# Research Papers — 2026-09 · bundle page 251 of 266 > Full-text markdown of papers #12501–12550 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/251 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/250.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/252.md) ### Loom: Weaving Diagnostic Strands into Free-Text Consensus via Embedding-Space Reweighting Authors: Ron Begleiter, Katya Egert Berg, Gilad Saban, Gil Shabat Published: 2026-09-02 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.02649v1 Aggregating noisy, conflicting textual hypotheses into a reliable consensus is a fundamental challenge when deploying NLP systems in real-world industrial settings. While monolithic Large Language Model (LLM) agents offer unbounded expressivity for tasks like Root Cause Analysis (RCA), they suffer from context limits, compounding hallucinations, and prohibitive inference latency. Traditional weak supervision offers statistical rigor but is mathematically restricted to discrete classes. We present Loom, a generative consensus framework deployed for real-world RCA that bridges these paradigms. Loom aggregates open-form hypotheses emitted by modular heuristics (diagnostic templates dynamically populated with episode-specific entities, times, and metrics) by projecting them into a continuous embedding space, and resolves conflicting signals with an iterative centroid-based reweighting algorithm. The resulting consensus weights ground a single lightweight LLM synthesis step. Evaluated on the OpenRCA benchmark, Loom occupies the accuracy--efficiency Pareto frontier: it matches a state-of-the-art autonomous agent on Bank and Market-2 and trails on Market-1 and Telecom, while using a single LLM call per incident on all four datasets ($\sim$26$\times$ faster; $\sim$33$\times$ with an 8B-parameter synthesizer). We discuss our deployment experience, highlighting lessons learned regarding the trade-offs between agentic depth and inference latency, negative results in redundancy detection, and how deterministic consensus fosters trust among Subject Matter Experts~(SMEs). --- ### Differentiable Electricity-Market Clearing for Gradient-Based Planning Authors: Luca Mungo, Maarten P. Scholl, Arnau Quera-Bofarull Published: 2026-09-02 Categories: cs.LG, cs.CY arXiv: https://arxiv.org/abs/2609.02646v1 Planning a large data center is difficult because a facility big enough to matter changes the electricity prices it will pay. Those prices are set by market clearing, a constrained optimization problem solved anew in every operating condition. However, simulating the market tells a planner how a candidate plan performs but not how to improve it. Here we treat market clearing as a differentiable optimization layer: each forward pass solves the market, and reverse-mode automatic differentiation propagates the planning cost back through the cleared prices to the plan. After validating these gradients against finite differences, we apply them to a concrete problem: allocating 50 MW of data-center load across six candidate buses in two synthetic networks, under a fixed cost per active site, evaluated over 36 operating states. Judged against exhaustive enumeration of all site combinations, gradient optimization recovers the continuous allocations almost exactly, with worst-case objective gaps of 2.3\% and 8.5\% of the cost difference between the best and worst single site. Its one systematic error is instructive: near the costs at which a site should close, the smooth relaxation of the discrete site count shrinks the site rather than closing it, so discrete transitions arrive late. Differentiable market clearing thus turns market-aware planning into a problem gradients can search. --- ### Learning to Attract and Repel: Dual Quality Margin Learning for Face Recognition (DQM-Face) Authors: El Ouanas Belabbaci, Bhavesh Wani, Philipp Terhörst Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02644v1 Face recognition in unconstrained environments remains highly challenging due to diverse and extreme variations encountered in real-world scenarios. To mitigate these effects, existing margin-based approaches model sample quality through feature magnitude. However, magnitude-based modeling alone is susceptible to identity-agnostic noise, which can degrade the reliability and discriminative power of learned representations. In this paper, we propose Dual Quality Margin Learning for Face Recognition (DQM-Face), a novel framework that enables refined attraction and repulsion dynamics during representation learning. Our approach unifies conventional magnitude-based quality estimation with a newly introduced semantic quality learning mechanism, realized via squeeze-and-excitation semantic attention. By jointly leveraging magnitude and semantic cues, we construct enhanced quality-aware margins that adaptively strengthen intra-class compactness through improved attraction during learning. To further enhance inter-class discrimination, we introduce a repulsion margin formulation that explicitly enlarges inter-class separation. The unified integration of semantic quality modeling with dual attraction-repulsion margin optimization results in a more structured and discriminative feature geometry. Extensive experiments on multiple challenging benchmarks demonstrate that DQM-Face consistently outperforms state-of-the-art face recognition methods. Moreover, we show that the quality learned for margin optimization is highly effective for face image quality assessment within the proposed framework, demonstrating that the learned quality signal is intrinsically aligned with the recognition objective. The code is publicly available: https://github.com/RAIB-group/DQM-Face --- ### From Detection to Localization: A Unified Forensics Framework for Fully Synthetic and Tampered Images Authors: Annalisa Gallina, Marco Fiorucci, Marco Brigo, Federica Battisti, Lamberto Ballan Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02640v1 The rapid advancement of generative models has significantly worsened the problem of manipulated image detection, as these methods are capable of producing highly realistic forgeries, reinforcing the importance of multimedia forensics. Conventional approaches typically frame image manipulation detection as a binary classification task (real vs. generated), which limits the capability to distinguish and localize different forms of manipulation. To address these constraints, this work extends an existing detector by introducing a unified multiclass framework (real vs. fully generated vs. tampered). In addition to classifying image authenticity, the framework incorporates a segmentation branch to enable pixel-level localization of tampered regions. The proposed approach outperforms selected recent benchmarks, offering an efficient solution with improved classification accuracy and higher IoU scores for the localization task. Find the code at https://github.com/anngal01/From-Detection-to-Localization-A-Unified-Forensics-Framework-for-Fully-Synthetic-and-Tampered-Images. --- ### TaRA: Training-Aware Low-Rank Adaptation Initialization Authors: Taehyeon Kim, Eunhyeok Park Published: 2026-09-02 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.02639v1 Low-Rank Adaptation (LoRA) has become a de facto standard for parameter-efficient fine-tuning (PEFT), yet its performance is highly sensitive to initialization due to the information bottleneck imposed by low-rank decomposition. Existing approaches attempt to construct high-quality LoRA initializations by exploiting principal components of pretrained weights, activations, or gradients. However, these methods do not directly account for the training dynamics of the full-rank model. In this paper, we propose Training-aware Low-Rank Adaptation Initialization (TaRA), a method that initializes LoRA such that the gradients induced by the low-rank factors closely approximate the gradient of the corresponding full-rank weight matrix. Derived from a mathematical formulation, TaRA improves gradient fidelity at the start of training while introducing negligible computational overhead. Across diverse and challenging fine-tuning tasks, TaRA consistently outperforms prior state-of-the-art methods, establishing a simple, robust, and scalable solution for effective LoRA initialization. --- ### Oracle, will I ever learn? A study of prediction convergence and complementarity across link prediction models Authors: Guillaume Méroué, Fabien Gandon, Pierre Monnin Published: 2026-09-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.02638v1 Knowledge graphs have become an important source of structured knowledge for Web applications, including search, question answering, and recommender systems. In these applications, link prediction can serve either as a prediction task itself or as a means to enrich incomplete knowledge graphs for downstream tasks. Interestingly, different link prediction models, or even different training runs of the same model, can produce substantially different predictions for the same query. This suggests a variability in the capture of the underlying knowledge by models, thus raising a fundamental question: to what extent do different models capture complementary knowledge, and how much of this knowledge could be recovered by combining them? We propose to measure model complementarity through the performance of an oracle that, for each query, selects the best prediction among a considered set of models, hence providing an upper bound on the performance achievable through model combination. Across several architectures and benchmarks, we find a substantial gap between individual models and their oracle, revealing that different models capture complementary knowledge. Yet, this complementarity rapidly saturates as more models are added, leaving a persistent subset of queries unsolved even by a large number of models. These findings reveal both the potential of model complementarity and a fundamental limit to what current link prediction models can collectively recover; thereby highlighting the need for further research to build robust Web applications. --- ### Automated Vulnerability Injection in Smart Contracts Using Large Language Models Authors: Luca Migliaccio, Roberto Natella, Naghmeh Ivaki, Nuno Laranjeiro, Marco Vieira Published: 2026-09-02 Categories: cs.SE, cs.AI, cs.CR arXiv: https://arxiv.org/abs/2609.02624v1 Assessing vulnerability detection tools for smart contracts requires datasets with known ground truth, yet such datasets are scarce and difficult to build by hand. We propose an approach that uses Large Language Models (LLMs) to automatically inject vulnerabilities into Solidity smart contracts, and demonstrate it in a case study targeting 49 vulnerability types from OpenSCV. Injected contracts are validated through a multi-step pipeline checking compilation, execution, business logic, and the presence of the intended vulnerability. Applied to real-world contracts from SmartBugs, LLMs generate nearly 1,000 candidate variants; after deduplication and validation, 32 confirmed vulnerable contracts spanning 25 vulnerability types survive (a 16.58% survival rate). Surviving contracts concentrate in structurally simpler targets and vulnerability types with localized syntactic patterns. We report practical challenges including LLMs' non-determinism and the difficulty of preserving contract semantics. We then use the validated contracts to assess three static analyzers, revealing complementary and incomplete coverage profiles. Results show that LLM-based vulnerability injection is feasible, while exposing key limitations in scalability and diversity. --- ### Scalable Direction-Following TTS via Voice Impression-Guided Pseudo Triplet Construction Authors: Kenichi Fujita, Yusuke Ijima Published: 2026-09-02 Categories: cs.SD, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.02623v1 Voice actors often re-read the same script while modifying their delivery in response to performance directions. We study this setting as direction-following TTS, where a system generates a new utterance that reflects a given direction relative to a reference utterance while preserving speaker identity and linguistic content. A key challenge is the lack of training data capturing such relative modifications. To address this, we propose a scalable pseudo-triplet construction pipeline that generates~(reference utterance, direction text, modified utterance) triplets. It generates controlled style variations using an impression-controllable TTS model and uses an LLM to produce natural language directions from estimated impression differences. Experimental results demonstrate that pseudo-triplets alone enable stable speaker-preserving modification, and that combining pseudo and recorded data further improves direction alignment while maintaining speaker similarity. Audio examples are available on our demo page https://ntt-hilab-gensp.github.io/IS2026pseudo/ --- ### Source Distribution Estimation by Posterior Averaging Authors: Trung-Dung Hoang, Lisa M. Koch Published: 2026-09-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.02622v1 Simulation-based science often requires a distribution over simulator parameters whose push-forward reproduces a set of real observations: this is the source distribution estimation (SDE) problem. Existing methods fit the source against a likelihood surrogate trained once from a fixed proposal prior. Their objective is therefore stated only in terms of the surrogate instead of the true simulator, which may fail for inaccurate areas in parameter space where the surrogate was never trained. We instead solve SDE by expectation maximization: an E-step trains an amortized posterior on fresh simulations from the current source estimate, and an M-step refits the source to the average of that posterior over the observed data. We give two parameterizations, (1) separate source and posterior flows and (2) a single shared conditional flow. We evaluate our method on three benchmark tasks under both broad and misspecified initial priors. Both improve on existing fixed surrogate approaches and on iterated variants of each, most clearly on Lotka--Volterra, where no baseline falls below 0.96 data-space C2ST while our methods reach 0.64-0.68 in three of four initial-prior settings. --- ### Collective creativity in hybrid societies Authors: Mason Youngblood, Katie Mudd, Manuel Anglada-Tort, Cameron Jones, Elena Miu, Diana Omigie, Margaret Schedel Published: 2026-09-02 Categories: cs.AI, cs.CY, cs.MA arXiv: https://arxiv.org/abs/2609.02620v1 Generative AI is changing how cultural artifacts are created and circulated, and with it our understanding of creativity itself. Researchers disagree about whether these tools enrich or impoverish culture, and we argue that much of that disagreement comes from conflating two distinct components of creativity: novelty, a property of single artifacts, and diversity, a property of populations. We argue further that creativity in the context of generative AI is best understood as a property of hybrid collectives, or populations of interacting people and algorithms, rather than of individuals. AI-assisted ideation reliably raises the novelty of individual output while narrowing diversity in the aggregate, but this is not an inevitable consequence of putting machines in the loop. Because humans and models search in complementary ways, mixed groups can outperform and out-diversify groups of either kind alone, and machine-discovered solutions can enter human culture and persist there. What decides the outcome is composition: which agents are present, in what proportion, and how they are connected. The question is no longer whether AI helps or harms creativity, but which mixtures let individual gains accumulate without eroding collective diversity. --- ### AffectDelta: Beyond Emotion Labels for Image Editing Authors: Xingzu Zhan, Lin Gu, Ruogu Fang Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02616v1 Emotion-driven image editing aims to evoke a specified target emotion by modifying emotion-relevant visual cues in a source image, while preserving the overall composition and semantic-structural coherence of the original scene. Existing scene-level editors typically specify the target with a single emotion category and often learn visual transformations from operation-level text instructions. A category collapses a mixed affective endpoint into one dominant label, while language cannot precisely quantify how coexisting emotions should increase, decrease, or remain stable. We introduce AffectDelta, a source-aware editor that treats editing as a transition between eight-dimensional emotion distributions. A frozen Emotion Distribution Predictor estimates the source state, and the signed source-to-target difference encodes the direction and magnitude of the requested transition. Within AffectDelta, an internal transition encoder and a source-aware diffusion backbone jointly translate this signal into context-dependent semantic and appearance changes. To train this formulation, we construct AffectPair-249K, comprising 248,841 source-target pairs with predicted eight-dimensional distributions and spanning both cross-category and within-category transitions. Experiments against six baselines, combining quantitative evaluation with qualitative comparisons, demonstrate improved affective alignment and content preservation, while ablations validate our design choices. Code and dataset will be made publicly available upon acceptance. --- ### Predictors of Loneliness in Older Adults Using Multimodal Analysis of Speech and Language Authors: Vinmay Khandode, Sai Karthik Kosuri, Neil K. R. Sehgal, Adam Greene, Elif Alpoge, Elana Duffy, Matthew Lee Smith, Thomas K. M. Cudjoe, Sharath Chandra Guntuku Published: 2026-09-02 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.02606v1 Loneliness is a critical public health issue among older adults, linked to higher risks of depression, cognitive decline, and mortality. Scalable, objective methods for its detection remain limited, particularly in natural conversational contexts. We analyzed speech and language markers of loneliness in 310 older adults using semi-structured telephone interviews to help understand how they process feeling lonely and how their language differs at different levels of feeling loneliness. Our multimodal framework combined linguistic features (psycholinguistic dictionaries, n-grams, and topic models) with acoustic features (pitch, tone, loudness) to examine associations with self-reported loneliness scores. Both predefined and data-driven methods captured patterns in verbal content and vocal delivery. Higher loneliness was associated with negations(r = 0.11), negative tone(r = 0.12), and conflict-related language. Lower loneliness was linked to social references(r = -0.18), motivational drives(r = -0.11), and emotional richness in speech(r = -0.12). We also found that the multimodal model (r = 0.298) outperforms the text-only and audio-only models. Findings suggest that loneliness manifests through both linguistic and acoustic cues, supporting the potential of speech-based analysis in psychological assessments and as an early indicator of emotional loneliness when used alongside existing assessments, rather than as standalone diagnostic tools. --- ### Generalizable Brain Tumor Segmentation with Self-Training and Tumor-Aware Deformations Authors: Henrique Zan Grande, Jeovane Honorio Alves, Rayson Laroca, Andre Gustavo Hochuli Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02600v1 This work presents an approach to the Generalizability Across Tumors (BraTS-GoAT) task of the BraTS 2026 Challenge, which focuses on robust segmentation of brain tumor sub-regions across a heterogeneous patient population. The proposed method employs the nnU-Net framework with a large residual encoder architecture, integrating a semi-supervised learning technique with pseudo-labels generated from the unlabeled training data and a tumor-aware deformable augmentation that locally deforms the lesion while preserving the surrounding anatomy. We evaluate the individual contributions of each component, as well as their combination, using varying proportions of the most confident pseudo-labeled cases. The submitted configuration for the generalization task achieves Dice and NSD scores of 0.881 and 0.473 for Whole Tumor, 0.817 and 0.490 for Tumor Core, and 0.775 and 0.533 for Enhancing Tumor on the BraTS-GoAT validation set, improving over the labeled-only baselines across all tumor regions and confirming that self-training and the proposed augmentation are complementary. Our source code is publicly available at https://github.com/Henrique-zan/brats-goat-2026/. --- ### Competitive Market Behavior of LLMs Authors: Pawel Struski, Jakub Swistak, Inez Okulska, Przemyslaw Biecek Published: 2026-09-02 Categories: cs.MA, cs.AI, econ.GN arXiv: https://arxiv.org/abs/2609.02580v1 Large language models (LLMs) are increasingly deployed as economic agents, yet there is little evidence whether LLM agents are suited for participating in market mechanisms designed for humans, and whether these mechanisms deliver desired outcomes when faced with LLM agents. We address this question by replicating seminal economic experiments, replacing human subjects with LLM agents. We place agents in a double auction environment, which is a widely-used market mechanism. We check whether such a market is able to deliver an efficient allocation of resources, thereby testing a novel dimension of alignment of LLM agents -- their compatibility with a fundamental market mechanism. We find that markets populated by LLM agents exhibit slower or no convergence towards market equilibrium, thus providing less efficient allocations than markets populated by humans. We then analyze agents' individual trading decisions and find substantial heterogeneity both across model families and market roles. We also run a lexical analysis of Chain-of-Thought (CoT) traces generated by the agents. We find that the decision to execute a trade rather than continue incrementally adjusting prices is associated with a shift from strategic considerations toward urgency. We publicly release our testing framework, which can be used for future evaluations. --- ### Deeply Interleaved Text-Image Contexts for Multimodal LLMs Assessment Authors: Zihao Wang, Xi Xiang, Yuwen Sun, Yingyu Li, Yabo Zhang, Yihan Zeng, Fan Li, Wangmeng Zuo Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02573v2 Current evaluations and training of multimodal models predominantly focus on multi-image tasks, largely overlooking interleaved text-image scenarios. In such multi-image tasks, text typically serves merely as task instructions, lacking deep semantic interaction with the visual content. In contrast, realworld applications like text-image co-creation, character tracking, and spatial reconstruction require constant interaction between text and images. Consequently, models must possess a deep understanding of these interleaved contexts. To bridge this gap, we introduce a novel benchmark, TIC-Bench (deeply interleaved Text-Image Contexts), designed to evaluate the capability of models to integrate text-image clues and recover the ground truth facts within deeply interleaved contexts. This benchmark encompasses three core domains: Logical, Temporal, and Spatial Association, which are further categorized into eight specific types, comprising a total of 2,280 questions. We evaluated 10 state-of-the-art MLLMs and observed a substantial performance gap compared to human experts, together with persistent difficulties in integrating evidence distributed across interleaved visual and textual inputs. Ultimately, this benchmark provides a valuable analytical tool for assessing and advancing the ability of multimodal models to effectively integrate text and image information in deeply interleaved contexts. TIC-Bench is publicly available at https://huggingface.co/datasets/pino10010/TIC-Bench --- ### Deeply Interleaved Text-Image Contexts for Multimodal LLMs Assessment Authors: Zihao Wang, Xi Xiang, Yuwen Sun, Yingyu Li, Yabo Zhang, Yihan Zeng, Fan Li, Wangmeng Zuo Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02573v1 Current evaluations and training of multimodal models predominantly focus on multi-image tasks, largely overlooking interleaved text-image scenarios. In such multi-image tasks, text typically serves merely as task instructions, lacking deep semantic interaction with the visual content. In contrast, realworld applications like text-image co-creation, character tracking, and spatial reconstruction require constant interaction between text and images. Consequently, models must possess a deep understanding of these interleaved contexts. To bridge this gap, we introduce a novel benchmark, TIC-Bench (deeply interleaved Text-Image Contexts), designed to evaluate the capability of models to integrate text-image clues and recover the ground truth facts within deeply interleaved contexts. This benchmark encompasses three core domains: Logical, Temporal, and Spatial Association, which are further categorized into eight specific types, comprising a total of 2,280 questions. We evaluated 10 state-of-the-art MLLMs and observed a substantial performance gap compared to human experts, together with persistent difficulties in integrating evidence distributed across interleaved visual and textual inputs. Ultimately, this benchmark provides a valuable analytical tool for assessing and advancing the ability of multimodal models to effectively integrate text and image information in deeply interleaved contexts. TIC-Bench is publicly available at https://huggingface.co/datasets/pino10010/TIC-Bench --- ### Learning-Based Reconstruction Attacks on Coordinate-Obfuscated Point Clouds Authors: Mohammad Waquas Usmani, Susmit Shannigrahi, Michael Zink Published: 2026-09-02 Categories: cs.CR, cs.LG arXiv: https://arxiv.org/abs/2609.02568v1 Volumetric video based on point cloud representations enables immersive virtual and augmented reality applications but introduces significant challenges for efficient and secure content delivery. Prior work proposed a selective coordinate encryption framework for point clouds that encrypts only a subset of coordinates, reducing computational costs while visually degrading unauthorized content. However, it remains unclear whether the remaining unencrypted information is sufficient to enable content reconstruction. In this paper, we evaluate the robustness of selective coordinate encryption against machine learning-based reconstruction attacks. We consider an attacker with access to selectively encrypted point clouds attempting to recover encrypted coordinates without decryption by exploiting spatial and geometric correlations in the unencrypted data. We evaluate PointNet and Random Forest models under two encryption granularities: \texttt{X}, where all $X$ coordinates are encrypted, and \texttt{2X}, where every second $X$ coordinate is encrypted. Our results show that reconstructing fully encrypted $X$ coordinates remains challenging, whereas the \texttt{2X} scheme leaks sufficient information through neighboring coordinates to enable accurate reconstruction. These findings demonstrate that the security of selective coordinate encryption depends strongly on encryption granularity. --- ### Online Reinforcement Learning in the Met Office Unified Model through Distributed Model-Agent Coupling Authors: Pritthijit Nath, Sebastian Schemm, Peter Haynes, Emily Shuckburgh, Mark Webb Published: 2026-09-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.02566v2 Machine-learnt corrections can complement numerical weather prediction provided that they operate stably within an evolving numerical model. In this study, we couple the Met Office (UKMO) Unified Model (UM) with distributed reinforcement-learning agents through rank-local tensors. A column-aware deep deterministic policy gradient (DDPG) actor uses local vertical structure together with full-column context to apply bounded corrections to potential temperature and horizontal wind. During training, we perform ten nudged 6-hr 12-min forecasts, with nudging towards the UKMO operational analysis providing an immediate counterfactual target from which the policy learns. The resulting actor is then frozen and applied to a non-nudged forecast without access to analysis inputs or further weight updates. Relative to a matched non-nudged native forecast at +6 h, the corrected forecast reduces global latitude-weighted MAE by 2.85% for $Z_{500}$, 2.16% for MSLP, 5.16% for $T_{500}$ and 2.27% for $T_{1.5\textrm{m}}$, with an observed 3.57% wall-time overhead compared to native execution. Even though training and inference share the same initialisation, this single-case experiment demonstrates significant promise and feasibility, laying the groundwork for RL-based bias correction and parametrisations within operational systems. --- ### Online Reinforcement Learning in the Met Office Unified Model through Distributed Model-Agent Coupling Authors: Pritthijit Nath, Sebastian Schemm, Peter Haynes, Emily Shuckburgh, Mark Webb Published: 2026-09-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.02566v1 Machine-learnt corrections can complement numerical weather prediction only if they adapt to the evolving model state while preserving dynamical consistency and numerical stability. To test this within a global forecasting model, we couple the Met Office (UKMO) Unified Model (UM) with distributed RL agents through rank-local tensors. A DDPG actor shares weights across the 70 vertical model levels of each atmospheric column and applies bounded potential-temperature corrections to the model tendencies. Across ten nudged training forecasts, nudging calculations towards the UKMO operational analysis provides an immediate counterfactual target. The frozen policy is then evaluated in a non-nudged forecast for inference. The coupled workflow successfully completes training and remains numerically stable in the evaluated case. Relative to a matched native UM forecast at +6 h, the learnt policy reduces Z$_{500}$ MAE in four of six latitude bands, including reductions of 45.8% and 40.8% in the northern and southern tropics. MSLP error too decreases in three bands, with a maximum reduction of 27.3% at 0-30°N. This single-case experiment demonstrates significant promise and feasibility of distributed online learning followed by non-nudged inference, laying the groundwork for RL-based bias correction and parametrisations within operational systems. --- ### MARS: What Retrieval Signals Are Hidden in Multimodal Large Language Models for Text-Video Retrieval? Authors: Uicheol Jung, Juyoung Hong, Geuntaek Lim, Yukyung Choi Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02565v1 Text-video retrieval requires representations that can distinguish videos with similar scenes, actions, and temporal patterns. Recent multimodal large language models have been adapted as embedding models, but they often represent each input using a single token from the final layer. This can compress diverse video-text cues into a single vector and limit fine-grained retrieval. To address this limitation, we propose MARS, a multi-layer and multi-slot embedding framework for text-video retrieval. MARS constructs multiple adaptive representation slots by combining hidden states from different decoder layers, compares corresponding text and video slots, and aggregates their similarities for retrieval. To better handle confusing candidates, we further introduce a hard-negative-aware slot specialization objective that encourages the slots to capture discriminative matching cues. Experiments on four text-video retrieval benchmarks show that MARS achieves state-of-the-art results in both direct similarity-based retrieval and reranking settings. Ablation studies and analyses demonstrate that multi-layer fusion, multiple slots, and hard-negative-aware slot specialization provide complementary gains. Code is available at https://github.com/sejong-rcv/MARS. --- ### Stereo 4D Radar for 3D Object Detection: Integrating Geometric Alignment and Absolute Velocity Estimation Authors: Seung-Hyun Song, Dong-Hee Paek, Woong-Chan Byun, Seung-Hyun Kong Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02560v1 Four-dimensional (4D) Radar is a powerful sensing modality capable of detecting surrounding three-dimensional (3D) objects under diverse weather conditions and providing Doppler-based motion information. However, raw 4D Radar signals contain significant clutter from road surfaces, guardrails, and surrounding vehicles, along with multipath-induced ghost reflections and the receiver's inherent noise floor. Consequently, preprocessing algorithms designed to remove such invalid measurements often make the Radar data excessively sparse. Moreover, the Doppler measurements provided by 4D Radar describe only the radial component of an object's velocity, limiting their ability to recover the full motion state. In this paper, we introduce a stereo 4D Radar-based 3D object detection framework that exploits the geometric disparity between left and right Radars to estimate the absolute velocity of objects and achieve more robust perception through the fusion of their complementary features. The effectiveness of the proposed framework is validated on our in-house stereo 4D Radar dataset, demonstrating performance gains of 8.82 points in AP 3D and 9.0 points in AP BEV over state-of-the-art mono 4D Radar baselines. These results demonstrate that absolute velocity estimation combined with stereo geometry-aware feature fusion leads to substantial improvements in 3D object detection. --- ### RGB-to-IR image translation for infrared vehicle detection in unseen UAV domains Authors: Thijs A. Eker, Ella P. Fokkinga, Jan Erik van Woerden, Elfi I. S. Hofmeijer, Sebastiaan P. Snel, Klamer Schutte, Friso G. Heslinga Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02556v1 Synthetic training data is crucial for developing vision AI when real-world data is scarce, as in thermal infrared (IR) aerial vehicle detection. While abundant UAV RGB imagery motivates RGB-to-IR translation for data augmentation, unobservable thermal traits (e.g., engine heat) make learning transferable mappings challenging. This work investigates whether modern generative translators can overcome this cross-modal gap to improve infrared vehicle detection on unseen UAV target domains. Translators are trained on paired RGB-IR source datasets and applied to RGB training images from held-out target datasets to generate synthetic IR data. Evaluated methods include supervised GANs, ControlNet-based diffusion models, and foundation-model editing via LoRA. The resulting synthetic IR imagery is used to train RF-DETR vehicle detectors, which are evaluated on unseen IR target test splits across five aerial datasets, with Kust4K and VTUAV serving as target domains. Synthetic IR consistently outperforms RGB and grayscale baselines. Stable Diffusion 3.5 with ControlNet yields the best results, improving mAP from 50.8 to 60.1 on Kust4K and from 25.6 to 38.4 on VTUAV compared to models trained only on source-domain IR data. Increasing output diversity via multiple seeds (+1.1 mAP) and prompt variations (+3.3 mAP) provides additional gains on VTUAV. Although a performance gap to real target IR data remains, generative RGB-to-IR translation effectively mitigates IR data scarcity and improves cross-domain aerial vehicle detection. --- ### ProbeMatchDTI: Probe-Driven Multi-Scale Biochemical Pattern Matching for Drug-Target Interaction Prediction Authors: Quan Hao, Mengyue Fan, Zifan Dong, Youru Li, Jianduo Zhao, Lechuan Xu, Hao Zhang, Fei Xia, Jigang Wang, Chong Qiu, Liguo Zhang Published: 2026-09-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.02549v1 Drug-target interaction (DTI) prediction is an important task in AI-driven drug discovery. Although recent biochemical representation learning methods have improved DTI prediction, their passive feature aggregation tends to favor dominant molecular patterns while suppressing weak yet binding-relevant signals, such as functional groups and residue-context patterns, limiting the modeling of multi-scale biochemical correspondences. To address this issue, we propose ProbeMatchDTI, a pattern-probe-driven framework comprising IterProbe and BindingProbe. IterProbe explicitly retains contextual states across refinement depths and uses learnable probes to select them at each position before cross-entity matching, thereby preserving weak biochemical patterns and strengthening associations among functional groups, local motifs, and molecular scaffolds. BindingProbe then characterizes cross-entity drug-protein complementarity at local biochemical-unit and whole-pair levels, jointly modeling fine-grained interactions and multi-scale correspondences while preserving weaker binding-relevant associations. Extensive experiments demonstrate the superiority of ProbeMatchDTI, achieving 2.0% and 0.5% higher AUC-ROC on BindingDB and DrugBank, respectively. Feature-level pattern analyses further characterize its probe-driven behavior in cross-scale biochemical pattern matching. We further connect ProbeMatchDTI predictions with an evidence-guided downstream drug-discovery workflow, demonstrating their utility for candidate refinement and validation planning. Our code is available at https://github.com/developer-hq/ProbeMatchDTI --- ### Learn from Whoever Is Right: Answer-Verified Multi-Teacher Distillation for Multi-Domain LLMs Authors: Xixiang He, Xingming Li, Baiqi Wu, Qiyao Sun, Xuanyu Ji, Ao Cheng, Qingyong Hu Published: 2026-09-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.02548v1 Modern large language models (LLMs) rely on reinforcement learning to build strong capabilities in individual domains, but integrating those capabilities into a single deployable model remains challenging. By routing each sample to the teacher whose domain matches it, existing approaches let a domain label decide which teacher provides supervision. However, domain expertise holds only on average: the matched teacher is not always correct on a given sample, while a teacher from another domain sometimes is. The reliable teacher therefore has to be identified per sample, not per domain. In this paper, we introduce Multi-Teacher Self-Distillation Policy Optimization (MT-SDPO), an on-policy distillation method that unifies several frozen teachers into one student model. MT-SDPO consists of three components: (1) self-anchors, where a rollout is supervised by a correct rollout from its own group; (2) answer-verified eligibility, where a teacher may supervise a sample only if its own answer passes a verifier; and (3) privileged distillation, which merges the anchor and all verified feedback into one context that an exponential moving average self-teacher reads and the student does not, thereby keeping one policy at deployment. Across five students from three model families, MT-SDPO lifts the weakest domain of Qwen3-8B by 14.79 points and narrows its domain gap by 74.7%, a better balance than serving one matched teacher per domain. Verified reliability, not domain membership, should decide who teaches. Code is available at https://github.com/hexixiang/MT-SDPO. --- ### TrajMind: Chaining Role-Specialized LoRAs for Fast-and-Slow Collective Trajectory Anomaly Diagnosis Authors: Jiahao Wu, Zhenqun Yang, Chen Jason Zhang, Qing Li Published: 2026-09-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.02540v2 Diagnosing collective anomalies from urban trajectories is increasingly important for traffic governance, as it reveals what happened, who was involved, and where and when the event occurred. Existing detectors efficiently produce scores or labels, whereas vision--language pipelines provide richer semantics; neither couples verifiable diagnosis with low-latency monitoring. The central challenge is to recognize collective patterns and recover exact event details from the source trajectories without running the full diagnostic pipeline for every monitored window. We therefore separate always-on screening from on-demand diagnosis: screening raises alerts, while diagnosis releases only source-verified what--who--where--when records. We present TrajMind, a fast-and-slow framework that switches three role-specialized LoRA adapters over one frozen vision--language backbone. Its slow path, \textit{TrajMind$_{\text{slow}}$}, chains canvas-based typing, type-conditioned localization over serialized trajectories, and executable verification, yielding structured, evidence-backed diagnoses. Additionally, the fast path, \textit{TrajMind$_{\text{fast}}$}, screens each window in a single text-only pass, delivering efficient structured alerts. Extensive experiments show that, TrajMind$_{\mathrm{slow}}$ outperforms the strongest baselines by at least $15.3$ percentage points in anomaly typing and $13.8$ percentage points in localization. These gains persist under cross-city transfer, and TrajMind$_{\mathrm{fast}}$ reduces latency by $41.1\%$ and maintains binary balanced accuracy of at least $93.5\%$. Together, TrajMind delivers accurate, evidence-backed diagnoses across cities and efficient front-line monitoring. --- ### TrajMind: Chaining Role-Specialized LoRAs for Fast-and-Slow Collective Trajectory Anomaly Diagnosis Authors: Jiahao Wu, Zhenqun Yang, Chen Jason Zhang, Qing Li Published: 2026-09-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.02540v1 Diagnosing collective anomalies from urban trajectories is increasingly important for traffic governance, as it reveals what happened, who was involved, and where and when the event occurred. Existing detectors efficiently produce scores or labels, whereas vision--language pipelines provide richer semantics; neither couples verifiable diagnosis with low-latency monitoring. The central challenge is to recognize collective patterns and recover exact event details from the source trajectories without running the full diagnostic pipeline for every monitored window. We therefore separate always-on screening from on-demand diagnosis: screening raises alerts, while diagnosis releases only source-verified what--who--where--when records. We present TrajMind, a fast-and-slow framework that switches three role-specialized LoRA adapters over one frozen vision--language backbone. Its slow path, \textit{TrajMind$_{\text{slow}}$}, chains canvas-based typing, type-conditioned localization over serialized trajectories, and executable verification, yielding structured, evidence-backed diagnoses. Additionally, the fast path, \textit{TrajMind$_{\text{fast}}$}, screens each window in a single text-only pass, delivering efficient structured alerts. Extensive experiments show that, TrajMind$_{\mathrm{slow}}$ outperforms the strongest baselines by at least $15.3$ percentage points in anomaly typing and $13.8$ percentage points in localization. These gains persist under cross-city transfer, and TrajMind$_{\mathrm{fast}}$ reduces latency by $41.1\%$ and maintains binary balanced accuracy of at least $93.5\%$. Together, TrajMind delivers accurate, evidence-backed diagnoses across cities and efficient front-line monitoring. --- ### A Comparative Study of Graph Representations for GNN-Based Power Grid Control in L2RPN Authors: Adrian Degenkolb, Qiong Huang, Benjamin Schäfer Published: 2026-09-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.02538v1 Graph construction is a critical but underexamined design choice in deep reinforcement learning for power grid control. We present a controlled experimental comparison of different graph representations, including physical topology, electrical-sensitivity, and hybrid variants for topology control in the Learning to Run a Power Network (L2RPN) environment. Our findings indicate that matching graph complexity to task granularity is more important than maximizing representational richness, and highlight the importance of controlled representation studies at scale. --- ### Spatially Aware World Action Model via Geometric Latent Diffusion Authors: Javier Alejandro Lopetegui Gonzalez, Paul Pacaud, Cordelia Schmid Published: 2026-09-02 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2609.02531v1 World Action Models (WAMs) leverage the capabilities of large-scale pretrained video diffusion models to jointly predict future observations and actions, inheriting rich visual and physical priors from internet-scale video. This has made them a promising paradigm for robot policy learning, yet the prevailing models operate exclusively on RGB observations and do not leverage 3D information. To bridge this gap, we introduce a Spatially Aware World Action Model (SA-WAM), which repurposes a pretrained video model for joint action, RGB, and depth prediction, enabling 3D-aware world modeling and action prediction within a single diffusion backbone. We use a nonlinear encoding that maps the unbounded depth signal into the bounded input domain expected by the frozen VAE tokenizer. This allows us to reuse the tokenizer without 3D-specific fine-tuning, incorporating geometric information without sacrificing the pretrained priors. SA-WAM achieves state-of-the-art results on the RoboCasa and LIBERO-Plus benchmarks, while simultaneously improving future-state predictions. Furthermore, SA-WAM outperforms strong baselines in real-world evaluation using a UR5 robotic arm, with strong gains in randomized environments. We analyze the correlation between world model prediction quality and rollout success, providing insights into WAM performance and avenues for its improvement. --- ### Fine-Grained Anomaly Perception in Wild UGC-Enhanced Images: A Comprehensive Dataset and Difference-Fusion Framework Authors: Yan Zhong, Gefei Chen, Qiufang Ma, Zhen Wang, Zhiwei Fan, Lei Shi, Tingting Jiang Published: 2026-09-02 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.02529v1 Image enhancement and restoration have become standard back-end operations on short-video and social media platforms to boost UGC visual experience. Yet these processes inevitably introduce visual anomalies--especially in faces, texts, and textures--that directly undermine perceptual fidelity and viewer trust. While existing IQA methods perform well on classic distortions, they target holistic quality assessment and fail to capture the specific, localized anomalies caused by enhancement algorithms in real-world UGC. To bridge this gap, we formally define a new task-quality Anomaly Perception for UGC image Enhancement (UEAP), and contribute the first UEAP benchmark dataset, named UEAP-4k, curated from the real business scenarios. It provides fine-grained annotations for anomaly categories, localization and severity levels. Furthermore, we propose a Difference-Fusion Anomaly Perception Method (DFAP-UGC) for wild UGC-enhanced images, which leverages explicit problem-reference difference fusion with dense spatial querying, regional verification, and quality-aware ranking, enabling robust anomaly identification in challenging scenarios. To handle the inherent coupling of subtasks in this new task, we propose a Locality-Aware Dynamic Task Prioritization (LADTP) training strategy that enables effective end-to-end learning and eliminates multi-stage overhead. Extensive experiments show that our method outperforms baselines adapted from classical approaches for this task, validating the value of this dataset and the superior of DFAP-UGC for robust UGC-enhanced image anomaly perception. Code and data will be public. --- ### Doppio: A Dataset for Contactless Weight Estimation of Falling Particles Authors: Simon Kiefhaber, Jan-Martin O. Steitz, Julia Grabinski, Christoph Reich, Paul Wagner, Max Zimmermann, Simone Schaub-Meyer, Stefan Roth Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02528v1 Measuring the mass of powder, including falling particles, is a common task in industrial applications. While scales are effective for static measurements, many applications require contactless sensing, where existing solutions are often costly, application-specific, and technically complex. In this work, we investigate computer vision as a practical alternative for contactless mass estimation. As an accessible real-world case study, we focus on coffee grinding and introduce \emph{Doppio}, a novel video dataset capturing videos of falling ground coffee, paired with precise, per-frame ground-truth weight measurements. To demonstrate contactless measuring, we evaluate deep learning-based approaches ranging from purely spatial feed-forward networks to recurrent spatio-temporal models. These models are analyzed with respect to their predictive accuracy and computational trade-offs. We demonstrate that deep learning-based computer vision models accurately estimate the cumulative weight of falling particles, establishing a solid foundation for future vision-based contactless measurement solutions. --- ### When Persona Attributes Improve Population Alignment in Large Language Models Authors: Leon Fröhling, Jens Rupprecht, Markus Strohmaier, Claudia Wagner Published: 2026-09-02 Categories: cs.CL, cs.CY arXiv: https://arxiv.org/abs/2609.02526v1 Large Language Models (LLMs) are increasingly used to predict the responses of human participants in survey panels. Towards that goal, persona prompting has recently emerged as a technique to inform and align large pretrained language models. Persona prompting refers to the practice of using short textual descriptions of 'personas' in prompts to steer the LLM's generations. Personas describe individuals through different attributes such as their socio-demographics, attitudes, or behaviors, with the aim of aligning LLMs to produce responses that correlate with the corresponding human responses. Yet, recent work has produced mixed and partly conflicting results of persona prompting without clear patterns of success and failure. Among the few consistent findings is that the selection of persona attributes matters, and that using more attributes does not necessarily lead to better performance. It remains unclear how different attribute selection methods perform and how to choose among them. In this paper, we propose that observed human response variation of a survey question is a potential explanation for the mixed performance observed so far. In addition, we compare the performance of persona prompting associated with different methods for selecting persona attributes. We evaluate these methods on four different (general) social surveys across two countries, six LLMs, and twenty prediction tasks per survey. Our work helps to identify when persona prompting can be expected to be useful in survey prediction tasks, and provides new insights on the effectiveness of different attribute selection methods for LLM-based survey prediction using persona prompting. --- ### Spectral Initialization and Scheduled Graph Smoothness for Uncertain Knowledge Graph Completion Authors: Md Abrar Jahin, Taufikur Rahman Fuad, Jay Pujara, Craig A. Knoblock Published: 2026-09-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.02519v1 Uncertain knowledge graphs (UKGs) extend knowledge graphs by assigning each triple a continuous confidence score. Since most possible triples lack observed confidences, recent methods rely on semi-supervised learning to generate pseudo-labels. These methods initialize entity embeddings without using the confidence-weighted graph, discarding its global community and hub structure. We introduce QUEST, which adds no trainable parameters to the standard confidence-distribution learning pipeline. First, QUEST initializes entity embeddings using the smallest non-trivial eigenvectors of the confidence-weighted graph Laplacian, incorporating community and hub structure before training. Second, QUEST applies an unbiased mini-batch Dirichlet energy regularizer to enforce early-stage structural consistency. On two UKG datasets, QUEST improves confidence prediction and link prediction on six of eight metric-dataset pairs over prior methods and matches the previous best on the remaining two, while removing the instability spike observed on dense graphs. These results indicate that spectral structural priors combined with a graph Dirichlet energy regularizer improve accuracy, training stability, and checkpoint reliability in UKG completion. --- ### Beauty is in the AI of the beholder: MLLMs systematically overrate facial attractiveness Authors: Santiago Grandas, Juan Sebastian Cely-Acosta, Mohit Mendiratta, Shafee Hassan, Macken Murphy Published: 2026-09-02 Categories: cs.CV, cs.HC arXiv: https://arxiv.org/abs/2609.02512v1 Beauty assessments from Multimodal Large Language Models (MLLMs) are increasingly popular amongst users, companies, and aestheticians. This raises the question of whether these AI models can accurately reflect human judgments of attractiveness. In a pre- registered exploratory study, we compared the attractiveness ratings of 2,513 human participants to four widely used commercial AI models: Claude, Gemini, GPT, and Grok. Results showed that MLLMs systematically rate faces more favourably and within a narrower range than humans and, at the time of study, do not reproduce human ratings in absolute terms. However, MLLMs exhibit strong correlations with human attractiveness judgments, accurately tracking the rank-ordering of faces. MLLMs may judge faces by different cues than humans; only face age was a predictor of facial attractiveness in both humans and MLLMs, with inconsistent patterns across models for ethnicity and gender. AI models strongly agree with one another, except for Grok, which also showed the lowest agreement with humans. Our findings suggest that while they may be able to approximate rank-orderings of human attractiveness, current off-the-shelf commercial MLLMs systematically overrate the beauty of human faces. --- ### Orthogonal Ensembles and Tested Explanations for Performer-Independent Body-Motion Emotion Recognition Authors: Naoto Nishida, Yoshio Ishiguro Published: 2026-09-02 Categories: cs.CV, cs.HC, cs.LG arXiv: https://arxiv.org/abs/2609.02510v2 We study body-only, 12-class acted-emotion classification from skeleton motion under leave-performer-out (LPO) evaluation, a hard, underdetermined setting: chance is 8.3%, and a protocol-matched reproduced STGCN++ baseline reaches only 25.73 +/- 4.03% Macro-F1. We show that reliable gains come not from a new architecture but from combining eleven models with orthogonal error modes: under 10-fold LPO cross-validation on the labeled training performers, an equal-weight logit-mean ensemble reaches 36.80 +/- 4.00% per-fold Macro-F1, a protocol-matched +11.07 pp (+43% relative) over the same-split reproduced baseline. Our central contribution is a tested explanation suite: for a strong ensemble member, part-masking and counterfactual edits show (rather than assert) that its decisions depend on motion-grounded body-region evidence, and this region saliency aligns with rule-based Laban Movement Analysis (LMA) attributes far more than with classical kinematics: region-level saliency-LMA Spearman rho = +0.500 versus +0.033, roughly 15x, and the alignment holds for the submitted 11-way ensemble itself at rho = +0.517; the audit is post hoc and needs no retraining. The same suite faithfully reports a negative: within-window temporal saliency is diffuse rather than localized. On the hidden challenge test set the submitted ensemble scored 37.23 % Macro-F1 and received the Best Performance Award of the MMAC Challenge 2026 (Human score is 39 %). Code is available at https://github.com/nawta/diema-challenge and the presentation at https://nawta.github.io/mmac2026/. --- ### Rethinking the Teacher-Student Framework for Test-Time Adaptation Authors: Damian Sójka, Marc Masana, Bartłomiej Twardowski, Sebastian Cygert Published: 2026-09-02 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.02507v1 Test-Time Adaptation (TTA) has recently emerged as a promising strategy that allows the adaptation of pre-trained models to changing data distributions at deployment time, without access to any labels. To mitigate error accumulation, researchers have widely adopted the teacher-student framework, though its long-term stability is often taken for granted. In this work, we challenge the common strategy of setting the teacher weights to an exponential moving average of the student by showing that error accumulation still occurs, although it is mostly apparent on longer sequences compared to those commonly utilized. We analyze the stability-plasticity trade-off within the teacher-student framework and propose to use an intransigent teacher that does not update its weights. Surprisingly, we show that this simple change allows TTA methods to significantly improve their performance on multiple datasets with longer scenarios and result in increased robustness to changes in hyperparameters. Finally, we show that those changes can be seamlessly and effectively applied to various architectures and experimental setups, including semantic segmentation. The code is available at https://github.com/dmn-sjk/intransigent_teacher. --- ### SR-Edit: Region-Aware Image Editing via Self-Refinement Authors: Andong Wang, Zehua Chen, Yuxuan Jiang, Jun Zhu Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02504v1 With the recent rapid progress in generative models, image editing has made remarkable advances, yet achieving faithful edits that precisely modify only the target regions while strictly preserving all other regions remains challenging. Since externally provided region annotations are often difficult to obtain in practice, a growing body of work seeks to improve preservation by automatically inferring edit and non-edit regions, and then enforcing consistency on the latter. However, these approaches still suffer from inaccurate region estimation and heuristic correction strategies that distort the native inference process, making methods designed for fidelity themselves a new source of artifacts. We propose SR-Edit, an image editing framework that overcomes these issues via iterative self-refinement. Specifically, at each iteration, SR-Edit first (i) extracts progressively precise and self-consistent region separation from the model's own predictions by lightweight post-processing, and then (ii) enforces preservation in non-edit areas through correction updates that remain aligned with the original sampling dynamics. Extensive experiments demonstrate that SR-Edit achieves superior preservation and overall image quality compared to existing editing techniques. --- ### Blending Concepts: Benchmarking Visual Metaphor Generation in Text-to-Image Models Authors: Chuer Chen, Zichen Wang, Yi He, Zhengxi Yu, Nan Cao Published: 2026-09-02 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.02502v1 Text-to-image (T2I) models have achieved remarkable success at faithfully rendering specified objects and attributes, yet their ability to produce visual metaphors, images that convey abstract ideas by combining elements from two distinct domains, remains largely unexamined. To bridge this gap, we introduce VMetaphor-Bench, the first benchmark for evaluating visual metaphor generation in T2I models. It comprises 1,500 visual metaphors curated from real-world creative imagery, organized into three levels and ten categories, with each sample paired with two prompts of differing specificity. For evaluation, we develop a hybrid framework within an MLLM-as-judge paradigm, combining a multiple-choice question (MCQ) based protocol of 9,594 questions across four levels of metaphorical fidelity with a dimension-based scoring protocol along three perceptual dimensions. Extensive evaluation of 11 representative T2I models reveals that even the strongest proprietary models struggle with compositional structuring and cross-domain mapping, key aspects of metaphorical expression, highlighting visual metaphor generation as an important frontier for future T2I research. --- ### Training seeds and model-selection stability in recommender-system evaluation Authors: Juan Manuel Rodriguez, Oleg Lesota, Antonela Tommasel Published: 2026-09-02 Categories: cs.IR, cs.LG arXiv: https://arxiv.org/abs/2609.02499v1 Recommender-system experiments often rely on a single random training seed, assuming that run-to-run stochasticity has limited impact on evaluation conclusions. This assumption is risky, as a training seed may influence several algorithm-dependent mechanisms, including parameter initialization, mini-batch ordering, dropout, masking, latent sampling, and training-time negative sampling. We examine this assumption by fixing the data partition and varying the training seed across hyperparameter configurations. We analyze seed effects at three levels: user-level metric sensitivity, validation-based model selection and recommendation-list agreement. Results show that seed variation is often detectable. Its impact depends on whether configurations are clearly separated, whether validation results transfer to test, and whether similar scores lead to similar top-$k$ lists. Findings suggest that reporting single-seed results can overstate the stability of recommender system evaluation, and that training seeds should be treated as part of the evaluation protocol rather than as incidental implementation noise. --- ### RINSE: Robust Target-Time Normality Estimation for Zero-Shot Graph Anomaly Detection Authors: Taufikur Rahman Fuad, Md Abrar Jahin, Amir Hussain Published: 2026-09-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.02497v1 Zero-shot graph anomaly detection seeks to deploy a detector trained on source graphs to unseen, unlabeled targets, yet domain shift can make source-derived notions of normality unreliable. We introduce RINSE (Robust Iterative Normality Self-Estimation), a gradient-free target-time framework that keeps the source-trained detector fixed while sequentially estimating target normality, representation calibration, and evidence reliability from the target graph. Its core idea is to identify a reliable subset of low-residual target nodes, use them to construct a trimmed target-aware normality model, and combine complementary anomaly evidence through reliability-gated rank fusion and encoder ensembling. Across eight unseen target graphs, RINSE achieves the highest average AUPRC among the evaluated methods under two separate preprocessing protocols, while block ablations and sensitivity analyses support the combined design. These results support robust target-time estimation as a practical approach to generalist graph anomaly detection without target labels, gradients, or per-target tuning. --- ### Debias-SparseGPT: Bias-Aware Pruning for Large Language Models Authors: Irina Proskurina, Guillaume Metzler, Antoine Gourru, Julien Velcin Published: 2026-09-02 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.02496v1 Model compression techniques such as pruning and quantization facilitate the efficient deployment and acceleration of Large Language Models (LLMs). However, recent studies show that weight sparsification methods, such as SparseGPT, can amplify existing biases in models, with outputs varying significantly depending on persona cues in the prompt. In this paper, we introduce Debias-SparseGPT, a post-training pruning method incorporating representational debiasing using a second-order term defined over demographically contrasting inputs. We perform empirical validation of our method over a wide range of generative LLMs. Across models and sparsity regimes (25%, 50%, and structured 2:4 sparsity), Debias-SparseGPT consistently reduces pruning-induced bias compared to SparseGPT while preserving model perplexity and zero-shot accuracy. Under the most restrictive 2:4 structured sparsity pattern, which most aggressively degrades model quality, augmenting the calibration set with long-context, content-rich examples further improves both downstream performance and fairness. Overall, Debias-SparseGPT advances the bias-performance trade-off while preserving the computational efficiency of sparse models. --- ### ViSAR: Training-Free Adaptive-$k$ Retrieval for Visual Document Question Answering Authors: Adrien Mialland, Marc Plantevit, Julien Gallois, Céline Robardet Published: 2026-09-02 Categories: cs.IR, cs.AI, cs.CL, cs.CV arXiv: https://arxiv.org/abs/2609.02486v2 Document Visual Question Answering (DocVQA) often leverages Retrieval-Augmented Generation (RAG), where late-interaction encoders are commonly used to identify document pages relevant to a user query, before answer generation by a Large Vision-Language Model (LVLM). Existing approaches typically retrieve a fixed top-$k$ number of pages regardless of query complexity, which increases LVLM latency and may degrade answer accuracy. We introduce ViSAR (Visual Semantic Activation Retrieval), a training-free adaptive-$k$ retrieval method for late-interaction visual document retrieval. ViSAR operates directly in the embedding space to construct a query-conditioned page-level similarity matrix that highlights query-relevant semantics and dynamically determines the number of pages to retrieve. Across multiple encoders and LVLMs, ViSAR retrieves compact, query-adapted page sets that reduce RAG latency by up to 58.7\%, while maintaining or improving answer accuracy compared with fixed top-$k$ and adaptive retrieval heuristics. Furthermore, we show that the similarity matrix structure correlates with answer accuracy, suggesting future directions for retrieval quality-aware document understanding. --- ### How LLMs Build Fictional Worlds: Setting and Narrative Space in AI-Generated Creative Storytelling Authors: Katrin Rohrbacher, Björn Nieth, Emmanuelle Salin, Bjoern Eskofier, Michaela Mahlberg Published: 2026-09-02 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.02482v1 In this paper, we analyze how Large Language Models (LLMs) employ worldbuilding strategies, focusing on setting as one measurable dimension of storyworld construction. We compare 1,000 AI-generated stories per model in English and German with human-authored fiction from Project Gutenberg. Building on prior work, we operationalize setting through five types of narrative space: "action", "perceived," "visual," "descriptive" and "no space", identified using fine-tuned BERT classifiers for German and English. We generate narratives using GPT 4.1, LlaMA 3.3, Mistral 3.2, and Gemma 3 and compare their spatial distributions to a human-authored baseline. We find that human-authored texts predominantly employ "action space," grounding narratives in embodied character-environment interaction, whereas LLMs systematically overproduce "perceived space," emphasizing atmosphere and affect. This divergence remains stable across narrative time. Overall, our findings show that LLMs exhibit worldbuilding patterns that differ consistently from human-authored fiction in ways that are both model-specific and language-sensitive. --- ### PragAlign: Feedback-Guided Pragmatic Alignment for Controlled Synthetic Dialogue Generation Authors: Smitha Muthya Sudheendra, Jaideep Srivastava Published: 2026-09-02 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.02480v1 Synthetic dialogue generation can support research in privacy-restricted service settings, but generated conversations must preserve communicative intent, affective meaning, and natural dialogue flow. We introduce PragAlign, a feedback-guided framework for controlled synthetic dialogue generation conditioned on service context, target intent, and target emotion, with auxiliary trait-style controls. PragAlign uses a generate--evaluate--revise loop in which an LLM-based evaluator scores intent alignment, emotion alignment, coherence, fluency, and aggregate quality, then provides criterion-specific feedback for up to three refinement rounds. On 800 matched dialogue specifications, PragAlign achieves 99.50\% evaluator-defined acceptance, compared with 72.25\% for one-shot generation and 95.88\% for repeated generation without structured feedback. This indicates that repeated attempts account for much of the gain over one-shot generation, while structured feedback primarily improves last-mile multi-constraint satisfaction rather than broad average quality. Refinement gains are concentrated in emotion alignment, which is also the dominant failure mode in ablations. A separate human evaluation of 1,200 generated dialogues shows that intent expression and dialogue flow are highly recognizable to annotators, while emotion appropriateness is less stable and more subjective. These results support PragAlign as a quality-control framework for improving evaluator-defined communicative constraint satisfaction, while showing that affective realization and independent human-perceived quality remain open challenges. --- ### UnCapsTSR: An Unsupervised Transformer-based Image Super-Resolution Approach for Capsule Endoscopy Images Authors: Anjali Sarvaiya, Shubh Kawa, Lalit Agrawal, Jagrit Joshi, Kishor Upla, Kiran Raja Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02476v1 Wireless Capsule Endoscopy (WCE) captures and streams video while passing through a patient's Gastrointestinal (GI) tract and is used to examine its irregularities. Although advantageous over conventional endoscopy, WCE suffers from limitations related to capsule size and wireless transmission, resulting in images with coarser resolution. This work presents UnCapsTSR, an unsupervised transformer-based Generative Adversarial Network (GAN) framework for improving the spatial resolution of Low-Resolution (LR) WCE images. The proposed method accomplishes SR without explicit degradation estimation of real-world LR data and eliminates the need for true LR-HR pairs. UnCapsTSR employs a Bilateral Total Variation (BTV) loss to ensure spatial continuity in SR images. A newly curated dataset from the Kvasir Capsule dataset is also presented for training WCE SR models. Generalizability is validated on KID and GIANA datasets that are not used during training. A new non-reference metric, Endoscopy Quality Metric (EndoQM), is introduced for quantitative evaluation of domain-specific WCE data. Experiments demonstrate consistent improvement over state-of-the-art unsupervised SR approaches using NIQE, BRISQUE, PIQE, and EndoQM. Statistical evaluation shows 40 to 80 percent improvement in EndoQM from LR to SR across the evaluated datasets. --- ### Learning to Fuse LLMs with Ontology Rankers for Rare-Disease Diagnosis Authors: Zhaoyang Jiang, Zhizhong Fu, Yunsoo Kim, Zicheng Li, Xuanqi Peng, Fei Teng, Jiacong Mi, Honghan Wu Published: 2026-09-02 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.02473v1 Ontology rankers remain useful for rare-disease diagnosis because each candidate can be traced to matched patient phenotypes. Large language models (LLMs) can generate differential diagnoses from the same patient description, but their predictions lack an equally clear evidence trail. Rather than asking which system should replace the other, we ask whether an LLM can improve the ranker without giving up its evidence. Our behavior-based fusion model examines the two ranked lists, their agreement, and the ontology support behind each candidate, and learns how much to rely on each system for the individual case. Before comparison, we remove a documented test-set leakage pathway caused by benchmark cases and ontology annotations being derived from the same publications. Across eight open LLMs, fusion improves Phenomizer Recall@1 by 7.86 percentage points on Phenopacket Store and 20.18 points on RAMEDIS. When paired with DeepSeek-V4-Flash through an API, a fusion model trained only on the other LLMs improves Recall@1 from 0.1657 to 0.2176, a 5.19-point gain, without retraining. For 90.8% of correct fused diagnoses, the disease retains candidate-level ontology evidence that can be inspected. These results show that LLMs can strengthen an established diagnostic tool without discarding the structured evidence that makes it useful. --- ### Learning to Track from Privileged Target Appearances Authors: Xin Chen, Jiao Xu, Dong Wang, Huchuan Lu, Kede Ma Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02471v3 Target templates define what a visual tracker searches for, yet the templates available at inference trade off localization certainty with appearance freshness: the initial ground-truth template is exact but becomes stale, whereas recent templates better reflect the current appearance but are cropped from uncertain predictions. We quantify this bottleneck with a non-deployable oracle that supplies an exact current-frame target crop, improving AUC on LaSOT by 15.2 percentage points. This gap reveals a training-only opportunity: frame-level ground truths provide exact current- and future-frame target crops, although such crops are unavailable at deployment. We introduce Privileged Appearance Transfer for Tracking (PATT), a teacher-student training framework that transfers these privileged appearances to a deployable tracker through multi-level representation prediction. The privileged teacher observes exact target crops from past, current, and future frames, whereas the student receives only past-frame templates and learns to predict the teacher's search representations. To avoid transferring unreliable teacher signals, PATT weights this transfer by the teacher's relative localization advantage over the student and its absolute localization accuracy. After training, the teacher, latent predictor, reliability weights, and privileged crops are removed, leaving standard student-only inference. Across seven benchmarks at two model scales, PATT achieves consistent gains under both long- and short-term tracking protocols. --- ### DeepAffinity: Long-Term Aspect Preference Prediction in eCommerce using Small Language Models Authors: Yotam Eshel, Guy Hadad, Guy Feigenblat, Yuri M. Brovman, Matt Gearhart, Bracha Shapira Published: 2026-09-02 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.02468v2 We explore predicting eCommerce user preferences for product aspects such as brand, size, and color - a task we define as Aspect Affinity. Solving this task improves customer understanding and enables fine-grained personalization in recommendation, search, and marketing. We frame Aspect Affinity as a temporal prediction task: forecasting a users future aspect choices from their time-ordered interaction history, capturing long-term preferences that evolve beyond the current session. To this end, we propose DeepAffinity, which leverages Small Language Models (SLMs) with structured prompts and specialized prediction heads fine-tuned for this task. We show DeepAffinity outperforms standard generative fine-tuning methods, while general-purpose open-source LLMs perform poorly without task-specific tuning, highlighting their limits in modeling nuanced behavior. Finally, DeepAffinity enhances recommendation quality on a large-scale multinational eCommerce platform. --- ### VIPS: Vehicle-Infrastructure Cooperative Planning Benchmark via Pseudo-Simulation Authors: Hoonhee Cho, Jae-Young Kang, Giwon Lee, Hyemin Yang, Heejun Park, Kuk-Jin Yoon Published: 2026-09-02 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.02462v1 End-to-end autonomous driving in urban environments requires robust decision-making under partial observability and complex multi-agent interactions. Severe occlusions and dense traffic at intersections limit the perception capability of single-agent systems, motivating recent efforts on Vehicle-to-Infrastructure (V2I) cooperation for perception and planning. However, existing evaluation protocols face a fundamental trade-off: open-loop evaluation fails to capture error accumulation and recovery from deviations, while closed-loop evaluation is costly, difficult to scale, and often relies on simulated environments that may suffer from domain gaps. To bridge this gap, we propose VIPS, a benchmark for cooperative autonomous driving in V2I settings based on pseudo-simulation. VIPS extends pseudo-simulation by integrating vehicle and infrastructure observations. This enables scalable yet realistic evaluation of robustness and error propagation without full simulation. We further present CoS-V2X, a cooperative planning framework based on sparse representations. CoS-V2X models vehicle-infrastructure interactions using compact features for efficient communication and robust decision-making under heterogeneous observations. Code and dataset are available at https://vips2026.github.io. --- ### Model-Free Surrogate-Assisted Neural Architecture Search for Evolving Variable-Length Dense Blocks Authors: Asif Ameer, Maryam Bashir, Irfan Younas, Muhammad Fayyaz Published: 2026-09-02 Categories: cs.NE arXiv: https://arxiv.org/abs/2609.02460v1 Neural Architecture Search (NAS) has emerged as a powerful paradigm for automatically designing deep neural networks; however, its practical adoption is often limited by substantial computational cost. To alleviate expensive full-training evaluations, surrogate-based methods have been introduced to estimate network performance efficiently. Nevertheless, existing approaches-particularly model-based surrogates-require training many candidate architectures and involve additional optimization overhead. In this work, we propose a Model-Free Surrogate PSO Network (MFSPNet) for evolving convolutional neural network architectures. The proposed method integrates a lightweight model-free surrogate predictor within a particle swarm optimization (PSO) framework, eliminating the need for pre-trained surrogate models. Specifically, MFSPNet introduces two key contributions: (1) a validation-loss-driven exponential moving average estimator (VLE-EMA) that captures early generalization behavior for reliable architecture ranking; and (2) a block-based dense connection strategy that enables effective stacking of evolved blocks while mitigating vanishing-gradient issues. This design also facilitates transferability of learned blocks across datasets. Extensive experiments demonstrate that MFSPNet achieves competitive performance with reduced computational cost. Under a consistent training protocol with ten independent runs, the proposed method attains error rates of 3.91%, 17.68%, and 1.91% on CIFAR-10, CIFAR-100, and SVHN, respectively, along with top-1/top-5 error rates of 28.29%/12.82% on ImageNet, while requiring less than three GPU days for architecture search. Due to computational constraints, the ImageNet result is based on a single run and should be interpreted as indicative of scalability. Overall, MFSPNet provides an efficient and reliable framework for cost-aware neural architecture search. --- ### CivBench: A Long-Horizon Benchmark for Tool-Mediated Agents in Civilization VI Authors: Austin Tudor David Andrews, Liam Wilkinson, Jamie Heagerty, Harry Coppock, Jakob Nicolaus Foerster, Rui Ponte Costa Published: 2026-09-02 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.02459v1 We present CivBench, an open-source benchmark for evaluating language model agents in long-horizon, tool-mediated environments through the Model Context Protocol (MCP). A single episode spans 300+ turns and produces thousands of tool calls over a large action space, requiring sustained planning, state monitoring, and execution under partial observability. The environment exposes 76 MCP tools and a narration layer that converts visual game state into structured text. We use CivBench to characterise agent behaviour across four model families in 23 admissible runs. The sample is a pilot, not a model ranking: aggregate outcomes do not reliably discriminate models at this scale. Instead, we introduce two interface-level metrics that the environment makes measurable: Proactive Monitoring Rate (PMR), capturing whether agents actively query latent strategic state, and RAG@10, capturing whether commitments stated in structured planning reflections are executed within ten subsequent turns. Across runs we observe two consistent patterns under a shared playbook protocol. Agents under-monitor strategically relevant state that is available but requires explicit querying: despite playbook guidance to query victory progress every 20 turns, agents do so only every 30 to 75 turns, and in 7 of 20 detectable defeats they failed to query within the 20 turn warning window before game end. Agents also frequently fail to execute near-term commitments stated in their own planning reflections (RAG@10 between 48.2% and 65.8% across models). Both patterns arise despite tool access and explicit guidance, and we interpret them as deviations under instruction rather than absences of capability. We release the environment, scenarios, logs, metrics, and analysis pipeline at https://github.com/lmwilki/civ6-mcp ---