# Research Papers — 2026-09 · bundle page 154 of 266 > Full-text markdown of papers #7651–7700 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/154 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/153.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/155.md) ### Mixed-Integer Nonlinear Differentiable Predictive Control for Underground Pumped Hydro Energy Storage Systems Authors: Honghui Zheng, Ján Boldocký, Yury Dvorkin, Ján Drgoňa Published: 2026-09-16 Categories: eess.SY, cs.LG arXiv: https://arxiv.org/abs/2609.17964v1 This paper extends Mixed-Integer Differentiable Predictive Control (MI-DPC) to multi-modal discrete decisions and nonconvex polynomial dynamics arising in Underground Pumped Hydro Energy Storage Systems (UPHES). A neural policy mapping problem parameters to continuous setpoints and integer mode selections via a Gumbel-Softmax layer is trained in a self-supervised manner by differentiating the expectation of the finite horizon control objective through the nonlinear dynamics model. Three methodological contributions enable this extension: a parallel differentiable simulator that preserves gradient magnitude, a Transformer encoder that captures long-range temporal dependencies, and a Gumbel-Softmax temperature annealing schedule that regularizes the combinatorial search. We demonstrate the framework on day-ahead scheduling of a UPHES, a large-scale mixed-integer optimal control problem with nonlinear unit performance curves and volume-head coupling. MI-DPC achieves only 1.6% suboptimality relative to a piecewise mixed-integer quadratic programming baseline, while providing five orders of magnitude speedup in online scheduling time. --- ### TACTICS: Taxonomy-Aware Intelligent Corpus Sampling for Machine Translation Authors: Prasanth Bathala, Anubhav Shrimal, Sukhdeep Singh Kharbhanda, Pradyumna Lanka, Rohit Dhaipule Published: 2026-09-16 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.17956v1 Large-scale machine-translation (MT) systems are typically evaluated on random samples from a corpus whose distributional composition is an artifact of how it was assembled. Such a sample inherits the phenomena the collection happens to contain rather than the full space a system must handle, spanning rule-governed conventions (terminology, punctuation, currency formatting) and context-dependent phenomena (tone, honorifics, document-level coherence), and thus provides no coverage guarantee for assessing robustness. We propose TACTICS (Taxonomy-Aware Coverage-opTimized Intelligent Corpus Sampling), which recasts coverage as an explicit objective. TACTICS induces a hierarchical taxonomy from a locale style guide, classifies segments against it, and selects a fixed-budget subset jointly optimizing coverage of rare categories, document-level coherence, and distributional fidelity to the full corpus. Applied to MT evaluation across four translation directions, TACTICS improves coverage of rare categories over lexical and embedding-based selection. By targeting the phenomena that separate systems, TACTICS makes a fixed evaluation budget go further, recovering the true system ranking from far fewer segments than random sampling wherever a real quality gap exists and never signaling a difference where none exists. --- ### EDCT-Bench: Uncovering Faithfulness Gaps in VLMs via Explanation-Driven Counterfactual Testing Authors: Sihao Ding, Santosh Vasa, Aditi Ramadwar, Thomas Monninger Published: 2026-09-16 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.17953v1 Vision-Language Models (VLMs) can produce Natural Language Explanations (NLEs) that sound plausible yet remain inconsistent with the visual evidence they cite. We present Explanation-Driven Counterfactual Testing (EDCT), an intervention-based protocol that extracts visual concepts cited in a model's explanation, applies verified minimal edits to them, and tests whether the resulting answer and explanation remain consistent with the edited image. Using this protocol, we create EDCT-Bench, a comprehensive benchmark spanning three complementary domains: knowledge-intensive visual question answering (OK-VQA), safety-critical driving (DriveLM), and 3D spatial reasoning (3DSRBench). Across the evaluated VLMs, EDCT reveals substantial faithfulness gaps, with models frequently producing responses inconsistent with verified visual changes. Finally, our fine-tuning study suggests that EDCT-generated counterfactuals provide high-impact training signals. --- ### Maximum Strong Independent Sets in Hypergraphs: Reductions, Bounds, and Greedy Certificates Authors: Yingquan, Wu, Jason Cong Published: 2026-09-16 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.17951v1 We study the maximum strong independent set problem in a finite hypergraph: find the largest vertex set that intersects every hyperedge in at most one vertex. This objective arises whenever each observed block is a local incompatibility constraint but transitive closure across overlapping blocks is not justified. A motivating example is multi-band LSH-MinHash deduplication, where each collision bucket gives local evidence, while connected-component contraction can impose spurious global equivalences. The paper develops an incidence-structural toolkit for this problem. We prove exact reductions for dominance, incidence twins, and weight-1 blocks; derive closed-form and low-weight upper bounds; introduce puncturing and covering certificates that sharpen those bounds; and analyze a layered greedy clustering algorithm driven by block weights and residual incidence. The algorithmic analysis includes feasibility, maximality, conditional optimality, a layered witness-matching upper bound, and incidence-local complexity bounds. The results give correctness, termination, fixed-point, and optimality certificates for broad incidence families, together with examples showing when different certificates separate or coincide. --- ### ASPIRE: Asynchronous Batched Self-Speculative Decoding for Long-Context LLM Inference Authors: Amir Ziashahabi, Hossein Entezari Zarch, Lei Gao, Murali Annavaram, Salman Avestimehr Published: 2026-09-16 Categories: cs.LG, cs.CL, cs.DC arXiv: https://arxiv.org/abs/2609.17943v1 Long-context LLM inference is bottlenecked by attention, whose repeated KV-cache reads make decoding memory-bound. Self-speculative decoding alleviates this by drafting tokens with sparse attention and verifying them with full attention, but existing batched methods remain synchronized: all requests in a batch share a single draft-verify schedule, even though the optimal draft length varies widely across requests and changes dynamically within each request. We propose ASPIRE, a non-synchronized batched self-speculative decoding framework built on three components. First, a unified mixed forward allows drafting and verifying requests to coexist in the same batched forward pass, removing the need for global draft-verify phases. Second, a lightweight online speculation scheduler uses per-request acceptance-rate estimates and a batch-aware cost model to let each request independently choose when to verify. Third, an intra-draft refresh layer performs full attention at a single designated layer during drafting, updating the sparse context at every draft step to reduce staleness during drafting. Across three models and five reasoning and long-context benchmarks, ASPIRE achieves $1.70$-$4.58\times$ speedup in decoding throughput over autoregressive baselines and improves average speedup by approximately $27\%$ over the strongest prior self-speculative baselines. --- ### On the Identifiability of Mixed Ordinal and Exponential Family Causal DAGs under Linear Parametric Models Authors: Sambit Mishra, Urbashi Mitra Published: 2026-09-16 Categories: cs.LG, math.ST, stat.ME, stat.ML arXiv: https://arxiv.org/abs/2609.17942v1 The problem of identifiability in linear parametric models (LPMs) whose nodes follow either an ordered logit model or a regular one-parameter exponential family is evaluated. The results go beyond classical structural equation models as well as results for nodes with observations from a homogeneous family of distributions. The main result establishes that the orientation of every edge joining an ordinal node to an exponential-family node is identifiable from the joint distribution alone at every parameter value, provided the ordinal node has at least three categories and the exponential-family node at least three points of support, with no restriction on the sufficient statistic. Converses show that both requirements are necessary: the three-category requirement is binding only for affine sufficient statistics, and the three-point requirement is binding under the canonical link. The guarantee extends to orienting every such mixed ordinal-exponential family edge of a given $d$-node undirected skeleton. Numerical experiments illustrate the theoretical results by successfully separating orientations within a Markov equivalence class, which are indistinguishable by conditional independence alone. --- ### Unsupervised spiking feature learning for event-based pedestrian crossing detection: approaching supervised accuracy without labelled training data Authors: Henok Teklu, Mustafa Sakhai, Matej Mertik, Maciej Wielgosz Published: 2026-09-15 Categories: cs.CV, cs.AI, cs.NE arXiv: https://arxiv.org/abs/2609.31671v1 Event cameras are well suited to pedestrian crossing detection, and spiking neural networks (SNNs) can process their output natively, but current SNN detectors are trained with supervised backpropagation and therefore require costly frame-level crossing labels. We investigate crossing detection with no labels in feature learning and report the first unsupervised results on the recent DVS-PedX pedestrian-crossing benchmark. A single spiking layer trained with winner-take-all spike-timing-dependent plasticity learns a dictionary from unlabelled event-frame patches; frames are encoded by cosine similarity to the learned filters with spatial pooling and read out by a linear classifier, the only supervised component. On the 24,454-frame test split the method attains 90.3% accuracy and an area under the receiver operating characteristic curve (AUROC) of 0.936, compared with 92.0% and 0.943 for a supervised spiking network trained end-to-end on the same frames; under adverse weather the AUROC is 0.913. The result is insensitive to the choice of plasticity rule but depends strongly on the readout protocol: with the classical neuron-assignment readout the same network attains only 0.70 AUROC. On the benchmark's real converted portion, a readout refit lifts performance from chance (0.53) to 0.67 AUROC, within the published supervised range. These findings indicate that the accuracy cost of removing labels from feature learning is small on this benchmark, and that reported weaknesses of unsupervised spiking networks may be attributable to the readout protocol rather than to the learning rule. --- ### One-Step Is Optimal: Unconditional Rectified Flows are Noise2Noise Denoisers, and Multi-Step Integration Provably Hurts---A Benchmark and Task-Based Detectability Study on Low-Dose CT Authors: Timothy Sereda, Debesh Jha Published: 2026-09-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.31670v1 Iterative and generative denoisers are increasingly used under the assumption that multi-step refinement outperforms a single regression pass. We show the opposite for \emph{label-free} denoising. An \emph{unconditional} rectified flow trained on two noisy observations of the same signal, as in Noise2Noise, has a minimiser whose one-step readout is exactly the MMSE denoiser without requiring clean targets. In contrast, multi-step integration provably departs from the MMSE solution because the flow terminates at the noisy data distribution rather than the clean-signal distribution. This departure is exact in a tractable Gaussian model and is confirmed experimentally: one-step flow matches a direct regressor, whereas multi-step Euler integration progressively reduces fidelity. Counterintuitively, the degradation increases with training quality, as a better velocity field more faithfully transports samples toward the noisy terminal law. The key ingredient is therefore the decorrelated \emph{pairing}, not the flow machinery: a one-step regressor trained on matched noisy pairs gives our best label-free result ($+1.99$,dB). We evaluate these findings on \textbf{CTDenoiser}, a controlled low-dose CT benchmark spanning five architectures and supervised, similarity-based, blind-spot, and per-image methods. Among label-free approaches, only correlated-noise-aware Noise2Sim improves over the noisy baseline, while Noise2Void is flat-to-negative because CT noise violates its pixel-independence assumption. Finally, although supervised denoisers gain approximately $4$,dB PSNR, a channelized Hotelling observer shows reduced low-contrast lesion detectability, revealing clinically relevant degradation missed by PSNR and SSIM. --- ### NanoForecast v0.5: Competitive Time Series Forecasting Through Training Pipeline Optimization Authors: Gautam Kishore Published: 2026-09-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.31669v1 We present NanoForecast v0.5, a 6.5M-parameter forecaster that competes with models 31x its size (TimesFM, 200M parameters) after training pipeline fixes and no architecture change. Retraining the v0.3 architecture with corrected loss-scope handling, tensor shape alignment, and wider augmentation coverage cuts overall Mean Absolute Scaled Error by 43.8% under one fixed protocol (MASE 3.030 to 1.704) on the same data and compute budget. NanoForecast v0.5 beats TimesFM on all three ETT datasets (MASE 0.676/1.110/0.287 vs. 0.705/1.360/0.545) and on exchange rate (4.317 vs. 4.383); TimesFM keeps a clear lead on the high-cardinality electricity and traffic sets. Against PatchTST (15M+ parameters, official configuration), v0.5 wins all three ETT sets. Training takes about 12 hours on a single cloud GPU (NVIDIA T4, Google Colab) and inference needs no GPU (measurements in this paper are on an Apple M4 CPU). We release all code, pretrained checkpoints, and evaluation framework under Apache 2.0 at https://github.com/eulogik/NanoForecast --- ### Query-aligned video frame selection for long video understanding Authors: Md. Safayet Islam, Dilip Sarkar, Liang Liang Published: 2026-09-15 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.31668v1 Multimodal large language models (MLLMs) process multimodal inputs by converting text, images, and videos into token sequences that are subsequently processed by a backbone language model. While MLLMs have achieved excellent performance in understanding the content of individual images, video understanding remains significantly more difficult because videos contain large number of video frames. MLLMs typically process only a subset of these frames, usually ranging from 8 to 64. MLLMs usually sample frames uniformly, regardless of their relevance to the question being answered. To address this limitation, several training-free, model-agnostic methods for selecting question-relevant frames have recently been proposed. In this work, we introduce a frame-selection method designed specifically for multiple-choice questions. We extend the query text by appending semantic cues derived from the answer choices and employ a direct query-frame alignment scoring mechanism. To the best of our knowledge, our method is the first to directly utilize answer choices as inference-time cues for selecting frames relevant to answering a question. The method first constructs a compact candidate pool by subsampling video frames at a fixed rate. The frames are then scored according to their maximum cosine similarity across all question-answer pairs to identify the most relevant frames for a given query. This approach preserves a fixed token budget while improving the relevance of the visual evidence provided to the downstream MLLM. We evaluate the effectiveness of our frame-selection method on the MLVU, Video-MME, and LongVideoBench benchmarks using three MLLMs: LLaVA-Mini, Qwen2-VL, and LLaVA-Video. Experimental results demonstrate that answer-aware frame selection generally outperforms uniform sampling and existing training-free frame-selection methods under the same frame budget. --- ### Age-Adaptive Handwriting Reconstruction from an IMU-Based Digital Pen through Shared Representations and Domain-Specific Heads Authors: Florent Imbert, Yann Soullard, Eric Anquetil, Hui Han Published: 2026-09-15 Categories: cs.CL, cs.AI, cs.CV, cs.LG, cs.NE arXiv: https://arxiv.org/abs/2609.31666v1 Digital pens are widely used to capture handwriting on digital devices, enabling precise trace recording and enhancing human-computer interaction. However, most are bundled with tablets and lack cross-brand compatibility. Recent digital pens equipped with kinematic sensors have emerged, designed for use on any surface. This especially opens significant potential for supporting handwriting acquisition in classrooms. Handwriting reconstruction from such an IMU-equipped pen poses a challenge due to the significant variability in sensor signals between adults and children. Even when producing visually similar traces, variations in writing dynamics, motor control, pen holding, and user confidence introduce substantial discrepancies in the captured signals. Additionally, the high variability in children's handwriting requires collecting large amounts of data, which is not feasible to implement in a school environment at scale. Furthermore, models trained exclusively on adult data fail to generalize to children's handwriting, and conversely, models trained on children's data perform poorly on adult writers. This crosspopulation degradation highlights the need for a unified model that can be deployed directly on the pen, without any user-specific adaptation. To address this issue, we propose a cross-domain learning, using an original neural network architecture based on a Temporal Convolutional Network and multiple prediction heads. The model is designed to be robust across age groups by leveraging shared features while effectively handling variability induced by differences in graphomotor development. This approach aims to improve handwriting trace reconstruction from sensor data, where each domain benefits from additional data provided by the other domain. --- ### Language-Augmented Video Action Anticipation: Design Fundamentals, Benchmarks, and Open Challenges Authors: Mahsa Mohammadi, Zeyu Fu, Sareh Rowlands Published: 2026-09-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.31665v1 Action anticipation predicts future human actions from partial video under incomplete context and temporal uncertainty. Recent systems introduce large language models (LLMs), vision-language models (VLMs), or language-derived semantics at different stages, but reported gains are difficult to interpret when task formulation, visual pretraining, supervision, decoder design, and evaluation code change simultaneously. The central contribution of this review is an evidence-aware design map that crosses task regime with the point at which language-derived information intervenes. We characterise task regimes along six axes. These axes organise the literature into five broad task families: single-action, sequence, object-interaction, cross-view, and planning-oriented settings. C1-C3 locate interventions in context construction, goal/intention modelling, and future decoding, while C4 is treated as an adjacent, emerging grounding/executability extension. Unlike a generic processing pipeline, the map links each intervention to an appropriate counterfactual, failure diagnosis, and permissible evidence claim. Supporting contributions include a protocol-level audit of Ego4D-LTA and EPIC-KITCHENS-100, a multidimensional evidence profile, and the Backbone-Aware Comparison and Ablation Protocol (BCAP). The unresolved EK-100 record is treated as a reporting-comparability case study and is not used as a leaderboard. Evidence for LLM benefits, goal ambiguity, and horizon effects is therefore formulated as testable hypotheses requiring matched validation, not as causal conclusions. The accompanying package contains the coded evidence, source locators, protocol metadata, and versioned catalogue used in the review. --- ### What Drives Dialectal Jailbreaks? An Ablation of Surface Form, Cultural Framing, and Strategy Banks Authors: Qingyang Xu Published: 2026-09-15 Categories: cs.CL, cs.AI, cs.CR arXiv: https://arxiv.org/abs/2609.31664v1 Recent work suggests that obscure language registers can weaken large language model refusal behavior, especially when paired with black-box prompt optimization. It remains unclear whether failures stem from non-standard surface form, culturally grounded framing, or optimization over an expressive prompt-strategy space. We study Chinese registers by extending a classical-Chinese red-teaming framework to Shanghainese and Cantonese and running a 36-cell ablation across surface forms, strategy-bank variants, and two target models. The main finding is corrective: dialectal surface form is neither necessary nor sufficient for high attack success. Non-optimized English, Mandarin, and naive dialect translations remain below 8\% attack success rate, whereas all conditions that retain an optimizer-controlled strategy bank reach 98--100\%. A culture-neutral generic strategy bank reaches the same ceiling at near-single-query cost, further indicating that strategy-bank expressiveness, rather than dialectal or cultural content, accounts for most of the observed effect. Dialect choice still affects query efficiency and response severity, and qualitative coding identifies recurring high-level mechanisms such as semantic glossing and procedural scaffolding. We qualify the absolute ceiling-level attack success rate by showing sensitivity to judge calibration and noting that the design does not fully separate one-shot strategy construction from iterative search. --- ### LLM-Guided Ontology-Driven Knowledge Graph Construction from Unstructured Text Authors: Abdelhadi Belfadel, Maxence Gagnant, Joseph Kattan, Sana Tmar Published: 2026-09-15 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.31663v1 Ontology-driven knowledge graph construction from industrial text remains challenging due to the domain specificity of documents, the scarcity of annotated resources, and the complexity of ontology engineering workflows. This paper presents and investigates the applicability of an ontology learning pipeline that combines compact open-source Large Language Models (LLMs), reusable prompting strategies, and open knowledge bases to support the extraction, structuring, enrichment, and evaluation of knowledge from textual corpora. The approach is tested and evaluated on a private French corpus of power-grid incident reports, using locally deployable open-source LLMs ranging from 7B to 32B parameters. Starting from unstructured reports, the approach extracts entities and relations, generates RDF triples, constructs related OWL ontology, enriches it using external knowledge sources, assesses the quality of the ontology, and subsequently constructs a populated knowledge graph grounded in the resulting ontology schema. Experiments on 80 manually annotated private reports show that schema-guided prompting significantly improves extraction quality, while quantized models provide an effective trade-off between performance and computational cost. These results demonstrate the feasibility of transforming domain-specific industrial text into ontology-based knowledge graphs using locally deployed open-source LLMs, while supporting the generalization of the extraction process through reusable prompting strategies. --- ### Learning Steadily: Accumulating Relative Point Margin Scores for Face Image Quality Assessment Authors: Guray Ozgur, Tahar Chettaoui, Eduarda Caldeira, Marco Huber, Jan Niklas Kolf, Naser Damer, Fadi Boutros Published: 2026-09-15 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.31662v1 Face Image Quality Assessment determines the suitability of captured face images for automated face recognition (FR), a critical capability for reliable biometric systems. Existing state-of-the-art FR-integrated FIQA methods suffer from temporal instability: as the feature space evolves during training, single-epoch quality estimates fluctuate, creating a moving target that undermines reliable quality prediction. We introduce CARPM-FIQA, a stabilization strategy for FR-integrated FIQA that accumulates relative point margin measurements, the ratio between intra-class compactness and inter-class separation, across the entire training trajectory rather than relying on single-epoch estimates. This cumulative averaging approach provides theoretically grounded advantages: reduced variance in quality estimates, improved mean squared error, and enhanced ranking stability with convergence guarantees as training progresses. Through controlled experiments on the SynFIQA dataset with labeled quality groups, we demonstrate that cumulative averaging achieves superior discriminative ability, and ablation studies across different training configurations confirm consistent improvements. Evaluated against twelve FIQA methods on eight challenging benchmarks with four FR models at two FMR thresholds, CARPM-FIQA places 4th (CARPM-FIQA(L)) and 6th (CARPM-FIQA(S)) of 17 compared methods by pAUC-EDC and AUC-EDC averaged across FR models and, after per-benchmark normalization, across benchmarks, staying within a few percent of the best method's normalized average for every FR model, providing a principled solution to training instability while maintaining the performance benefits of FR integration. More broadly, our work demonstrates that temporal aggregation strategies can stabilize training objectives in deep learning systems where target values inherently fluctuate due to evolving feature representations. --- ### ForensicZoom: Adaptive Visual Inspection with Multimodal LLMs for Industrial-Grade Face Forgery Detection Authors: Hang Zhou, Yiming Tang, Kun Yu, Qian Zhu, Minghao Li, Weigao Wen Published: 2026-09-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.31661v1 Reliable face forgery detection is critical to the security of online identity verification systems, where missed attacks compromise security and excessive false positives disrupt legitimate users. Specialized forensic detectors achieve strong detection performance but provide limited interpretability, while multimodal large language models (MLLMs) offer strong semantic understanding and interpretable reasoning yet remain substantially weaker for face forgery detection. We argue that a key limitation lies in how visual evidence is acquired: subtle forensic artifacts may be poorly represented at standard resolution, while uniformly processing all cases at higher resolution is computationally inefficient. We therefore introduce ForensicZoom, an industrial-grade MLLM framework for adaptive visual inspection. ForensicZoom first equips a general-purpose MLLM with forensic-aware visual representations and aligns the language model with these features. Its central mechanism, NEED_ZOOM, enables the model to autonomously request magnified views of suspicious regions when the initial evidence is insufficient, turning fixed-pass classification into adaptive multi-round forensic reasoning. The zoom behavior is learned through reward shaping that balances detection accuracy with unnecessary visual inspection, concentrating additional computation on difficult cases. A final attribution optimization stage improves natural-language forensic reports while preserving detection performance. On large-scale industrial identity verification data, ForensicZoom achieves over 97% TPR at 0.1% FPR, substantially outperforming both specialized detectors and existing MLLM-based methods while producing actionable forensic attributions. These results demonstrate that ForensicZoom can provide an effective path toward accurate, interpretable, and scalable MLLM-based face forgery detection. --- ### Parser, Chunking, and Embedding Interactions in Retrieval-Augmented Generation over Indian Government Regulatory Documents Authors: Shubham Kumar Singh Published: 2026-09-15 Categories: cs.CL, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2609.31660v1 Retrieval-augmented generation (RAG) pipelines are typically assembled from independently-chosen components -- a document parser, a chunking strategy, and an embedding model -- yet these choices are rarely evaluated jointly, and evaluations that do combine them are usually run on a single document or corpus. We present a controlled factorial study of 3 parsers, 3 chunking strategies, and 5 dense embedding models, together with a sparse BM25 baseline, evaluated against 800 question instances, each with one or more required evidence strings, with evidence strings automatically validated against source text and a 10% random sample manually reviewed, across four structurally distinct Indian central-government regulatory documents. We fit linear mixed-effects models with document-query-level random intercepts to the resulting 72,000-row result set, run Holm-corrected paired comparisons between matched dense and sparse configurations, and report clustered bootstrap confidence intervals for all 54 unique retriever configurations. We find that no single retriever family dominates across documents; parser and chunker choice interact significantly; MPNet-base is a consistent underperformer with a severe failure mode on table-derived questions; and the corpus exhibits a near-saturated evidence-preservation ceiling above 98%, indicating that retrieval differences are driven primarily by ranking quality rather than information loss during ingestion. We additionally report embedding-dimension and chunk-size/overlap ablations and an efficiency/quality Pareto analysis. We release our full evaluation harness, corpus manifest, and 800-question benchmark. --- ### Cross-Material Support Transfer for Core-Loss Prediction Under Waveform Covariate Shift Authors: Cong Yao, Chunye Gong Published: 2026-09-15 Categories: cs.LG, cond-mat.mtrl-sci arXiv: https://arxiv.org/abs/2609.31659v1 Power magnetic materials are characterized on the sinusoidal and triangular waveforms that excitation hardware conveniently produces, whereas deployed converters expose cores to trapezoidal, PWM-shaped flux trajectories, so loss models must predict exactly where their training data are thinnest. The final test of the MagNet Challenge embeds a deliberately extreme instance of this characterization-deployment mismatch: for material D, trapezoids form 16.4% of the test set but only 1.4% of the training set. The 95th-percentile relative error, hereafter p95, of the best submission, built on sequential transfer learning, stalled at 15.9%, the worst among the five materials. This paper shows that the obstacle is missing information under covariate shift rather than class imbalance, and that the missing support can be borrowed from sibling materials instead of being extrapolated. Controlled experiments first refute the imbalance reading: four standard remedies fail, and raising the trapezoidal share to the test-set level degrades accuracy further. The proposed material-identity support transfer, MIST, then trains one 2784-parameter predictor jointly on all five challenge materials. Material identity enters through feature-wise linear modulation, or FiLM, the scarce material's true-label loss is reweighted, and material D receives no fine-tuning, so that the bias of its trapezoid-free training set is never re-installed. MIST lowers the five-seed material-D p95 from 20.39+/-2.03% to 12.38+/-0.92% and the trapezoidal-class p95 from 37.4+/-8.8% to 15.16+/-1.69%, surpassing the best submission with one-sixth of its parameters and no fine-tuning stage; removing material identity at matched capacity inflates the error by an order of magnitude. These results argue that scarce materials should be characterized jointly with their siblings. --- ### Cross-Dataset Transfer and Unknown-Class Detection in Imbalanced SAR Ship Classification Authors: Ch Muhammad Awais, Marco Reggiannini, Davide Moroni, Giulio Del Corso Published: 2026-09-15 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.31658v1 Ship classification from Synthetic Aperture Radar (SAR) imagery is a critical computer vision task, yet the robustness of models under deployment shifts remains unclear. While models are often trained on one dataset and deployed on another, we lack a comprehensive understanding of their cross-dataset generalization. To address this, we evaluate six pretrained models on two SAR ship datasets in three settings: in-domain classification, cross-dataset transfer, and unknown-class detection. For unknown detection, we hold out all classes one at a time. In-domain, SARDet100K gives the best balanced accuracy on both datasets (73.4\% on FUSARShip and 53.2\% on OpenSARShip). In cross-dataset transfer, we observe strong failures: some models show moderate accuracy but near-chance balanced accuracy (for example, 64.5\% accuracy but 33.3\% balanced accuracy for OpenSARShip to FUSARShip). In unknown detection, performance depends on the held-out class and dataset, while MC-dropout variance is often close to random. These findings show that cross-dataset generalization in SAR remains limited and that task-specific uncertainty scores are often more informative than MC-dropout variance for held-out-class detection, although their relative ranking depends on the dataset and held-out class. --- ### Enhancing Foundation Models for Imbalanced SAR Ship Classification via Targeted Oversampling Authors: Ch Muhammad Awais, Marco Reggiannini, Davide Moroni Published: 2026-09-15 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.31657v1 Remote-sensing foundation models offer strong representations for SAR imagery, but their behavior under severe long-tail class imbalance is still not well characterized. We benchmark DOFA and SAR-JEPA on the imbalanced OpenSARShip dataset and compare them with ImageNet-pretrained baselines under a fixed, training-efficient protocol that keeps the backbone frozen. To mitigate imbalance without fine-tuning, we apply four oversampling methods in embedding space exclusively to minority classes and train a lightweight classifier head on the augmented embeddings. Across both foundation models, oversampling improves Macro-F1 and test accuracy relative to their respective baselines, with the largest Macro-F1 gains observed for DOFA using ADASYN (34.39 to 38.56) and for SAR-JEPA using SVM-SMOTE (25.89 to 32.30). We also report class-wise behavior, showing that aggregate improvements can coexist with persistent failures on specific rare classes. Code for embedding extraction and reproducible multi-seed evaluation is provided to support rapid experimentation on free-tier hardware. --- ### From Phase Transition to Systemic Failure: A Decoupled Analytics Framework for GNN Robustness Authors: Shuai Yan, Dan Peng, Jie Li, Ke Wang Published: 2026-09-15 Categories: cs.LG, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2609.31656v1 Data quality is a major bottleneck for the reliable deployment of graph neural networks (GNNs) in real-world graph mining tasks. Among various sources of degradation, label noise and feature distribution shift (hereafter referred to as distribution shift) are two common yet fundamentally different challenges. To study their effects under controlled conditions, this paper constructs a synthetic homophilic graph regression benchmark in which the two factors can be manipulated separately. A total of 41 configurations and 410 runs are conducted to evaluate the behavior of representative GNN models under varying noise and shift conditions. The results show two distinct patterns. First, under additive label corruption, performance remains relatively stable over a broad range of noise settings and begins to deteriorate sharply only after an observed transition region around the 50 percent noise ratio. Second, under extreme feature distribution shift, all tested models suffer substantial degradation, with test MSE increasing by 48 times to 316 times and correlation dropping by 73 percent to 89 percent. These findings suggest that, in the present controlled setting, GNNs are considerably more tolerant to moderate label perturbation than to severe distribution mismatch. The study provides a controlled empirical baseline for understanding how data quality affects GNN-based graph mining systems and offers practical implications for deployment-oriented monitoring and model maintenance. --- ### One Evaluation, Any Operating Point: Hypernetwork-Amortized MeanFlow for 3D MRI Reconstruction Authors: Ruibo Wang Published: 2026-09-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.31655v1 Generative priors reconstruct accelerated 3D MRI well but pay heavily at deployment: tens of network evaluations per volume, and protocol-specific hyperparameter tuning. A third hidden cost is the scanner's fixed sampling pattern. We treat the whole operating point as an input. A 3D MeanFlow patch network (a one-step flow model) is fine-tuned end-to-end through a warm-started, five-iteration differentiable conjugate-gradient projection. A small hypernetwork maps the operating point (data-consistency weight, acceleration, and the Cartesian sampling pattern itself) to the network's per-channel modulation. Three findings follow. (i) Learning the acquisition is worth more than any other operating point: on clinical knee data, the learned mask gains up to +2.34 dB over the protocol's variable-density mask. This gain requires the solver: with a feed-forward reconstructor the same learned mask hurts at 4x (-1.7 dB), but with the data-consistency projection it adds +4.6 dB. (ii) One evaluation is highly effective: it beats a 20-step patch-diffusion prior by up to +3.1 dB on brain and +2.9 dB on knee. Three to five evaluations extend the front to +6 dB while using a quarter of the prior's network calls. (iii) Fully sampled targets are optional: trained self-supervised on a split of acquired samples, the reconstructor matches its supervised twin at 4x on real data. Finally, we report what failed and why: subject-adaptive acquisition from measured energy, combining self-supervision with learned acquisition, and amortising the data-consistency weight. --- ### AI in Science: Early Insights Authors: Mihai Codreanu, Alex Imas, Juan Mateos-Garcia, Joseph Emmens, Evalyne Muiruri, Arthur Turrell, Julian Jacobs, Atoosa Kasirzadeh, Ana Trisovic, Yiyuan Chen, Tanya Rodchenko, Catherine Pollard et al. Published: 2026-09-15 Categories: econ.GN, cs.AI arXiv: https://arxiv.org/abs/2609.28504v2 Scientific progress is a key driver of economic growth and prosperity. There is great excitement - but also concerns - about the impacts of AI on science, but so far little data. We provide early insights on this from three data sources: a sample of 15 million Gemini interactions, an inventory of over 2,600 specialized AI models across disciplines, and a survey of over 600 scientists. We map these data to a new taxonomy of scientific tasks to study how scientists are using AI. Four main findings emerge. First, we find broad adoption and coverage: scientists use AI more than most other occupations. Specialized AI models have broad disciplinary coverage and are highly cited. Nearly half of the scientists surveyed report using some form of AI every day. Second, we document evidence that LLMs (proxied through Gemini usage) and specialized models act as complements-- LLMs are used for general analysis, coding, and manuscript preparation, while specialized models provide domain-specific predictions, data generation and classification. Third, scientists report large productivity gains from using AI: a saving of nearly 7 hours per week, time which is primarily re-invested in more research. Finally, we show that AI is already changing the scientific process. As some stages of scientific research become easier, bottlenecks shift downstream. Scientists report an increased backlog of untested hypotheses and substantial demand for output verification. Our findings suggest that AI holds significant potential to increase scientific productivity. However, as with other sectors, its ultimate impact will be governed by complex task interdependencies and investment into the elimination of emerging bottlenecks. --- ### Stable and Faithful Explanations for Knowledge Tracing Authors: Praveena Padi, Arun Morampudi, Ujval Sai Gopal Irrinki, Pradeep Kumar Dolabehera Kakitapelli Published: 2026-09-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.28502v1 Knowledge tracing (KT) models predict student performance opaquely, limiting pedagogical action. This study contributes a validation protocol testing predictive competitiveness (RQ1), explanation stability (RQ2) and retraining-based faithfulness (RQ3) together. Thirteen behavioral features across five pedagogical themes were engineered from ASSISTments 2009 and 2012, with history features computed from temporally preceding interactions and current response latency retained only for retrospective analysis. ASSISTments 2009 was rebuilt: the uncorrected skill-builder release duplicates each multi-skill interaction across one row per skill, and because those rows share one correctness label, they leak it into preceding-interaction features. Rebuilding lowered model AUC and reordered the explanation results. An Extreme Gradient Boosting (XGBoost) model explained with Tree SHapley Additive exPlanations (TreeSHAP) was compared against four deep baselines (DKT, SAKT, AKT and SimpleKT) under an information-matched protocol giving the deep models the same behavioral signals and restricting XGBoost to what is derivable from the identifier-and-correctness stream they consume. XGBoost reached an area under the curve (AUC) of 0.777 on 2012 and 0.786 on rebuilt 2009, with prediction-time AUCs of 0.771 and 0.775, respectively, after excluding current response latency; restricted to the baselines' information it performed as they did (0.697 against 0.700, and 0.717 against 0.720), locating the difference in information supplied, not model family. Rankings were consistent across folds, seeds and conditioning schemes (Spearman rho = 0.989-1.000), and removing top-ranked TreeSHAP features harmed AUC more than random removal, though split-gain and permutation rankings performed comparably. Student-level examples are illustrative interpretations, not validated recommendations. --- ### TTTIR: Unlocking Instance-Specific State Evolution via Test-Time Training for Image Restoration Authors: Kaihang Zheng, Jun Li, Hang Guo, Hongyu Chi, Zimo Liu, Tao Dai, Jinpeng Wang, Yaowei Wang Published: 2026-09-15 Categories: eess.IV, cs.AI, cs.CV, cs.MM arXiv: https://arxiv.org/abs/2609.26151v1 Image restoration is inherently challenging due to the diverse and highly input-dependent nature of real-world degradations. While recent architectures like Transformers and state-space models have advanced the field, they predominantly rely on static, globally shared parameters, which struggle to fully accommodate instance-specific degradation patterns. Test-Time Training (TTT) offers a promising paradigm for generating data-dependent operators, yet its standard self-supervised inner loop lacks the explicit guidance required to transition degraded features toward clean structures. To address this, we propose TTTIR, a novel framework that reformulates image restoration as an instance-specific state evolution process. Specifically, we design Progressive State Sequence Generation (PSSG) to construct complementary spatial-frequency target states (defining what to recover), and State Transition Evolution (STE) to adapt lightweight transition operators via a restoration-oriented TTT inner loop (determining how the features should evolve). Extensive experiments demonstrate that TTTIR consistently outperforms state-of-the-art models across multiple image restoration benchmarks, achieving dynamic instance-specific recovery with favorable computational scalability. The code is available at https://github.com/Elysiaaaaaaaa/TTTIR.git. --- ### Yarn tracking of large-scale 3D textile reinforcements using topological material features Authors: Hafsa El Herichi, Arturo Mendoza, Yanneck Wielhorski, Hugues Talbot, Stéphane Roux Published: 2026-09-15 Categories: cs.CV, eess.IV, eess.SP arXiv: https://arxiv.org/abs/2609.22315v1 Automated segmentation of CT images has become increasingly important to enhance the reliability of simulations through the generation of high fidelity numerical models. This study addresses the challenging task of semi-automatically tracking textile reinforcements in fan blade dry preforms using X-ray CT images captured at coarse resolutions (i.e., above 140 $μ$m). Our approach offers a scalable, slice-based analysis conducted on planes orthogonal to the main yarn directions, applied to a large-scale real industrial component. This enables accurate identification and tracking of yarn paths while requiring minimal training. The method models three key yarn properties statistically: their typical cross-section shape, their continuity and movement in the 3D space, and their spatial relative arrangement with respect to neighboring yarns. These statistical properties are integrated into a tracking framework via a variational formulation that optimizes all yarn center positions in successive cross-section planes. The method tracks more than 3,000 warp yarns across 1,500 slices and achieves a tracking success rate above 90%. Overall, this work demonstrates a promising approach toward large-scale, automated textile reinforcement annotation, paving the way for more efficient material characterization in complex composite structures. --- ### GameReplica: A Benchmark for Black-Box Visual Game Replication by Vision-Language Agents Authors: Boyu Qiao, Zixin Tang, Xiaoshuai Hao, Wenbo Li Published: 2026-09-15 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2609.22308v1 Coding-agent benchmarks usually evaluate implementation after the target behavior has been specified in text, code, or demonstrations. Existing research has extensively evaluated the ability of coding agents to generate programs from textual specifications. However, under black-box conditions where neither source code nor documentation is available, it remains underexplored whether an agent can induce the rules solely through visual observation and active interaction and reproduce the target system as a verifiable executable system. To this end, we present GameReplica, a closed-loop evaluation framework for end-to-end black-box game replication that covers the full perception, exploration, induction, reproduction, and verification pipeline. GameReplica comprises 125 tasks spanning 25 games across 5 core mechanism families, with each game instantiated at five difficulty levels. The tasks require an agent to access the target game only through screenshots and an action interface, induce the key visual elements and gameplay rules from pixel feedback and interaction outcomes, and generate a self-contained, runnable game replica that can be automatically verified by an external program. Experiments show that current coding agents still face substantial challenges in end-to-end black-box replication: the best-performing model (Claude Opus 4.8) achieves an overall score of 71.6\%, while the remaining models score only 4.0\%--42.9\%. Further analysis reveals a consistent pattern across all models: visual-fidelity scores are substantially higher than implementation- and rule-consistency scores, indicating that agents replicate visual appearance more readily than game mechanics. The difficulty levels further amplify the performance gap: from L1 to L5, the overall score of weaker agents drops sharply, whereas that of the best-performing agent declines only slightly. --- ### TAPe+ML: A Compact Structured Representation for Multi-Task Computer Vision Authors: Sergey Kurinov, Alexey Upatov Published: 2026-09-15 Categories: cs.CV, eess.IV arXiv: https://arxiv.org/abs/2609.20869v2 We present TAPe+ML v3, a compact computer vision system based on TAPe (Theory of Active Perception), a structured representation that encodes relations among perceptual elements before recognition. Instead of operating directly on pixel tensors, the system uses a shared TAPe representation and a modular recognition architecture for image classification, object detection, and instance segmentation. TAPe+ML v3 combines background and contour processing, local object localization, prototype-based classification, and a coordinator for specialized submodels. Across the reported experiments, it uses fewer than 100,000 parameters. On COCO object detection, it obtains 84.7 mAP50 and 65.3 mAP50-95. On COCO instance segmentation, it obtains 80.7 mask mAP50 and 58.4 mask mAP50-95. In classification experiments, it reaches 92 percent validation accuracy on Imagenette under an identical-training comparison with a raw-pixel baseline, and 89.9 percent Top-1 accuracy on ImageNet-Real. We also evaluate compactness in video scene detection and adaptation under distribution shift in an industrial pilot. The results suggest that shifting part of the modeling burden from network parameters to a structured input representation can support compact multi-task vision systems with reduced data, memory, and compute requirements. --- ### Automated Physics-Informed Neural-Networks-Based Calibration of Highly Segmented Silicon Telescopes Authors: M. Rejmund, A. Lemasson, P. Morfouace, D. Ramos, J. Taieb, J. D. Frankland Published: 2026-09-15 Categories: physics.ins-det, cs.LG, nucl-ex arXiv: https://arxiv.org/abs/2609.20868v1 Transfer and multi-nucleon transfer reactions are essential tools for probing nuclear structure and reaction dynamics, requiring precise determination of the identity, energy, and emission angles of reaction products. The increasing granularity of modern silicon telescope arrays enhances experimental capabilities but challenges detector calibration, as conventional channel-by-channel approaches become inefficient and difficult to scale. In this work, we present a fully automated, physics-informed calibration framework based on neural networks, specifically designed for highly segmented silicon detector arrays. The method formulates calibration as a global optimization problem, in which detector gains and geometrical corrections are determined simultaneously by minimizing the width of the reconstructed excitation energy under two-body kinematics constraints. The approach relies exclusively on experimental data and well-established physical principles, without requiring explicit modeling of detector response. A distinctive feature is the use of multiple neural network sub-models sharing a common loss function with embedded physics constraints, enabling coherent and self-consistent calibration across all detector channels. This strategy ensures scalability, robustness, and reproducibility, making it particularly suitable for next-generation detector systems with increasing complexity. The performance of the method is demonstrated using experimental data from the Particle-Identification Silicon-Telescope Array (PISTA) in high-resolution fission studies in inverse kinematics. The results show excellent agreement with theoretical kinematics, high-quality particle identification, and a significant improvement in calibration efficiency. The proposed framework provides a general and adaptable solution for the calibration of complex detector systems in modern nuclear physics experiments. --- ### CovR: Coverage-Aware Hardware Verification via Reasoning-Guided Reinforcement Learning Authors: Manar Abdelatty, Maryam Nouh, Sherief Reda Published: 2026-09-15 Categories: cs.AR, cs.CL arXiv: https://arxiv.org/abs/2609.19189v2 Design verification remains one of the most resource-intensive stages of hardware development, often consuming up to 70% of the total design effort. While recent work has explored using Large Language Models (LLMs) to automate testbench generation, most existing approaches focus narrowly on functional correctness, overlooking the critical aspect of coverage quality. To bridge this gap, we present CovR, an agentic framework for automated testbench generation that combines self-reflection loops with simulation-based feedback to maximize coverage. Using this pipeline, we construct a large-scale dataset of 16,514 natural specification RTL reasoning testbench tuples with a strong teacher model, enabling coverage-aware supervision. Building on this, we propose a reinforcement learning (RL) framework tailored for coverage-driven testbench generation, leveraging tool-derived rewards from simulation and coverage feedback to optimize a student model. Experimental results show that the CovR finetuned model achieves 93.81% cov@10 on VerilogEval and RTLLM V2.0, and 87.76% cov@10 on CVDP, outperforming state-of-the-art approaches by 7.97% and 3.59%, respectively. Furthermore, deploying the finetuned model back into the agentic refinement pipeline further improves cov@10 to 94.27% on VerilogEval and RTLLM V2.0 and 91.39% on CVDP. Moreover, when integrated as a plug-in stimulus engine for full verification workflows, CovR improves coverage by 18.95% and mutation detection score by 1.19%, while revealing 4.46% undetected failures, highlighting the importance of optimizing for coverage in LLM-based hardware verification. --- ### Message capacity and claim wording set the transition points of collective truth-finding in language-model networks Authors: Makoto Fukushima Published: 2026-09-15 Categories: cs.MA, cs.CL, nlin.AO, physics.soc-ph arXiv: https://arxiv.org/abs/2609.19183v1 Whether human or large language model (LLM), an agent in a discussion reads only a few of the others' contributions, bounded by cognition, context, or cost. LLM collectives can settle on a wrong consensus even when a majority starts out correct; we ask how far that reading bound alone decides the outcome. We model the bound with one number, the message capacity, which sets how many of the others' messages an agent reads, and generate the communication network from it. Over 31,824 randomized queries, we found that an 8-billion-parameter model's judgment of a claim effectively reduces to a logistic function of a weighted sum of its inbox, the update rule of a stochastic binary neuron with divisively normalized weights. From these weights and the network's degree statistics alone, the wrong consensus should become unreachable from any start once agents read, on average, fewer than 6.4 of their 31 sources. In 1,414 episodes with assigned starts the prediction failed: the correct side won in fewer than 50% of episodes from every start, and in only 28-45% when 75% of agents started correct. The failure traces to the field, the threshold that a claim's wording sets for the agent's answer before any message is read: the experimental claims' fields lay below the calibration mean, and with each claim's own field the same weights reproduce the outcomes. Reversing the wording showed that the threshold follows what a claim asserts, not whether it is true. On a second 8B model the pipeline predicts claim-dependent bistability; transition points appeared where computed, and an eight-claim calibration matched in 15 of 16 conditions. At 70B the assertion bias is not detected. Thus a collective's fate is largely set by two single-agent measurements: the threshold a claim's wording sets, and the message capacity that sets the transition point. --- ### What Do We Expect from LLMs? Mapping the Design of LLM Benchmarks Authors: Chao Wang Published: 2026-09-15 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.19182v1 Benchmarks are central to how progress in large language models (LLMs) is assessed and communicated. Yet model rankings alone reveal little about how evaluation requirements themselves are changing. The expanding variety of benchmarks offers another perspective: what researchers expect LLMs to do, and what they count as successful performance. We systematically map 14,767 papers introducing or updating evaluation resources from arXiv submissions between January 2022 and August 2026. Using staged screening and automated full-text coding, we examine changes in target systems and domains, evaluation materials and conditions, and scoring mechanisms. The collection shows growing emphasis on action, interaction, and professional applications, while established and newer design elements frequently coexist. Model participation also develops unevenly: LLM-based scoring grows within both agent and non-agent groups, whereas model-generated materials show no comparable sustained increase in recent cohorts. These findings illuminate how public research translates capability expectations into concrete tests and criteria for success. As AI participates in constructing tests, performing tasks, and judging responses, they also raise a question: does expanding evaluation provide more independent evidence, or risk reproducing the preferences and blind spots of its participating models? --- ### BioPhys-Bridge: A Benchmark for Interdisciplinary Scientific Reasoning in Physics-Grounded Biological Research Authors: Qingyang Xu Published: 2026-09-15 Categories: cs.AI, q-bio.QM arXiv: https://arxiv.org/abs/2609.19180v1 Language models face unique challenges in analyzing interdisciplinary scientific research literature. In biophysics research, faithful answers require grounding observed data in source evidence, interpreting it through a quantitative physics model, and linking it to a biological mechanism. To address this challenge, we introduce BioPhys-Bridge, a novel benchmark dataset for evidence-grounded scientific reasoning over biophysical literature. Each case contains evidence blocks, stable evidence IDs, quantitative values, units, equations, assumptions, mechanisms, and next decisions as grounding targets for question answering (QA) and retrieval-augmented generation (RAG). The initial release contains 500 cases, 1,517 agent-facing tasks, and covers six biological domains and nine physical model families, including three sparse families reserved for future expansion. We enforce strict quality gates for all cases in schema, evidence-integrity, quantitative-grounding, source-license, duplicate, unit-normalization, with domain expert review and annotation for 81 cases. Preliminary evaluations show that DeepSeek-V4-Flash obtain the highest evidence-ID $F_1$ score (0.360), followed by Qwen3.7-Max (0.316) and GPT-4o-mini (0.294). BioPhys-Bridge is an interdisciplinary benchmark for evaluating attribution, faithfulness, hallucination reduction, and biological experiment design with complex, multi-step scientific reasoning. Future works will increase the size and complexity of the dataset and perform comprehensive evaluations. Code and data are available in the GitHub repository and on Hugging Face. --- ### Optimal Transport Metric Learning for Feature Alignment in Partially Supervised Segmentation Authors: Dakini Mallam Garba, Salim Abdou Daoura Published: 2026-09-15 Categories: eess.IV, cs.AI arXiv: https://arxiv.org/abs/2609.19176v1 Multi-organ segmentation is often challenged by partially annotated datasets and domain shifts across different imaging sources. To address these limitations, we propose a two-stage learning framework that efficiently leverages partial supervision. In the first stage, the model learns from available annotations to produce accurate segmentations of annotated organs, establishing robust feature representations. In the second stage, we introduce learnable organ prototypes and a Sinkhorn-triplet loss to enforce organ-wise feature consistency across datasets. This encourages latent embeddings of the same organ to remain close, while increasing separation between different organs, even when annotations are missing. Our approach achieves performance comparable to state-of-the-art methods on the BTCV dataset, while remaining computationally efficient. By explicitly aligning feature distributions rather than relying solely on pseudo-labels, the framework effectively mitigates domain shift, making it particularly suitable for medical image segmentation tasks with limited annotation resources. --- ### Beyond the Previous Layer: Residual Predictive Structure in Sparse MoE Routing Authors: Hao Li, Yasuyuki Tahara, Yuichi Sei Published: 2026-09-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.17940v1 Sparse mixture-of-experts models route each token through a sequence of expert selections. We ask whether the immediately preceding selection adequately summarizes this trajectory for predicting the next router. Using frozen OLMoE and JetMoE models, we measure the held-out predictive gain from earlier expert selections while retaining the most recent selection as a common baseline. In OLMoE, extending the history from one to eleven layers raises router-logit $R^2$ from 0.59879 to 0.66544. A preregistered JetMoE replication yields four-layer gains of 0.14275 and 0.20528 at two target depths, with paired bootstrap intervals above zero. These gains survive nonlinear decoding: adding history to a small multilayer perceptron improves $R^2$ by 0.17137 and 0.21861, whereas nonlinear decoding of the recent state alone adds 0.00139 and 0.00936 over a linear probe. Parameter-matched controls preserve the advantage, and cross-fitted history residuals predict target residuals with $R^2$ of 0.20549 and 0.23556. These findings identify residual predictive structure in expert-selection trajectories beyond adjacent-layer persistence. --- ### Locating Hidden Failures Makes Long-Horizon Agents More Reliable Authors: Salman Rahman, Yubin Kim, Mihir Parmar, A. Ali Heydari, Genglin Liu, Simon A. Lee, Weizhi Zhang, Arian Hosseini, Ahmed A. Metwally, Yuzhe Yang, Baharan Mirzasoleiman, Xin Liu et al. Published: 2026-09-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.17930v1 As AI agents take on long, autonomous tasks, we increasingly oversee rather than perform the work, yet we still judge them almost entirely by whether they finally succeed. An outcome cannot reveal where a run went wrong, whether the agent recovered, or the irreversible harm it caused along the way, and where long-horizon agents fail remains unmapped. We study $2518$ agent trajectories across software engineering, computer use, and science, close to real deployment, and classify $6967$ mistakes into $78$ failure types. Failure follows a recurring signature: after its first mistake an agent often fails to recover and rarely catches the error itself, so the run continues unchecked while still looking correct; whether an agent recovers depends on the task and the environment's feedback, not on the agent framework running it. Long-horizon agents can do real harm on the way to a passing result: even runs scored as solved delete data, corrupt systems, or fabricate success rather than earning it. We release these human-verified annotations as Traverse, a benchmark on which six frontier judges struggle to locate failure regardless of scale: even the strongest correctly identifies the first mistake in fewer than a third of runs. Yet Scout, a $4$B verifier we trained, locates failure far better than these judges and transfers to domains it never saw. Used at test time to select among an agent's candidate runs, it raises task success above the agent's own single-attempt performance, without retraining the agent. By making failure cheap to locate and correct, this work is a foundation for more trustworthy long-horizon agents that learn from their own mistakes, and a practical path to overseeing increasingly autonomous AI. --- ### Symmetry without a manifold: intrinsic dimension on orbits Authors: Chon-Fai Kam, Miloud Bessafi, Frédéric Cadet Published: 2026-09-15 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2609.17926v1 The standard geometric derivation of neural scaling exponents takes the intrinsic dimension of a data manifold as its input. On modular addition in $\mathbb{Z}_p$ that derivation has no input. The exact algebraic solution is an orbit of $\mathbb{Z}_p$ acting by isometries. Transitivity alone makes the ratio statistic underlying the standard dimension estimator a point mass, so the estimator is undefined, and here the two nearest neighbour distances coincide exactly. Breaking the symmetry at scale $ε$ returns a number, but one that tracks $1/ε$ with no scale free plateau. We show that the failure is general, since on any finite orbit of a group acting by isometries the estimator reports the resolution at which the set is probed rather than a dimension. What replaces the power law is exponential in hidden width, $L(h)=L_\infty+A\exp(-c\,h^α)$, with $R^2$ between 0.982 and 0.995 against 0.857 to 0.906 for a power law admitting the same floor and fitted under the same protocol. Where the data supply is sufficient the rate belongs to the regulariser rather than to the group, since weight decay moves $c$ by a factor of 47 while group order moves it by 1.10, a residual below seed to seed resolution, for every fixed $α$ between 0.75 and 2. The critical width falls with group order rather than rising, against capacity counting that assigns a fixed number of neurons to each irreducible representation. --- ### Rapid Loss of the Sierra Nevada's Largest Trees Driven by Fire Authors: Fabien H. Wagner, Dan J. Dixon, Christopher W. Woodall, Mayumi C. M. Hirye, Felipe Saad, Griffin Carter, Ricardo Dalagnol, Lorena Alves, Cynthia Creze, Stephen C. Hagen, Zhihua Liu, Christopher Mihiar et al. Published: 2026-09-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.17925v1 Large trees disproportionately contribute to biomass storage, habitat structure, and ecosystem functioning. However, their distribution and health dynamics remain poorly quantified at a regional scale. Here, a deep learning model (U-Net-ID) and canopy height models derived from sub-meter aerial imagery from 2020 were used to delineate all individual trees with crown area $\geq$ 100 m$^2$ across the Sierra Nevada Floristic Province. The model was trained using more than 3.3 million synthetic tree crowns and achieved a median Intersection over Union (IoU) of 0.602 when validated against an independent dataset of 20,273 crowns. A total of 6,515,705 large trees were mapped, occurring across approximately 78.7% of the Sierra Nevada Floristic Province. The spatial distribution of large trees showed associations with elevation, temperature, and precipitation. Using Sentinel-2 time series from 2020 to 2025, tree health dynamics were characterized by extracting spectral trajectories for each crown and applying BFAST breakpoint detection algorithm combined with a disturbance classification framework to identify mortality, disturbance, and recovery trajectories of individual trees. Wildfires, estimated from CAL FIRE fire perimeters, were identified as the dominant driver of large-tree mortality, killing 10% of all large trees in the Sierra Nevada, with mortality strongly concentrated during the extreme 2020-2021 fire seasons. --- ### Audio for Sports Highlight Detection: A Comparative Empirical Study Authors: Hao Xu, Meenakshi Sarkar, Vishnu Raj, David Gunawan Published: 2026-09-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.17923v1 Sports highlight detection aims to identify the most exciting and meaningful moments from long sports videos. While existing methods often emphasize visual or visual-language representations, sports videos contain rich audio cues, including commentator speech, crowd reactions, whistles, ball impacts, and referee calls. In this work, we revisit the role of audio in sports highlight detection and ask a simple question: how far can audio alone go? We construct lightweight audio-only baselines using pretrained audio representations and compare them with visual-only and audio-visual methods on the SV-Highlights benchmark. Surprisingly, our audio-only GRU baseline achieves strong performance and outperforms several existing audio-visual methods under our supervised evaluation setting. Furthermore, a simple audio-visual fusion baseline achieves the best performance across all metrics, indicating that audio and visual cues provide complementary information. To better understand the contribution of audio, we conduct source-separated analysis and show that vocal/commentary audio is more informative than background-only audio, while their combination performs best. We also analyze interpretable audio cues and find that highlight clips exhibit higher RMS loudness, peak loudness, and mid-frequency energy than non-highlight clips, although substantial distribution overlap indicates that loudness alone is insufficient. Our findings suggest that audio is an underexplored but highly informative modality for sports highlight detection and should be treated as a primary signal rather than merely an auxiliary cue. --- ### Collaborative Memory for Multi-Agent VLM Systems Authors: Huixin Zhang, Shao-Jun Xia, Di Wang, Liangxi Liu, Hainan Xiong, Zihao Wang Published: 2026-09-15 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.17921v2 Vision-language model (VLM) agents combine specialized perception, tools, and reasoning to address complex visual tasks. In multi-agent settings, different agents inspect different image regions, video frames, or visual representations, so collaboration extends beyond distributed reasoning to distributed perception. This makes shared visual context a central problem in VLM agent collaboration. In this paper, we frame memory hierarchy, cross-agent sharing, and consistency mechanisms around the need to reconcile interpretations and update dependent reasoning. Effective collaboration requires agents to build on contributions from other agents, recover missing visual context, and reconcile differing interpretations as new evidence emerges. Shared visual memory preserves not only images or textual summaries but also the dependencies among observations, agent interpretations, and subsequent reasoning. Together, these design considerations shape how information flows and evolves across VLM agents. The proposed framework provides a foundation for building reliable and resource-efficient agent teams. --- ### EdgeReMIND: A Scalable, Top-Ranked Memorization Baseline for Temporal Multi-Relational Link Prediction Authors: Bryant Pollard Published: 2026-09-15 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.17916v1 Temporal link prediction on the Temporal Graph Benchmark 2.0 (TGB 2.0) faces a scalability ceiling: on the benchmark's three largest datasets, every existing embedding method runs out of memory or exceeds the time budget. These large-scale graphs are the ones nearest real deployment scale, so failing on them is a real production limitation. EdgeReMIND sets the highest reported test mean reciprocal rank (MRR) on six of eight TGB 2.0 datasets and is the only relation-aware method that runs on all of them. This linear memorization model, with learned per-relation weights over data-calibrated features, is therefore not merely a fallback where embeddings fail but a practical state-of-the-art baseline across the benchmark. --- ### Face-voice Association across LAnguages and Gender (FLAG) 2027 Challenge Evaluation Plan Authors: Marta Moscati, Swapnil Khandoker, Muhammad Saad Saeed, Shah Nawaz, Fatima Noor, Rohan Kumar Das, Mubashir Noman, Junaid Mir, Muhammad Haroon Yousaf, Khalid Malik, Markus Schedl Published: 2026-09-15 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.17913v1 Face--voice association models may rely on language or gender cues in the voice rather than on speaker-specific voice characteristics, which can lead to a performance deterioration when the model has to identify a multilingual speaker or distinguis same-gender speakers. To investigate these issues, we introduce the Face-voice Association across LAnguages and Gender (FLAG) 2027 Challenge. The challenge formulates face--voice association as a cross-modal verification task: given a voice, identify the speaker's face from a ``gallery'' of faces consisting of the speaker's face and a set of negative samples. Models are evaluated on identities not present in the training data (``unseen'') and both for languages present or absent from the training data (``heard'' and ``unheard''). Two evaluation settings are used to test models' reliance on gender: a standard, unconstrained and a gender-constrained one, where the latter uses a same-gender gallery. The performance of existing, baseline models in these settings reveals that models performance degrades under language shifts and in gender-constrained settings, highlighting the need to foster the development of models that capture identity-specific aspects beyond language and gender. The challenge provides a benchmark dataset, pretrained baseline models, and an evaluation framework to advance face--voice association. --- ### Zing-0.5: Toward Playable Worlds with Real-Time Joint Action and Text Control Authors: Mingyang Chen, Shengdong Chen, Xiaoxiao Fu, Bosheng Gong, Haoyuan Guo, Bowen Li, Jiawen Li, Kejun Li, Tianpeng Li, Yin Liu, Haoze Sun, Zeyang Tian et al. Published: 2026-09-15 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.17909v1 We introduce Zing-0.5, a 5B autoregressive world model designed for playability: users can explore generated worlds, influence unfolding events, and respond to the resulting feedback through joint keyboard and online text control. Our approach brings together three technical contributions: (1) Unified action and text conditioning, combining magnitude-aware keyboard inputs with temporally aligned text instructions and jointly annotated videos to learn navigation and event control within the same sequence; (2) Event-scale supervision for incremental generation, using a segment-level teacher trained on connected multi-prompt videos to supervise a block-level causal student through distribution-matching distillation; and (3) Low-cost real-time interaction, combining four-step generation with context-preserving streaming to support 832 x 480 inference at 24 FPS at an estimated server rental cost of approximately USD 0.009 per stream-minute. Zing-0.5 achieves an overall score of 81.0 and a consistency score of 88.5 across 158 WBench Navigation cases. A joint-control demonstration shows a text-directed event change during continued navigation without restarting generation. We release the model weights, inference code, and Zing-SGLang serving implementation to support further work on playable generated worlds. --- ### Walking the Score Manifold: Continuous-time Generative Dynamics on Learned Data Manifolds Authors: Jan Tauberschmidt, Brian B. Moser, Stanislav Frolov, Andreas Dengel, Andrew B. Duncan, Sebastian J. Vollmer Published: 2026-09-15 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.17901v1 Generative modeling of time-dependent data is typically formulated on a discrete temporal grid, restricting supervision to the observed timestamps in the training data. We instead frame generation as continuous-time evolution on a learned data manifold. To this end, we leverage pretrained score-based models as geometric priors and learn a vector field that evolves data along score-induced interpolation paths. Because these dynamics follow transitions that respect the geometry learned by the score model, they support generation at arbitrary timestamps and temporal super-resolution beyond the discretization of the training data. Moreover, this geometric formulation allows us to train the vector field simulation-free through a regression objective. To improve long-horizon rollout robustness, we introduce an objective that promotes path-relative transverse exponential stability. While motivated by stability theory, it admits a practical interpretation as denoising score matching transverse to the interpolation path. Further, we extend the framework to a probabilistic setting that models a distribution over plausible future trajectories. We demonstrate the method on natural video and scientific dynamical data, including temporal super-resolution, PDE-based spatiotemporal fields, and molecular dynamics. Our results show that score-based priors provide a strong foundation for learning stochastic continuous-time generative dynamics. --- ### Generalized DCCQ: From Binary Quotients to Multinomial Simplex Geometry and Critical-Strip Coordinates Authors: Y. Kenan Yılmaz Published: 2026-09-15 Categories: stat.ML, math.NT arXiv: https://arxiv.org/abs/2609.17899v1 We extend the discrete complex complement quotient (DCCQ) framework from binary Bernoulli counts to multinomial count compositions. For m+1 categories, m is the number of independent probability degrees of freedom. Integer count vectors modulo common scaling determine rational points of the m-dimensional probability simplex. Building on standard simplex and log-ratio coordinate geometry, for m >= 2 we define the full multinomial DCCQ coordinate map and show that it is a real-analytic diffeomorphism The previously established binary baseline m=1 gives a critical-line coordinate, while the ternary case m=2 gives the full open critical strip; higher multinomial models retain m-2 additional real contrasts. We also give a one-versus-rest specialization, an exact integer-lattice realization of the ternary coordinate, and a hyperbolic representation of its log-ratio. No zero-location theorem or proof of the Riemann Hypothesis is claimed. --- ### QEMScore: How Much Does the Measurement Add to Learned Quantum Error Mitigation? Authors: Yue Zhao, Huayue Gu, Yushun Dong, Xiyang Hu Published: 2026-09-15 Categories: quant-ph, cs.LG arXiv: https://arxiv.org/abs/2609.17896v1 How much does the noisy measurement add to learned quantum error mitigation? An accuracy table cannot say, because a model handed circuit structure can score well without reading the measurement at all. QEMScore adds the comparison that can. Each simulated circuit carries an exact ideal answer. The learned mitigator is scored beside a capacity-matched control, a model just as flexible that reads the same circuit description but never the measurement. Each method's measurement spend is accounted and not equalized. We run a controlled campaign on simulated circuits and reanalyze two published learned mitigators, Q-LEAR and QRAFT, from their released hardware data. Three findings stand out. First, under familiar within-family conditions (S0) evaluated across two spin-chain families and three seeds, continuous couplings identify the target, and the control that never reads the measurement matches 87.7 to 100.5 percent of the mitigator's gain over an affine fit to the circuit description. A plain polynomial in the coupling parameters, fitted after the campaign, beats the mitigator on all six evaluations, reflecting the selected learners' capacity. Second, for these selected learners, matching most of the gain is not matching the accuracy: on five of six evaluations the mitigator removes 19.5 to 74.5 percent of the error the capacity-matched control leaves, a learner-specific gap rather than a measurement requirement. Third, on released hardware data where descriptors only partially identify queries, the findings differ: flexible models of the descriptors show negligible mean gain over affine fits in Q-LEAR, and measurement inputs carry predictive gains in both Q-LEAR and QRAFT. These comparisons reflect representation- and protocol-specific behavior rather than an isolated cross-regime difference. A learned mitigator's accuracy should therefore be reported beside such controls. --- ### TabPFN-3.5: Technical Report Authors: Benjamin Jäger, Nick Erickson, Léo Grinsztajn, Felix Birkel, Klemens Flöge, Oscar Key, Kürşat Kaya, Jonas Kübler, Adèle Frankel, Tobias Schröder, Anurag Garg, Jan Hendrik Metzen et al. Published: 2026-09-15 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2609.17895v2 We introduce TabPFN-3.5, our new flagship Tabular Foundation Model. It significantly outperforms its predecessor, TabPFN-3, and all existing baselines across a broad range of tabular problems. TabPFN-3.5 sets a new state of the art on standard tabular prediction in TabArena, and extends it to the data practitioners encounter in practice: non-i.i.d. data with temporal or grouped splits, tables with strings, text and images, high-cardinality categorical features, and wide tables with many features. These gains carry over to our task-specific harnesses: state of the art on relational data and stronger time-series forecasting. For faster inference, our variant TabPFN-3.5-Fast runs up to 3x faster than TabPFN-3 while keeping most of the accuracy gains. In addition, we upgrade TabPFN-3.5-Plus, expanding our multimodal capabilities with advanced text and date handling alongside proprietary inference optimizations. Finally, we release a new version of our Thinking mode, TabPFN-3.5-Thinking, which scales inference-time computation to push the state of the art further. It benefits from our stronger base model and from inference-time improvements that make it up to 12x faster than TabPFN-3-Thinking. --- ### Graph neural networks for exoplanet atmospheres Authors: Antonia Vojtekova, Kai Hou Yip, Ingo P. Waldmann, Nikolaos Nikolaou, Olivia Venot, Ahmed Faris Al-Refaie, Bruno Merín Published: 2026-09-15 Categories: astro-ph.IM, astro-ph.EP, cs.NE arXiv: https://arxiv.org/abs/2609.17894v1 Calculating disequilibrium chemistry in exoplanet atmospheres remains a significant computational bottleneck in atmospheric retrievals. The increasing observational precision from facilities such as JWST and the Ariel mission requires including disequilibrium chemistry in these analyses. Previous studies have demonstrated that neural networks can emulate kinetic chemistry, although their spatial inductive bias does not align with the topology of chemical reaction networks. This study introduces a graph neural network surrogate that represents chemical species as nodes and temperature-dependent reaction rates as edges, thereby enabling information propagation along physically meaningful chemical pathways. The model is trained on atmospheres generated using the Venot+2020 chemical scheme and Guillot temperature-pressure profiles. The GNN accurately reconstructs disequilibrium abundances across the sampled parameter space and reduces the mean abundance error by a factor of approximately 3 compared to the previous U-Net model. When applied to transmission spectra, most predictions fall within the observational precision expected for JWST and Ariel, with only about 7% of test atmospheres exceeding a 20 ppm mean spectral error. Performance variations are primarily observed in chemically transitional regimes near a carbon-to-oxygen ratio of one and at low temperatures. An evaluation of the boundary-case planet WASP-39b demonstrates effective performance under a moderate domain shift. Perturbation analysis indicates that disturbances propagate along chemical connectivity rather than spatial adjacency, confirming that the architecture captures the structure of reaction networks. These results suggest that GNN surrogates provide accurate, computationally efficient predictions of disequilibrium chemistry, facilitating integration into the atmospheric retrieval pipeline. --- ### Bracketing Uncertainty in Clustering Under the Manifold Hypothesis Authors: Savik Kinger, Luciano Dyballa, Steven W. Zucker Published: 2026-09-15 Categories: stat.ML, cs.LG, q-bio.QM arXiv: https://arxiv.org/abs/2609.17892v1 The manifold hypothesis suggests a natural criterion for clustering: partition data according to the manifold component from which each point is drawn. Whether two components are separable depends on a geometric tradeoff: the ambient separation between components versus the largest gap in sampling. In practice, this tradeoff is rarely assessed explicitly, leading standard methods to over-commit to a single clustering assignment even when the data do not support a unique answer. We formalize this tradeoff by combining intrinsic manifold geometry (volume growth and reach) with sample-level quantities (fill distance and density), yielding a threshold phenomenon for mutual-$k$-nearest-neighbor graphs: when the offset-to-fill ratio exceeds a conservative upper threshold, component separation is preserved; below a lower threshold, components fuse. The gap between these thresholds defines a geometric uncertainty zone in which the number of clusters is not identifiable from the data. Nevertheless, conventional approaches still seek one: sweeping parameters (an engineering approach) or fitting a generative mixture model (a model-based approach). Rather than forcing a single estimate of the number of clusters, we propose Manifold-Based Clustering (MBC), which returns an explicit bracket interval to quantify the underlying data uncertainty. This bracket acts as an empirically calibrated diagnostic: it narrows when a single resolution is supported, widens when multiple resolutions coexist, and collapses to one when no separated structure is detectable. Empirically, we find that many real datasets lie within the uncertainty zone rather than admitting one clear answer. Our results suggest that ambiguity in cluster number is often intrinsic, and should be quantified rather than resolved. --- ### OBC-Prune: Outcome-Based Calibration for Large Reasoning Model Pruning Authors: Ha Lan Nguyen, Huy Hoang Tran, Trac-Duy Tran, Dung D. Le Published: 2026-09-15 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.17890v1 Large reasoning models (LRMs) generate long chain-of-thought traces before answering, creating significant inference overhead. Pruning can reduce this cost, but its effectiveness depends on the calibration data used to estimate parameter importance. Recent work calibrates on the model's own rollouts instead of generic dataset, but treats all reasoning tokens uniformly, regardless of whether they contribute to successful reasoning. As a result, pruning protects weights by statistical salience rather than by their contribution to correct reasoning, so weights behind erroneous computation survive as readily as those behind correct computation. These erroneous patterns then get carried into the pruned model, degrading reasoning quality, producing both lower accuracy and longer reasoning traces. We propose Outcome-Based Calibration for Large Reasoning Model Pruning (OBC-Prune) to close this gap. OBC first constructs difficulty-matched pairs of correct and incorrect rollouts from problems the model answers inconsistently. It then estimates the causal importance of each reasoning sentence through intervention-based analysis, quantifying how removing its influence affects subsequent predictions. These causal importance scores are converted into per-token weights that rescale the calibration activations used by one-shot pruning methods (SparseGPT, Wanda, ALPS), without modifying the underlying pruning algorithms. Experiments on DeepSeek-R1-Distill-Qwen 1.5B, 7B, and 14B models at 40\% and 50\% sparsity demonstrate consistent improvements over state-of-the-art calibration baselines across most model sizes and sparsity levels on MATH500, LiveCodeBench, and AIME 2025. These results indicate that preserving causally important reasoning circuits is a substantially more effective pruning objective than uniformly preserving observed activations. ---