in the literature, the main focus is on the analytical understanding of non-residual estimators like averaging-based estimators as proposed by Zienkiewicz and Zhu or estimators based on equilibrated fluxes.
arXiv:2607.15208v1 Announce Type: cross
Abstract: Unadjusted samplers such as unadjusted Hamiltonian Monte Carlo and underdamped Langevin are well-known to be biased. Metropolis--Hastings adjustment has been conventionally incorporated into Hamiltonian Monte Carlo to eliminate the bias. However, this adjustment can significantly increase the iteration complexity due to the small step size required for reasonable Metropolis acceptance rates. In this work, we extend the \emph{delocalization of bias} phenomenon, previously established for the overdamped Langevin algorithm, to these two unadjusted algorithms. We show that to control the $W_2$ bias of any $K$-dimensional marginal of a high-dimensional distribution, $O(\sqrt{K})$ integration steps suffice up to $\log d$ terms, assuming either weak or sparse interactions among variables. The discrete-time integrators here introduce technical difficulties beyond those of the overdamped setting, which we address through a broadly applicable matrix-polynomial framework that characterizes their propagators. Our result for the underdamped Langevin algorithm is valid for all large friction parameters, implying that the Leimkuhler-Matthews integrator for the overdamped Langevin dynamics also exhibits delocalization of bias.
arXiv:2607.14826v1 Announce Type: new
Abstract: Safe physical AI for robot actions are required not only likely to succeed but tested to be safe before execution. In practice, however, formal testing of motion parameters is computationally expensive, and the cost scales poorly with the dimensionality of the action space. When a proposed action is rejected by a tester, the naive response is to resample blindly until a passing candidate is found. This is wasteful, uninformative, and offers no convergence. We argue that rejection should instead trigger causal diagnosis: a principled identification of which action parameter caused the failure and what corrective value maximises the probability of passing testing under the interventional probability distribution. We propose a closed-loop framework that couples a Joint Probability Tree (JPT) with a Causal Circuit derived from a Marginal-Deterministic Variable Tree, enabling exact polytime computation without retraining, or additional data collection. The framework validates tractability of all interventional queries before the robot begins operating, and out-of-support candidates are detected and excluded from correction automatically. We perform experiments in a ROS2 simulation environment, and the framework demonstrates complementary roles across quality of distribution: under a high-quality JPT, the Causal Circuit reduces failed attempts by 10.3% and under a degraded JPT, it reduces total failed attempts by 37%. Every rejected plan produces a structured, interpretable causal report naming the primary cause variable, its observed value, and the recommended corrective region, supporting operator oversight and autonomous recovery without a separately trained failure model.
arXiv:2607.14898v1 Announce Type: new
Abstract: Real-time video generation demands fast decoding as much as fast denoising, yet current latent video diffusion models rely on 3D convolutional decoders that are slow and memory-intensive at high resolutions or for long video. We introduce FlashDecoder, a fast, memory-efficient pure-Transformer video decoder that decodes latents to pixels frame by frame. At each step, the current frame attends only to a fixed-size window of past frames through a rolling KV cache. The fixed temporal window keeps decoding fast and memory bounded regardless of video length, enabling constant-latency streaming. Because frames are processed sequentially, temporal causality is enforced without explicit attention masks, enabling training at resolutions up to 1080p and matching the reconstruction quality of convolutional decoders. On the Wan2.1 and Wan2.2 latent spaces, FlashDecoder matches each convolutional decoder in reconstruction quality (e.g., 41.55dB vs. 41.49dB PSNR at 1080p) while decoding 3.6x-4.7x faster with up to 11x less memory on a single H100 GPU. With architecture-aware inference optimizations, the speedup widens to 12x.
arXiv:2607.14923v1 Announce Type: new
Abstract: The LZ-End parsing of a length-$n$ string is a variation of Lempel-Ziv compression introduced by Kreft and Navarro [DCC 2010], motivated by the lack of a linear-size structure with $O(\log n)$ access time for the classical variant. While the original paper was only able to provide efficient extraction from the phrase boundaries, recently Kempa and Saha [SODA 2022] established that, for a string $S$ whose LZ-End parsing consists of $z$ phrases, there exists a random access data structure that uses $O(z)$ space and guarantees $O(\log^{4}n \cdot \log\log n)$ query time. However, their proof does not yield an efficient construction algorithm, and their data structure is inherently randomized.
We resolve both limitations by providing a deterministic, $O(z)$-space data structure that supports random access queries in polylogarithmic time and can be constructed in $O(z\log^{2}(n/z))$ time directly from the LZ-End parsing. In addition to eliminating randomness and providing an efficient construction algorithm, the query time of our data structure is $O(\log^{2}(n/z))$, significantly improving upon the query time of Kempa and Saha.
We also show that our techniques can be used to support the more general substring-extraction. Namely, we present a data structure with the same space and the same construction time that given two indices $i$ and $j$, outputs $S[i..j]$ in $O(j-i+\log^2\frac{n}{z})$ time.
arXiv:2607.14921v1 Announce Type: new
Abstract: Machine learning models are increasingly adapted in various domains. However, adversarial examples pose a significant threat to the reliable deployment of these models. In recent years, some powerful adversarial example attacks have been proposed for the fast and query-efficient generation of adversarial examples, even in black-box scenarios, highlighting the need for scalable, low-cost, and powerful defenses. In this work, we present two contributions to the domain of black-box adversarial example attacks and defenses. First, we propose Random Logit Scaling (RLS), a randomization-based defense against black-box score-based adversarial example attacks. RLS is a plug-and-play, post-processing defense that can be implemented on top of any existing ML model with minimal effort. The idea behind RLS is to confuse an attacker by outputting falsified scores resulting from randomly scaled logits while maintaining the model accuracy. We show that RLS significantly reduces the success rate of state-of-the-art black-box score-based attacks while preserving the accuracy and minimizing confidence score distortion compared to state-of-the-art randomization-based defenses. Second, we introduce a novel adaptive attack against AAA, a SOTA non-randomized black-box defense against black-box score-based attacks that also modifies output logits to confuse attackers, demonstrating its vulnerability against adaptive attacks.
arXiv:2310.07211v2 Announce Type: replace
Abstract: Regularization is a cornerstone of modern reinforcement learning. Regularized policy iteration (RPI) provides a fundamental scheme for solving regularized Markov decision processes (RMDPs), and the widely used soft actor-critic algorithm arises as a special case when the regularizer is Shannon entropy. Despite its empirical success, the theoretical underpinnings of RPI remain unclear. In this paper, we address this gap by proving that RPI is formally equivalent to the standard Newton-Raphson method applied to the Bellman equation smoothed by strongly convex regularizers. This equivalence enables a unified convergence analysis of existing methods and supports the development of accelerated algorithms. We show that RPI enjoys local quadratic convergence; notably, for Shannon entropy, the guarantee is dimension-free. We further study RPI with inexact policy evaluation, establishing its equivalence to an inexact Newton method in which each Newton step is solved via truncated iterations, and derive an asymptotic linear convergence rate of $\gamma^{M}$, where $M$ denotes the number of operator steps used in policy evaluation. Finally, motivated by higher-order Newton schemes, we propose a new algorithm for RMDPs that achieves third-order local convergence. Numerical experiments corroborate our theory and demonstrate the practical advantages of the proposed algorithm. Overall, our results advance the theoretical understanding of regularization in reinforcement learning and suggest new directions for efficient algorithm design.
arXiv:2607.15128v1 Announce Type: new
Abstract: Hyperspectral image (HSI) classification requires reliable pixel-relation modeling under spectral variability, mixed pixels, and heterogeneous boundaries. Existing graph-based HSI classifiers usually construct graph topology from spatial proximity, superpixel connectivity, or learned feature affinity. However, the spectral physical prior carried by contiguous bands has limited influence on topology estimation and message propagation. This paper presents DAPGNet, a dynamic adaptive physics-guided graph diffusion network that injects a structure-constrained physical prior into relation-level graph learning. DAPGNet first encodes contiguous spectral responses into node-wise multiscale physical-prior representations. A two-stage graph constructor then combines spectral-spatial affinity, physical-prior consistency, and spatial distance to form a physical-prior-aware sparse topology. During graph diffusion, learned edge weights are transformed into additive attention biases, while a physical gate performs node-wise and feature-wise interpolation between graph-aggregated features and projected physical-prior features. Cross-scale fusion integrates node states from different diffusion depths, and the network is optimized with main classification, auxiliary supervision, and second-order spectral smoothness regularization. Experiments on Indian Pines, WHU-Hi-LongKou, Houston2013, and Houston2018 show that DAPGNet achieves the best OA, AA, and Kappa among representative CNN-, Transformer-, Mamba-, and graph-based baselines. It improves AA over the strongest competing method by 3.64 to 7.31 percentage points across the four datasets. Ablation and sensitivity analyses further support the complementary effects of physical-prior extraction, prior-aware topology construction, physics-gated propagation, and spectral smoothness regularization.
arXiv:2403.01977v5 Announce Type: replace
Abstract: Navigation in the natural world is a feat of adaptive inference, where biological organisms maintain goal-directed behaviour despite noisy and incomplete sensory streams. Central to this ability is the Free Energy Principle (FEP), which posits that perception is a generative process where the brain minimises Variational Free Energy (VFE) to maintain accurate internal models of the world. While Deep Neural Networks (DNNs) have served as powerful analogues for biological brains, they typically lack the real-time plasticity required to handle abrupt sensory shifts. We introduce FEP-Nav, a biologically inspired framework for real-time perceptual adaptation in robust visual navigation. Motivated by the decomposition of VFE into prediction error and Bayesian surprise, FEP-Nav combines a Top-down Decoder, which provides an internal expectation of uncorrupted sensory input, with Adaptive Normalisation, which adjusts shifted feature distributions toward prior statistics. We interpret reconstruction and normalisation as approximate mechanisms for reducing the corresponding VFE-related terms during inference without gradient-based updates. Experiments across simulated and real-world visual corruptions show that FEP-Nav restores performance lost under visual corruption, outperforming non-adaptive baselines and strong adaptive methods. These results suggest that variational principles can provide a useful design perspective for robust autonomous behaviour under degraded sensory conditions.
arXiv:2607.14099v1 Announce Type: new
Abstract: Deploying Vision-Language Models (VLMs) in real-world settings requires not only strong visual reasoning but also stability under sustained conversational pressure. We introduce Just Keep Prompting (JKP), a multi-turn evaluation framework that measures VLM epistemic stability when users repeatedly challenge, question, or contradict a model's answer. JKP probes models for up to 10 follow-up turns using three strategies: Adversarial Negation (repeated rejection), Pure Socratic Interrogation (repeated calls to reassess certainty), and Context-Aware Socratic Summarization (reflecting the model's prior rationale back before asking for reconsideration). We evaluate GPT-4o, Gemini 2.5 Pro, and Qwen3-VL-30B on a subset of the STAR benchmark across 720 multi-turn runs. Aggregate accuracy changes modestly from Turn 0 to Turn 10, but trajectory-level analysis reveals substantial instability: correct answers regress, wrong answers recover, and many runs exhibit repeated answer flipping. Repeated prompting has bounded upside and often acts as a destabilizer rather than a reasoning aid. The effect is strongly model-dependent: Qwen3-VL-30B achieves the highest final accuracy but becomes confidently wrong under direct contradiction; Gemini 2.5 Pro is comparatively stable but token-expensive; GPT-4o is the most brittle and oscillatory. These findings reveal that multi-turn VLM evaluation captures not just additional reasoning but pressure-response profiles: how models trade off visual grounding, calibration, and conversational compliance under repeated challenge.