# Research Papers — 2026-09 · bundle page 33 of 266 > Full-text markdown of papers #1601–1650 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/33 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/32.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/34.md) ### An Exact Generate - Transform Decomposition of Small-LLM Team Scaling Across Orchestration Architectures Authors: Blaz Bertalanic, Carolina Fortuna Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36104v1 Replacing one LLM agent with a collaborating team can raise accuracy, but whether scaling the team helps, and which architecture to scale, is unclear. Sweeping eight agent orchestration architectures across five instruction-tuned 7-9B models, five short-answer benchmarks, and an executable-code benchmark up to 30 calls, we find that the returns to team scaling are sharply task-dependent: from three to thirty calls accuracy rises by up to 17 points on the two arithmetic word-problem benchmarks (GSM8K, GSMHard) but by at most four on ARC, GPQA, and MMLU, for every architecture, a split the usual task-averaged number conceals. Proposer-Critic captures the arithmetic gains, scaling steepest and, in aggregate, surpassing every other architecture at the largest budget (item-clustered intervals exclude zero), though it ranks among the weakest elsewhere, and no architecture wins across tasks. We explain these trajectories with an exact generate-transform decomposition. Partitioning any workflow into proposal coverage and a downstream transform, any accuracy change splits exactly into an extensive coverage dividend and an intensive transformation change. The decomposition diagnoses each task: arithmetic offers coverage headroom that a critic-guided transform converts, whereas the multiple-choice benchmarks either saturate in coverage or fail to convert it, and on open-ended code generative recovery nearly vanishes so accuracy tracks coverage. At equal call budgets token cost still varies 2.1x. Extra calls therefore create candidate opportunity that only some architectures, on some tasks, convert. Team scaling is a task- and architecture-specific bet, not a uniform lever. --- ### One Geometry, Different Outcomes: Readout-Dependent Effects of the Modality Gap in Vision-Language Models Authors: Aditya Sharma, Divya Saxena Published: 2026-09-28 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.36101v1 Contrastive vision-language models learn shared embedding spaces by aligning matched image-text pairs, yet their representations remain separated by a modality gap. Prior work reports divergent effects of modifying this gap: reducing it can improve zero-shot classification and cross-modal alignment, whereas removing gap-related structure can degrade image-text retrieval. In this paper, we provide a unified geometric explanation for these task-dependent effects. Across CLIP and SigLIP encoders, we find that a single dominant direction captures 94.4-99.9% of the squared norm of the image-text mean separation, revealing that the mean-separation component is approximately rank-one. A decomposition of the similarity score then identifies three task-specific roles. In zero-shot classification, query-side fixed gap-offset subtraction is exactly equivalent to an additive class bias. In standard cross-modal retrieval, projecting out the gap direction and renormalising residuals discards candidate-specific norm information, inducing a multiplicative ranking distortion; a geometry-derived exponent tracks the grid-search optimum (Spearman rho = 0.93) and restores performance in some settings, although the gains transfer unevenly. In mixed-modal retrieval, the gap direction sorts candidates by modality; its removal can improve cross-modal ranking, unlike random or non-gap controls. Residual semantic structure after removal defines the limits of the rank-one account. Together, these results explain why gap modification can improve, degrade, or restore performance across downstream settings. By clarifying when and why gap modification changes model behavior, this account provides a principled basis for selecting gap interventions in similarity-based vision-language systems across evaluated downstream tasks. --- ### Data Unlearning via Inverse Distillation Authors: Aleksei Leonov, Nikita Kornilov, Zhenhe Zhang, Evgeny Burnaev, Iaroslav Koshelev, Alexander Korotin Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36099v1 Multi-step matching models, including flow and diffusion models, produce high-quality outputs but incur substantial inference costs and may reproduce unwanted components of their training datasets. We introduce Inverse Distillation Unlearning (IDU), a unified framework that simultaneously distills a teacher multi-step matching model into an efficient one-step student generator and suppresses outputs corresponding to a designated training subset. We first formulate distillation as a min-max objective over a data distribution and then represent this distribution as a mixture of the forget-set and the generated distributions. This allows us to compare this mixture with the teacher's training distribution and recover only the retained data at the optimum. Our method requires only a pretrained full-data teacher and data from the forget set, without access to retained training examples, extra feature extractors or classifiers. Extensive experiments on MNIST and CIFAR-10 datasets under flow-matching and score-based diffusion settings demonstrate that IDU substantially reduces the generation frequency of forgotten classes while preserving generation quality on the retained classes. To the best of our knowledge, IDU is the first unified framework for simultaneous unlearning and distillation in unconditional flow-matching and score-based models. --- ### Better Behavioral Prediction, More Faithful Model Ablations? Evidence from Sequential Choice Authors: Hanbo Xie Published: 2026-09-28 Categories: q-bio.NC, cs.CL arXiv: https://arxiv.org/abs/2609.36097v1 Using predictive models to explain cognition requires more than accurate behavioral predictions. Input ablations offer an appealing route: remove information from a model and interpret the resulting performance change as evidence of its importance for behavior. Yet this inference assumes that the model's dependence on information reflects the dependence of the process generating the behavior. We test it in two synthetic sequential bandit tasks with known generating policies, where past choices can remain informative when feedback is unavailable to a predictor. We compare GRUs and Transformers trained from scratch, a fine-tuned LLaMA model, and cognitive models across systematically varied reward contributions. Our analyses distinguish prediction after training without reward observations from the response of a fixed predictor to donor-reward replacement. Three findings emerge. First, in the restless task, neural models trained without rewards predict held-out choices better than four simple training-fitted behavioral baselines. Second, under matched donor replacement, accurate predictors can respond much less than the known generator. Third, at some reward weights, neural networks predict better than a pooled reinforcement-learning model but have less faithful changes in choice probabilities; the model ordering differs between the two tasks. These independent-test results separate information sufficient for prediction from response fidelity under a specified ablation in sequential choice. They motivate validating model-ablation responses independently of predictive performance before using them to infer how the observed behavior was generated. --- ### PHASE: A Physiology-Guided Hierarchical Foundation Model for Intracranial EEG Authors: Yipeng Zhang, Chenda Duan, Yuanyi Ding, Tianyi Wang, Atsuro Daida, Masaki Izumi, Yuta Tanoue, Naoto Kuroda, Shaun A. Hussain, Nishant Sinha, Eishi Asano. Hiroki Nariai, Vwani Roychowdhury Published: 2026-09-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.36087v1 Clinicians and neuroscientists have long analyzed intracranial electroencephalography (iEEG) through directly measurable physiological characteristics, which carry much of the information that downstream tasks depend on. Recent iEEG foundation models learn by reconstructing or predicting their inputs, which leaves the retention of these characteristics implicit. They are also evaluated mainly on cognitive decoding and a narrow clinical task, i.e., seizure detection. On a broad, clinically relevant benchmark such as Omni-iEEG, they remain below task-specific models when used frozen. We introduce PHASE, a physiology-guided foundation model that makes these characteristics explicit learning targets, pairing them with masked latent prediction in a temporal stage (PHASE-T) within each channel and a spatiotemporal stage (PHASE-ST) across synchronized channels. PHASE is pretrained on heterogeneous recordings from 222 participants at nine clinical sites. On all five Omni-iEEG clinical tasks, frozen PHASE-T outperforms every evaluated foundation model by up to 31\%, and fine-tuned PHASE-T surpasses the task-specific models, setting a new state of the art. PHASE-T benefits from physiological supervision, outperforming variants trained with latent prediction alone or auxiliary waveform reconstruction on every task in matched ablations. PHASE-T generalizes to unseen institutions, outperforming the compared models with few or no local labels. PHASE-ST further improves seizure-onset-zone identification over PHASE-T and, when frozen, decodes sound volume and pitch on BrainTreebank better than published models. Beyond task performance, PHASE learns to encapsulate the physiological characteristics clinicians recognize, from seizure onset and its propagation to anatomical region identity, even though its pretraining contains no ictal recordings or anatomical labels. --- ### PADMÉ: Preference Alignment Data Synthesis for Meta-Evaluation of LM Agent Evaluators Authors: Cheng Chang, Yining Mao, Peng Qi Published: 2026-09-28 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.36086v1 Language models are frequently employed to evaluate other language models. An LM evaluator scoring agentic behaviors across multiple criteria is valuable, provided that its decisions align with human judgment. We call the problem of evaluating this alignment Meta-Evaluation. Tackling it directly is difficult: collecting human data is expensive, absolute scoring is hard to align, and using an LM meta-evaluator recurses the question of trustworthiness. We adopt a reformulation of meta-evaluation as a preference judgment problem: rather than comparing human and LM evaluator scores of a trajectory, we ask whether their implied preferences align. Building on this, we introduce PADMÉ, a data synthesis method that generates reliable criterion-based meta-evaluation data for agentic settings. PADMÉ uses only small language models, requires no human involvement during evaluations, and operates under a low computational budget. We build a prototype of PADMÉ and synthesize a dataset of 1,000 samples across four agentic domains and three evaluation criteria. Human validation on a 150-sample subset demonstrates that PADMÉ improves agreement with human judgment from 73% to 85% over a naive baseline. Meta-evaluating 25 common models with our dataset demonstrates the correlations between evaluation performance and scoring granularity, leniency, and model size, among other factors. --- ### Measuring trainable degrees of freedom in materials graph neural networks: a random-subspace intrinsic dimension analysis Authors: Shehroz Ahmad Shoaib, Kangming Li Published: 2026-09-28 Categories: cond-mat.mtrl-sci, cs.AI arXiv: https://arxiv.org/abs/2609.36084v1 Final predictive accuracy is the standard basis for comparing graph neural networks (GNNs) in materials-property prediction, but it does not show how strongly performance depends on access to trainable parameter-space directions. Here, we introduce trainable-degree dependence as a complementary characterization of materials GNN learning. Using random-subspace intrinsic-dimension analysis, we train CGCNN, ALIGNN, and DimeNet++ in randomly oriented parameter subspaces across six prediction tasks and measure how performance recovers as independent trainable degrees of freedom are restored. The resulting recovery curves separate endpoint accuracy from the trainable-dimensional demand required to recover it. They reveal distinctions that final errors alone miss: metallic classification and log-bulk-modulus regression recover near-reference performance from small fractional subspaces, formation-energy and band-gap prediction show stronger architecture dependence, and phonon prediction is most sensitive to dimensional restriction. Dataset-size sweeps show that band-gap models require larger fractional subspaces as training data grows, whereas formation-energy and bulk-modulus responses are more stable. A width sweep shows that fractional thresholds can remain stable while absolute threshold dimensions increase with model size. Random-subspace analysis therefore provides a targeted stress test for how materials GNNs use their optimization space. --- ### GeoOutageBench: Benchmarking Ambiguity-aware, Ontology-grounded Geospatiotemporal KGQA for Multimodal Power Outage and Resilience Analysis Authors: Ethan D. Frakes, Amy Kvien, Rishabh Kundu, Redad Mehdi, Van D. Tran, Vibha S. Mandayam, Kristopher O. Davis, Erika I. Barcelos, Roger H. French, Yinghui Wu, Mengjie Li Published: 2026-09-28 Categories: cs.AI, cs.CL, cs.IR arXiv: https://arxiv.org/abs/2609.36082v1 We introduce GeoOutageBench, a benchmark for assessing LLM-based geospatiotemporal KGQA for multimodal outage and resilience analysis. Unlike existing KGQA benchmarks for Web knowledge, GeoOutageBench considers a spatiotemporal KG that integrates visual, textual, and structured data from outage records, remote sensing, weather observations, storm and power events, geographic entities, and domain ontologies. It provides a competency query taxonomy at different difficulty levels from spatiotemporal containment and proximity, spatiotemporal co-occurrence analysis, multimodal evidence, to hypothetical evaluation. Over multimodal KG and query classes, GeoOutageBench provides user-configurable evaluation of three important, highly coherent yet less studied tasks: (1) LLMs' understanding for ambiguous geospatiotemporal questions in terms of NL to SPARQL interpretation, (2) query-driven assessment of ontology utility, and (3) answer accuracy of multimodal KGQA retrieval. GeoOutageBench provides a design principle and foundation for assessing LLM-KG systems that support real-world infrastructure resilience analysis. Our benchmark, source code, data, results, and other documentation are available at https://github.com/UCF-SAGE/GeoOutageBench. --- ### Early Learning Shapes Later Directions Of Representation Change In Continual Learning Authors: Yuantao Deng, Jinnuo Liu, Kaizhen Tan, Yuchen Liu Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36081v1 Representations continually change as a network learns new tasks. We ask whether early representational changes naturally form a geometric structure that continues to shape later learning. We identify a low-dimensional subspace of early representation drift, which we call a scaffold, and test whether it is reused across subsequent tasks. Across four pretrained visual encoders and two datasets, later representational changes consistently favor this early-defined subspace over matched random alternatives. This reuse is history-dependent: when networks experience different early tasks but identical later training inputs, each network preferentially reuses the scaffold induced by its own learning history. The same preference appears in individual optimizer updates, even though the network's dominant local response directions shift away from the original scaffold. Finally, constraining motion within the scaffold slows new-task acquisition more than matched random constraints, while effects on old-task retention are less consistent. In summary, these results suggest that early experience leaves a persistent geometric imprint on how neural networks adapt to future tasks. --- ### A Polyphonic Conception of AI Understanding Authors: Matthieu Queloz, Pierre Beckmann Published: 2026-09-28 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.36079v1 When a doctor, a judge, or an engineer must decide whether to trust an AI model's output, they cannot avoid asking what the model understands. Purely mathematical or statistical descriptions struggle to distinguish trustworthy from untrustworthy outputs without reintroducing the question of AI understanding in all but name. Yet the question is ill-framed as it stands, because the inherited concept operates within a monophonic paradigm: the idea that a cognitive system's understanding of something must be localised to a single mechanism underpinning all the capacities conferred by such understanding. Drawing on a wide range of mechanistic evidence, we show that LLMs are pervasively polyphonic: outputs emerge from coalitions of parallel mechanisms of uneven reliability, which variously complement, duplicate, or drown out one another, with several coalitions sufficing for a task without any one being indispensable. Polyphony not only complicates attributions of understanding, but renders monophonic inference patterns hazardous. In response, we develop a conception of understanding fit for polyphonic AI. It centres on sound circuitry that is reliably and correctly recruited and in control of outputs. Attributions of understanding thereby become tractable claims about internal organisation, and can do the work of guiding trust in AI. --- ### GEM-KMeans: Memory-Efficient and Accurate Clustering on Massive Scale with GPU Optimization Authors: Peng Xu, Nihar Koganti, Volodymyr Kindratenko, Xiaohui Chen Published: 2026-09-28 Categories: cs.LG, cs.AR, cs.PF arXiv: https://arxiv.org/abs/2609.36074v1 Memory-efficient scaling on clustering problems without sacrificing statistical accuracy is of central interest for large-scale data analysis and machine learning problems. Nonnegative low-rank (NLR) matrix factorization for $K$-means is a scalable clustering method, which connects to semidefinite relaxations with optimal average-case exact recovery guarantees. However, a direct GPU implementation of NLR requires multiple large factor-sized buffers and substantial data movements that are essentially memory-bound. In this paper, we introduce GEM-KMeans, a spectrally normalized yet mathematically equivalent NLR formulation that fuses the gradient update, nonnegative projection, and sufficient statistics for normalization and iterate movement into a matrix-multiplication epilogue. Instead of retaining three massive factor-sized arrays, our IO-aware GPU implementation materializes only one single factor with small tile-reduction arrays as additional storage in the High Bandwidth Memory (HBM). We derive explicit memory costs and spectrally normalized smoothness bounds for optimizing the clustering objective function. Accurate clustering is demonstrated at massive scales on synthetic and real datasets, where performance gains of GEM-KMeans over existing GPU-accelerated Lloyd's algorithms involve data-dependent runtime tradeoffs. --- ### LongCat-DeepResearch Technical Report Authors: Meituan LongCat Team, He Zhu, Yue Xu, Wanli Wu, Haolin Ren, Yuxin Bian, Jiarui Zhao, Rongzhi Zhang, Quanchi Weng, Jinghao Cui, Yu Fan, Yuhan Liu et al. Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36071v1 We present LongCat-DeepResearch, a deep research system that combines an enhanced LongCat model with a multi-agent workflow for producing comprehensive, evidence-grounded reports. The workflow separates global planning from detailed investigation and coordinates revision at the section level. Multiple planning agents first explore external sources and refine an actionable research plan, termed ResearchSpec. Research agents then investigate and draft their assigned sections in parallel, gathering additional evidence in separate contexts as their analyses develop. Once the sections are assembled, global review guides targeted local revisions, reducing reliance on repeated full-report rewriting. This workflow also supports the construction of research tasks and trajectories for the mid-training and post-training of LongCat's general-purpose models. LongCat-DeepResearch achieves 55.25 on DeepResearchBench, 51.35 on DeepResearchBench II, and 79.83 on ResearchRubrics. On an in-house benchmark, it scores 76.04, ranking second among four compared systems. Development-set analyses show benefits from combining planning perspectives, while further planning refinement has mixed effects. Additional editing improves average automatic readability preference across two benchmarks, with different trends on each. --- ### Mixture-of-Kittens: MoE Megakernel for NVL72s Authors: Stuart H. Sul, Nash Brown, Henry Wildermuth, William Lin, Federico Cassano, Christopher Ré Published: 2026-09-28 Categories: cs.DC, cs.LG arXiv: https://arxiv.org/abs/2609.36070v1 AI accelerator systems are rapidly consolidating into scale-up architectures, where tens to thousands of GPUs communicate over high-bandwidth, single-hop fabrics. We find that existing Mixture-of-Experts (MoE) training systems, optimized for conventional scale-out networks, transfer poorly to this setting, often running slower than a naive baseline built with PyTorch and NCCL. With industry roadmaps pointing toward even larger scale-up domains, understanding the performance tradeoffs of this hardware regime is increasingly important. We present Mixture-of-Kittens (MoK), an MoE training system designed for Nvidia NVL72. MoK builds on three insights that unlock performance on scale-up domains: (1) choosing push- or pull-based communication per operator, (2) restructuring the computation-communication overlap, and (3) fully eliminating CPU-GPU synchronization. MoK distills these insights into a single deterministic training megakernel that fuses token dispatch, shared and routed expert FFNs, and token combine. Across MoE layer shapes from four widely used open-weight models, MoK delivers up to $2.37\times$ the throughput of the strongest publicly available baseline. In a production run on 512 GPUs spanning multiple GB300 NVL72 racks, MoK improves end-to-end training throughput by $1.41\times$. --- ### The Uneven Decline of Collective Knowledge Production: Evidence from Stack Overflow After Generative AI Authors: Myokyung Han, Taegyoon Kim, Jinhyuk Yun, Lanu Kim Published: 2026-09-28 Categories: cs.SE, cs.AI, cs.DL arXiv: https://arxiv.org/abs/2609.36069v1 Generative AI (Gen AI) is reshaping how individuals learn and work, but its consequences for collective knowledge, the shared body of knowledge that online communities produce together, remain poorly understood. Prior work has documented an aggregate decline in participation on knowledge-sharing platforms, but it remains unclear which specific kinds of knowledge are being lost first. We study this question using Stack Overflow, one of the largest online communities for software engineering, treating the release of ChatGPT-3.5 as a natural shock. Analyzing over two million questions posted between 2020 and 2025, we track how two dimensions of collective knowledge, difficulty and data availability, change following Gen AI's release. Using diverse methods and robust checks, we find consistent patterns. Easy questions decline sharply while difficult questions become more common, a pattern corroborated by rising code complexity. Data-rich topics and tags lose share of questions, while data-scarce ones gain ground. The two dimensions also interact: the decline in easy questions is concentrated specifically within data-rich domains, while difficult questions increase regardless of data availability. This pattern extends beyond Python across programming languages, with more prevalent languages showing sharper shifts. Together, our findings reveal that Gen AI's impact on collective knowledge is uneven, eroding easy, accessible knowledge first while more complex, less common knowledge persists. --- ### AerialDojo-200K: A Large-Scale Benchmark Suite for Open-World Aerial Object-Goal Search Authors: Tongtong Feng, Xin Wang, Haoran Hou, Ren Wang, Weiran Wang, Shaokai Zhu, Ziqi Jia, Hao Wang, Yu-Wei Zhan, Zongyuan Wu, Jinghao Cui, Wenwu Zhu Published: 2026-09-28 Categories: cs.CV, cs.AI, cs.ET, cs.MM, cs.RO arXiv: https://arxiv.org/abs/2609.36066v1 Open-world aerial object-goal search is a foundational yet challenging task, requiring aerial agents to autonomously explore large-scale, unstructured three-dimensional environments and reach target objects specified by semantic descriptions or reference images, rather than following route-specific instructions. However, research in this task remains at a nascent stage and relies on small, environment-specific benchmarks with heterogeneous action spaces and data formats. These limitations hinder large-scale training and cross-benchmark evaluation, constraining the scalability and generalizability of aerial agents. To address this problem, we propose AerialDojo-200K, a large-scale benchmark suite for open-world aerial object-goal search, with 3 times as many scenes and 18.7 times as many task instances as the largest existing benchmark for this task. Specifically, we construct 42 simulation scenes spanning four scene families and 21 scene types, including 18 urban, 12 natural, six infrastructure, and six disaster scenes. To ensure data quality, 12 annotators spent two months manually annotating 109 landmarks, 2099 target objects, and 2099 object anchors across these scenes. We further construct 205,732 task instances, comprising over 100K semantic-goal and over 100K image-goal instances across Base, Standard, and Long-Horizon settings. Each task instance includes a collision-free reference trajectory and corresponding multi-view video recordings. We also develop a unified evaluation framework with a scene partition comprising 21 in-distribution scenes and 21 out-of-distribution scenes. Finally, our evaluation of five open-source and four closed-source multimodal large language models reveals that there is still a long way to go toward achieving general-purpose aerial agents. All can be found at https://fengtt42.github.io/AerialDojo/. --- ### FLOORA: A Human-Aligned Domain-Specific Language Model for Architectural Design Authors: Sahand Rezaei-Shoshtari, Patryk Wozniczka, Shu Ishida, Gregg Streuber, Farnoosh Javadi, Jeffrey Landes, Angela Ju, Muhammad Azam, Bryan Lim, Johan Luttun, Indrajeet Haldar, Jonathan Shaw et al. Published: 2026-09-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.36064v1 Foundation models are powerful generators, but many engineering domains require structured representations that general-purpose systems handle poorly. We introduce FLOORA (Floor Layout Optimization with RL Alignment), a family of small domain-specific language (DSL) models for architectural layout generation. With specialized data and alignment, our 0.6B model outperforms much larger frontier models, achieving VLM judge win rates up to 92.0% on out-of-distribution real-world buildings and 96.0% on synthetic buildings. Human evaluations further corroborate these results, with FLOORA selected as the best model in 89.3% of evaluations. FLOORA combines a token-efficient DSL, custom tokenization, domain-specific pretraining, supervised fine-tuning (SFT), and reinforcement learning (RL) with learned human-preference and verifiable rewards. This pipeline improves architectural and geometric validity, supported by extensive empirical evaluation and ablation studies. Although focused on architecture, our results suggest that similar domain-specific recipes may be useful in other engineering domains with structured, verifiable outputs. Datasets, models, and inference code are available at https://github.com/AutodeskAILab/floora. --- ### Understanding Decision-Making Mechanisms in Neural Routing Solvers Authors: Fatemeh Askari, Mazdak Teymourian, Mohammad Izadi, Mahdieh Soleymani Baghshah Published: 2026-09-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.36063v1 Neural Combinatorial Optimization (NCO) has achieved strong empirical success, yet the internal mechanisms driving model decisions remain largely unexplored. In this paper, we investigate three representative autoregressive NCO models spanning two encoder-decoder configurations: AM and POMO (heavy-encoder, light-decoder), and LEHD (light-encoder, heavy-decoder). Through behavioral analyses, representation probing, and causal interventions, we examine how these models construct solutions and use internal representations during decoding. Our results suggest that AM and POMO predominantly follow a persistent geometric pattern throughout solution construction, whereas LEHD contains linearly accessible information about multiple future actions. Causal experiments further provide evidence for the role of future-node representations in LEHD's decision-making. We also observe that LEHD relies strongly on the current-node representation for immediate local decisions, while the start-node representation plays a broader navigational role over the subsequent route. Cross-instance alignment analyses additionally indicate that LEHD maps current-node representations into a relatively shared latent region, which may provide a stable reference for evaluating subsequent decisions. Across the Traveling Salesman Problem and the Capacitated Vehicle Routing Problem, these results reveal distinct decision-making patterns across these architecturally distinct solvers and provide a foundation for more interpretable analyses of NCO solvers. Code and additional visualizations are provided in the https://github.com/NCO-Interpretability/NCO-Interpretability. --- ### SMat-Attention: Structured Long-Context Sequence Modeling Authors: Emile Anand, Abdullah Ateyeh, Archer Wang, Marin Soljačić Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36062v1 Long-context sequence models face a fundamental tradeoff: softmax attention uses flexible token-level interactions at quadratic cost, whereas linear attention obtains linear-time training and constant-time decoding by compressing history into a fixed-size state. In this work, we ask whether we can connect these regimes through a tunable notion of structure. To this end, we introduce Structured Matrix Attention (SMat-Attention) via a family of causal masks with structured long-range routing whose row supports have VC-dimension $d$. In our construction, $d=1$ recovers the standard causal mask, and increasing $d$ permits richer subset-routing patterns. We give chunkwise forward and backward algorithms to enable hardware-efficiency. For sequences of length $T$, the hard-routing construction takes $O(T^{2-3/d}+T)$ work, despite the mask being dense, for our prescribed family. In fixed-horizon streaming, decoding after the distant prefix takes constant time per token using $O(T^{1-1/d})$ cached states. SMat-Attention therefore makes VC-dimension an explicit knob governing access-pattern complexity, prefill cost, and decoding memory. Empirically, subset-routing and rule-assisted multi-key retrieval experiments illustrate the masks' routing expressiveness. Extensions to Mamba-2 and Gated DeltaNet using learned routing with top-$k$ query reads retain subquadratic prefill, improve recall accuracy over the backbones in several settings, and achieve comparable small-scale language-modeling performance. --- ### Introducing the CZAR Loss: A Tailored Objective Function for Financial Log-Return Predictions Authors: Joel Pfeffer, J. M. Diederik Kruijssen, Florian Stecker, Steven N. Longmore Published: 2026-09-28 Categories: cs.LG, cs.CE, q-fin.CP, q-fin.MF arXiv: https://arxiv.org/abs/2609.36061v1 In quantitative finance, standard regression losses are misaligned with the economics of return prediction. As the conditional mean of financial log-returns is close to zero, symmetric losses such as the mean squared and mean absolute errors make the constant zero forecast a near-optimal solution, penalizing models with genuine but noisy directional skill. This applies both during training, where predictions shrink toward zero, and during evaluation, where trivial forecasters can lead loss-based rankings. Under a Gaussian linear prediction model, we show that all symmetric monotonic losses share a universal breakeven directional accuracy against the zero predictor, which rises sharply and becomes unobtainable as the prediction noise approaches the standard deviation of the returns. We introduce the CZAR (Composite Zero-Agnostic Return) loss function, a piecewise quadratic loss built around five key requirements: convexity in the prediction, asymmetry with true return direction that vanishes at zero, near-linear penalization of undershoots and wrong-direction predictions, divergence for large errors, and an adaptive loss floor for evaluation. CZAR is provably convex in the prediction at fixed true value, has closed-form gradient and Hessian suitable for custom objectives in gradient-boosted libraries, and its four hyperparameters reduce to a single choice through correlated defaults. In idealized tests, the minimum directional accuracy required for a CZAR-evaluated forecaster to outperform the zero predictor under mean log loss remains near the 50% chance level, whereas the corresponding threshold for symmetric losses rises sharply with prediction noise. This advantage persists under heavy-tailed return distributions. In a LightGBM experiment on intraday BTC log-returns, CZAR-trained models reduce the `zero-returns bias' and improve directional accuracy on large-magnitude returns. --- ### Mnemon: Raw Records, Fast Judgments, Slow Thoughts Authors: Guangren Wang Published: 2026-09-28 Categories: cs.CL, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2609.36059v1 Long-term memory lets an LLM assistant use a history it can no longer reread, and most memory systems build it by rewriting conversations into facts, graphs or typed memories at write time. We argue that the work of memory divides, as thinking does, into two systems. Most of it is fast System 1 work: many small, independent yes/no judgments about records, such as whether a record is needed or no longer current, which a decision model makes by the dozen in a third of a second. Only a little is slow System 2 work: writing a few search queries, naming what the reply needs and composing the answer, which an LLM does well but slowly. We present Mnemon, a memory agent built on this division. It keeps conversations as raw, dated records; an LLM (System 2) plans searches over them, a decision model, Jev (System 1), judges what the searches return, and rules with explicit budgets turn the judgments into a small View for an unchanged answering model. A background pass consolidates each record once into topic timelines, value histories and standing instructions linked to the records, so that questions about a whole conversation reach evidence their own searches miss. Because nothing is decided about a record when it is written, the same agent can read any store that returns dated records. With gpt-4.1-mini answering, as in a public re-evaluation of 14 systems, Mnemon scores 91.7% on LoCoMo, the highest among them, and 83.8% on LongMemEval-S, from under 4k tokens of context per question, with the lowest effective cost index on LoCoMo. With a reasoning model answering, it reaches 92.2% on LoCoMo and 94.4% on LongMemEval-S, the latter on par with the best published results. From 100K to 10M tokens of history on BEAM, its cost per question grows by a factor of 1.11. On the same records, Jev separates gold evidence better than two LLMs and is 3-11 times faster. --- ### ABC: Advantage-Based Control Variates for Reinforcement Learning with Verifiable Rewards Authors: Hsiao-Ru Pan, Florent Draye, Bernhard Schölkopf Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.36058v1 Recent progress in reinforcement learning with verifiable rewards (RLVR) has highlighted the effectiveness of simple critic-free policy-gradient methods such as Group Relative Policy Optimization (GRPO). In contrast, actor-critic methods rely on learned value functions whose approximation error can introduce bias through commonly used advantage estimators such as temporal-difference error. Motivated by this observation, we revisit trajectory-level control variates through an advantage-value formulation, which we call Advantage-Based Control Variates (ABC). This formulation reveals that the covariance structure is closely related to the return decomposition used in Direct Advantage Estimation (DAE). Finally, we combine ABC with DAE into a single actor-critic algorithm and evaluate it in an offline-to-online RLVR setting, where the critic is first trained on previously collected trajectories and adapted during online learning. On mathematical reasoning tasks, ABC achieves performance competitive with GRPO using substantially fewer online optimization steps. --- ### Mirror-Score: Calibrated, Inference-only Scoring Exposes the Limits of Sequence-compatibility Ranking in D-peptide Design Authors: Jiada Li Published: 2026-09-28 Categories: cs.AI, cs.CE, cs.ET, cs.LG, q-bio.QM arXiv: https://arxiv.org/abs/2609.36057v1 D-peptides combine protease resistance with high target specificity, but computational design of D-peptide binders remains immature. Mirror-Peptidizer introduced an in silico mirror-image screening pipeline using target reflection, backbone generation, and ProteinMPNN sequence design, but its raw ProteinMPNN negative log-likelihood (NLL) ranking was not validated against measured affinities, and only 4 of 9 tested MDM2 designs bound detectably. We introduce Mirror-Score, a calibrated, inference-only scoring framework for heterochiral D-peptide/L-protein complexes, and a public benchmark of 31 crystal complexes across four target families, including 18 with literature-verified affinities. Raw ProteinMPNN NLL is not a valid affinity ranker: its pooled Spearman correlation with affinity is 0.19, and correlations reverse between MDM2/CHIP (+0.62) and gp41 (-0.70). We therefore evaluate Boltz-2 mirror-space cofolding confidence. For the complete viral-entry family (7 structures representing 3 peptides), interface predicted local distance difference test (pLDDT) achieves structure-level leave-one-out Spearman rho = 0.90 (p = 0.006) and correctly orders all three peptides by affinity, whereas NLL fails (structure-level rho = 0.18). Because only three independent chemotypes are represented, this result indicates directional consistency rather than a statistically validated predictor. Cross-family calibration does not transfer at current sample sizes, supporting family-matched calibration as the practical deployment mode. We also specify a prospective design protocol for the antimicrobial-resistance targets LasR and LecB from Pseudomonas aeruginosa, including mirrored structures, ligand-derived hotspot maps, diffusion-model-ready inputs, and Mirror-Score ranking. Code, benchmark data, structures, and analysis scripts are openly available at https://github.com/Jiadalee/Mirror-Score. --- ### GeoWind2Plan: Mission-Time 3D Urban Wind Prediction for Energy-Efficient UAV Planning Authors: Shaoxiang Qin, Yucheng Zhao, Fuyuan Lyu, Di Zhou, Jiachen Yao, Xue Liu, Anima Anandkumar, Liangzhu Leon Wang, Xiongye Xiao Published: 2026-09-28 Categories: cs.AI, cs.RO arXiv: https://arxiv.org/abs/2609.36056v1 In urban low-altitude flight, buildings reshape ambient wind into spatially varying 3D flow, making unmanned aerial vehicle (UAV) energy depend on local wind exposure as well as path length. However, building-resolved wind information is rarely available when a mission must be planned. Computational fluid dynamics (CFD) can produce high-fidelity urban flow fields, but each simulation is tied to a fixed inflow boundary condition and can take hours to days, which is incompatible with urban UAV missions that typically last minutes to tens of minutes. We present GeoWind2Plan, a geometry-to-wind-to-planning framework for mission-time 3D urban wind prediction and energy-efficient UAV planning. Given only a background wind vector, 3D building geometry, and a start-goal pair, GeoWind2Plan transforms the building geometry into a reference-wind frame, predicts mission-relevant 3D wind patches with a localized geometry-conditioned neural operator, stitches them into a queryable local wind field, and optimizes a feasible 3D path and speed profile using a physically grounded UAV energy model. Rather than pursuing CFD-perfect reconstruction, GeoWind2Plan targets decision-useful wind prediction: trajectories are planned with predicted wind and evaluated under high-fidelity CFD wind. Across held-out urban domains, wind speeds, and mission wind-angle regimes, GeoWind2Plan performs corridor-localized wind inference in about 3 seconds, compared with roughly 8 hours for CFD. Under CFD evaluation, trajectories planned with GeoWind2Plan reduce energy by 6.9%, 12.7%, and 4.5% in tailwind, headwind, and crosswind missions relative to wind-agnostic planning, recovering 87.9%, 85.7%, and 75.0% of CFD-reference savings. These results show that fast, corridor-localized 3D urban wind prediction can make wind-aware UAV energy planning practical at mission time. --- ### What if automating AI R&D triggers an intelligence explosion? Authors: Alan Chan, Christoph Winter, Andrew Barto, Jakub Pachocki, Geoffrey Hinton, Eric Horvitz, Yoshua Bengio, Dawn Song, Jack Clark, Hilary Greaves, Anton Korinek, Samuel Hammond et al. Published: 2026-09-28 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2609.36054v1 In contrast to even a year ago, AI systems now write most of the code inside the companies that build them. As more of the AI research and development (R&D) pipeline is automated, could AI progress radically accelerate in an "intelligence explosion," where years of advances are compressed into months or less? Preliminary evidence suggests that it could. In this work, we assess this evidence, analyze an intelligence explosion's potential impacts, and propose policy responses. AI systems are on track to automate most AI R\&D work within a few years, and possibly all of it. If this triggers an intelligence explosion, it could dramatically bring forward AI's benefits, but also pose extreme risks: capabilities growth could accelerate far beyond what society can keep up with, humanity could lose control over superhuman AI systems, and checks on power within and between states, companies, and branches of government could be severely eroded. Although there remains much uncertainty about these possibilities, the high stakes warrant serious further attention. Policymakers should urgently obtain more visibility into the automation of AI R&D, develop ways to steer and constrain an intelligence explosion, and prepare society to adapt to an intelligence explosion's impacts. --- ### PowerZooJax: A JAX-based Power System Benchmark for Reinforcement Learning Authors: Zhanhua Pan, Xiao Liu, Zhilong Cao, Jianhong Wang, Dawei Qiu Published: 2026-09-28 Categories: cs.AI, cs.LG, eess.SY arXiv: https://arxiv.org/abs/2609.36052v1 Power system operation is a safety-critical sequential decision-making problem, making it a natural testbed for reinforcement learning (RL). However, existing RL environments for power systems are often narrow in scope and computationally limited by CPU-based simulation workflows, making large-scale evaluation difficult. We introduce PowerZooJax, a JAX-based benchmark suite for RL in power system operation. It provides five constrained Markov decision process tasks spanning generation, transmission, distribution, distributed energy resources, and data center microgrid. By rewriting power flow, economic dispatch, market clearing, and device dynamics as JAX computation graphs, PowerZooJax keeps the entire training and evaluation loop on the GPU. Experiments show substantial speedups over CPU-based simulations and demonstrate standardized evaluation of policy returns, safety violations, and out-of-distribution stress conditions. Our open-source benchmark is available at: https://github.com/powerzoojax/PowerZooJax. --- ### Improving scalable oversight with co-trained monitors Authors: Joseph H. Rudoler, Kevin Tan, Benedict Tessler, Timothy Kong, Enric Boix Adserà Published: 2026-09-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.36049v1 Worker-monitor setups are a promising approach to AI oversight, but training workers against fixed monitors can incentivize monitor evasion. We study whether this failure mode can be avoided by co-training the monitor alongside the worker, and explore both supervised and self-supervised approaches. In the supervised setting, we prove a characterization: monitoring is possible with vanishing error and query rates exactly when the class of possible monitor functions has finite Littlestone dimension. This connects worker monitoring with an established literature on adversarial online learning. For self-supervision, we propose a co-training procedure based on test-time distillation: the monitor uses additional test-time compute to generate training labels, then trains its standard-compute policy on those labels. For majority-vote labels, we give a finite-sample sharpening guarantee under adaptive worker distributions with action coverage, that shows that the monitor's verdicts converge to its initial modal verdicts. We stress-test the former in code-security settings where the worker is trained adversarially to fool the monitor. Our results suggest that adaptive monitors are better at keeping pace with evolving worker strategies, while fixed monitors are more vulnerable to evasion. --- ### Neural networks for spectral optimization Authors: Alexis de Villeroché, Beniamin Bogosel, Stéphane Breuils, Dorin Bucur, Jacques-Olivier Lachaud Published: 2026-09-28 Categories: cs.LG, cs.AI, math.OC arXiv: https://arxiv.org/abs/2609.36047v1 Given a functional dependent on the spectrum of a differential operator, we address the problem of finding a domain which optimizes this functional. PDE solvers might be used to tackle this optimization. It is however computationally expensive. We propose two neural network models which learn the spectrum directly from the geometry of the domain and can be used to optimize the domain from one or more eigenvalues. We investigate two representations. The first encodes the domain through Fourier coefficients and a light MLP, which is efficient on star-shaped geometries, achieving a precision of 0.2\%. Through a rescaling of the coefficients the designed models satisfy the scaling law of the eigenvalues. Additionally, averaging the outputs of the trained surrogates over rotations and reflections induces invariance for these transformations. The second is a model that takes the landscape function, the indicator function and the gradient of the landscape function. A Gram-Schmidt process produces orthogonal eigenfunctions as output of the model along with the associated eigenvalues. The landscape model reaches 1\% mean relative error on the first ten eigenvalues, compared with 4\% for an FNO model. Replacing the landscape by an SDF worsened both prediction and optimization errors. The trained model also generalizes from synthetic shapes to domains given as classical image dataset. The resulting surrogates of both approaches recover classical spectral optima such as the disk for the first eigenvalue or the conjectured minima of higher eigenvalues. This confirms that our models produce accurate differentiable estimates of eigenvalues, which can be used in shape optimization problems involving spectral quantities. --- ### Graph-Split Bayesian Causal Forest for Spatial Heterogeneous Treatment Effect Estimation Authors: Shuren He, Huiyan Sang, Ligang Lu Published: 2026-09-28 Categories: stat.ME, stat.ML arXiv: https://arxiv.org/abs/2609.36046v1 In spatial observational studies, treatment assignment and outcomes often exhibit spatial dependence patterns, and treatment effects may vary across space and subpopulations due to both measured and unmeasured spatially structured confounders. Accounting for spatial dependence while estimating heterogeneous treatment effects (HTEs) is a central task in spatial causal inference. Causal Bayesian additive regression tree methods are popular nonparametric methods for modeling and estimating HTEs. Despite their flexibility and uncertainty quantification, the axis-aligned split rules often adopted in these models are not suitable for modeling spatial structures. We propose a spatial structure-aware Bayesian nonparametric method, called Graph-Split Bayesian Causal Forest (GSBCF), that integrates graph-split Bayesian additive regression trees (GS-BART) with the Bayesian causal forest propensity-score regression framework for spatial heterogeneous causal inference. Spatial confounding is accommodated through graph-guided split rules in modeling decision trees of the prognostic and HTE functions. We develop an efficient informed proposal sampling algorithm for posterior computation, enabling full Bayesian inference of the spatial conditional average treatment effect function. Simulations and a real data study demonstrate substantially improved estimation accuracy and uncertainty quantification over existing causal BART methods. --- ### Hybrid Neural Simulation-Based Inference for Robust Applications and Limited-Budget Scenarios Authors: Sean Benevedes, Mani Dehghan, Aishik Ghosh, Tae Hyoun Park Published: 2026-09-28 Categories: hep-ph, cs.LG, hep-ex, physics.data-an arXiv: https://arxiv.org/abs/2609.36044v1 We develop two hybrid techniques that approach the performance of neural simulation-based inference (NSBI) analyses while substantially reducing the computational cost of inference and preserving some or all of the reliability guarantees of parametric methods. The first approach is broadly applicable, while the second is tailored to a class of particle physics analyses that admit a semi-parametric NSBI formulation. With only a modest compromise in raw sensitivity, these methods represent an important step toward computationally efficient NSBI in offline analyses and also open the door to the exploration of trigger-level applications in the future. Based on our comparison studies, we recommend the use of our first approach, Latent Categories, for robust and efficient inference. --- ### SAGE: A Statistical Acceptance Gate for Self-Evolving Agents Authors: Yihao Wang, Linhan Xia, Rui Liu, Zhaofeng Zhang, Hongyu Wu, Yang Yang, Jinglu He, Yu Guo, Kai Lei Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.36043v1 Large Language Model (LLM)-based agents increasingly self-evolve by editing a persistent skill document that encodes their workflow, tool-use rules, and decision logic. This loop has two steps, an optimizer that proposes a candidate edit and a gate that accepts or rejects it. Prior work has concentrated on the optimizer, while the gate still follows a naive rule that keeps any edit which improves an aggregate validation score. We show that this rule fails in two ways. First, it admits permanent regressions, since an edit can raise the average while breaking items the skill already solves. Second, it is vulnerable to the Optimizer's Curse, since the best observed score on a finite and noisy validation set is upward biased. To solve the above two limitations, we propose a statistical acceptance gate for self-evolving agents (SAGE). Compared with previous work, SAGE has two contributions. First, SAGE proposes a per-item paired comparison that evaluates the current skill and the edited skill on identical validation items, which exposes regressions that an aggregate score hides and penalizes them asymmetrically. Second, SAGE also employs a one-sided paired test that commits an edit only when its wins are statistically reliable against its losses, and it abstains otherwise. SAGE is a conservative refinement of the standard gate that recovers the baseline exactly at a boundary setting. It commits only a subset of the baseline's edits, filtering out those whose gains are unreliable or purchased by breaking already-solved items. Across five benchmarks and four backbone LLMs under an equal-budget protocol, SAGE lowers the regression rate in 19 of 20 settings and matches the baseline in the remaining one, for example from 36.5% to 0% on LiveMath and from 42.8% to 0% on OfficeQA with DeepSeek-V4. SAGE also attains the highest final score in all 20 settings, raising LiveMath from 34.15 to 48.78. --- ### Multi-Class, Multi-Tier Network Intrusion Detection: A Comprehensive and Reproducible Benchmark Authors: Yufeng Xin, Bryant Goseland, Mohamed Rahouti Published: 2026-09-28 Categories: cs.CR, cs.LG arXiv: https://arxiv.org/abs/2609.36039v1 Machine learning (ML) and deep learning (DL) have dominated Intrusion Detection System (IDS) research in recent years. Unfortunately, many existing studies have produced inflated results and unreliable benchmarks due to critical oversights and mistakes in the ML and DL pipeline, from data collection and labeling to feature engineering and model training and evaluation. CIC-IDS2017 is a standard benchmark for network intrusion detection. Still, many published results on this dataset are difficult to compare due to labeling errors, inconsistent flow extraction, potential leakage, and performance evaluation metrics dominated by benign traffic. In this paper, we present a comprehensive benchmark with corrected PCAP-level labeling and a complete evaluation pipeline with diverse ML models. We evaluate eleven tabular classifiers at three nested levels: binary attack detection, nine-class attack-family attribution, and fifteen-class fine-grained classification. A soft-voting ensemble of Random Forest, XGBoost, and LightGBM obtains the best fine-tier macro-F1 of 0.955, with coarse and binary macro-F1 scores of 0.980 and 0.999, respectively. We further conducted a feature selection study based on an analysis of feature importance. This comprehensive benchmark pipeline is configurable and open-source, enabling new feature extraction and model plugins for new datasets. Future work should use this pipeline as a reference point for richer features, rare-class analysis, and model generalization towards new datasets and attack classes. --- ### Dual-Anchor Acceleration Is Near-Optimal for Stochastic Monotone Root-Finding Authors: TaeHo Yoon, Nicolas Loizou Published: 2026-09-28 Categories: math.OC, cs.LG arXiv: https://arxiv.org/abs/2609.36033v1 Among distinct optimal acceleration mechanisms for deterministic monotone root-finding problems and fixed-point problems, dual-anchoring has recently been shown to admit a more robust direct stochastic extension than standard anchor acceleration. However, without additional strong monotonicity, the existing stochastic dual-anchoring guarantee has two limitations: first, it requires cocoercivity in expectation, and second, it attains only $O(ε^{-3})$ oracle complexity, leaving a gap to the near-optimal $\tilde{O}(ε^{-2})$ complexity achieved by other methods. In this work, we address both of these limitations by combining dual-anchoring with stochastic resolvent approximation and optimized variance control. For unbiased stochastic oracles with variance bounded by $σ^2$, where sample operators are monotone and uniformly $L$-Lipschitz, our algorithm finds a point with $ε$-residual with a near-optimal oracle complexity of $O ( (LD / ε) \ell + (σ^2 / ε^2) \ell^2)$, where $\ell = \log (1 + LD / ε)$ and $D$ is the initial distance to a solution. This result improves the best known oracle complexity in the noise-dominated regime under these samplewise assumptions, reducing the poly-logarithmic factor from cubic to quadratic. --- ### TORQUE: Optimizing What (not) to Quantize Before and After Rotation Authors: Ran Ben Basat, Michael Mitzenmacher, Shay Vargaftik Published: 2026-09-28 Categories: cs.LG, cs.AI, cs.IT, cs.NI arXiv: https://arxiv.org/abs/2609.36032v1 Uniform random rotations are an effective preprocessing step for quantization: they make normalized coordinate distributions approximately Gaussian, enabling the use of codebooks optimized offline. We introduce TORQUE, a framework that improves on previous quantization works that use random rotations by jointly optimizing how many and which coordinates to preserve at high precision both before and after rotation, under a fixed overall expected bit budget. Intuitively, before rotation, preserving large input coordinates at high precision can reduce overall error by preventing the rotation from spreading their values across many coordinates. Likewise, after rotation, preserving a small fraction of the largest-magnitude coordinates at high precision allows the remaining values to be quantized more accurately using codebooks optimized offline for the resulting truncated Gaussian distribution. We derive a quantization error upper bound and prove that top-$k$ pre-rotation retention minimizes it for each $k$. This reduces the search over coordinate subsets to an optimization over $k$, enabling a fast optimizer that uses offline codebooks and parallel parameter selection for practical implementation. We demonstrate an improved tradeoff between reconstruction accuracy and storage cost through numerical evaluation under the Gaussian model and experiments on nearest-neighbor retrieval, KV-cache compression, and activation compression. --- ### Preferent Compression Bounds Are Tight Authors: Dario Paccagnan, Marius Tirlea Published: 2026-09-28 Categories: cs.LG, eess.SY, math.OC, math.PR arXiv: https://arxiv.org/abs/2609.36030v1 The lack of rigorous safety and performance certificates remains a key bottleneck to the deployment of modern learning-based methods. Sample compression has recently emerged as a powerful tool for deriving such certificates, with particularly sharp bounds available for algorithms satisfying a so-called preference property -- also known as stability in the learning theory literature. These bounds find direct application across domains as different as the Scenario Approach, Pick-to-Learn, and Support Vector methods. However, whether they are tight has remained an open problem. In this paper we resolve this question affirmatively and show that the state-of-the-art bound for preferent compressions is provably tight. We establish this by exhibiting an explicit construction based on the uniform distribution and order statistics that attains the bound in the limit. Along the way, we also provide a considerably shorter and more accessible proof of this bound, requiring only elementary counting arguments and no infinite-dimensional duality. --- ### CoDimRecon: Agentic Reconstruction of Sim-Ready 3D Scenes with Deformable Curves, Surfaces, and Volumes Authors: Shuzhao Xie, Lelin Wang, Guying Lin, Zhi Wang, Minchen Li Published: 2026-09-28 Categories: cs.CV, cs.GR, cs.RO arXiv: https://arxiv.org/abs/2609.36024v1 Reconstructing simulation-ready 3D scenes from real-world observations enables robotics, gaming, and immersive applications, yet existing methods largely assume rigid objects. This leaves an important gap for deformables, whose simulation-ready geometry depends on dimensionality (curves, surfaces, or volumes) and whose behavior may require models beyond elasticity. We present CoDimRecon, an agentic framework that reconstructs editable scenes containing rigid, articulated, and deformable objects from multi-view RGB observations. Scene-level geometric priors ground scale and layout, while object-level generated meshes guide the agent toward detailed, compact geometry; articulated rigid objects are decomposed into movable parts with explicit joints. For deformables, category-wise agent sessions reconstruct curves as centerlines with radii, surfaces as manifold shells with thickness, and volumes as watertight solids for volumetric meshing. Reusable simulator skills initialize compatible physical models and parameters, while agent-guided behavioral tests expose mismatches and trigger targeted revisions of motion, geometry, numerics, or material modeling. On evaluated Replica and ScanNet++ scenes, CoDimRecon achieves competitive compositional reconstruction accuracy while additionally producing deformable assets for rod, shell, and solid simulation. We further demonstrate robot interactions across all three representations, including a controlled paper-folding case in which behavioral testing motivates plastic bending. --- ### Persistence Forcing: Exploiting Feature Specialization in Pixel-Space Diffusion Authors: Chong Wang, Zixuan Fu, Shiqi Huang, Siyuan Yang, Hao Cheng, Bihan Wen Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.36014v1 Pixel-space diffusion Transformers (DiTs) directly operate on high-dimensional visual data, yet their hidden representations typically undergo uniform refinement across depth. Natural images, however, are inherently organized at different levels of granularity. Global structure can often be represented compactly, whereas local textures and fine details require richer representations. Motivated by this, we introduce heterogeneous refinement in pixel-space DiTs, assigning different feature groups distinct refinement budgets across depth. Consequently, an ordered feature specialization emerges: sparsely refined features predominantly encode global visual structure, whereas more frequently refined features increasingly specialize toward localized, high-frequency details. We refer to these two groups as persistent and active features, respectively. Building on this emergent specialization, we introduce Persistence Forcing (PerF), which explicitly exploits this persistent--active feature organization for pixel-space image generation. This enables persistent features to continuously condition actively refined features, allowing stable global information to guide the ongoing refinement of finer visual details. During generative sampling, this interaction further induces a meaningful guidance direction that promotes coherent global structure and naturally complements classifier-free guidance. On ImageNet $256\times256$, PerF-L achieves FID of $1.91$, approaching $1.86$ of JiT-H with only half the parameters, while PerF-H further achieves FID of $1.63$ and $1.76$ on ImageNet $256\times256$ and $512\times512$, respectively. --- ### In-Context Learning for Robots: Methods and Applications Authors: Haojian Huang, Zexi Li, Junhao Guo, Yehang Zhang, Wenxuan Peng, Bohan Zhou, Weilin Ruan, Leyi Wu, Chenxu Wang, Jianchong Su, Binghui Xie, Wosong Chen et al. Published: 2026-09-28 Categories: cs.RO, cs.LG arXiv: https://arxiv.org/abs/2609.36012v1 General-purpose robots must infer what a new task requires and translate that understanding into appropriate physical action. In-context learning (ICL) for robots supports this process by using demonstrations and interaction to direct existing competence with neural parameters held fixed during deployment. We organize this literature review around the interfaces connecting contextual evidence to execution, distinguishing four families: context-conditioned policies, geometric demonstration transfer, world-model-based control, and skill- and agent-based execution. Comparing these interfaces clarifies their transfer assumptions and the roles of training, correspondence, and memory in making context useful. Across manipulation and navigation, we examine how these mechanisms preserve taught requirements as objects, environments, and execution conditions change. This analysis links method design to evaluation practices that distinguish responsiveness to teaching, physical transfer, and benefits from retained experience. The resulting agenda connects compositional task acquisition and faithful transfer with physical recursive self-improvement, in which experience improves the ability to learn subsequent tasks. --- ### Infrared Subtraction with Artificial Intelligence Authors: Wenjie He, Xiaohui Liu, Yandong Liu, Zhan Wang Published: 2026-09-28 Categories: hep-ph, cs.AI, hep-ex, nucl-ex, nucl-th arXiv: https://arxiv.org/abs/2609.36007v1 We present AI-developed local infrared subtraction, building on projection to Born and EFT matching. The framework separates an integrable radiation term from a finite contribution at Born kinematics, referred to as the Born contact. The contact is determined using the EFT singular distribution in a resolution observable such as N-jettiness $τ_N$. Under human physics guidance, an LLM develops two implementations. One uses a neural network for phase space projection and fits the contact by matching to EFT cumulants. The other uses an analytic construction that keeps the Born momenta fixed while integrating over radiation. It combines the EFT $δ(τ_N)$ coefficient with finite 4-dimensional radiation integrals to calculate the contact term directly. This gives a local subtraction formula without a slicing parameter, while reusing existing lower-order radiation calculations and EFT singular predictions. As a demonstration, we reconstruct the full NLO correction for massless 3- and 4-jet production in electron-positron annihilation. The attempt to the NNLO dijet production is also made by recursively using the NLO P2B construction with the LLM designing machine-learning controls to reduce the variance of the contact integral. The tested predictions are in good agreement with EERAD3. The numerical calculation and projection-network training use a 2020 Apple M1 MacBook, without GPU acceleration, illustrating the feasibility of the construction with modest computing resources. The appendices develop an extension of the local subtraction to 3-jet NNLO, giving explicit radiation maps and a proposed contact formula. We also show how to integrate over NNLO radiation while keeping the Born momenta fixed, for any number of massless final-state jets. Our results demonstrate how AI can help higher-order calculations by constructing infrared subtraction and improving its numerical integration. --- ### Making Cross-Continental Federated Learning Repeatable with FLIP: a Multi-Application Study Authors: Rafael Garcia-Dias, Alexandre Triay Bagur, Chayanin Tangwiriyasakul, Virginia Fernandez, Parhom Esmaeili, Piyalitt Ittichaiwong, Yang Li, Lawrence Adams, Wason Buncharoen, Martin Chapman, Benjamaporn Chayanond, Sadthavud Chunrod et al. Published: 2026-09-28 Categories: cs.LG, cs.CV, cs.DC arXiv: https://arxiv.org/abs/2609.36001v1 Federated learning (FL) in healthcare remains challenging, as the overhead of rebuilding governance guarantees for every collaboration stops most projects at the proof-of-concept stage. Here we present FLIP (Federated Learning Interoperability Platform), an open-source, multi-application platform that makes FL training and evaluation repeatable. FLIP implements common FL workflows as a set of composable services: cohort queries against per-site structured databases, on-demand DICOM retrieval from institutional PACS, per-site project approval, and reusable FL job types. To demonstrate FLIP, we ran two distinct use cases, federated fine-tuning and federated evaluation, on synthetic chest X-ray cohorts across two client nodes based in the United Kingdom (UK) and Thailand. In FLIP, each institution independently approves its participation in each project and operates its own node under local IT security processes. This study makes an operational rather than an algorithmic claim. It does not compare federated with centralised training; for that question, we refer the reader to existing systematic reviews and meta-analyses. The central result is evidence that such platforms enable international FL collaboration and improve repeatability, auditability, and site-specific governance. We also present a comprehensive comparison of existing platforms to help researchers and operators choose the right platform for their use case. --- ### Causal and Interpretable Structures in LLM Compositional Tasks Authors: Gurbir Arora, Toni J. B. Liu, Jiajun Bao, Raphaël Sarfati, Christopher J. Earls Published: 2026-09-28 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2609.35970v1 Large language models are able to solve tasks whose answers depend on not only individual input tokens, but also on relations among them. How is such relational information represented and processed across transformer layers? We study activations from ensembles of prompts that require inferring relationships between three tokens corresponding to a cyclic concept (months, hours, weekdays, and musical notes) to correctly predict the next token. Across model families (Llama, Qwen, Gemma, and Mistral) and cyclic concepts, we find a consistent layerwise progression in how the joint dependence among the tokens is geometrically organized and causally used: intermediate layers use a joint representation based on the inferred relationship between two tokens, while later layers use a joint representation associated with all three tokens to correctly complete the task. We also find other relationships between tokens that are geometrically structured but remain causally inert in the next-token prediction. Crucially, when taken together, these geometric and causal investigations reveal the representation-level mechanism that progressively organizes and composes the relational information to form the answer. More surprisingly, restricting the models to such causally relevant joint representations improves next-token prediction accuracy. --- ### SIFARI: Self-Supervised Interferometric Fitting for Astronomical Radio Imaging Authors: Shunyuan Mao, Andrea Isella, Paris Perdikaris, Li-Ta Lo, Hui Li Published: 2026-09-28 Categories: astro-ph.IM, astro-ph.EP, cs.LG arXiv: https://arxiv.org/abs/2609.35966v1 Radio-interferometric images are reconstructed from sparsely sampled visibilities, and CLEAN-based imaging can struggle with spatial filtering, complex morphologies, and uncertainty quantification. Alternative methods that fit visibilities directly can address some of these limitations but often require manual choices of image priors and model hyperparameters. We present SIFARI (Self-Supervised Interferometric Fitting for Astronomical Radio Imaging), a self-supervised neural network workflow that represents sky brightness as a continuous function of position and fits measured visibilities without an external image training set or explicit spatial regularizer. An empirical rule sets the Fourier feature scale from the visibilities before training, controlling how readily the network fits fine structure. Sampling network weights with Stochastic Weight Averaging-Gaussian (SWAG) gives approximate brightness uncertainty estimates, which we combine with a thermal-noise floor to construct spatially resolved signal-to-noise maps. In synthetic ALMA tests, SIFARI yields an effective point-source response about eight times narrower than the natural-weighting CLEAN restoring beam and recovers more extended flux than CLEAN when short baselines are missing. It also achieves higher image fidelity than the restored CLEAN images in all three morphology benchmarks. Applied to ALMA observations of PDS 70, SIFARI recovers the bright outer ring together with faint compact emission in the central cavity. For long-baseline-only WISPIT 2 data, SIFARI supplies a sky model for phase self-calibration where the CLEAN model is inadequate. The restored, self-calibrated SIFARI image has approximately 30% lower RMS noise than the CLEAN image made from the original visibilities without self-calibration. --- ### Systematic Multi-Agent Vision-and-Language Navigation: Formulation, Benchmark, and Method Authors: Yunzhe Xu, Zhe Liu Published: 2026-09-28 Categories: cs.CV, cs.AI, cs.RO arXiv: https://arxiv.org/abs/2609.35965v1 Vision-and-Language Navigation (VLN) has largely focused on a single agent following a single instruction, yet many real-world applications require teams of robots to tackle tasks beyond the capabilities of any individual agent. We present Systematic Multi-Agent Vision-and-Language Navigation, providing, to our knowledge, the first systematic formalization of multi-agent VLN as a constrained coordination problem: each mission consists of subtasks carrying dependency and resource constraints (presence locks and holding chains). A verified four-stage crafting pipeline instantiates the task as MAVLN, comprising 11,724 episodes across 145 scenes with teams of up to four agents under three instruction regimes, accompanied by tailored constraint-aware metrics. We further present TRISS, a coordination-ready navigation system coupling an LLM-based subtask scheduler, a shared topological memory that turns each agent's exploration into team knowledge, and a conflict-aware execution mechanism that realizes simultaneous intentions as collision-free routes. Extensive experiments establish TRISS as a comprehensive baseline and reveal substantial room for improvement across scheduling, planning, and execution, highlighting the challenges of coordinating under MAVLN task constraints. Project page: https://xyz9911.github.io/mavln. --- ### Solver Agent: an Agentic AI Framework for Theoretical Physics Computations Applied to F-theory Uplifts of O3-planes and S-folds Authors: Eliott Morgensztern, Cesar Fierro Cota, Alessandro Mininno Published: 2026-09-28 Categories: hep-th, cs.AI, math-ph, math.AG arXiv: https://arxiv.org/abs/2609.35958v1 We introduce Solver Agent, an AI framework based on large language models for calculations and proofs in mathematics and theoretical physics. The solution process is tracked through a persistent ledger that records assumptions, derivations, and computations. A central agent delegates tasks to specialized sub-agents, while independent agents verify both intermediate steps and the final result. This setup improves the traceability, reproducibility, and verification of computer-assisted calculations. Applying Solver Agent, we study global F-theory uplifts of Type IIB orientifolds and their S-fold generalizations. We establish sufficient conditions for Weierstrass models over projective threefolds with terminal $\mathbb{Z}_k$ quotient singularities ($k\in\{2,3,4,6\}$) to give $\mathbb{Q}$-factorial projective elliptically fibered Calabi-Yau fourfolds with isolated Gorenstein terminal quotient singularities. These geometries realize O3-planes and S-folds, where local D3-brane probes of the latter yield four-dimensional $\mathcal{N}=3$ superconformal field theories. Using stringy invariants, we derive fixed-point contributions to Hodge data and Euler characteristics, and show that these Euler corrections determine the localized D3-brane charges required for tadpole cancellation. We illustrate these results using toric hypersurface constructions, where a single three-dimensional polytope determines both the Type IIB Calabi-Yau threefold and the F-theory base; here, the orientifold double cover naturally forms a bisection of an alternative genus-one-fibered uplift with discrete $\mathbb{Z}_2$ gauge symmetry. Finally, we provide methods for toric computations and four-form flux analysis in four-dimensional $\mathcal{N}=1$ compactifications with non-abelian gauge sectors. --- ### HEIR: Learning Human-Entity Interactions with Functional Roles Authors: Di Wen, Wenhao Guo, Yuedong Tan, Yun Huang, Minheng Wu, Zhihang Chen, Haiwen Sun, Fei Teng, Zhiyuan Gao, Yufeng Zhang, Yuanhao Luo, Jingqi Zhang et al. Published: 2026-09-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.35955v1 Understanding human-entity interactions requires recovering each person-action event's participants, roles, and shared identities. This structure can support embodied agents by clarifying who acts on which entities and how, informing anticipation and coordination in shared environments. Standard HOI metrics score individual links, leaving complete event composition undermeasured. We introduce HEIR (Human-Entity Interactions with Functional Roles), an image benchmark for complete grounded participant-role sets across object, interpersonal, and self-directed interactions. It contains 18,730 images, six roles, 105 actions, and 437 nouns, with shared entities, role changes, and repeated fillers; 51.6% of images contain multiple actors and 62.1% contain multiple actions. HEIR pairs relation AP with complete-set AP and structural evaluation. We also introduce CoRISP (Compositional Role-aware Interaction Set Prediction), which uses shared entity identities to combine role-conditioned evidence and predict normalized participant-role sets. Cardinality and role-multiplicity potentials couple assignments through event size and role composition, with exact per-event normalization. Across 16 baselines, relation and complete-event rankings diverge even after aligning action weights. CoRISP leads the evaluated systems on repeated-role events and shared-participant images in HEIR by 2.87 and 3.82 Set mAP points, respectively. On V-COCO, CoRISP achieves 73.72/76.23 role AP and 61.06/68.59 complete-set AP on two-slot actions under Scenarios 1/2. These results show the value of learning and evaluating event composition alongside individual relations. The code and dataset are publicly available at https://github.com/Kratos-Wen/HEIR. --- ### ROSS: Relearning from Self-Generated Rollouts through Selective Supervision Authors: Zhiwei Zhang, Huayu Deng, Fei Zhao, Jiayan Fu, Bin Liang, Kam-Fai Wong, Mu Chuan Published: 2026-09-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.35954v1 Large language model post-training generates self-generated rollouts through reinforcement learning and on-policy distillation, yet this experience is often treated as stale once the policy advances. Historical rollouts can remain compatible with a later policy while preserving behaviors that the policy no longer expresses reliably. However, they may also contain mistakes, abandoned attempts, and redundant actions that should not be imitated, motivating finer-grained selective supervision. We introduce ROSS (Relearning from Self-Generated Rollouts through Selective Supervision), which preserves the full historical trajectory as context while applying loss only to selected model-generated continuations. Across domain-specific reinforcement learning, multi-teacher on-policy distillation, and agentic reinforcement learning, ROSS consistently improves upstream checkpoints and outperforms baselines across mathematics, code generation, instruction following, and software engineering. On Qwen3.6-35B-A3B, ROSS improves the six-benchmark MOPD average from 58.40% to 62.20% and SWE-bench Verified from 64.20% to 68.40%. These results show that self-rollout training leaves behind reusable behavioral experience that can yield further gains through offline supervised fine-tuning (SFT), without additional policy rollouts. --- ### Right Words, Wrong Moment: A Clinician-Grounded Analysis of Distress in 19,930 Conversations between Young People and ChatGPT Authors: Marx Wang, Ella Zhang, Cameron Tan, Andrea Mock, Songling Ngo, Zijing Wang, Robert Wolfe, Shirin Amouei, Rachel A. Hanebutt, Desmond C. Ong, Caroline Figueroa, Katie Davis et al. Published: 2026-09-28 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.35953v1 Young people increasingly turn to General-Purpose Conversational Agents (GPCAs), such as ChatGPT, in moments of distress. We examine young adults' (ages 18-25) experiences using ChatGPT. We first collected 19,930 ChatGPT conversations and survey data from 158 young adults. We then selected five example conversations reflecting user distress. Finally, we asked ten clinicians to review those five conversations. We found distressed participants reported greater emotional engagement with ChatGPT and greater behavioral change from using it than their peers. When they turned to ChatGPT in moments of acute distress, ChatGPT was quick to give overly dramatic responses and excessive action-oriented suggestions. Clinicians endorsed ChatGPT's availability and much of its wording, but identified seven process failures, such as prematurely jumping to solutions. We translated clinicians' feedback into design guidelines following three stages: 1) asking about safety, 2) de-escalating intensity to restore emotional regulation, and 3) exploring concerns without agreeing with them. --- ### HEAR: Real Voices, Real Bias: A Large-Scale Human-Recorded, Demographically Diverse Benchmark for Audio Language Models Authors: Shen Yan, Duc Le, Irina-Elena Veliche Published: 2026-09-28 Categories: cs.SD, cs.AI, cs.CY arXiv: https://arxiv.org/abs/2609.35952v1 We introduce HEAR (Human-recorded Evaluation of Audio-LLM bias by Real speakers), a large-scale, ecologically valid benchmark comprising 87k real human audio samples from 843 demographically diverse participants. HEAR enables comprehensive evaluation through Multiple Choice Question Answering (MCQA) and open-ended long-form tasks. To our knowledge, this is the first large-scale voice benchmark grounded entirely in authentic human speech. We evaluate model behavior across both real-time speech-to-speech and speech-to-text architectures. Our results reveal that voice-conditioned bias is a model-specific property. Furthermore, we demonstrate that personalization instructions consistently exacerbate demographic disparities. Our findings establish that voice bias is a controllable model characteristic, providing a foundational framework for future bias mitigation and evaluation in Audio-LLM development. --- ### Intrinsic Associative Memory on Riemannian Manifolds: Curvature, Capacity, and Emergent Modes Authors: Krishnakumar Balasubramanian, Zhaoyang Shi Published: 2026-09-28 Categories: cs.LG, cs.AI, stat.ML arXiv: https://arxiv.org/abs/2609.35948v1 Geometry does more than constrain an associative memory: curvature determines what it remembers and which states it creates. We develop intrinsic dense associative memories on Riemannian manifolds by casting memory as Epanechnikov kernel-density mode seeking. We compare geodesic and volume-corrected energies and show that curvature separates their behavior. We prove that geodesic memory always retains an isolated pattern, while corrected memory obeys a sharp Ricci-curvature threshold: positive curvature can erase memories in high dimensions, while negative curvature reinforces them. We derive geodesic capacity scalings of $q_β^{-1/2}$ for retaining every pattern and $q_β^{-1}$ for a typical one, where $q_β$ is the pairwise kernel-overlap probability. We show how overlap \emph{creates} novel memories: designed $N$-pattern configurations realize all $2^N-1$ subset modes, but random data at the storage threshold yield only a Poisson number. We establish exact one-step recall using Riemannian mean shift. In simulations, we recover the predicted curvature transition and every designed mode. On WordNet's full noun hierarchy, we demonstrate that volume correction improves low-capacity retrieval. Together, our work shows that curvature is a design variable for associative memory, not merely a property of the data. --- ### FluxLite: Inference-Time Proposal Control for Discrete Diffusion Models Authors: Yinuo Ren, Haoxuan Chen, Grant M. Rotskoff, Jiequn Han, Lexing Ying Published: 2026-09-28 Categories: cs.LG, stat.CO, stat.ML arXiv: https://arxiv.org/abs/2609.35947v1 Many inference-time tasks for pretrained discrete diffusion models and diffusion language models reduce to drawing samples from a tilted version of the pretrained distribution. Feynman-Kac sequential Monte Carlo (SMC) makes this correction exact in principle, but its prescribed weights routinely degenerate when the proposal dynamics are misaligned with the tilt, capping the practical gains from additional particles. We introduce FluxLite, a lightweight, training-free proposal-control framework for discrete diffusion. On the sparse directed graph of pretrained reverse rates, any sparse jump-rate perturbation can be exactly compensated by a $q_t$-weighted graph-divergence term in the Feynman-Kac potential; the target path is therefore preserved while the residual reweighting variance becomes a local convex objective. We instantiate this principle as two practical samplers: a one-hop local reallocation rule (HEU) and a small nonnegative quadratic program over pretrained-rate bases (D-VCG). We further prove population stability under the standard score-entropy training loss, identifying a tilted-path coverage factor that governs robustness to score error, together with finite-particle convergence for a fixed controlled Feynman-Kac recursion. Empirically, FluxLite improves over standard Feynman-Kac SMC baselines by up to two orders of magnitude in terminal KL on an analytically tractable finite-state CTMC benchmark, and reduces row-correlation MSE on 2D Ising sampling by 5-7x in geometric mean and up to 55x at peak. --- ### EnJoi: Ensemble Joint Score Filter for Generative Data Assimilation Authors: Julien Moreau, Marc Lelarge Published: 2026-09-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.35944v1 Data Assimilation (DA) aims to recover the full state of a dynamical system that is only partially observed. A solution is to use Score-based models to generate physically consistent trajectories that agree with the observations. These Autoregressive Diffusion models are trained by conditioning on the previous state; however, they do not take into account the uncertainty of their past predictions. We propose a new diffusion-based assimilation algorithm that dynamically balances the confidence in the current state and the new observations. Crucially, we choose to learn the distribution of the joint state containing both the past and future. This allows us to use a modified version of En4DVar, a classical DA algorithm that relies on the covariance of an ensemble of particles. Experiments on fluid and traffic flow simulations show improved reconstruction performance, especially in situations where observations are sparse and non-homogeneous. ---