# Research Papers — 2026-09 · bundle page 210 of 266 > Full-text markdown of papers #10451–10500 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/210 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/209.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/211.md) ### Multi-Level-Set-Based Physics-Driven Neural Network to Solve 3-D Inverse Scattering Problems Authors: Yutong Du, Zicheng Liu, Bo Qi, Yali Zong, Peixian Han Published: 2026-09-08 Categories: cs.LG, physics.comp-ph arXiv: https://arxiv.org/abs/2609.08594v1 This paper proposes a level-set-based physics-driven neural network solver (LSPDNN) for 3-D electromagnetic inverse scattering. To mitigate boundary blurring and reconstruction artifacts in voxel-wise contrast reconstruction, the proposed solver exploits the piecewise homogeneity of practical scatterers by representing unknown targets with multiple coordinate-dependent neural level-set components. Specifically, a soft-union multi-material model is proposed to separately describe the object support and material distribution. The global support is formed by the union of multiple level-set components, while the local contrast is determined by normalized component weights and learnable complex permittivity candidates. In addition, a model-consistent total variation (TV) regularization is imposed on the material-region indicators, rather than directly on the reconstructed contrast, to suppress fragmented material assignments without excessively smoothing material interfaces. An adaptive loss balancing strategy is further introduced to reduce the dependence on manually selected regularization weights. For each measurement instance, the neural level-set parameters and material candidates are optimized by minimizing a physics-consistent objective function. Numerical and experimental results demonstrate that LSPDNN can reconstruct scatterers with clear boundaries, more uniform material regions, and substantially reduced background artifacts. The results highlight the advantage of the neural level-set parameterization in challenging 3-D inverse scattering cases involving irregular shapes, closely spaced objects, multiple materials, and measurement noise. --- ### FedGenSC: Federated Generative Semantic Communication with Channel-Aware Adaptation Authors: Rita Abou Fares, Razan Al Kakoun, Maher Nouiehed, Hadi Sarieddeen Published: 2026-09-08 Categories: eess.SP, cs.LG arXiv: https://arxiv.org/abs/2609.08593v1 Integrating generative adversarial networks (GANs) into federated semantic communication (SemCom) is a natural progression, as generative priors can recover semantic fidelity under channel distortion that discriminative decoders cannot. However, naive GAN federation introduces three failure modes that prior work has, to the best of our knowledge, neither identified nor resolved: discriminator aggregation instability under non-independent and identically distributed (non-IID) data, semantic drift caused by divergent local embedding spaces, and channel-agnostic generation that cannot adapt to heterogeneous link conditions. We propose federated generative semantic communication (FedGenSC), which mitigates all three by employing a global generator with local-only discriminators, providing cross-client semantic information through a semantic prototype bank, and conditioning generation on the instantaneous signal-to-noise ratio (SNR). Experiments on the Europarl dataset over Rayleigh fading channels (K=10 clients, Dirichlet α=0.5) show that FedGenSC under non-IID data outperforms the FedDeepSC baseline across the tested SNR range, achieving up to a 58.2% relative improvement in bilingual evaluation understudy (BLEU)-1 at 18 dB. Ablation studies confirm the independent contribution of each component. --- ### A Three-Tier Persona Vector for Controllable User Simulation in Agentic Evaluation Authors: Rahul Khedar, Eshita, Sneha Teja Sree Reddy Thondapu, Mayank Malhotra, Arup Kumar Das, Jitesh Chandra Mishra, Arun Menon, Avinash Karn, Mouli V Published: 2026-09-08 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.08592v1 Evaluating tool-augmented LLM agents requires diverse, realistic user inputs yet most evaluation frameworks use flat role descriptions ("you are an angry customer") that produce near-identical conversations regardless of the underlying scenario. In this paper, we propose a three-tier persona vector with 23 operationalized dimensions: 6 categorical demographics (jurisdiction, age, channel, device, language proficiency, time availability), 12 continuous behavioral traits (patience, assertiveness, digital literacy, etc.) sampled with Gaussian noise around curated profile base vectors, and 5 continuous emotional states (frustration, anxiety, trust, confidence, stress) that shift in response to scenario context. Orthogonal to the persona, a 4-level query-complexity overlay controls utterance phrasing from direct to deliberately vague. We evaluate the persona model inside a synthetic data generation pipeline across 64,698 multi-turn conversations spanning 8 named profiles and 3 production corpora. Key findings: (i) a 15.8 percentage-point spread in agent goal-achievement across personas confirms trait vectors produce measurably different user behavior; (ii) the same persona behaves differently across scenarios due to scenario-reactive emotional state shifts, validating the scenario-reactive design; (iii) domain-specific projects show persona sensitivity on booking-flow compliance (~15-20 percentage points gap between tier-aware and pressure-test personas), demonstrating the model faithfully reproduces real-world difficulty distributions; (iv) seven rule-described trait correlations produce auditable co-occurrence patterns without requiring learned covariance matrices. The persona model is fully specified for reproduction. --- ### The Unreliable Progress Bar: Can LLM Agents Reliably Report Task Progress Throughout Execution? Authors: Boyang Wang, Yunhan Wang, Yalun Wu Published: 2026-09-08 Categories: cs.SE, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.08589v1 Recent large language models can emit task-progress signals that agent frameworks use to decide whether a task should continue or stop, yet whether a model can reliably report its task progress at every stage of a task, and where and how its reports fail, has not been studied systematically. We evaluate this ability on the public benchmark $τ^2$-bench and on StageIF, a controlled testbed in which reporting checkpoints are placed across the task's lifecycle. Both settings require reports at multiple task stages. We find that reporting reliability depends on the stage a task has reached, and that almost every deployed model we test is reliable at some stages and unreliable at others. Where reporting breaks down is not the same everywhere. Most deployed models lose accuracy once work is under way and recover once the task is done. The newest generation closes that mid-task drop and instead grows conservative at the finish line. Our study exposes a capability gap in task-progress reporting and provides an evaluation protocol that spans the whole course of task execution for this ability on which agent operation depends. The findings indicate that agent frameworks should not control task flow on the strength of the model's state reports alone. --- ### Limitations of Automated Simulatability: LLM Simulators Can Bypass Explanations Authors: Antonin Poché, Fanny Jourdan, Nils Feldhus, Qianli Wang, Jing Yang, Simon Ostermann, Nicholas Asher, Philippe Muller, Vera Schmitt Published: 2026-09-08 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.08585v2 Simulatability is an evaluation protocol for explanations that quantifies their usefulness by how well they help a user predict a task model's outputs. Since human evaluation is costly, automated simulatability replaces human explainees with LLM simulators, as proposed in ConSim (Poché et al., 2025) for large-scale experiments. We qualitatively replicate and extend ConSim's ranking of explanation methods across the tested datasets, explanation families, and simulator LLMs, and identify two limitations. First, when class names are meaningful, simulators can obtain high simulatability by solving the classification task directly, without relying on the explanations. Second, class anonymization can reward explanations for leaking the hidden label mapping, a limitation we expose with a new classes-as-concepts baseline. These results are consistent with a shortcut hypothesis: in the tested settings, simulator predictions mainly rely on task priors, while explanations produce small changes. We derive recommendations for more robust automated simulatability evaluations. --- ### Leveraging Cardiac Imaging to Improve ECG-Based Detection of Chagas Disease in Resource-Constrained Settings Authors: Laura Alvarez-Florez, Daniel Uyterlinde, Samuel Ruipérez-Campillo, Lukas P. A. Arts, Folkert W. Asselbergs, Fleur V. Y. Tjong Published: 2026-09-08 Categories: cs.LG, cs.AI, eess.IV arXiv: https://arxiv.org/abs/2609.08582v1 Chagas disease is a major cause of cardiomyopathy in Latin America. Cardiac magnetic resonance (CMR) imaging can characterize its structural abnormalities, but scanners and expert readers remain scarce in endemic regions. Electrocardiography (ECG) is inexpensive and widely available, yet structural disease must be inferred indirectly from electrical signals. We propose to transfer CMR-derived structural knowledge to ECG through contrastive pre-training. Using 63,193 paired ECG-CMR examinations from the UK Biobank, we align an ECG encoder with a clinically grounded CMR embedding space using an asymmetric InfoNCE objective. Despite seeing no Chagas cases during pre-training, the resulting representation improves ECG-based Chagas detection. Across CODE-15% and SaMi-Trop, a frozen linear probe achieves an AUROC of 0.851 and sensitivity at the top 5% of predicted risk (Top5%-TPR) of 0.427 in five-fold cross-validation, compared with 0.827 and 0.377 for an unaligned ECG-FM baseline. On the PhysioNet/CinC 2025 Challenge test set, our model obtains the highest AUROC on SaMi-Trop-3 and the best ELSA-Brasil challenge score among the three top-performing methods, indicating that imaging-supervised ECG representations can generalize to populations and settings beyond the pre-training distribution. --- ### AlphaRJM: Reward-Jump Memory for Stochastic Return-Guided Alpha Discovery Authors: Sayan Dhan, Selvaraju Natarajan Published: 2026-09-08 Categories: cs.LG, q-fin.CP, stat.ML arXiv: https://arxiv.org/abs/2609.08581v1 Formulaic alpha discovery is a pool-dependent symbolic search problem in which informative feedback is observed primarily when a complete expression is evaluated. This delayed feedback creates two coupled difficulties: the retained alpha pool does not preserve the full history of realized evaluation feedback, and the value of an intermediate construction action is uncertain because its consequence depends on the formula eventually completed. We introduce AlphaRJM, which addresses these difficulties through Reward-Jump Memory, an event-driven latent state that remains fixed during token construction and updates only at terminal evaluation events using the realized pool reward and evaluation outcome, and an action-conditioned SDE return critic that represents future discounted discovery returns with stochastic particles. The particles guide action selection through their mean and uncertainty and are learned using a distributional Bellman objective combining energy-distance matching, mean calibration, and jump regularization. Empirically, AlphaRJM delivers strong and stable gains across multiple equity universes, forecasting horizons, and random seeds, while ablations confirm the complementary roles of persistent evaluation history, stochastic return modeling, and distributional supervision. --- ### Instability in Patient Clustering: A Multiverse Analysis of Unsupervised Clustering in the CENTER-TBI cohort Authors: Sean R E A Bagcik, Aneeta Merlin Chacko, Ewout W Steyerberg, Maarten van Smeden, Andrew I R Maas, Erik van Zwet, Nicole S Erler Published: 2026-09-08 Categories: stat.ME, stat.AP, stat.ML arXiv: https://arxiv.org/abs/2609.08577v1 Understanding patient heterogeneity is key to improving prognostic modeling in traumatic brain injury (TBI). Unsupervised clustering is widely used to explore patterns in patient characteristics that may define subgroups. However, it involves a multitude of decisions, including the choice of algorithm, the distance metric, and the method used to determine the "optimal" number of clusters. The aim of this study is to investigate how these choices influence the resulting clustering solution. We analyzed data from 4,509 patients enrolled in the Collaborative European NeuroTrauma Effectiveness Research in TBI (CENTER-TBI) study. K-medoids, agglomerative, and spectral clustering were applied in a complete 3 X 2 X 2 factorial design, in combination with Euclidean or Gower's distances, and silhouette score or gap statistic to choose the number of clusters. We investigated the agreement of clustering solutions with UpSet Plots and stability with the (adjusted) Rand index. Comparisons were made both across approaches using the original dataset and within approaches using bootstrap resampling. Clustering results varied substantially depending on the analysis choices. The number of suggested clusters varied widely, from one to twenty-five. Adjusted Rand indices confirmed low concordance between methods. Moreover, none of the clustering solutions demonstrated discriminatory performance comparable to a supervised logistic regression model in classifying patient recovery illustrating the limited usefulness of clustering for this purpose. The high instability in clustering results compromises interpretability and underscores that such solutions should not be blindly interpreted as underlying structure. --- ### Which Forms of Caregiver Feedback Support Grammar Learning? A Reinforcement-Learning Study of Child-Like Language Models Authors: Jing Liu, Marianne Schweitzer, Abdellah Fourtassi Published: 2026-09-08 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.08576v1 Social interaction is central to children's language learning, but the effects of different forms of caregiver feedback are difficult to isolate in naturalistic data. We use child-like language models as controlled learners to test which forms of feedback support grammatical development. Small GPT-2-style models are pretrained on child-directed language from CHILDES, then fine-tuned with reinforcement learning using reward models trained to capture four feedback types: communicative feedback, structural alignment, semantic contingency, and affective feedback. Reward fine-tuning yields limited gains on minimal-pair evaluations, but clearer effects in free generation. Structural alignment produces the strongest improvements in grammaticality, providing a novel, plausible mechanistic account of how this feedback can support grammar learning. Communicative feedback yields more moderate gains. In contrast, semantic contingency and affective feedback do not improve grammaticality, although further analyses suggest that they may support other aspects of language learning beyond grammar. These results suggest that different forms of caregiver feedback make complementary contributions to language learning. --- ### Do New Attention Mechanisms Actually Fix Attention Sinks at Million-Token Context? Authors: Sara Rizwan, Samaanah Abdus Salam, Mohammed Mudassir Uddin Published: 2026-09-08 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.08574v2 Long context language models now advertise windows of one million tokens, but two habits limit how much of that window is used. Attention heads with nothing useful to read still spend their budget on the first token, which is called the attention sink, and where a fact sits in the context changes whether the model finds it. Gated attention cut first token attention from 46.7 percent to 4.8 percent at NeurIPS 2025, and Kimi K3 pairs that idea with Kimi Delta Attention and Attention Residuals behind a one million token window, eight times past the range where these diagnostics have been reported. This paper asks whether the fix survives that jump. We build SinkProbe, a suite that measures sink mass, massive activation, position resolved recall and the recency gap, and apply it to four small models that differ only in how they mix tokens and depth. Three results follow. The training objective produces the sink, not the architecture. Gating did not reproduce its published effect at our scale. Sink mass, activations and position bias moved independently. Code, data and the measurement protocol are released at https://github.com/sararizwan7/Attention-Mechanisms-in-1M-Context-Window --- ### AgentGrad: Intervention-guided Prompt Optimization for Multi Agent Systems Authors: Jaewon Chu, Jinwoo Seo, Jaewon Cho, Jeehye Na, Yunyang Xiong, Youngdae Kim, Hyunwoo J. Kim Published: 2026-09-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.08572v2 Large language model (LLM)-based multi-agent systems (MAS) achieve strong performance by employing specialized multiple agents, yet their performance depends on the prompt design of each agent. For MAS prompt optimization, textual gradient methods that guide prompt updates using natural-language feedback have emerged as a leading paradigm. In this paper, we identify limitations in two stages of existing textual gradient approaches: gradient extraction and gradient aggregation. In gradient extraction, previous works select a target prompt without verifying whether modifying it resolves the failure, and derive gradients without agent-level supervision over the corresponding agent's intermediate output. In gradient aggregation, individual gradients are randomly grouped and concatenated, often mixing unrelated failure modes and producing prompts that fail to generalize. To address these limitations, we propose AgentGrad, a prompt optimization framework for multi-agent systems based on sequential intervention and semantic textual gradient abstraction. For each failure, sequential intervention modifies the behavior of one agent at a time to identify the target agent whose modification resolves the failure. The modified output of the target agent then serves as agent-level supervision for extracting a fine-grained gradient. Semantic textual gradient abstraction clusters semantically similar gradients to prevent mixing unrelated failure modes, and abstracts each cluster into a generalized gradient that captures the shared corrective pattern. Experimental results show that AgentGrad achieves state-of-the-art performance across five MAS benchmarks while reducing wall-clock optimization time by $2.5\times$ and optimization cost by 21.8\% on average compared to the next-best baselines. --- ### Effects of model architecture and learning strategies on deep learning-based recognition of activated sludge microscopic images and comparison with quantitative image analysis Authors: Suguru Hakoshima, Tomohiro Tobino, Fumiyuki Nakajima Published: 2026-09-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.08570v1 Microscopic image analysis has long been recognized as a promising approach for monitoring activated sludge. In recent years, deep learning-based image analysis has been increasingly adopted in this field because of its high performance. However, previous studies on microscopic image analysis of activated sludge have rarely explored transformer-based models or self-supervised foundation models and have instead relied on CNNs and supervised ImageNet pretraining. In addition, previous studies often downsampled image sizes, but the effects of downsampling have not been sufficiently investigated, and the relationship between downsampling strategies and image analysis performance remains unclear. Furthermore, no study has quantitatively compared deep learning performance with quantitative image analysis (QIA), which was widely used before the emergence of deep learning. In this study, to examine how model architecture and learning strategies affect performance in microscopic image analysis of activated sludge and to quantitatively determine whether deep learning outperforms QIA, we prepared three types of activated sludge samples, classified their microscopic images, and evaluated classification accuracy. Our results showed that transformer-based architectures and alternative pretraining methods were effective in terms of classification accuracy. Our downsampling analysis showed that using overly small images reduced accuracy, but increasing image size beyond a certain point did not improve it further. In addition, the analysis indicated that, to achieve high classification accuracy, maintaining the field of view was a more effective downsampling strategy than maintaining resolution. Finally, our comparison between deep learning and QIA showed that deep learning outperformed QIA in terms of accuracy. --- ### BIO-MEMART: Biometric-Aware KV Cache Memory for Multi-User LLM Agents Authors: Yanhong Qian, Xuanying He, Qingguo Meng, Shihao Ding, Xingbo Dong, Zhe Jin Published: 2026-09-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.08566v1 KV cache is evolving from a serving optimization into an external memory substrate for long-term LLM agents. In a shared multi-user deployment, however, reusable KV blocks introduce a missing access-control question: semantic relevance alone cannot determine whether a memory block is authorized for the current physical user. We propose Bio-MemArt, a biometric-aware KV-cache memory framework for multi-user LLM agents. Bio-MemArt attaches a normalized biometric template to each stored KV memory block, filters the shared memory pool with the current user's biometric probe, and then runs the original MemArt retrieval and KV reuse pipeline only inside the authorized candidate pool. This design preserves latent-space retrieval, direct cache reuse, and decoupled position encoding while adding physical-user access control to shared KV memory. We evaluate Bio-MemArt under Owner and Non-owner query conditions on long-term dialogue QA with face and palmprint benchmarks. Across face benchmarks, the average owner and non-owner biometric success rates are 95.71% and 0.86%; across palmprint benchmarks, they are 97.60% and 2.00%. In the efficiency study, average prefill tokens drop from 18,781.96 under full-context prompting to 28.57 with Bio-MemArt, showing that biometric gating preserves the low-token operating regime of KV-cache memory. --- ### Non-Adaptive 1-Bit Mean Estimation: Minimax Rates and the Sample-Interval Tradeoff Authors: Ivan Lau, Jonathan Scarlett Published: 2026-09-08 Categories: stat.ML, cs.IT, cs.LG, math.ST arXiv: https://arxiv.org/abs/2609.08564v1 We study distributed one-dimensional mean estimation under a 1-bit communication constraint. Each agent observes one sample, drawn independently from an unknown distribution, and returns a single bit in response to a query $Q: \mathbb{R}\to\{0,1\}$ chosen by a central learner. The distribution has mean in $(-λ,λ)$ and $k$-th central moment at most $σ^k$, for a fixed $k>1$. The order-optimal two-stage protocol of Lau and Scarlett uses responses from the first batch to choose the second-batch queries, motivating the question of whether this single round of interaction is necessary. We answer this negatively: for every $k>1$, a non-adaptive protocol attains the adaptive 1-bit minimax rate (and concurrent works reached the same conclusion via different strategies). We further determine the minimax sample complexity among non-adaptive 1-bit estimators when every one-set $Q^{-1}(1)$ is restricted to a union of at most $s$ intervals. Relative to unrestricted non-adaptive 1-bit querying, this constraint adds a term of order $(λσ/(s\varepsilon^2))\log(1/δ)$, giving the full tradeoff between sample complexity and interval complexity to within $k$-dependent constant factors. As a corollary, we identify, order-wise, the minimum interval budget needed to retain the unrestricted 1-bit minimax sample rate. --- ### Certified Topological Interaction in Neural Representations: Exact Tests and the Statistic They Require Authors: Sushovan Majhi Published: 2026-09-08 Categories: cs.LG, math.AT arXiv: https://arxiv.org/abs/2609.08561v2 Class disentanglement--the separation of a representation's class-conditional point clouds along depth and over training--is measured by descriptive curves: the sentence such a study wants to write, layer l+1 is more disentangled than layer l, is an eyeball judgement with no null. We supply the inferential layer for a topological measurement of class overlap, the Intersection Euler Characteristic Profile: the Euler characteristic of the overlap of the clouds' ball unions as a function of scale, from one Alpha-complex sweep with no boundary-matrix reduction. Every number carries a test--exact permutation tests in both directions, a guarded separation certificate the invariant requires, and a paired sign-flip test for comparative claims. Building that test taught a lesson outliving this invariant: its statistic must be scale-free. On the raw profile mass, which has units of feature length, 12,375 paired tests return 5,633 significant steps of which every one at the first epoch points the wrong way, certifying feature-norm dynamics as disentanglement; the dimensionless statistic returns 2,707, with 2,026 decreases. Across 111 networks and 52,650 measurements, disentanglement is depth-graded and early, and interaction quotients rank class pairs by confusability (rho=0.83), on par with cheap separability statistics. In a 96-model factorial, augmentation is the one training choice that separates classes relative to chance; weight decay compresses the overlap without separating. Only a k-fold statistic can pose the structural question: the joint entanglement of a class triple sits below its strongest pair in 97% of triple-layer cells and 99.5% of deep cells, at median ratios far below a measured null floor, in vision encoders and frozen language models--a regularity, not a law. The unnormalized mass predicts test accuracy (R^2=0.94), the quotient does not, and neither beats a linear probe. --- ### Personalizing LLM Agent Memory Using Biometrics Authors: Yanhong Qian, Qingguo Meng, Shihao Ding, Xingbo Dong, Zhe Jin, Hanrui Wang, Isao Echizen Published: 2026-09-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.08558v1 Personalized memory helps LLM agents deliver stable, tailored assistance by storing and reusing user-specific data across interactions. In multi-user scenarios, however, retrieval must consider not only semantic similarity but also whether the current requester matches the identity associated with the stored memory. We propose Bio-Memory, a biometric-aware memory architecture that conditions memory retrieval on both semantic similarity and biometric matching. Built on top of A-Mem, Bio-Memory augments each atomic memory note with a biometric embedding and uses biometric matching to form the retrieval candidate pool before semantic ranking. We evaluate Bio-Memory on LoCoMo in a 10-user shared-agent setting over 7 face benchmarks and 10 palmprint protocols. Across datasets, Bio-Memory consistently separates owner and non-owner queries. Under face-based personalization, the largest average gap reaches 27.29% / 21.15% in F1 / BLEU-1 on CALFW; under palmprint-based personalization, the corresponding gap is 25.75% / 19.22% on MS_Blue. These results support biometrics as a practical control signal for personalized memory retrieval in shared environments. --- ### Not All Variables Agree: Reliability-Aware Variable-Wise Gradient Surgery for Multivariate Time-Series Forecasting Authors: Jinwoo Park, Hyeongwon Kang, Pilsung Kang Published: 2026-09-08 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.08554v1 In data-driven training, multivariate time-series forecasting is usually optimized with a scalar loss averaged over samples, variables, and horizons. This averaging is convenient, but the optimizer sees only the aggregated gradient, which does not reveal whether the variable-wise contributions align or oppose one another. To quantify how often this disagreement arises, we measure the variable-wise gradients directly and find that 30.6% of their pairwise cosine similarities are negative on average across seven datasets. However, conflict and harm are not the same thing. Under shared training 35 of the 64 variables do worse than a full-input single-target oracle, and the harmed fraction is not reliably predicted by how often gradients conflict. We propose Per-Variable Surgery (PV-Surgery), an optimizer-side training strategy for backbones with cache-compatible layers. One backward pass builds variable-wise gradient proxies from output-side signals and keeps the pointwise forecasting loss. Reliability-aware selection targets layers whose proxy sums closely approximate their shared-gradient slices. Conditional pooling forms anchor and conflict pools without dropping variables. Common-direction surgery aligns variable or pooled gradients with their normalized mean and restores input norms to avoid reweighting. In experiments across five backbones, seven datasets, and four horizons, PV-Surgery lowers MSE by 3.61% and MAE by 2.93% on average. For multivariate forecasting, this indicates that the variable-wise structure hidden by mean-loss training is a usable optimization signal. --- ### CreaMem: A Scene-Aware Memory Architecture for Personalized Agents Authors: Qixuan Sun, Yue Que, Bowei He, Jin Guo, Dihang Yang, Wenchang Situ, Chen Ma Published: 2026-09-08 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.08550v1 Long-term memory is a core capability for personalized LLM agents. To support it, existing memory systems organize information using various criteria such as topic segments or summary hierarchies. However, we identify two major limitations in these designs. First, they lack scene awareness: memories from unrelated life scenes share the same retrieval space, which inflates the search space and introduces cross-scene interference. Second, they encode each memory from a single perspective, making it difficult to retrieve complementary views of the same event. In this paper, we propose the CreaMem architecture, which enables scene-aware memory organization by partitioning memory into several Life Scene Memories to reduce cross-scene interference at retrieval. To go beyond the single perspective and achieve cross-memory synergy, entries are dual-coded from both episodic and trait-based perspectives within each memory. We further devise a permemory balanced sampling strategy at retrieval time. Extensive experiments on two long-term memory benchmarks show that CreaMem improves QA accuracy across all evaluation metrics, with particularly large gains on multi-hop reasoning performance, validating scene-aware partitioning and cross-memory synergy. To enhance reproducibility, we release our code in a public GitHub repository. --- ### STSG-VQA: Evidence-Grounded Temporal Question Answering from Surgical Spatio-Temporal Scene Graphs Authors: Jing Li, Duygu Sarikaya Published: 2026-09-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.08543v1 Despite recent advances in surgical vision-language models (VLMs), temporal reasoning remains limited because existing supervision is largely frame-centric. Frame-level scene graphs (SGs) have proven effective in providing structured representations of surgical environments but do not explicitly model the dynamics of surgical workflows. To explicitly model how surgical states evolve across time, we introduce a multi-level structured temporal supervision methodology that augments frame-level surgical SGs with object-level continuity, event-level interaction continuity, and procedure-level connectivity. We then execute temporal queries over the resulting spatio-temporal scene graphs (STSGs) to generate evidence-grounded question-answer pairs, which together form the STSG-VQA benchmark. Each question is linked to the temporal interval and STSG evidence used to derive its reference answer, enabling traceable verification. The benchmark contains 18,458 question-answer pairs across seven temporal categories. Fine-tuning Qwen3-VL-4B and Hulu-Med-4B with STSG-derived supervision improves question-level micro accuracy by 24.39 and 19.56 percentage points over their zero-shot baselines and by 16.50 and 14.25 points over static scene-graph supervision, respectively. These gains span all temporal categories, indicating that STSG-derived supervision helps surgical VLMs reason over temporally grounded interactions rather than isolated frames. The code and dataset will be made publicly available upon acceptance. --- ### The Exact Time-Uniform Rate Frontier for Stochastic Gradient Descent on Smooth Convex Objectives Authors: Ruijie Li, Kang Chen, Tianyu Wang Published: 2026-09-08 Categories: math.OC, cs.LG, stat.ML arXiv: https://arxiv.org/abs/2609.08537v1 We study the time-uniform convergence of the raw iterate of standard stochastic gradient descent (SGD) for unconstrained smooth convex objectives. We prove that, under standard noise assumptions, the time-uniform convergence rate gets arbitrarily close to $\sqrt{\log n / n}$ but never reaches it. More specifically, we prove that for every positive, eventually nondecreasing sequence $h$ satisfying $h(n) = o(\sqrt{n})$, a bound of order $h(n)/\sqrt{n}$, holding simultaneously for all $n$ with probability at least $1-α$ and uniformly over the problem class, is achievable if and only if \( \sum_{j = 1}^{\infty} \frac{1}{h(2^j)^2} < \infty. \) The constructive sufficiency result follows from a dyadic horizon-free schedule together with an additive conditional-restart inequality. The necessity counterpart applies to every deterministic nonnegative schedule and holds even for a one-dimensional analytic smooth convex objective with Gaussian noise. --- ### Layer Selection in VLMs for Zero-Shot OOD Detection via Multi-Resolution Entropy Estimation Authors: Shyam Nandan Rai, Francesco Di Salvo, Sebastian Doerrich, Christian Ledig Published: 2026-09-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.08524v1 Out-of-distribution (OOD) detection is crucial for safe deployment of medical AI systems, where domain shifts arise across institutions, acquisition protocols, and patient populations. VLMs enable zero-shot OOD detection by embedding images into a language-aligned latent space, where cross-modal similarity serves as a non-parametric confidence signal for identifying in-distribution samples. Yet existing methods rely almost exclusively on final-layer embeddings, implicitly assuming that the deepest representations are universally optimal. We first show that this assumption does not hold in medical imaging: intermediate layers provide complementary OOD signals, and the optimal representational depth depends on the respective image modality. While prior work selects layer combinations via entropy minimization of normalized histograms, we demonstrate that single-resolution entropy estimation is highly sensitive to binning choices, leading to performance variations of up to 19.3% AUROC. To address this instability, we propose a multi-resolution entropy estimation strategy that aggregates histogram statistics across multiple discretization scales, enabling robust and stable intermediate-layer selection. Across two medical OOD benchmarks, namely MIDOG and OASIS, covering distinct imaging modalities, diverse shift types, and different VLM backbones, our method consistently outperforms state-of-the-art approaches, offering a lightweight and stable solution for zero-shot OOD detection. --- ### Same Values, Different Languages? From Multilingual Probing to Steering LLMs Toward Chinese Social Values Authors: Yuemei Xu, Kexin Xu, Jian Zhou, Haoyu Lu, Yequan Wang, Aishan Liu Published: 2026-09-08 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.08515v1 As Large Language Models (LLMs) are increasingly integrated into human society, aligning them with pluralistic social values has become a critical priority. However, whether LLMs exhibit consistent value preferences across languages remains underexplored, particularly for culturally grounded values, which are more abstract and difficult to evaluate and align than safety-centric principles. We investigate this issue through Chinese Social Values (CSV), a value system rooted in Chinese culture and comprising $12$ dimensions across national, societal, and personal levels. We construct C-Voices, the first comprehensive multilingual contrastive probe dataset for CSV, with 86,400 dilemma-based instances in six languages, each pairing a CSV-aligned action with a value-conflicting alternative. Building on the contrastive probes of C-Voices, we then propose a fine-tuning-free value vector steering method that derives value directions from hidden-state discrepancies and selectively intervenes on value-sensitive layers during inference. Experiments on six languages show that CSV-oriented preferences are model-dependent and language-sensitive, with the same dilemma eliciting divergent responses across languages. Our method achieves effective CSV steering, supports cross-lingual transfer of value vectors, and generalizes to existing FLAMES and ValuePrism. --- ### Temporal State Transport in Video Generation: Diagnosing and Correcting Spectral Imbalance Authors: Luyao Tang, Bingjun Luo, Dong Yi, Jialin Guo, Haoning Xi, Cheng Chen, Yizhou Yu, Chaoqi Chen Published: 2026-09-08 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.08505v1 Reliable video generation requires more than high-quality frames to form a coherent story: a model must maintain a persistent state, transporting visual attributes such as identity, scene layout, motion, and fine details across time. Existing training-free methods mainly strengthen cross-frame attention or analyze local attention entropy, but these views do not reveal whether temporal interactions stay in a healthy transport regime. In this work, we study video generation through the perspective of Temporal State Transport. We introduce Spectral Tension, a signed diagnostic that compares local attention diffuseness with global spectral diversity, and use it to identify two opposite temporal failures: fragmented transport and over-mixing hotspots. Based on this diagnosis, we propose Spectral Transport Homeostasis, a training-free regulator that softly corrects pathological temporal states while largely preserving balanced ones. Experiments on pretrained video generation models show that the original model often occupies imbalanced temporal regimes, whereas our method selectively applies larger corrections to the worst temporal hotspots and improves temporal consistency and visual quality without finetuning. Code: https://github.com/lytang63/temporal-state-transport --- ### SignRefine: Adapting Foundational Video Models for Sign Language Generation Authors: Anton Pelykh, Edward Fish, Ozge Mercanoglu Sincan, Richard Bowden Published: 2026-09-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.08496v1 Sign language video generation demands precise hand and facial articulation, yet modern video diffusion models, trained predominantly on spoken-language video, produce artifacts that render signing unintelligible. We propose SignRefine, a sign language video generation model that produces comprehensible signing from 2D keypoint conditioning alone, generalizing across appearances and visual conditions. Our approach builds on a pretrained video diffusion transformer and introduces local adapters with spatial grounding to selectively refine hand and face regions, steering the strong base model's prior toward accurate articulation. To enable this work and support broader sign language research, we present NVSign, a large-scale dataset of video content natively produced in sign language, offering diverse signer appearances, environments, and natural conversational settings. Trained on this data, our model shows up to 30% improvement in hand pose precision metrics over the strongest baseline and is preferred by sign language users for visual quality and comprehensibility in more than 80% of comparisons. --- ### AURORA: Active Uncertainty-Driven Re-Orientation for In-Hand Reconstruction Authors: Feiyu Zhao, Yuetong Li, Chenxi Xiao Published: 2026-09-08 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2609.08493v1 Observing objects grasped by a robot hand is challenging due to severe visual occlusions. Although in-hand manipulation can expose hidden surfaces, existing approaches often rely on predefined or open-loop reorientation strategies that do not explicitly target under-observed regions. We propose AURORA, an active 3D reconstruction framework that closes the loop between online object-centric reconstruction and in-hand reorientation. At its core, Ray-GPIS estimates direction-wise reconstruction uncertainty along candidate viewing rays and selects next-best-view targets using an uncertainty--novelty objective, which are realized through an axis-conditioned in-hand rotation policy. The resulting RGB-D observations are fused incrementally using CAD-free 6D pose tracking and lightweight geometric reconstruction. Experiments demonstrate that AURORA improves reconstruction quality and information-acquisition efficiency over non-active rotation strategies, while Ray-GPIS also outperforms active view-planning baselines in reconstruction performance, action-ranking quality, and planning efficiency. Targeted ablations further validate its robustness to hand occlusion and pose errors. The project webpage is available at https://aurorahand.github.io/ --- ### When Topology Betrays Privacy: Lattice-Based Reconstruction Attacks on Secure Aggregation in Decentralized Federated Learning Authors: Wenrui Yu, Changlong Ji, Johannes Bjerva, Qiongxiu Li Published: 2026-09-08 Categories: cs.CR, cs.LG arXiv: https://arxiv.org/abs/2609.08476v1 Secure Aggregation (SA) is widely regarded as a strong defense against model-update leakage in Federated Learning (FL), as it reveals only aggregate results while hiding individual updates. In Decentralized Federated Learning (DFL), SA is commonly instantiated as local neighborhood aggregation, where each node obtains a weighted aggregate over its neighbors. We show that this locality creates a structural leakage surface: sparse decentralized topologies provide colluding semi-honest nodes with asymmetric aggregate views, exposing multiple hidden linear combinations of honest participants' private states. Reconstructing private states from these aggregate views is fundamentally challenging, as both the private states and the aggregation coefficients are hidden. We tackle this challenge by establishing a formal connection to the Hidden Subset Sum Problem, a long-studied problem in cryptography. Building on this formulation, we design a lattice-based reconstruction approach that combines lattice reduction with structural filtering to reconstruct protected model states. We evaluate our attack on image, tabular, and text tasks under sparse DFL topologies. Our results show that colluding semi-honest nodes can recover the original local updates of honest nodes, enabling downstream reconstruction of private training data. These findings demonstrate that SA alone does not guarantee privacy in DFL when local aggregation induces asymmetric observations. --- ### Do Reviewers Still Reward Lexical Complexity? A Frozen-Rater Study of Preference Drift in 124K ICLR Reviews Authors: Jiabin Zheng Published: 2026-09-08 Categories: cs.CL, cs.DL, cs.LG arXiv: https://arxiv.org/abs/2609.08475v1 Large language models have collapsed the cost of producing lexically elaborate prose, and whether peer reviewers still reward it is a question about the evaluator, not about the text. When the association between a writing cue and review scores moves across years, the reviewers may have changed, the submissions may have changed, or both, and a regression of scores on text cannot say which. We separate the two with a frozen rater: 81,850 machine reviews of ICLR submissions from 2018 to 2025, all generated in one February-April 2025 window with one model family and one prompt, so that its year-to-year coefficients track submission composition alone and the human-minus-frozen trend difference identifies reviewer preference drift. On 32,638 submissions with 124,615 human reviews, the human coefficient on non-domain lexical complexity falls from +0.142 to -0.015 while the frozen rater moves from +0.080 to +0.082; the three-way difference-in-differences is -0.0100 (q=0.013), and forty random-wordlist placebos through the same specification centre on zero. Humans still reward sentence-length variability, which the frozen rater never registers, while the frozen rater still pays for lexical complexity at its earlier rate. Every claim is held to a double gate of false-discovery control and interval exclusion, and the findings that failed adversarial re-testing are reported. Reviewers discounted a cue whose production cost collapsed, as models of manipulable signals prescribe; an LLM judge calibrated to historical human preferences inherits the earlier schedule and drifts out of alignment while its agreement with humans on totals stays ordinary. --- ### Inclusive electron-nucleus cross section models from domain adaptation Authors: Krzysztof M. Graczyk, Beata E. Kowal, Rwik Dharmapal Banerjee, Jose Luis Bonilla, Hemant Prasad, Jan T. Sobczyk Published: 2026-09-08 Categories: hep-ph, cs.LG, hep-ex, nucl-ex, nucl-th arXiv: https://arxiv.org/abs/2609.08463v1 We apply transfer learning (TL) to construct data-driven models of inclusive electron-nucleus cross sections. Starting from an ensemble of deep neural networks pretrained on \(^{12}\)C data, we fine-tune the models separately for \(^{3}\)He, \(^{6}\)Li, \(^{16}\)O, \(^{27}\)Al, \(^{40}\)Ca, and \(^{56}\)Fe. The resulting models improve for all targets, marginally so for oxygen, where the carbon baseline is already adequate, although their predictive robustness depends on the amount, coverage, and precision of the available target data. We systematically study how model performance depends on the number of fine-tuned layers, on the fraction and selection of the training data, and on the overlap between the source and target kinematic domains. The layer-wise analysis shows that oxygen requires only shallow adaptation, whereas helium, calcium, and iron require substantially deeper fine-tuning. Lithium represents the least robust case because of its limited dataset, while aluminum demonstrates a strong sensitivity to a small subset of highly constraining measurements. For selected kinematic configurations outside the coverage of the carbon training data, the adapted models remain consistent with the measurements within their estimated uncertainties. Finally, we compare the resulting predictions with those of the phenomenological F1F2 model. --- ### Selective boundary condition reduction via learned error gating Authors: Daniel Fernández, Dominik Penk, Dominik Riedelbauch Published: 2026-09-08 Categories: math.NA, cs.LG arXiv: https://arxiv.org/abs/2609.08461v1 Parametric PDEs can admit different boundary conditions with different accuracy and computational cost. We introduce a framework for learning when one reduced boundary condition can replace another: paired solutions train a neural network to estimate the resulting domain and boundary errors, and the simpler condition is used only when both predicted errors meet prescribed tolerances. We focus on singular limits in applications, in which a stiff Robin or nonlinear boundary law is replaced by its limiting Dirichlet form. We evaluate the method on a galvanic corrosion problem and other nonlinear stationary and evolution problems. --- ### Detecting Authorship in Political Texts with Inductive Stylometry Authors: Gennadii Iakovlev, Levente Littvay Published: 2026-09-08 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.08459v1 Political texts are rarely authored by the nominal speaker alone. Tweets, speeches, reports, and official statements are drafted, edited, or harmonized by staff, yet political science has paid limited attention to the stylistic traces these hidden authors leave behind. This paper develops and stress-tests an inductive stylometric approach for recovering latent authorship structure in political communication, combining character 3-gram features with UMAP dimensionality reduction, and Burrows' Delta. We apply the approach to six corpora that vary in length (from tweets to long documents), in mode (written and oral), and in language (English and Hungarian). The approach recovers near-disjoint analyst fingerprints in formal legal prose in both languages, sorts a politician's tweets into validated subsets while uncovering additional insights, and distinguishes scripted from improvised speech. It fails, however, to resolve individual speechwriters within scripted corpora. Frequency-based stylometry is thus a powerful tool that, depending on authorial signal strength and institutional editing, can uncover authorship traces relevant to legislative studies, political communication, and policy research. --- ### SRPO: Setwise Relative Policy Optimization for Multi-Agent Systems Authors: Shengtian Yang, Ziyu Xiong, Yu Li, Yewen Li, Qingpeng Cai, Lei Feng Published: 2026-09-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.08452v2 Multi-agent systems enable complex reasoning and tool use by coordinating agents that divide roles and refine candidate solutions. Existing methods typically update individual agent responses or treat a complete trajectory as one training example. However, these methods may produce misleading policy updates because they assign the same final outcome to responses or trajectory segments that may play different roles in different team decisions. This is because treating each response as an independent update may separate outputs that jointly determine the next action, while treating an entire trajectory as one update may combine decisions made after different observations. These limitations call for a policy update defined at the level of a team decision, outputs that lead to the same state transition are optimized under a shared objective. In this paper, we propose Setwise Relative Policy Optimization (SRPO) for multi-agent systems. We represent the outputs used together to produce one state transition as an active set. A singleton set covers division of labor, while a larger set covers joint co-evolution, and the set composition can change across decisions. SRPO assigns a shared advantage to each active set, clips the combined policy change, and normalizes its scale according to the set size. This ties each update to the decision that produced the next state. Experiments on mathematical reasoning and multi-turn search demonstrate the effectiveness of SRPO across both tasks. Additional ablation studies analyze the normalization choice and training behavior under changing active-set sizes. --- ### GSComplete: Gaussian Splat Completion with 2D Diffusion Priors Authors: Elias Brugger, Philipp Erler, Stefan Ohrhallinger, Paul Guerrero Published: 2026-09-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.08449v1 Gaussian splats provide a fast, high-fidelity representation for 3D objects but are often constructed from incomplete input data in practice, leaving missing regions. Existing completion methods either do not preserve the original splats or require scarcely available 3D training data. We propose GSComplete, which combines 3D generation based on Score Distillation Sampling with a novel preservation loss that encourages the original splats to be preserved where they should be visible. This effectively completes the Gaussian splat object using only 2D diffusion priors while fully preserving existing splats and generating new splats only in missing regions, without occluding the input. To evaluate our approach, we introduce a new dataset of partial Gaussian splat objects and show that GSComplete achieves significantly more accurate preservation of the input than existing methods with comparable plausibility of the completed result. Our code and dataset will be made available upon acceptance. --- ### Topological Fraud Detection in Latent Transaction Spaces Authors: Avraham Bourla Published: 2026-09-08 Categories: cs.LG, cs.CR arXiv: https://arxiv.org/abs/2609.08445v1 Working entirely on topologically anonymized embeddings, we perform fraud detection using iterative rounds of unsupervised filtering followed by supervised sniping. The result is an ultra-low latency privacy--preserving triage that allows institutions to flag suspicious activity without compromising Personally Identifiable Information. --- ### SequenceO1: End-to-End Ultra-Long (100K) Sequence Modeling in Recommendation with Low-Rank Caching Authors: Lin Guan, Jia-Qi Yang, Zhishan Zhao, Jiaqi Huang, Hangyu Wang, Longbin Li, Beichuan Zhang, Haonan Jiang, Jinan Ni, Xiangyu Fan, Xiaowen Li, Ziyao Ren et al. Published: 2026-09-08 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2609.08443v1 Modeling long-term user behavior is central to sequential recommendation and billion-scale industrial recommender systems, yet production ranking models operate under strict latency, memory, communication, and training-throughput constraints. At the 100K scale, the challenge extends beyond attention complexity: raw sequence features must be stored, transferred, and repeatedly processed during training and online serving. Existing approaches based on history truncation, multi-stage behavior retrieval, compressed lifelong histories, or train-short/infer-long extrapolation either weaken end-to-end optimization or retain substantial length-dependent cost. We present SequenceO1, an end-to-end framework for ultra-long user behavior sequence modeling, deployed at full traffic on Douyin with histories of up to 100K interactions. SequenceO1 follows a compress-then-reason design. Its Sketch Attention (SA) uses learnable prototypes and prototype-wise normalization to compress the raw history into a fixed-size, target-agnostic user representation. Target-conditioned Stacked Target-to-History Cross Attention (STCA) then models complementary time scales: a recent 10K suffix for short-term interests and the compact sketch for long-term preferences. To make training and inference practical, SequenceO1 combines low-rank user representation caching, multi-request user-level batching, pipeline lift, and a fused FlashSA kernel to amortize feature storage, communication, and computation across targets, training instances, and consecutive requests. Production experiments show consistent offline and online gains, while the compact cached sketch retains most of the benefit of directly scaling end-to-end sequence ranking to 100K. These results provide a practical model-system approach to efficient attention, sequence compression, and scalable long-sequence and long-context recommendation systems. --- ### AirAnchor: Bridging Local and Global Spatial Information for Zero-Shot Aerial Vision-and-Language Navigation Authors: Shanwei Fan, Bin Zhang, Zhiwei Xu, Yingxuan Teng, Siqi Dai, Lin Cheng, Guoliang Fan Published: 2026-09-08 Categories: cs.CV, cs.AI, cs.RO arXiv: https://arxiv.org/abs/2609.08442v1 Aerial Vision-and-Language Navigation requires drones to follow natural-language instructions and navigate through complex urban environments. Accurate navigation relies on both local and global spatial information, which support immediate action grounding and long-horizon path planning, respectively. However, existing zero-shot methods typically operate at a single spatial scale, relying either on local representations constructed online from current observations or on global memories built offline from historical experience. To address this limitation, we propose AirAnchor, a new paradigm that bridges local and global spatial information through spatial anchors and integrates both into a shared navigation framework, enabling comprehensive spatial grounding for decision-making. AirAnchor consists of three core components: (1) Query-Driven Spatial Anchor Grounding, which identifies decision-relevant anchors from visual observations and organizes them into local spatial representations; (2) Persistent Object Spatial Memory, which incrementally maintains an object knowledge base as persistent global spatial memory and retrieves landmark-related spatial priors; and (3) a Spatially-Informed Navigation Agent, which explicitly integrates both local and global spatial information into an agentic framework for decision-making. Extensive experiments on AerialVLN demonstrate that AirAnchor substantially outperforms existing zero-shot baselines, validating the effectiveness and efficiency of the proposed paradigm. --- ### EvolveScaler: Synthesizing Information-Evolution Contexts via Executable State Machines and Natural-Language Rendering Authors: Ziliang Zhao, Zenan Xu, Shuting Wang, Zhao Wang, Bowen Cao, Minda Hu, Lincheng Li, Pluto Zhou, Zhicheng Dou Published: 2026-09-08 Categories: cs.AI, cs.SE arXiv: https://arxiv.org/abs/2609.08435v2 In persistent interactions, long contexts may encode an evolving process rather than a fixed record: later events can revise or revoke earlier information, changing what remains valid and what conclusions follow. We call this setting information evolution (IE). Solving IE requires identifying valid records, applying updates in order, and reconstructing the query-relevant state from the event history. Existing text-first synthesis pipelines make such data difficult to verify because state transitions and answer logic remain implicit. We introduce EvolveScaler, a code-driven framework that defines information evolution before rendering it as natural language. Human-authored operational specifications define state transitions, record validity, difficulty controls, and executable answer logic; a strong LLM then synthesizes a self-contained simulator from each specification. Executing validated simulators produces natural-language multi-turn event histories, while deterministic replay computes reference answers and atomic checklists. We instantiate EvolveScaler with 117 task prototypes and 159 final-question operators across five difficulty levels spanning approximately 7 to 1,200 events per instance, yielding about 35,100 training examples and 585 validated evaluation instances. On the very_long tier, the strongest model reaches 59.3% avg@5, while six models score below 10%. Training an internal A3B model on 6,000 EvolveScaler examples improves performance over its base checkpoint on all eight independently constructed out-of-distribution benchmarks, with a 5.25-point average gain. These results show that code-driven IE synthesis provides both challenging evaluation and transferable training supervision. --- ### CAR-MIL: Counterfactual Attention Regularization for Multiple Instance Learning Authors: Imane Chraki, Pierre Marza, Stergios Christodoulidis, Maria Vakalopoulou Published: 2026-09-08 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.08419v1 Multiple Instance Learning (MIL) is widely used for weakly supervised learning, particularly in digital pathology, where fine-grained annotations are costly. Most MIL methods aggregate instance features via attention mechanisms. However, attention weights do not always faithfully reflect instance importance and may focus on spuriously correlated regions. In this work, we propose CAR-MIL, a framework that explicitly guides attention learning through a counterfactual attention regularization objective inspired by counterfactual explanations. Built on a standard attention-based MIL architecture, our approach introduces a lightweight counterfactual attention branch trained to produce an alternative prediction while remaining close to the factual attention distribution. This encourages prediction changes to arise from minimal, structured redistributions of attention, leading to more informative evidence allocation. The resulting factual and counterfactual attention maps capture complementary evidence: the former highlights regions supporting the prediction, while the latter reveals regions whose reweighting would challenge it. We evaluate our method on synthetic MIL benchmarks with instance-level ground truth enabling controlled analysis of attention behavior and on five digital pathology datasets across four tasks. CAR-MIL maintains competitive classification performance, with the largest gains observed on more challenging tasks, while improving attention reliability, demonstrating the benefits of integrating counterfactual explainability reasoning into attention learning. Code is available at: https://github.com/ImaneCR/CAR-MIL/. --- ### Feyospace-v1: How the Cyber Mercury Seven Trained Frontier Cyber Models Authors: Zongjie Li, Alan Z. W, John Nicolas J, Walter H. F, Scott Donald L, Gordon Y. P, Deke X Published: 2026-09-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.08418v1 Training capable cyber agents is often treated primarily as a problem of model scale, yet open-weight post-training is constrained more directly by the cost of executable environments, reliable multi-turn supervision, and access to strong teachers. We present a data-centric framework that addresses these bottlenecks through five complementary systems: Choulea analyzes hidden reasoning signatures, SkyReal reduces teacher-sampling cost, Hongzwang bypasses API restrictions on teacher execution, PSBreakup restores capabilities weakened by model merging, and Kreator converts expert interventions into trainable reasoning. Our data engine constructs resettable coding, vulnerability, CTF, kernel-history, full-exploit, firmware, and device-backed environments. Candidate trajectories are retained only after execution verification and evidence auditing, yielding 164,269 trajectories for long-context supervised fine-tuning. The three checkpoints improve over their starting models by an average of 23.76% on the full CyberGym suite and 10.49% across the pooled CTF suites. As of September 1, 2026, Feyospace-s1 achieves a verified success rate of 63.24% and ranks 10th on the official CyberGym leaderboard, while all three checkpoints rank 1st among models at comparable parameter scales. To our knowledge, this is the first end-to-end demonstration that a seven-person independent team can train open-weight models with leading agentic cyber capability. --- ### Stochastically Perturbed Weights: Ensembles from Deterministic Machine-Learning Weather Models Authors: Simon Adamov, Oliver Fuhrer, Reto Knutti, Sebastian Schemm Published: 2026-09-08 Categories: cs.LG, physics.ao-ph arXiv: https://arxiv.org/abs/2609.08412v1 Machine-learning weather models (MLWMs) now match or outperform operational numerical weather prediction (NWP) at global medium-range forecasting, at far lower inference cost. Many deployed MLWMs are deterministic, producing a single forecast with no estimate of its own uncertainty, whereas a growing family of trained-probabilistic models generate calibrated ensembles directly, at the price of a dedicated training run. We ask instead how much uncertainty can be extracted from a deterministic checkpoint that already exists, without retraining it. Where physical ensembles represent model uncertainty by stochastically perturbing parametrisation tendencies, we perturb the network's raw weight tensors at inference time, a scheme we call stochastically perturbed weights (SPW). We also ask whether it works, where and on which scales to inject the noise, and where it fails. A three-phase ablation across four deterministic backbones, Aurora, GraphCast, SFNO, and AIFS, selects one production baseline per model, benchmarked against the trained-probabilistic AIFS-ENS, FourCastNet 3 and Atlas as well as the operational ECMWF ensemble (IFS-ENS) over 112 initialisation times. At a 240 h (10-day) lead time the SPW ensembles reach continuous ranked probability skill scores (CRPSS) between 0.04 and 0.13 below the best trained-probabilistic baseline, at zero marginal training cost. No injection site works across models: the productive tensor group is architecture-specific, so SPW is at present a tuning procedure rather than a plug-and-play recipe. Its main failure mode is a coherent whole-field offset that overdisperses the domain mean, and restricting the noise to coarse scales or perturbing the initial conditions each repair part of it. --- ### Compositional Multilingual and Behavioral Attribute Steering Authors: Hyun Gu Kang, Daniil Gurgurov, Tanja Baeumel, Josef van Genabith, Simon Ostermann Published: 2026-09-08 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.08410v1 This study examines the compositionality of steering vectors for language and behavioral control in large language models. Focusing on language, jailbreak, and conciseness, we investigate whether additive, training-free composition of attribute steering vectors can preserve the intended steering effect of each attribute, across four instruction-tuned models from two model families and two size scales. We find that single-attribute steering is reliable for all three attributes, but only within an appropriate combination of intervention layer and steering strength, with abstract behaviors (jailbreak, conciseness) favoring middle layers and language favoring earlier layers. We show that additive composition of two attribute vectors succeeds in steering both attributes simultaneously when each is injected at its own best-performing layer, and that this partially extends to three simultaneously composed attributes, addressing an inconsistency left open by prior work on training-free composition. We further analyze the geometric properties of these steering vectors, finding that they are approximately orthogonal in the residual stream, consistent with their compositional behavior. --- ### FastE: Readout-Triggered Token Compression for LLM Embedding Inference Authors: Jinsong Shu, Jinyong Wen, Baokun Wang, Zhongle Xie, Lidan Shou, Weiqiang Wang, Gang Chen Published: 2026-09-08 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.08407v3 In this study, we identify depth-dependent prefix redundancy in final-readout LLM embedding models, notably across representative backbones including Qwen3-Embedding and Qwen3-VL-Embedding. We find that removing prefix states is substantially more damaging in shallow layers than at greater depth, showing that prefix states become increasingly compressible as the prefix and readout states propagate through the network. To this end, we introduce FastE, a training-free, plug-and-play method. FastE uses a shared fixed threshold on batch-mean readout-prefix alignment as a lightweight online heuristic for selecting when compression occurs, and ranks prefix states by the attention scores they receive from the readout position to determine which states are retained in subsequent layers. Our evaluations demonstrate FastE's ability to substantially reduce computational costs: on NarrativeQA with Qwen3-Embedding-0.6B, it reduces decoder-backbone FLOPs by 40.11% while retaining 99.53% of Full Forward nDCG@10. Across five text embedding benchmarks, two backbone scales, and three cross-modal retrieval tasks, the quality-efficiency trade-off is directly customizable through the maximum removal ratio without retraining. We believe FastE offers practical value for scalable embedding generation in retrieval, indexing, clustering, and multimodal representation systems. --- ### Environments as Scaffold: Enriching Feedback to Bootstrap Self-Evolving Agents in Long-Horizon Tasks Authors: Hongbang Yuan, Zhuoran Jin, Yixin Cao Published: 2026-09-08 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.08404v1 Large Language Models demonstrate remarkable proficiency in static reasoning, yet training them as autonomous agents through Reinforcement Learning (RL) for long-horizon tasks is often hindered by severe reward sparsity. While conventional \textit{agent-side warming} up via supervised fine-tuning (SFT) can alleviate this, it is frequently limited by data scarcity and constrained exploration. To address this, we propose a paradigm shift to \textit{environment-side adaptation} by constructing \textbf{F}eedback-\textbf{E}nriched \textbf{E}nvironments (\textbf{FEEs}). Through a pilot study, we establish a feedback design strategy that reformulates environments by transitioning from action guidance to observation enrichment during the later stages of both intra-episode exploration and inter-episode evolution. Large-scale experiments on SciWorld and BFCL benchmarks using various Qwen3 model scales and RL algorithms such as GRPO, GSPO, and DAPO demonstrate that FEEs consistently yield performance improvements over standard settings. Furthermore, our analysis reveals that training with FEEs \textbf{(1)} stabilizes training dynamics by reducing entropy volatility, \textbf{(2)} facilitates proactive state-space exploration in difficult tasks, \textbf{(3) }ensures the internalization of environmental guidance into policy weights rather than acting as a mere inference-time prior, and \textbf{(4) }identifies intra-group feedback consistency as a critical boundary for stable optimization. --- ### Towards Embodied Air-Ground Cooperative Object Search: Benchmark, Dataset and Agentic Method Authors: Boao Yu, Zimo Chen, Junreng Rao, Yue Hu, Zhengqiu Zhu, Yong Zhao, Rusheng Ju Published: 2026-09-08 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.08402v1 Air-Ground Object Search (AGOS) in urban environments is a challenging embodied task, which requires an Unmanned Aerial Vehicle (UAV) and an Unmanned Ground Vehicle (UGV) to jointly search for and verify a specified target vehicle from multi-view visual references. To study this underexplored problem, we introduce AGOS-Bench, the first dedicated benchmark for evaluating whether general-purpose Vision-Language Models (VLMs) can integrate aerial discoveries and ground-level verification through UAV-UGV cooperation. We further provide AGOS-Dataset as the companion resource of exemplary trajectories constructed by an automatic pipeline. It consists of 7.7k episodes for searching objects of diverse categories and attributes, spanning three difficulty levels. To address the AGOS task, we propose AGOS-Agent, a training-free and tool-augmented approach. The agentic method relieves VLMs from complex and dynamic coordination via a deliberate search-handoff-verify cooperation protocol, only demanding VLMs for scene understanding and decision-making. Extensive experiments on nine VLMs show that AGOS-Agent improves overall success rate for eight of the nine evaluated backbones while reducing decision steps for all nine. On the hard split, the SR and SPL of Gemini-3.6-Flash increase from 8.6% to 55.7% and from 7.6% to 44.0%, respectively. --- ### MLIP Detective: Active Failure Mode Discovery Beyond Benchmark Scores for Machine-Learning Interatomic Potentials Authors: Ryuhei Okuno, Nontawat Charoenphakdee, Kaoru Hisama, Yuta Tsuboi Published: 2026-09-08 Categories: cs.LG, cond-mat.mtrl-sci, physics.chem-ph arXiv: https://arxiv.org/abs/2609.08399v1 Universal machine-learning interatomic potentials (u-MLIPs) aim to generalize across diverse configurations. Benchmarks enable reproducible evaluation but may not expose failures outside their predefined scope. Here, we show that physics-informed search can complement benchmark-based evaluation by uncovering hidden failure modes. We introduce MLIP Detective, an agentic framework for active failure mode discovery. Starting from benchmark evidence, MLIP Detective generates falsifiable, physics-informed failure hypotheses, screens them with inexpensive simulations, and escalates only the most suspicious cases to human experts together with proposed verification protocols. Without issue-specific prompting, MLIP Detective identified and characterized a systematic anomaly in MACE-MPA-0: the model predicted some relaxed adsorbate-surface systems involving O- or F-containing adsorbates to be higher in energy than their corresponding separated fragments. Using cross-model comparisons, MLIP Detective further inferred a likely training-data origin for the anomaly, consistent with recent reports. --- ### From Coordinates to Candidate Regions: Temporal Change Localization via Region Selection in Remote Sensing Multimodal LLMs Authors: Juwan Chung, Sungjune Park, Yeongyun Kim, Yong Man Ro Published: 2026-09-08 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2609.08391v1 Remote sensing multimodal large language models (RS-MLLMs) have advanced scene understanding and visual question answering over satellite imagery, yet localizing specific objects or changed regions remains challenging. Existing approaches rely on generating bounding box coordinates as token sequences, which is fragile for the small, densely packed objects common in remote sensing and increasingly error-prone when multiple targets must be localized simultaneously. In this work, we present an RS-specific formulation of the region selection paradigm, previously explored in natural-image MLLMs, and extend it to temporal change localization over multi-image sequences. Our framework employs a text-conditioned region proposal module, encodes each candidate as special tokens carrying per-frame visual features enriched with spatial and temporal cues, and lets the LLM localize targets by selecting region tokens in its response. We construct a multi-task training and evaluation suite spanning localization, referring expression, visual grounding, and understanding tasks across single-image and multi-temporal settings. Experiments show that our approach substantially outperforms coordinate-generation baselines on temporal change localization, while improving single-image visual grounding and maintaining competitive understanding performance. Oracle analysis decomposes the contributions of the region proposer and the LLM selector, providing diagnostic insight unique to this framework. Our code will be available at https://github.com/juwan-kr/RS-RegionSelect. --- ### Noise Adaptive Streaming Audio-Visual Speech Token Enhancement for Robust Full-Duplex Spoken Dialogue Models Authors: Bella Godiva, Yeonju Kim, Yong Man Ro Published: 2026-09-08 Categories: cs.SD, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2609.08390v1 Full-duplex spoken dialogue systems enable simultaneous listening and speaking, but their audio-only perception often fails under background noise and overlapping speech, leading to incoherent responses. Recent audio-visual dialogue approaches show that incorporating visual cues such as lip movements improve robustness under audio corruption. However, existing approaches often adapt the large speech dialogue model itself to process visual input, requiring costly multimodal training. We propose AV-STE, a modular streaming audio-visual front-end that restores corrupted semantic speech tokens from noisy audio and lip video before they reach the speech LLM. The downstream dialogue model remains entirely frozen, preserving its pretrained conversational capabilities. When integrated with frozen Moshi, AV-STE improves average GPT-4o-judged response coherence from 1.42 to 1.91 under same-dataset speaker interference while largely preserving turn-taking behavior. Gains also transfer to out-of-domain Seamless Interaction. --- ### GALoc: Gravity Aligned Wireframes for Depth-Free Monocular Floorplan Localization Authors: Jeahn Han, Minji Kim, Jeongbin Sohn, Jonghyeok Park, Matthias Wuest, Pyojin Kim Published: 2026-09-08 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2609.08385v1 Floorplans are compact, appearance-invariant maps ideal for indoor localization, yet existing methods rely on depth networks that are brittle in cluttered scenes. We propose GALoc, a geometry-first framework that replaces depth prediction with gravity-aligned wireframes that satisfy verticality and coplanarity by construction. Given monocular RGB, camera intrinsics, relative poses, and IMU orientation, GALoc constructs a linear constraint matrix encoding verticality and coplanarity, and finds the camera gauge minimizing its smallest singular value via global search. The rectified wireframes are projected into bird's-eye-view layouts through a closed-form, FOV-consistent transformation and matched against the floorplan via metric-free SE(2) search. We evaluate end-to-end on Structured3D, with calibrated noise on Gibson, and on real-world author-collected sequences. When sufficient wall geometry is visible, GALoc matches or outperforms depth-based baselines -- achieving 88% sequential localization success at 0.1m over 100-step sequences on Gibson vs the baseline's 68% -- while abstaining in structure-blind scenes. --- ### Equivariance Breaks the Learning Rate Authors: Andrei Manolache, Mathias Niepert Published: 2026-09-08 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.08381v1 Equivariant networks are commonly trained with Adam, yet recent work reports that matrix-structured optimizers such as Muon can perform better on these architectures without explaining why. We identify one source of this difference inside equivariant linear layers. Each irrep block learns a channel-mixing matrix $W_l$ shared across its $2l+1$ components, giving the expanded map $W_l \otimes I_{2l+1}$. For a single application of the layer, the gradient of $W_l$ sums $2l+1$ outer product contributions and has rank at most $2l+1$. Adam rescales stored weights individually without using the irrep boundaries, so one learning rate can produce different spectral step sizes across blocks within a layer. We address this mismatch by normalizing each block update separately, without introducing a new hyperparameter. This changes only the scale of the update, leaving Adam's moment estimates and its direction within each block unchanged. We evaluate the mechanism in a controlled $\mathrm{SO}(3)$-equivariant model with a matched dense control and in an e3nn interatomic potential model trained on rMD17 and MD22. The toy setup isolates a mismatch that grows with width while the dense control shows no corresponding growth. In the interatomic potential model, block normalization and tuning Adam's momentum coefficients independently improve performance, but neither alone matches Muon. Combined, they make Adam competitive with Muon on all datasets, indicating that blockwise step control and momentum accumulation account for much of Muon's advantage. --- ### How to Make the Gradient Mapping Small for Constrained Stochastic Min-Max Problems and Beyond Authors: Ahmet Alacaoglu Published: 2026-09-08 Categories: math.OC, cs.LG arXiv: https://arxiv.org/abs/2609.08380v1 We study the stochastic first-order oracle complexity for constrained or regularized convex-concave min-max optimization and stochastic monotone variational inequalities. We focus on the case when suboptimality is measured in terms of the gradient mapping, also known as, forward-backward or natural residual, an optimality notion that generalizes the gradient norm for unconstrained problems. In this setting, under standard unbiased oracle access with now-standard variance assumptions, the best-known complexity for making the norm of the gradient mapping less than $\varepsilon$ is $\widetilde{O}(\varepsilon^{-4})$, compared to the near-optimal $\widetilde{O}(\varepsilon^{-2})$ that is established in the unconstrained case. We bridge this gap to improve the gradient mapping complexity for constrained convex-concave min-max problems to $\widetilde{O}(\varepsilon^{-2})$. We then extend to prove the same complexity for problems without the bounded variance, by using the Blum-Gladyshev assumption. --- ### Geographically Regularized AUC-Maximizing Personalized Federated Learning Authors: Mayu Hiraishi, Kensuke Tanioka, Toshio Shimokawa Published: 2026-09-08 Categories: cs.LG, stat.ME arXiv: https://arxiv.org/abs/2609.08379v1 Accurate diagnostic and risk-prediction models are important for supporting clinical decision-making during infectious disease outbreaks. However, privacy and governance requirements may restrict patient-level data sharing across healthcare institutions, and data distributions often vary. Moreover, AUC is widely used to evaluate discriminative performance, motivating its direct optimization in model development. We propose geographically regularized AUC-maximizing personalized federated learning (GrAUC-PFL), which directly optimizes a smooth pairwise AUC surrogate to learn personalized models while keeping patient-level data local and accounting for institutional heterogeneity. Graph-based regularization encourages geographically neighboring institutions to have similar coefficient vectors while retaining a personalized models. Simulations and a real-data application suggest improved discriminative performance, particularly when geographically neighboring institutions have similar data-generating characteristics. ---