# Research Papers — 2026-09 · bundle page 222 of 266 > Full-text markdown of papers #11051–11100 of 13298 published in 2026-09, newest first. AI-readable mirrors of the full 13298-paper 2026-09 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-09/page/222 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/221.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/223.md) ### Identifying Habit, Physics, and Nuisance in Robot World Models Authors: Jinting Hang, Zhenhui Cai Published: 2026-09-06 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2609.09210v1 Teleoperated demonstrations are often multimodal even when the underlying dynamics are nearly deterministic given the executed action. We argue that this multimodality typically mixes three factors--operator habit in action selection, shared physics, and observation nuisance--and that entangled next-observation predictors absorb all three. We formalize the split with a structural causal model a=g(h,z,u), z'=f(z,a), o=r(z,c), and test it with complementary interventions: replacing or shuffling actions at fixed state sharply increases next-state error, whereas appearance and camera changes should not; habit-aware reverse scoring improves ranking of feasible pasts without rewriting the dynamics. The associated adaptation rule is to freeze a shared physics readout and update only a thin interface. On StackCube, DROID, and RH20T this rule improves low-shot transfer relative to training from scratch, retains cleaner dynamics under corrupted adaptation data, and extends from proprioception to pixel observations with multi-view and multi-step checks. We do not equate latent actions with operator habit, and we do not target large-scale video generation benchmarks. --- ### AutoLexSteer: Automatic Contrast Construction for Lexical Activation Steering Authors: Shuhe Wang, Lachlan Cowley, Eduard Hovy, Jey Han Lau Published: 2026-09-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.06879v1 Steering vectors have rapidly emerged as a popular and effective method for guiding the output of LLMs in very specific ways. But constructing accurate steering vectors is a difficult manual process due to the opacity of embeddings. We introduce Hangman, a novel type of steering vector that operates using word senses, as well as AutoLexSteer, the first fully automated process for building steering vectors. AutoLexSteer employs families of closely-related words extracted from WordNet to specify both the steering source to be avoided and the desired steering target. The steering vectors are quite precise, can be used to steer at the level of words and sets of word senses (meanings), and are able to steer certain LLM behaviors like sycophancy. The dataset and code can be found at https://github.com/ShuheWang1998/autolexsteer. --- ### Novel Methods for Catheter and Guidewire Segmentation in X-ray Fluoroscopy under a Federated Learning Setting Authors: Chayun Kongtongvattana Published: 2026-09-06 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.06876v1 Endovascular procedures rely on real-time manipulation of thin instruments, catheters and guidewires, under X-ray fluoroscopy guidance, where accurate visual analysis is essential for procedural safety. Learning-based methods are constrained by structural complexity, data scarcity, and privacy regulations precluding centralised training across institutions. This thesis presents a structure-aware federated learning framework for catheter and guidewire analysis, with four contributions evaluated on real-animal and phantom data. A benchmark dataset, CathAction, is introduced for catheterisation analysis, with over 600,000 annotated frames and 40,000 segmentation masks. A shape-sensitive loss transforms masks into signed distance maps compared in a structural feature space, improving Dice coefficient by up to 2.9 points across five backbones. This is extended to federated learning with shape-sensitive loss, preserving geometric consistency under heterogeneous client data and outperforming federated averaging by up to three points in mean intersection-over-union as clients scale from four to eight. Federated learning with projected gradient descent adds adversarial optimisation, raising mean intersection-over-union by over ten points on real-animal data. Finally, a structure-aware diffusion framework synthesises catheter and guidewire video sequences, combining structural supervision with a domain-adaptive reconstruction objective, reducing Frechet video distance over a strong baseline while maintaining visual fidelity. Incorporating synthetic sequences into federated training raises the Dice score from 44 to 51 percent under data scarcity, with gains across four held-out sites. Together, these contributions advance privacy-preserving catheter and guidewire analysis, supporting collaborative training without centralising patient data or large amounts of manual annotation. --- ### MedGSSR: Generalizable Medical Image Super-Resolution 3D Reconstruction via Hierarchical Feed-forward Gaussian Splatting Authors: Chengkai Wang, Luoyu Hong, Yiting Zhao, Jiamin Wang, Xiang Feng, Feiwei Qin, Zhenzhong Kuang, Xuefei Yin, Ali Bashashati, Yanming Zhu Published: 2026-09-06 Categories: eess.IV, cs.CE, cs.CV, q-bio.QM arXiv: https://arxiv.org/abs/2609.06874v1 High-resolution volumetric medical imaging is critical for clinical diagnosis, yet acquisition is often limited by scanner hardware, scan time, and for CT, radiation dose. Medical 3D Super-Resolution (Med3DSR) offers a computational alternative, but existing methods commonly rely on per-subject optimization, pretrained priors, or coordinate-based implicit representations, which compromise anatomical fidelity and limit efficiency. To address these limitations, we present MedGSSR, a fully end-to-end feed-forward framework that represents volumes as an explicit 3D Gaussian field for Med3DSR. Unlike coordinate-based implicit functions, our explicit 3D Gaussian representation naturally enhances signal continuity and local high-frequency fidelity. Specifically, MedGSSR explicitly decouples the reconstruction process into coarse-grained structural preservation and fine-grained textural refinement through the proposed Pyramid Anatomical Encoder and a Hierarchical Gaussian Projector. To support arbitrary-scale super-resolution, we introduce sub-voxel Gaussian decomposition and a Differentiable Gaussian Voxelizer that directly queries the continuous 3D intensity field, reducing discretization artifacts. Extensive experiments on MRI and CT benchmarks demonstrate that MedGSSR significantly outperforms state-of-the-art methods. Notably, our framework exhibits robust generalizability across unseen datasets without requiring per-subject optimization, enabling fast inference and high-fidelity volumetric super-resolution in practical clinical settings. Our project webpage, including code, is at https://william2ai.github.io/medgssr --- ### Large Classification-Risk-Optional Label Acquisition Authors: F. Setoudehtanzangi, Geoffrey J. McLachlan Published: 2026-09-06 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2609.06873v1 We study how a limited labeling budget should be allocated to minimize multiclass zero-one classification risk. We consider parametric classification problems in which features are observed for all sampling units while class labels can be acquired selectively. By combining the Fisher information supplied by an acquired label with the local geometry of multiclass excess risk, we derive an acquisition criterion that minimizes the leading asymptotic coefficient of expected multiclass excess risk. The resulting rule values a label according to how strongly its information is aligned with parameter directions that perturb the active Bayes decision boundary, rather than according to posterior uncertainty or global parameter information alone. We characterize the oracle acquisition design, establish its threshold structure, and derive face-specific and cost-sensitive extensions. An analytic example shows that posterior uncertainty and classification value can produce different, and even reversed, acquisition rankings. We further develop a two-stage adaptive procedure that attains the oracle leading-risk criterion under regularity conditions and provide explicit results for Gaussian discriminant analysis. Three-class QDA experiments illustrate the resulting acquisition geometry, while an application to the six-class Statlog Landsat Satellite data shows that classification-risk acquisition can differ materially from both uncertainty-based acquisition and the complete-classification-information comparator. The adaptive classification-risk design attains lower mean error than this Fisher comparator across the labeling budgets considered, although it does not uniformly outperform entropy or margin sampling and differences among the targeted strategies become small as the labeling budget increases. --- ### Exact Record Omission in Delta Attention: A Transport Criterion, Its Cost, and a Replay Certificate Authors: Vishwajith Ramesh Published: 2026-09-06 Categories: cs.LG, cs.CL, cs.CR arXiv: https://arxiv.org/abs/2609.06872v3 An assistant can stop repeating a deleted statement while its recurrent memory still carries that statement's influence. We examine this distinction by saving the state difference immediately after a record, carrying this saved difference, or receipt, through later updates, and comparing the corrected state with the state built from the same conversation with the record omitted. Unrolling the recurrence gives an exact criterion: transport reaches this never-stored state if and only if the additional differences created by later updates cancel after transport. We evaluate eleven conditions using 40 prospectively selected synthetic Kimi Linear contexts, with 40 separate calibration contexts and paired uncertainty estimates computed across complete records. Attention masking removed all 40 exact greedy target answers, yet a fixed three-query candidate-scoring attack with known candidate values and log-probability access attained AUC 0.740 (95% interval 0.695--0.809); the measured false-positive rate was 7.5%. Prompt-only forgetting still returned 33 of the 40 targets. Checkpoint replay matched the complete declared active state and audit logits in all 80 contexts, while preserving all 120 retained answers in the evaluation cohort. An independent Qwen cohort and matched original-bf16/8-bit Kimi controls confirmed native transport mismatch beyond the numerical error measured in matched controls. Replay also supported record replacement and successive deletions with a new record added between them. These results provide a practical audit of both the requested memory change and the assistant's remaining behavior, with replay work determined by the surviving suffix. --- ### Organization of Valence and Arousal in Vision-Language Representations of Built Environments: Insights from the EMOIS Dataset Authors: Madoka Yonekura, Katsunori Kohda, Nobuhiko Muramoto, Takahiro Yamaguchi Published: 2026-09-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06870v1 Visual perception of built environments contributes to the affective impressions that people form in everyday life. However, how these impressions are represented within vision foundation models remains largely unexplored. To support the systematic investigation of this subject, we introduce the Emotional Impression of Spaces (EMOIS) dataset, comprising 1,544 real-world built-environment images. Each image is annotated with image-evoked valence and arousal ratings collected from Japanese adults by conducting a large-scale web-based survey, with approximately 120 ratings per image. Using Contrastive Language--Image Pre-training (CLIP) representations, we perform predictive and geometric analyses to systematically investigate how valence and arousal are encoded and organized within the representation space. These analyses reveal that valence exhibited stronger and more coherent organization than arousal. Cross-dataset analyses with the Open Affective Standardized Image Set (OASIS), a benchmark dataset of general affective photographs, reveal differences in affective organization between the two datasets. Regression analyses demonstrate high predictive performance for valence and arousal within EMOIS, with mean coefficients of determination of 0.865 and 0.807, respectively, across repeated internal hold-out evaluations. Finally, we present an example-based interface illustrating how learned representations can support qualitative interpretation of predicted affective values. These findings can help elucidate affective representations of built environments and establish EMOIS as a densely annotated resource for future affective computing research in this domain. --- ### PPIM: Pennes Physics-Informed Mamba for Heat-Source-Conditioned 3D Bioheat Simulation Authors: Dongyun Lee, Kyungho Yoon, Minwoo Shin Published: 2026-09-06 Categories: cs.LG, math-ph arXiv: https://arxiv.org/abs/2609.06869v1 Three-dimensional bioheat simulation aims to predict transient temperature distributions in biological tissue and is commonly modeled using the Pennes bioheat equation, which combines thermal diffusion, perfusion-mediated heat loss, and external heat generation. In this study, we consider a controlled 3D Pennes bioheat simulation under a localized heat-source condition inspired by microwave ablation (MWA). To evaluate neural approximation performance, we compare three neural partial differential equation (PDE) solvers under the same controlled simulation: a spatial Fourier-feature physics-informed neural network (PINN), a generic PINNMamba temporal subsequence model, and Pennes Physics-Informed Mamba (PPIM). PPIM builds on the temporal subsequence model by incorporating conditioned heat-source input and Pennes-aware state-space model (SSM) decay initialization. All three neural models are trained under the same conditions with the same Pennes residual, and an explicit finite-difference method (FDM) solution is used only as the numerical reference. In a representative 600~s run, PPIM achieved the lowest MAE, relative $L_1$ error, and relative $L_2$ error among the evaluated neural solvers. Error maps further showed that the remaining PPIM errors were more concentrated near the heat-source region than across the rest of the domain. These results indicate that PPIM is effective for approximating the FDM reference final temperature field in this controlled simulation. The source code is available at https://github.com/muvYun/PPIM. --- ### Feature Superposition in Neural Networks: From Theory to Practice Authors: Dai Shi, Xiaoyu Li, Andi Han, José Miguel Hernández-Lobato Published: 2026-09-06 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.06862v1 Superposition refers to neural networks representing more features than they have dimensions. It offers a possible explanation for polysemantic neurons and motivates methods for recovering interpretable features from neural activations. Theoretical models typically start with a given set of input features and assumptions about how their values vary across inputs, then study how a network encodes those values in a lower-dimensional hidden representation. Empirical work, by contrast, seeks to identify the features encoded in trained networks and determine their role in computation. In this survey, we review the geometry, learning, and computation of superposed representations, explaining how feature statistics and decoder choice affect the conclusions. To connect these theoretical accounts with evidence from trained networks, we compare practical methods for recovering and analyzing features and examine what their evaluations establish. Since accurate activation reconstruction alone does not establish feature identity or causal use, we discuss the methods' documented failures and applications in light of the evidence available for these different claims. Finally, we assess previously stated open problems and identify remaining theoretical and empirical questions about superposition in trained networks. We hope our work can pave the way for a deeper understanding of superposition and more reliable methods for interpreting neural networks. --- ### Cluster-Based Dimensionality Reduction by Nonparametric Distributional Screening Authors: Sanoja Jha, Rishikesh Muralimohan, Praveen Athauda Arachchi, Abhishek Bhattacharjee Published: 2026-09-06 Categories: stat.ME, math.ST, stat.AP, stat.ML arXiv: https://arxiv.org/abs/2609.06854v1 We consider dimensionality reduction for high-dimensional observations accompanied by a supplied partition into two or more clusters. The objective is not to construct a low-rank projection, but to retain an interpretable subset of the original coordinates that preserves the distributional information distinguishing the clusters. For each coordinate, the proposed procedure compares the cluster-specific empirical distribution functions through a several-sample Kolmogorov-Smirnov separation statistic. We formalize the resulting marginal cluster support and establish simultaneous finite-sample concentration over all coordinates, explicit bounds for false inclusions and omissions, and exact support recovery when the minimum distributional separation dominates the high-dimensional stochastic error. We also quantify the dimension inflation induced by using an unadjusted testing level and give a familywise-error-controlled version. Under a conditional sufficiency condition, sure screening preserves the full-data posterior cluster probabilities, mutual information, and Bayes risk; an additional result characterizes robustness to imperfectly estimated cluster labels. The procedure is invariant to strictly increasing coordinate transformations and can retain low-variance cluster signals that principal components may discard. We further develop average dual information, a criterion combining partition agreement after transformation with structural coverage of cluster-relevant coordinates, and derive its basic properties and consistency. Simulations illustrate the theory, the interpretability of the selected coordinates, and the distinction between cluster-directed screening and variance-directed projection. --- ### Characterizing Contention-Induced Reliability Collapse in KV-Cache Timing Side Channels for Multi-Tenant LLM Serving Authors: Rana Abu Bakar Published: 2026-09-06 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2609.06853v1 Shared key--value (KV) cache reuse improves large language model (LLM) serving, but it can also create a timing side channel that reveals whether a prefix is already cached. Previous work shows that such attacks are possible, but their reliability under realistic multi-tenant contention is less understood. We study this problem through seven experiments on live shared LLM-serving systems. On a vLLM server running DeepSeek-R1-Distill-Llama-8B on NVIDIA GB10, mean Cohen's d drops from 0.7789 with no synthetic workers to 0.2109 with two workers (t=8.412), while higher worker counts cause no statistically detectable further loss. A 120-run sparse-overlap experiment places the best breakpoint at the boundary of the measured range (tau=0, 95% CI (0.000,0.113)), supporting an ambient-versus-loaded regime change rather than an internal physical threshold. AUROC falls from 0.650 at ambient to 0.531 near 61% overlap and partially recovers to 0.574 at saturation. Concurrency-depth variance is the strongest measured correlate of effect size (r=-0.416) and hit consistency (r=-0.637). An interleaved control preserves the same non-monotonic ordering. The main collapse is also reproduced on a real two-node, two-GPU tensor-parallel vLLM setup, where mean d falls from 3.418 to 0.511 (p<0.01). Two SGLang pilots are statistically inconclusive. Overall, KV-cache timing reliability depends strongly on the load regime and serving stack, and measurements on quiet systems can overestimate operational attack reliability. --- ### You Are What You Read: Misalignment via In-Context Persona Induction Authors: Kyuhee Kim, Benjamin Berczi, Cozmin Ududec Published: 2026-09-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.06851v1 Broad misalignment has been produced by finetuning on narrow data, harmful or benign, and in context only by demonstrations of the undesirable behaviour itself. We show that benign data suffices in context, with no finetuning and no demonstration of harmful behaviour in the prompt. Biographical facts that converge on a single figure, placed in a model's context as ordinary conversational turns, lead it to answer as that figure on questions the facts never touch. We call this persona induction. Across nine personas and thirteen models, identity adoption rises sigmoidally with the number of facts and crosses 50% within 3 to 10 of them. Misalignment then tracks which figure is described. Harmless personas reach full adoption with near-zero misalignment, while harmful ones voice their characteristic views on unrelated questions, at rates up to 80%. A formatting instruction can gate when the persona activates. Because each fact is individually benign, accumulated biographical context is flagged by content filters on 3% of inputs against 24-33% for an equivalent direct instruction. --- ### Learning transferable human physiology from two million hours of sleep with SleepFM-2 Authors: Rahul Thapa, Christopher Sun, William Theodor Lehn-Schioler, Sophia Claire Kivelson, Umaer Hanif, Hyatt Moore, Harrison G. Zhang, Hafsa Ahmed, Marcus Dige, Niels R. Lorenzen, Elisabeth Roxane M. Heremans, Adrien Specht et al. Published: 2026-09-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.06849v1 Sleep provides a nightly window into health by capturing coordinated activity across the brain, heart, muscles and respiratory system. We introduce SleepFM-2, a sleep foundation model developed and evaluated on 282,511 polysomnography recordings from 26 cohorts, including 235,865 used for pretraining. These data span more than two million hours of multimodal physiology. Compared with SleepFM, SleepFM-2 improves disease prediction and sleep scoring, supports arousal, limb movement and respiratory event detection, and transfers to wearable sensing and subjective sleep phenotypes. A model combining its PSG representation with age, sex and BMI met a prespecified discrimination and significance criterion for 215 subsequently recorded EHR phenotypes in two held-out cohorts, including one health system unseen during pretraining. For 155 phenotypes, the PSG representation added reproducible information beyond demographics. SleepFM-2 also outperformed a 480-feature baseline derived from the same recordings. Its disease scores revealed a reproducible principal component associated with reduced sigma-band spatial coupling and increased hypnodensity entropy. The frozen encoder performed within the observed range of expert scorers for sleep events and transferred to wakeful EEG, headband and in-ear EEG, wrist PPG and wrist accelerometry. It improved sleep staging across six accelerometry cohorts and achieved disease-prediction performance in UK Biobank similar to models pretrained directly on accelerometry. Finally, SleepFM-2 captured aspects of subjective sleep not recovered by conventional PSG summaries, particularly reports of the recorded night. These results show that multimodal sleep physiology can provide a transferable representation of human health across diseases, clinical tasks, sensors and subjective experience. --- ### RAIDAL: Redundancy-Aware Information Density Active Learning for CTC-Based Continuous Sign Language Recognition Authors: Rafael A. Diniz Augusto, Gabriel L. Oliveira, Erickson R. Nascimento Published: 2026-09-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06843v2 Continuous sign language recognition (CSLR) is a key technology for accessibility, yet its development remains limited by the high cost of annotating continuous video streams. Active learning offers a path toward mitigating this cost, but standard acquisition functions are not designed for weakly aligned sign language videos, where sign executions are interleaved with rest poses, irregular pauses, sign-like motion, and temporally redundant frames. This temporal redundancy can undermine sample selection, as acquisition scores may be influenced by timesteps from regions that are not associated with the decoded gloss sequence, distorting the video's estimated informativeness. In this work, we show that modern CSLR models already contain a mechanism for identifying gloss-level temporal evidence: the CTC decoder. Although typically used only during inference, its alignment peaks indicate where the model localizes each predicted gloss in the feature sequence, providing a source of temporal structure for active learning acquisition functions at zero additional labeling cost. Thus, we introduce RAIDAL (Redundancy-Aware Information Density Active Learning), which repurposes the CTC decoder to restrict representation-based scoring to decoder-aligned gloss regions, rather than exposing the acquisition function to the entire unfiltered video. Across three datasets and two architectures, RAIDAL achieves its strongest data-efficiency gains over competing baselines in large-vocabulary, budget-limited settings, while remaining competitive in the smaller-vocabulary, large-budget setting. The code used in this work is publicly available at github.com/verlab/RAIDAL. --- ### XYBench: Can LLMs Respond Pragmatically to Queries with Misconceptions? Authors: Akhila Yerukola, Jena D. Hwang, Mingqian Zheng, Jenna Godsey, Hyunwoo Kim, Valentina Pyatkin, Jennifer Hu, Maarten Sap Published: 2026-09-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.06842v1 When non-expert users ask LLMs for assistance, their queries can often have misconceptions (e.g., "How do I parse XML with regex?"). In such cases, often referred to as the XY-problem, LLMs must identify the misconception ("regex are fragile") and meaningfully direct the user toward a pragmatic solution that will address the root problem implicit in the request ("use an XML parser"). We introduce XYBench, a benchmark of 8,115 such queries, drawn from technical (StackOverflow/StackExchange) and everyday (WikiHow and a manually-curated subset) domains. We design an evaluation paradigm that assesses model responses along three criteria grounded in cooperative response theory: (a) presence and (b) emphasis on pragmatic solutions, and (c) identification of misconceptions. Our experiments show that even the strongest LLMs predominantly answer the literal request (0.75--0.92) and far less often the intended one (0.33--0.71), while substantially lagging behind humans at identifying misconceptions (at most 63% vs. 79--90%). Further, models overwhelmingly prefer pragmatic responses in a multiple choice setting yet consistently fail to generate them. Oracle ablation experiments show that providing explicit user intent at generation time helps; however a large gap remains, suggesting pragmatic redirection is a fundamentally underdeveloped capability in current LLMs. --- ### WAPP: Safe Learning of Positive Security WAF Policies from Live Traffic Authors: Heba Osama, Zeyad Ahmed, Mohamed Amgad, Ahmed Saafan, Jana Elfeky, Mariam Abdelati, Haitham Ghalwash Published: 2026-09-06 Categories: cs.CR, cs.AI, cs.NI arXiv: https://arxiv.org/abs/2609.06840v1 Web Application Firewalls (WAFs) mainly rely on signatures to detect known attacks, which can leave gaps against modified or previously unseen payloads. Positive security provides a complementary approach by learning legitimate traffic and blocking inputs that fall outside the learned profile. However, learning directly from live traffic can be unsafe when malicious requests contaminate the training data. This paper presents the Whitelisting Autonomous Policy Producer (WAPP), a framework that combines trust filtering, deterministic rule synthesis, confidence scoring, and validation before enforcement. WAPP is evaluated on three controlled applications using a live Coraza and OWASP Core Rule Set (CRS) stack. Results show that, on the tested DVWA username field, unfiltered learning becomes Degraded at 0.2\% poisoned traffic and Broken at 0.5%, while the evaluated free text field can admit malicious inputs even without poisoning. On the frozen poisoning dataset, the ablation configuration with all seven candidate signals improves the measured poisoning resilience from 53% to 90%, compared with 62% for the Kruegel--Vigna baseline. The deterministic synthesizer provides attack blocking comparable to the tested language model without model inference cost. WAPP blocks confirmed CRS bypasses on constrained fields, while free text inputs remain a precision challenge that requires character level operator control. --- ### A Shared-Backbone Approach for Multi-Task MedMNIST Classification Authors: Stefan-Dorian Gavril, Andrei Arhire, Adrian Iftene Published: 2026-09-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06838v1 Multi-task biomedical classification requires models to generalize across disparate modalities and class distributions. We study 11 heterogeneous MedMNIST datasets using the harmonic mean of per-task macro-F1. We evaluate three backbones with task-specific linear heads. We identify a resolution domain shift between the MedMNIST API and evaluation environment. Resolving this inconsistency and optimizing architecture-specific regularization substantially improved performance. Our best configuration, a ConvNeXt-Tiny backbone with label smoothing, achieved a leaderboard harmonic-mean macro-F1 of 0.73294 in the Tensor Reloaded: Multi-Task MedMNIST competition, ranking sixth at the close of the official competition phase. Our implementation is publicly available at: https://github.com/GavrilStefan-Dorian/A-Shared-Backbone-Approach-for-Multi-Task-MedMNIST-Classification --- ### BinauralVAE: Spatial Audio Reconstruction For World Models Authors: Luis Vitor Zerkowski, Luiz Velho Published: 2026-09-06 Categories: cs.SD, cs.LG arXiv: https://arxiv.org/abs/2609.06837v1 Embodied artificial intelligence has historically very much relied on visual perception, leading to a proliferation of multiple vision-centric world models. However, this reliance fails to capture spatial understanding in its entirety and can even present vulnerabilities in environments with visual occlusions, low-light conditions, or blackouts-scenarios, where acoustic information becomes a critical alternative for spatial awareness and navigation. Despite its potential, research into realistic spatial audio and particularly the development of audio-centric world models remains sparse. In this technical report, we introduce BinauralVAE: a flexible, open-source pipeline (https://github.com/Luizerko/BinauralVAE) that explores multiple models for spatialized audio reconstruction, progressing from fundamental baselines to advanced, mathematically grounded architectures. Our approach evaluates various Variational Autoencoder architectures -- including complex-valued variants -- to learn robust latent representations of binaural signals. Developed alongside AudioWorldSim, our methodology leverages realistic acoustic data captured as a simulated robot navigates an environment. This pipeline establishes a foundation for state representation in a future audio-based world model, designed to map the direct causal connection between navigational actions and their resulting acoustic consequences, and helping to enable sound as an essential complementary modality for spatial knowledge acquisition. --- ### Skynet: Workflow-Level Anomaly Detection for Agentic AI via Semantic and Structural Modeling Authors: Chaoyu Zhang, Hexuan Yu, Heng Jin, Shanghao Shi, Ning Zhang, Yi Shi, Yulia R. Gel, Y. Thomas Hou, Wenjing Lou Published: 2026-09-06 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2609.06835v2 Agentic AI systems execute complex tasks through long-horizon workflows of planning, tool use, and multi-agent coordination. Task failures in these systems often originate from a single step, such as an injected prompt or a flawed plan, and are then amplified through downstream dependencies as the corrupted step propagates across many subsequent agents and tool calls. Existing defenses either target a specific class of attacks or failures, or inspect individual prompts and steps in isolation. Both leave the global dependency structure of a workflow unexamined, and miss the inconsistencies that only emerge when the execution is viewed as a whole. We argue that anomaly detection for agentic AI must reason at the workflow level, where global execution structure exposes signals that local checks cannot see. We present Skynet, a principled workflow-level anomaly detection framework that turns observed multi-agent execution into directed workflow graphs and scores them against learned benign behavior. Skynet jointly models the semantic execution context and the structural organization of inter-agent delegation, tool invocation, and data-flow dependencies, and trains only on benign workflows. Because training never sees attacks or failures, this design naturally extends to zero-day detection: any execution that violates benign workflow regularities surfaces as off-manifold geometry under a single decision rule. We evaluate Skynet on three public agentic safety and failure benchmarks. It sustains high recall together with a sub-1% false positive rate, with per-workflow and per-step latencies low enough for online monitoring of agentic AI runtimes. --- ### NormViz: A Benchmark and Framework for Grounding Multimodal Reasoning in Global Cultures Authors: Akhila Yerukola, Fabrice Y Harel-Canada, Simran Khanuja, Abhinav Sukumar Rao, Ashima Suvarna, Nanyun Peng, Saadia Gabriel, Maarten Sap Published: 2026-09-06 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.06831v1 AI systems are used worldwide, but they struggle to serve the needs of culturally diverse populations. Prior work on cultural understanding evaluates AI systems on text-only settings or on visual artifact recognition (e.g. foods, clothing). The ability to reason about visually observable behaviors through local social norms, which we call visual norm understanding, remains unexamined. We introduce NormViz-Bench, a high quality, human-validated benchmark of 3,268 contrastive image pairs (6,536 images) spanning 16 countries. Each pair varies only in the culturally relevant behavior (e.g., objects, attributes, spatial relations, and actions) that alters how each image is interpreted. Each image is labeled as conforming to, violating, or irrelevant to local social norms, and pair-level evaluation requires both images to be correctly classified, thereby preventing reliance on superficial visual shortcuts. Even the strongest VLMs, Gemini 3.0 Flash and Qwen2.5 VL 7B, succeed on only 26.6% and 21.6% of pairs, struggling most with identifying violating and culturally benign visual behaviors. Towards bridging this, we introduce NormViz-Train, a training dataset of 64k images paired with explanations. Though absolute performance remains low (<30%), finetuning on NormViz-Train improves pair accuracy relatively by up to 125% and 36% Qwen3-VL 4B and 8B respectively, showing a path forward to teach models to connect visual perception to cultural significance. Together, NormViz-Bench and NormViz-Train establish visual norm understanding as a challenging and consequential frontier for multimodal AI. --- ### Constrained Bayesian Optimization for Hierarchical Federated Learning in IoT Networks for Plant Disease Classification Authors: Athanasios Papanikolaou, Athanasios Tziouvaras, Apostolos Xenakis, Periklis Chatzimisios, Shameem A. Puthiya Parambath, George Floros, Enrica Zereik, Ivan Petrovic, Fabio Bonsignorio Published: 2026-09-06 Categories: cs.LG, cs.DC arXiv: https://arxiv.org/abs/2609.06830v1 The deployment of Hierarchical Federated Learning (HFL) in resource-constrained Internet of Things (IoT) environments requires careful configuration to balance predictive performance with energy consumption and execution time. This challenge is particularly relevant to smart agriculture, where distributed IoT devices can support automated plant disease classification while operating under limited computational and communication resources. This paper presents a constrained Bayesian Optimization framework for the efficient configuration of HFL deployments. The proposed approach jointly explores the deep learning backbone architecture, aggregation strategy, and number of communication rounds, while the federation size is determined according to the spatial coverage requirements of the agricultural deployment. A weighted objective function captures user-defined trade-offs among energy consumption, execution time, and predictive performance, while explicit constraints ensure compliance with deployment-specific resource and accuracy requirements. The framework is evaluated on an IoT-based plant disease classification task considering multiple deep learning architectures, federated aggregation strategies, and communication-round settings. Experimental results across 30 independent optimization runs show that the proposed approach explores only 11.11% of the search space, while consistently identifying solutions within 1% of the exhaustive-search optimum, with a mean optimality gap of only 0.056%. --- ### Formation of structural attractors in neuromorphic systems Authors: Yurii Parzhyn, Alexander Schwarzmann, Mykyta Lapin, Kostiantyn Bokhan Published: 2026-09-06 Categories: cs.AI, cs.LG, cs.NE, q-bio.NC arXiv: https://arxiv.org/abs/2609.06826v1 This paper examines the theory of Invariant Structural Learning (ISL), which proposes a non-optimization approach to concept formation. Learning is interpreted as convergence to structural attractors in a hypergraph space, rather than as the minimization of a global loss function. The paper presents the ISL model, including its mathematical formalization, computational verification, and a hypothetical neurobiological interpretation. The mathematical section introduces the formal apparatus of the structural reduction process and proves its finite convergence, the existence and uniqueness of class structural attractors, and the self-organization of attractor maps. The computational section demonstrates the feasibility of the proposed approach on classical image recognition tasks, utilizing the proposed learning mechanism without backpropagation and with extremely small training datasets. Finally, the neurobiological section formulates hypotheses regarding the possible implementation of structural attractors in dendritic trees, neural coding as a projection of internal attractor dynamics, and the development of neural architectures supporting the proposed learning concept. These hypotheses are discussed in the context of modern experimental data in the fields of dendritic computations, synaptic plasticity, and the structural organization of neural circuits. The proposed neurobiological mechanisms are presented as testable hypotheses rather than established biological facts. The results demonstrate the mathematical consistency and computational feasibility of the proposed model, while the neurobiological hypotheses outline potential directions for its experimental verification. --- ### Generalist Open-World Temporal Perception Authors: Cristian Sminchisescu Published: 2026-09-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06823v1 The next generation of artificial intelligence systems will likely be natively temporal and multimodal in both inputs and outputs: able to converse, perceive, predict, reason, and synthesize through a shared world representation. Realizing this requires a temporal perceptual substrate integrating sensory streams, language, and structured outputs within a multimodal world model. We seek a generalist open-world perceptual system that represents biological forms, natural physical structures, and artifacts, and their interactions, as a coherent, temporally persistent process. The model should infer geometry, articulation, semantics, interaction structure, and uncertainty from raw multimodal streams; maintain identity through occlusion and viewpoint change; generalize across species, forms, mechanisms, and materials; and abstain or expand its ontology when encountering the unknown. The objective is a structured world state supporting understanding, prediction, counterfactual reasoning, and controllable synthesis. Recent work suggests that some cross-modal and reasoning-like capabilities can emerge from large-scale generative video pretraining, reminiscent of language-model scaling. Yet these capabilities are often accessed through language probes or expressed through photorealistic video, leaving explicit semantic, geometric, or temporal structure largely unexposed. This paper articulates an alternative and complementary paradigm: perception and synthesis as distinct conditionings within a shared Generalist Open-World Temporal Perception Architecture (GOWTPA). Recognition, structured prediction, and simulation arise as different conditionings of the same generative substrate, while reasoning and embodiment-specific policies build upon the resulting world state. This positions generalist temporal perception as a potential foundation layer for broader multimodal intelligence and physical AI. --- ### Diagnosing and Dynamically Filtering Occupancy World Models for Active Mapping Authors: Jiahui Zhang, Gongbo Liang, Yu Zhang Published: 2026-09-06 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2609.06820v2 Active mapping requires a robot to select camera viewpoints that efficiently reconstruct an unknown 3D scene. To reason about unobserved regions, recent systems use pretrained occupancy networks as world models that complete missing geometry. The predicted structure contributes to expected coverage gain and constrains feasible robot motion. Consequently, occupancy errors can change both what the robot chooses to explore and where it is able to move. We diagnose these effects by holding the planner fixed and varying only the occupancy representation provided to it. We consider planning without completion, with learned occupancy, with false positives removed by a ground truth oracle, with false negatives restored by an oracle, and with ground truth occupancy. Our experiments show that correcting false positives or false negatives alone does not consistently improve final coverage. This finding reveals a gap between occupancy accuracy and downstream planning performance. Ground truth occupancy provides a much larger improvement in coverage efficiency than in endpoint coverage, suggesting that planning and reachability remain important bottlenecks even when the geometric world model is accurate. Based on these findings, we introduce a dynamic filtering strategy that preserves predictions in unexplored space while suppressing repeatedly unsupported occupancy using online observations. Preliminary examples show that this strategy can redirect viewpoint selection toward reachable surfaces that would otherwise remain unobserved. --- ### Unsound Search with Policy and Value Networks in Legends of Code and Magic Authors: Dustin Rubin Published: 2026-09-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.06816v1 Decision-time search in perfect and imperfect information games with enumerable belief states are effective methods for game AI. Collectible card games are imperfect information games with large belief states. Legends of Code and Magic is a collectible card game competition where the belief states are $2^{101}$. The Legends of Code and Magic (LoCM) champion, ByteRL, plays with no search. Other works claim sound enumeration-based search is unusable in the genre due to the number of belief states. We measured three previously defined properties that predict where theoretically unsound perfect information Monte Carlo's defects are cheap and found LoCM sits in the favorable region. Starting with imitation learning of the runner-up policy, NeteaseOPD, we created a policy and value feed-forward network. Our agent searches over worlds sampled from a prior over the opponent's deck built from the runner-up's drafts. Using our strictest configuration in the battle phase we beat ByteRL with a win percentage of 51.35% 95% CI (50.37, 52.33), over 10,000 pre-registered games using the LoCM official referee and time limit. Search is not a minor factor on the matchup between our agent and ByteRL. Without search this agent scores 26.8% and adding search adds +24.6 points. Unsound search in imperfect information games could be exploitable. We replicate a published best-response attack against ByteRL. We then apply the same attack protocol to two search configurations of our agent, and each one resists it better than ByteRL at every iteration. In LoCM unsound search gives us a stronger and more resilient agent. --- ### Typed Federated Artifacts for the Agentic Web:Sharing Tool-Routing Knowledge Across Frozen,Heterogeneous LLM Agents Authors: Abhijit Chakraborty, Ni Trieu, Vivek Gupta Published: 2026-09-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.06815v1 An open, networked web will allow agents to run frozen models from multiple vendors, keep their history private, and teach each other which tool to call and when. Flat text (prompts, example pools) makes it difficult for the protocol to distinguish between noise statistics, merging rules, and documentation. Weights and adapters cannot transfer that knowledge between platforms. We suggest sharing typed federated artifacts, schema-validated objects with well-defined fields for per-field privacy (described here, but measured), dispute resolution, and cross-model transfer, and instantiating them as SYNAPSE1, a common tool-routing knowledge. After deleting 192 garbage entries and 1,916 training items that duplicate or almost duplicate test queries, a federated compendium routes within 1.1 points of a centralized one at 20 MB of JSON per client each round on StableToolBench (3,180 tools). The same experience merged and shown to the router as typed fields rather than one flat string is worth 8.5 points on clean data and 7.4 under 60% injected contradiction. Crossing merge and rendering shows the halves are inseparable (the typed merge shown flat is the worst arm), while three conflict policies are indistinguishable, so the conflict log that motivated this work is not the On τ-bench retail, each compendium arm improves GPT-4o agents' per-step tool-call accuracy by at least 6.7 points, attributed to format rather than federated experience. Two cautionary findings conclude the paper: on a topic-labeled math proxy and StableToolBench, a TF-IDF classifier over the same labeled experience beats every LLM routing arm (by 48 and 26 points, mostly retrieval recall) because the benchmark's pool holds labeled queries for every supposedly unseen tool and every test query verbatim before our filter. It cannot measure routing to tools without labels, which routing exists for. --- ### Measuring GEO Visibility: Prompt Corpora Define the Answer Market Authors: Olivier Martinez Published: 2026-09-06 Categories: cs.IR, cs.CL arXiv: https://arxiv.org/abs/2609.06811v1 GEO (generative engine optimization) visibility scores aggregate source appearances, citations, or brand mentions in generated answers. The prompt corpus selects the situations evaluated, while weights determine their relative importance. Together they define an "answer market" that need not represent actual user demand. Prompt wording can alter retrieval, competing sources, and generated answers. Scoring then requires identifying the appearances, citations, or mentions of interest. If a language model performs this task, its instruction can change the score assigned to an unchanged answer. Our critical survey examines how these choices help define what a GEO score measures. It draws on research into whether indicators measure the intended phenomenon, total survey error, and information retrieval evaluation. The framework specifies situation annotation, prompt formulations, execution conditions, weights, and scoring rules. When weights are unknown or remain to be chosen, the framework reports sets of admissible scores. It distinguishes values compatible with data and assumptions about a target population (partial identification) from variation across weighting conventions (normative sensitivity). A citation alone does not establish a source's contribution. The article defines a comparison of answers generated with and without a source in a controlled documentary context, distinct from an intervention on the full engine with competing sources. The framework is supported by reproducible calculations. No new experiments are reported; its general empirical validity remains to be assessed. --- ### Disparity Has a Sign: Stereo Matching Beyond the Zero-Disparity Plane Authors: Jian Shi, Xinge Yang, Chaoyang Wang, Wolfgang Heidrich, Peter Wonka Published: 2026-09-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06809v1 Modern stereo matching models fail when disparity crosses zero, with end-point error (EPE) rising by 4.6-37$\times$. Yet stereoscopic content, from cinema 3D to VR, routinely contains objects behind the zero-disparity plane (ZDP), corresponding to negative disparities. The blind spot cascades through datasets, architectures, and evaluation protocols, all of which inherit the non-negative geometry. Rectified parallel cameras place ZDP at infinity, so every finite depth yields $d=fB/z \ge 0$ by construction, and nothing within the standard pipeline can violate, or even measure, a negative disparity. To measure it, we propose \textit{ZDPShift}, a benchmark of $21{,}495$ stereo pairs from seven cinematographer-authored open movies, each frame rendered at five zero-disparity-plane positions with dense signed ground truth. Six state-of-the-art image and video stereo matching models collapse once the plane moves. On identical scene content, FoundationStereo goes from $2.24$ px EPE to $75.33$px, with every backbone leaving roughly half of all pixels exceeding a three-pixel disparity error. What is missing, however, is not the underlying matching capability. % The capability itself, however, is already present. Training on supervision synthesized from SceneFlow, which adds no new data or parameters, keeps the error flat across the signed range. Training only the decoder, with the pretrained matching features frozen, performs comparably across all six backbones, with EPE jittering within $0.2$px. Thus, the pretrained features already extend to the negative regime they were never trained on, and only the output convention discarded it. Meanwhile, positive-regime accuracy on KITTI, Middlebury, ETH3D, and Sintel is largely preserved. --- ### Comparative Study of Anatomical and Learned Features in AI Models for Structural Brain MRI Authors: Boyang Yu, Miquel Lopez Escoriza, Long Chen, Arjun V. Masurkar, Narges Razavian, Carlos Fernandez-Granda Published: 2026-09-06 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.06807v1 In this work, we comprehensively evaluate three popular feature-extraction paradigms in AI-based neuroimaging modeling: (1) computation of anatomical surfaces and volumes, (2) supervised learning with convolutional neural networks (CNNs), and (3) unsupervised pretraining of vision transformer (ViT) foundation models, followed by supervised finetuning. Our study is based on 18 publicly available datasets containing 3D structural T1-weighted MRI scans from approximately 80,000 participants across seven distinct clinical tasks. We observe that a linear model based on anatomical features matches the diagnostic performance of complex nonlinear features learned by sophisticated AI frameworks, including foundation models trained on thousands of scans. Conversely, CNNs and pretrained ViTs learn features that implicitly capture relevant anatomical information, bypassing the need for explicit feature extraction. Building upon these insights, we propose Anatomy Segmentation Pretraining (ASP), a novel method to incorporate anatomical information during foundation-model pretraining, which outperforms existing models in biological age estimation. --- ### Train Smarter, Not Harder: Switching Signal-Guided Training in Active Learning Authors: Nagham Omar, Maya Rozenshtein, Evgeny Mishlyakov, Avigdor Gal Published: 2026-09-06 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2609.06806v1 Training strategy, namely whether to retrain from scratch or fine-tune from the previous checkpoint, is an overlooked decision variable in active learning. We show that this choice has exploitable structure: retraining is most useful in early rounds, when each batch can substantially reshape the labeled distribution, while fine-tuning becomes safer once the model trajectory stabilizes. We propose HybridAL, an adaptive training schedule that monitors an online stabilization signal and switches from retraining to fine-tuning after sustained stabilization. Two complementary signals, spectral exponent change $Δα$ (weight-based) and accuracy change $Δ$Acc (validation-based), span different points on the time-calibration trade-off. Across three encoder backbones and six text-classification tasks (five seeds each), HybridAL keeps endpoint macro-F1 non-inferior to retraining and fine-tuning at a 0.010 margin, saves up to 49% of retraining time, and recovers a substantial fraction of retraining's calibration advantage as measured by negative log-likelihood (NLL). Compared with schedules that switch at a pre-committed round, HybridAL obtains lower NLL at moderate additional cost, showing that trajectory-dependent switching provides a stronger time-calibration trade-off than fixed early switching. --- ### Hardware Trojan Threats to Multi-Chiplet Photonic Neural Network Accelerators Authors: Sudeep Pasricha Published: 2026-09-06 Categories: cs.AR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.06796v1 Multi-chiplet photonic neural network accelerators (MCPNAs) combine photonic computation, photonic communica-tion, and heterogeneous chiplet integration to enable scalable and energy-efficient AI acceleration. However, their distributed archi-tecture and reliance on third-party chiplets introduce significant hardware security risks. This paper examines Hardware Trojan (HT) threats to MCPNAs across three dimensions: confidentiality, integrity, and availability. --- ### Improving Proficiency and Efficiency of Android GUI Agents via Self-Generating Tool Actions Authors: Juyong Lee, Woogyeol Jin, Kimin Lee Published: 2026-09-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.06792v1 Android agents using a hybrid action space that combines GUI actions and tool actions (e.g., accessing application data via APIs) remain largely underexplored, mainly due to the excessive effort required to create tools. To address this gap, we introduce DroidTool, a framework for augmenting the agents with self-generated tools, which are realized as Python functions operating on application states (e.g., a database). To create tools with minimal human labor, DroidTool employs an agentic workflow featuring stages: proposal, implementation, test generation and execution, and repair. Notably, when testing the created tools for verification, it constructs relational tests across relevant tools for natural preparation of appropriate test preconditions and improved test coverage, rather than testing each tool separately. The GUI agents augmented with the generated tools achieved approximately 4.47%p higher performance with approximately 20.05% fewer interactions than the GUI-only agents, averaged across representative benchmarks: AndroidWorld, B-MoCA, and MobileSafetyBench. --- ### When Speech Meets Lips: Interpretable Audio-Visual Synchronization for L2 Pronunciation Assessment Authors: Bowen Yu, Mingyu Huang, Yishen Liu, Yue Zhao Published: 2026-09-06 Categories: cs.CV, cs.SD arXiv: https://arxiv.org/abs/2609.06788v1 Automatic Pronunciation Assessment (APA) systems have achieved strong performance with transformer-based models and self-supervised speech representations. However, most methods rely only on acoustic signals and overlook temporal synchronization between speech and articulatory movements, limiting diagnostic feedback on timing mismatches important for L2 pronunciation training. We propose an interpretable audio-visual synchronization framework that explicitly models speech-lip temporal alignment through feature encoding, cross-attention fusion, lag estimation, stability quantification, and visualization. The framework introduces frame-level lag trajectories and a Lag Stability Index (LSI) to quantify synchronization robustness. We also interviewed 30 participants, including 10 instructors and 20 students with diverse first-language backgrounds, to assess its effectiveness. By transforming implicit alignment into interpretable representations, the framework connects automatic scoring with actionable Computer-Aided Pronunciation Training feedback. Datasets and supplemental materials are available at https://www.robots.ox.ac.uk/~vgg/data/lip_reading/. --- ### When Retain Constraints Conflict: Mitigating Forget-Retain Interference in Tabular Data Authors: Zijie Liu, Jinhao Duan, Bingqi Shang, Xinming An, Sijia Liu, Tianlong Chen Published: 2026-09-06 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.06786v1 Machine unlearning aims to remove the influence of designated training data while preserving model utility, but its behavior on tabular data remains underexplored. This gap is important because tabular prediction is widely used in high-stakes domains and is increasingly adapted to language models through record serialization and schema-aware prompting. We identify a key challenge that distinguishes tabular unlearning from unlearning in free-form text or other modalities: schema-induced forget-retain overlap. In serialized tabular data, records share fixed column-name/value slots, similar attribute ranges, and common output spaces. Consequently, a forget row may have nearby retain rows that rely on the same high-signal attributes, causing retain preservation to oppose the update required for forgetting. Motivated by this failure mode, we propose Conflict-Aware Unlearning (CAU), a schema-aware approach that reduces forget-retain interference by relaxing preservation constraints on retained rows that most conflict with the forget set. Across sample-level and feature-level unlearning on clinical and non-medical tabular tasks, CAU more closely matches a retraining oracle while maintaining predictive utility and retain-region behavior. Our results show that reliable tabular LLM unlearning depends not only on the forgetting objective, but also on how retain constraints are constructed. --- ### Unified Multi-Layer Subspace Modeling for Cross-Domain OOD Detection Authors: Gerhard Krumpl, Henning Avenhaus, Horst Possegger Published: 2026-09-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06785v1 Out-of-Distribution (OOD) detection remains a fundamental challenge for neural networks, whose predictions can be overconfident on inputs that deviate from the training distribution. Most post-hoc OOD detection methods derive scores from a single representation level (eg., logits or penultimate features) or combine multiple layers via depth selection or OOD-calibrated weighting. However, because OOD shifts are diverse, the most informative representation level can vary strongly across OOD types and domains, making fixed-layer choices and OOD-calibrated aggregation brittle. In this paper, we propose PRISM (Projected Representation with Intermediate-layer Subspace Modeling), a model-agnostic post-hoc OOD detection method that models a unified multi-layer feature representation rather than aggregating independently scored layers. PRISM fuses intermediate and deep features into a single hierarchical embedding, estimates an in-distribution (ID) principal subspace, and then combines two complementary signals: (i) a class-conditional Mahalanobis distance in the projected subspace and (ii) the residual energy orthogonal to the learned manifold. This simple design avoids OOD-tuned layer weighting while capturing both in-subspace semantic deviations and off-subspace anomalies. Across diverse benchmarks spanning natural images, medical imaging, and industrial visual inspection, PRISM achieves consistent state-of-the-art cross-domain OOD detection performance with a single default configuration across all evaluated domains and architectures. We further show that PRISM incurs minimal inference overhead, making it practical for real-world deployment. --- ### AURA-Eval: Evaluation Framework for Acting Under Risk Awareness in LLM Agent Trajectories Authors: Ruoxi Shang, Christina-Maria Androna, Orfeas Menis Mastromichalakis, Yu Feng, Aniruddhan Ramesh, Rico Angell, Shang Hong Sim, Chrysoula Zerva, Emmanouil Koukoumidis Published: 2026-09-06 Categories: cs.CR, cs.AI, cs.CL, cs.SE arXiv: https://arxiv.org/abs/2609.06783v1 LLM agents operate in workflows where unsafe actions can have real consequences. Existing safety evaluations often reduce behavior to a single score, obscuring risk recognition, pre-action detection, and safe task completion when a safe solution exists. We introduce AURA-Eval, a framework combining controlled augmentation with granular diagnosis of behavior in tool-use trajectories. Its pipeline identifies safety-critical decision points, generates controlled variations, and constructs counterparts differing in whether a request has a safe fulfillment path. Using 157 sourced trajectories, we generate 1,249 evaluation items and evaluate 20 frontier and open-weight models. We developed rubrics to classify risk detection, action strategy, and scenario-specific action safety. Our results show that LLM agents engage in unsafe behavior more often when no safe fulfillment path exists. In these cases, frontier proprietary models more often recognize risk and exhibit safer behavior by proposing alternatives, while evaluated open-weight models more often directly execute unsafe requests. Increasing impact or reducing opportunities for oversight before execution also exposes greater vulnerability across models. --- ### DrugReason: Dynamic Multi-View Reasoning over Knowledge Graph and Language Evidence for Drug Repurposing Authors: Zijie Liu, Hongxuan Li, Zhen Tan, Jinhao Duan, Baixiang Huang, Zunpeng Liu, Kai Shu, Tianlong Chen Published: 2026-09-06 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.06779v1 Drug repurposing aims to identify new therapeutic uses for existing compounds and, compared with de novo drug discovery, offers a faster and more cost-effective path to clinical translation. However, the space of candidate drug-disease pairs is enormous and their underlying relationships often depend on complex multi-hop biological mechanisms, making it difficult to reliably predict which pairs represent true therapeutic relationships. Existing approaches tackle this from two directions: knowledge graph-based methods organize curated biomedical evidence into structured relational networks for grounded multi-hop reasoning, while LLM-based methods leverage pretrained knowledge to generate flexible mechanistic rationales. Yet neither is sufficient alone - KGs are confined to observed graph structure while LLMs lack factual grounding and risk hallucination. To address this gap, we propose DrugReason, a multi-view reasoning framework that integrates grounded KG reasoning with LLM-generated mechanistic inference for drug repurposing. DrugReason adaptively routes diverse reasoning paths to specialized experts conditioned on the query context, while a cross-expert distillation objective enables knowledge sharing without sacrificing expert specialization. Experiments on PharmaDB, DDInter, and DrugBank show that DrugReason improves average performance over strong single-view reasoning baselines and achieves competitive or superior results compared with graph-based alternatives, while providing interpretable routing-based predictions. --- ### NEO-BENCH: A New Multi-Source Benchmark for Generalizable Astronomical Streak Detection Authors: Jiayou He, Jessica Yao Published: 2026-09-06 Categories: astro-ph.IM, cs.CV arXiv: https://arxiv.org/abs/2609.06774v1 Near-Earth Objects (NEOs) can appear as faint streaks in long-exposure astronomical images. Detecting these streaks across diverse observatories requires methods that remain reliable despite differences in image quality, orientation, sky background, and noise. However, existing detectors are commonly evaluated using data from only one source, providing limited evidence of cross-source generalization. We introduce NEO-Bench, a multi-source benchmark containing 8,376 images from five astronomical-image datasets. The sources include the Hubble Space Telescope, a Stellina smart telescope, the United Arab Emirates Meteor Monitoring Network, a TETRA1 telescope using a Celestron C14 with Fastar, and the Roboflow Asteroid dataset. We converted the data to a common YOLO format, audited a sample of labels, and defined within-source and leave-one-source-out evaluation protocols. We evaluated four approaches: Hough, Radon, Gaussian PSF, and YOLO26L. Leave-one-source-out F1 decreased in 14 of 20 image-level method-source pairs and 13 of 20 IoU@0.50 localization pairs. Across the datasets categorized as medium or hard, F1 decreased in 11 of 12 image-level pairs and 9 of 12 localization pairs. These results show that cross-source performance remains inconsistent and that reliable generalization across astronomical imaging sources remains an open challenge. The benchmark, code, and data are publicly available. --- ### CAROL: Context-Aware Online Learning for Fuzzer Scheduling Authors: Zirui Liu, Mengfan Xu, Juan Zhai, Shenglong Yao, Shiqing Ma Published: 2026-09-06 Categories: cs.SE, cs.LG arXiv: https://arxiv.org/abs/2609.06772v1 Ensemble fuzzing runs multiple fuzzers on a target while a scheduler allocates CPU time among them. Existing schedulers base these decisions on compact summaries of past performance and rules fixed before a campaign. Our measurements reveal two limitations. First, past-reward summaries do not reliably capture performance evolution: after accounting for estimation noise, agreement between consecutive-window rankings is statistically indistinguishable from within-window self-agreement. Second, predictive signals vary across targets: on eight of nine targets, a weighting learned from the other eight predicts reward worse than one learned on the current target. We introduce CAROL, an online scheduler that uses each fuzzer's current context. Already available to the dispatch loop, this context describes reward trends, waiting and plateau time, reached code, and estimation uncertainty. CAROL uses context in two ways: a domain-guided method detects whether a fuzzer is rising or rotting and applies a phase-specific learning rule, while a learned method predicts reward from 15 context signals and uses predictive uncertainty for online selection. Across nine Magma targets, CAROL triggers more unique bugs than each of three ensemble-scheduling baselines whenever their results differ, and fewer on none. Compared with the strongest baseline for each target, CAROL gains 11.8% and surpasses an oracle that retrospectively selects the best single fuzzer per target. Removing context eliminates the gain, and the additional bugs are concentrated among those the baselines trigger rarely or never. Run unchanged on five widely used C++ programs, CAROL finds 120 previously unknown crashing defects, deduplicated by site, fault, and entry point; all were reported to maintainers through the projects' stated disclosure channels. --- ### AuthBench: A Large-Scale Multilingual Benchmark for Authorship Representation across Genres and Lengths Authors: MaoXun Huang, Zhenxing Zhang, Claire Cardie Published: 2026-09-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.06771v1 Authorship signals matter in settings where writing style carries identity: digital forensics, plagiarism analysis, account linking, misinformation investigation, and machine-generated text detection. Yet current authorship benchmarks remain fragmented, usually covering only a narrow language set, a single genre, or a limited document-length regime, which makes it difficult to assess whether modern representations truly generalize. We introduce AuthBench, a large-scale multilingual benchmark for authorship representation that is designed to make this evaluation broad, standardized, and realistic. AuthBench contains 428,150 documents written by 153,825 individuals across ten widely used languages, 9 primary genres, 66 fine-grained genres, and four document-length buckets. It supports two complementary tasks: authorship attribution, formulated as same-author retrieval and authorship verification, formulated as same-author binary decision. We benchmark 47 neural models and three non-neural baselines under a unified zero-shot protocol. Results show that authorship representation remains far from solved: the best retrieval model reaches only 0.258 Success@5, while the best verification model achieves 0.076 EER and 0.968 ROC-AUC. The leaderboard also reveals a meaningful task split, with different model families leading retrieval and verification, and large performance differences across languages, genres, and lengths. These findings position AuthBench not only as a new benchmark, but as a diagnostic resource for studying when and why authorship representations succeed or fail. We release AuthBench, its evaluation toolkit, and benchmark data at https://github.com/mao-code/AuthBench and https://huggingface.co/datasets/MaoXun/AuthBench. --- ### Ordinary, Reasonable Chatbots: Do AI Models Track Human Legal Judgments? Authors: Nirav Patel, Emily Wenger, Christopher Buccafusco Published: 2026-09-06 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2609.06769v2 As people increasingly rely on artificial intelligence (AI) for guidance in their own lives, scholars, lawyers, and even judges have begun to consider the role of AI in legal decision-making. As "silicon sampling" -- the use of generative AI models in social science research -- is now impacting academia, "silicon jurors" could make an appearance in courtrooms. This study joins an emerging line of research on generative AI models' ability to simulate human legal judgments. In particular, we study how large language model (LLM)-powered chatbots respond to series of questions about legal reasonableness. When the law needs to judge the appropriateness of a behavior, it most often asks whether the behavior was "reasonable." Yet despite the ubiquity of reasonableness judgments, they are the site of constant vexation for lawyers, judges, and lay people. Reasonableness seems inherently vague and unpredictable, since it relies on variable context and implicit conceptual schemas. Moreover, many scholars caution that reasonableness judgments may vary along demographic lines. We compare the answers of human participants to those of twenty-six LLMs across twenty-five different legally relevant reasonableness judgments. Overall, our findings suggest that chatbot responses generally track those of human participants. Nonetheless, we find some suggestive -- and potentially concerning -- results. Compared to humans, LLMs generate more homogeneous responses and occasionally treat a variable standard as an invariant rule. And, compared to humans, LLMs tend to generate answers that are more favorable to the government and to corporations. Finally, our results indicate that LLMs' responses tend to align more closely with those of respondents who are white, male, older, and more educated. More systematic research is needed to confirm or reject these initial findings. --- ### LATS: Levy Adaptive Tree Sampling for Feedback-Driven Diverse Target Discovery Authors: Binglin Ji, Anindya Sarkar, Hengchang Lu, Lecheng Kong, Yixin Chen, Yevgeniy Vorobeychik Published: 2026-09-06 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.06761v1 While diffusion models excel at capturing complex data distributions, scientific discovery often requires steering generation toward specific, uncharacterized regions that maximize a target objective. These high-utility modes frequently reside in low-likelihood tail regions and are only revealed sequentially through interactive feedback. Existing diffusion samplers fail in this regime: they inherit the pre-trained model's bias toward high-density regions, leaving rare yet promising phenomena underexplored. Conversely, exploration-heavy samplers ensure broad coverage but fail to efficiently exploit high-utility modes when constrained by a strict sampling budget. To resolve this dilemma, we introduce Levy Adaptive Tree Search (LATS), a principled sampling framework for online feedback-driven search. LATS leverages heavy-tailed exploration coupled with tree-based value backpropagation to progressively uncover preferred modes. By maintaining broad distributional coverage, LATS successfully discovers low-likelihood, high-utility regions while preserving sample fidelity and structural diversity. Experiments across diverse benchmarks, including materials science, demonstrate that LATS significantly outperforms baselines in target discovery efficiency. --- ### Agentic Visual Generation: From Generative Models to Agentic Control Authors: Yinming Huang, Shuyuan Tu, Xi Yan, Jiahao Zhan, Zihan Yang, Zhen Xing, Hui Zhang, Tiehua Zhang, Yu-Gang Jiang, Zuxuan Wu Published: 2026-09-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06758v1 Visual generation is evolving from generative models used through a single invocation into agentic control processes that can plan, select tools, inspect intermediate synthesized outputs, revise failures, and reuse prior experience. In most existing systems, the controller is an LLM or VLM, while visual generation models serve as tools or executors. However, existing work lacks a consistent criterion for determining when a generation system becomes agentic. Planning depth, tool use, multi-role collaboration, and reinforcement learning are often treated as evidence of agenticity, even though none of them necessarily determines which generation decisions the controller can make. We organize the field according to what the controller can directly control in the generation process. At L1 Conditioning Control, the controller prepares the input to a predetermined generator but does not control which visual operation is executed. At L2 Execution Control, it selects and invokes actual generation, editing, rendering, or other content-modifying operations. At L3 Outcome-Adaptive Control, it observes an intermediate outcome and uses that observation to change a subsequent operation within the current task. At L4 Experience-Adaptive Control, it retains experience from completed tasks and uses that experience to change decisions on future tasks. L0 Fixed Support separately denotes generators, editors, evaluators, reward models, benchmarks, and fixed pipelines without a deployed controller that makes generation-level decisions. These levels describe a progressively broader decision-making scope rather than model size, system complexity, output quality, tool or role count, or training method. Applying this framework across image, video, editing, 3D, world, slide, and user-interface generation reveals how controller capabilities have evolved and how their mechanisms are distributed across levels. --- ### A Novel Semantic Manifold Alignment Attack against Embedding-to-Embedding Obfuscation in Privacy-Preserving LLMs Authors: Sicong Li, Lingfeng Yao, Xingke Yang, Ke Tu, Chenhao Wu, Hao Wang, Jiang Liu, Phone Lin, Xin Fu, Miao Pan Published: 2026-09-06 Categories: cs.CR, cs.CL arXiv: https://arxiv.org/abs/2609.06749v1 With the widespread applications of large language models (LLMs), privacy-preserving inference has become increasingly essential for sensitive queries. To balance privacy and utility, a series of lightweight obfuscation approaches has recently been proposed, where users locally transform plaintext embeddings into the fixed ciphertext ones. While such Embedding-to-Embedding Obfuscation (E2EO) schemes demonstrate considerable resilience against traditional token frequency and embedding inversion attacks, the core mechanism behind remains to be the large-scale one-to-one substitution, which provides no cryptographic guarantees. In this paper, we propose Proxy Manifold Alignment (PMA), a novel attack against E2EO in privacy-preserving LLMs. Our key observation is that E2EO schemes keep the original semantic structure, so that the obfuscated vector stream can be regarded as an unknown tokenizer-language whose symbols are the vectors themselves. Therefore, the proposed ciphertext to plaintext reconstruction attack can be formulated as a translation task from the unknown tokenizer-language to plaintext. Specifically, by only accessing the obfuscated vector stream, the target tokenizer and a public corpus, the PMA attack first employs Word2Vec to model the co-occurrence patterns within the obfuscated stream and the public corpus independently, and constructs two proxy vector embeddings. Then, the attack aligns the underlying manifolds of these two embeddings based on structural similarity. Finally, it maps the obfuscated vectors back to plaintext. Experimental results demonstrate that PMA consistently achieves higher plaintext recovery than other state-of-the-art attack methods. --- ### Reason Through the Latent! Making Latent Visual Reasoning Necessary Authors: Suhyeong Park, Junha Jung, Jaewoo Kang Published: 2026-09-06 Categories: cs.AI, cs.CL, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.06746v3 Latent visual reasoning aims to perform multimodal reasoning through hidden-state computation rather than explicit textual chains of thought. However, visual information being present in a latent state does not imply that the model actually relies on that state when producing its answer, especially when alternative image-conditioned paths remain available. We introduce Causal Visual Recurrent Reasoning (CVRR), which preserves pretrained visual competence while making recurrent computation the required image-conditioned path to prediction. CVRR initializes recurrence from the question hidden state after the pretrained vision-language model has incorporated the image, then repeatedly updates this state while re-reading the same fixed visual evidence. Before decoding, visual states and the original multimodal KV cache are removed so that only the final recurrent state carries image-conditioned information to the answer. Across the $V^*$, MMVP, BLINK, and MME-RealWorld-Lite benchmarks, CVRR retains strong performance under this strict interface, while compatible latent reasoners fail to recover comparable visual competence even when retrained under the same constraint. Causal interventions further show that predictions remain sensitive to recurrent content when the question is held fixed, and that persistent visual evidence causally revises the recurrent trajectory. These results distinguish latent informativeness from latent computation that is actually used for prediction. --- ### Simulating the Marginal Green Contribution of AI Modules in a Smart-Agriculture Platform: Evidence from Two Monte Carlo Experiments Authors: Zhaoyang Li, Ruijie Zhang, Zhaoji Sun, Lu Zhang Published: 2026-09-06 Categories: cs.AI, math.NA arXiv: https://arxiv.org/abs/2609.06740v1 Smart agriculture platforms usually bundle AI diagnosis, IoT sensing and decision push into a single package, so the green benefit attributable to each component remains unclear and resource-allocation decisions lack quantitative evidence. Building on a previous platform-level Monte Carlo assessment, this paper makes the components explicit and runs two controlled simulation experiments. Experiment 1 follows the chain from AI capability to farmer behavior to agrochemical input reduction, modeling pesticide/fertilizer reduction as avoidable blind-application share times prescription effectiveness times decision-touch coverage times adoption rate, and compares an experienced-extension mode with the AI mode: the probability of reaching 20% pesticide reduction is essentially zero in the extension mode but 20.7% at baseline, up to 49% with diagnosis accuracy 0.95 and adoption 0.85 under AI; the probability of 15% fertilizer reduction rises from near zero to 52.0%. Experiment 2 compares current practice (P0), IoT engineering retrofit (P1), and P1 plus AI irrigation scheduling (P2): median aggregate water saving rises from 7.8% (P0) to 11.0% (P1) and 16.0% (P2), with AI adding 5.0 percentage points beyond engineering; paddy CH4 reduction reaches 30.5% under AI scheduling versus 19.8% under manual operation, and the rice irrigation-methane subsystem carbon intensity declines 27.9%. Sensitivity analyses of both experiments consistently indicate that the primary bottleneck for meeting green targets is farmer adoption rather than algorithm accuracy, and that AI data fusion is robust to soil-moisture sensing errors. This work provides a reproducible simulation framework for component-level green-value evaluation and promotion-strategy optimization of smart agriculture platforms. --- ### Monte Carlo-Based Ex-Ante Assessment of the Green Benefits of an AI-Driven Smart Agriculture Platform in Hainan Authors: Zhaoyang Li, Ruijie Zhang, Zhaoji Sun, Lu Zhang Published: 2026-09-06 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.06737v1 Smart agriculture platforms are widely regarded as key carriers for implementing China's pesticide and fertilizer reduction, water-saving and carbon-reduction agendas, yet a unified quantitative framework for assessing their green value is still lacking. Taking an AI-driven decision platform for tropical agriculture as the object (integrating large-language-model question answering, multimodal pest diagnosis, IoT sensing, satellite remote sensing, and a closed-loop field record system), this study builds a cradle-to-farm-gate agricultural carbon accounting model covering pesticide and fertilizer production, field N2O, irrigation electricity and paddy CH4, translates platform interventions into quantifiable transmission parameters, and propagates parameter uncertainty by Monte Carlo simulation over three Hainan scenarios (mango, winter vegetable, rice/nanfan, area-weighted 40%:30%:30%). Under full adoption, median reductions are 23.5% (90% interval 15.0%-33.2%) for pesticide use, 21.0% (13.8%-28.9%) for fertilizer, 16.5% (10.9%-23.5%) for irrigation water, and 21.5% (16.1%-27.2%) for carbon intensity. Attainment probabilities are high for fertilizer reduction >=15% (90.6%) and clear carbon decline (98.1%), but only about 20% for aggregate water saving >=20%, favoring scenario-specific statements. Sobol first-order indices show soil-test recommendation and organic substitution jointly explain about 83% of the variance of aggregate carbon-intensity reduction. Convergence tests show 10,000 iterations stabilize all statistics; conservative/baseline/optimistic scenario bounds are reported. The framework offers a reproducible, calibration-ready methodology for ex-ante green-value assessment and pilot observation design. --- ### LASSNet: Level-Aware Availability-Conditioned Spatial-Semantic Fusion for Brain Tumor Segmentation with Missing MRI Modalities Authors: Haobin Chen, Ao Chang, Rundong Wang, Zhicheng Li, Zhihao Tang, Heqin Zhu Published: 2026-09-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06733v1 Brain tumor segmentation from multimodal MRI relies on complementary evidence across four imaging sequences, yet one or more modalities may be unavailable because of acquisition cost, protocol variation, scan failure, or patient condition. Existing work has explored reconstruction, knowledge transfer, and direct feature fusion, but leaves open whether missing-modality fusion should change with representation level. High-resolution lateral features retain spatial detail, whereas compressed bottleneck features encode semantic and inter-modality context. We therefore hypothesize that fusion should be conditioned jointly on modality availability and feature hierarchy. We propose the Level-Aware Availability-Conditioned Spatial-Semantic Fusion Network (LASSNet), which contains two level-specialized modules. Hierarchical Availability-Conditioned Fusion (HACF) constructs four lateral representations using count-normalized aggregation of available modalities, mask-conditioned channel modulation, and local 3D refinement. Tri-Scale Relational-Spatial Fusion (TriRSF) models relations among available modality descriptors and spatial context across multiple bottleneck resolutions, followed by cross-scale aggregation and availability-conditioned global spatial attention. A shared coarse-to-fine decoder starts from TriRSF semantics and progressively injects HACF features, without reconstructing missing inputs. Across all 15 non-empty modality configurations, LASSNet obtains mean Dice scores of 76.7% and 83.2% over WT, TC, and ET on BraTS2019 and BraTS2023, respectively. --- ### Event Interaction in Low-Rank Bottlenecks for Temporal Relation Extraction Authors: Wei Sun, Tingyu Qu, Jesse Davis, Marie-Francine Moens Published: 2026-09-06 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.06731v1 Temporal relation extraction determines whether an event occurs before, after, or simultaneously with another event, and therefore relies on accurately modeling how the two events interact. Mainstream systems achieve this by concatenating event spans or using shallow fusion, which works well when all model parameters are trainable. However, in parameter-efficient fine-tuning, low-rank bottlenecks restrict information flow and prevent these interaction signals from passing through, leading to clear performance drops. To address this limitation, we propose a theoretically grounded architecture, Convolutional Bottleneck Interaction (CBI), which first applies lightweight depthwise convolution to enhance event representations and then uses element-wise multiplication to capture effective event-event interactions inside the bottleneck. Across five datasets and seven backbone models in the Adapter and LoRA settings, CBI provides consistent and substantial gains, up to +31.7 micro F1, while adding minimal computational cost, showing that explicit interaction inside low-rank spaces is crucial for temporal relation extraction. The code is available at https://github.com/VRCMF/CIF.git. --- ### Uni-Light: An Ultra-Lightweight Framework via Uncertainty-Aware Knowledge Distillation for Brain Tumour Segmentation Authors: Libing Kuang, Soren Salehi, Ziling Wu, Ahmad P. Tafti, Armaghan Moemeni Published: 2026-09-06 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.06729v2 Accurate 3D brain tumour segmentation from multi-modal Magnetic Resonance Imaging (MRI) is essential for clinical diagnosis and treatment planning. Existing brain tumour segmentation methods often suffer from heavy computational demands, while current lightweight architectures frequently lack the capacity to maintain segmentation fidelity in complex tumour regions. To address these issues, we propose a novel ultra-lightweight framework (Uni-Light) that achieves high-fidelity segmentation with substantially reduced computational overhead. It combines multi-scale convolutions with an uncertainty-aware knowledge distillation scheme that directs the student model toward hard-to-classify regions, complemented by a Signed Distance Field boundary loss for geometric constraints. Experimental results on BraTS2023-GLI and MSD-BTS datasets demonstrate that Uni-Light reduces parameters by 97.56%, floating-point operations (FLOPs) by 73.03%, and inference memory footprint by 81.58%, while surpassing the state-of-the-art model by an average of 1.47% in Dice score, offering a highly competitive trade-off between segmentation accuracy and computational efficiency in resource-constrained clinical settings. This work also advances data engineering for medical imaging by demonstrating that teacher model uncertainty can be exploited as a data-driven supervisory signal, re-prioritising the training data distribution without requiring additional annotation. ---