# Research Papers — 2026-09 · bundle page 238 of 266 > Full-text markdown of papers #11851–11900 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/238 [← Previous bundle](https://aipod.dev/papers/archive/2026-09/page/237.md) [Next bundle →](https://aipod.dev/papers/archive/2026-09/page/239.md) ### SocioGesture: Real-Time and Adaptive Social Gesture Perception for Human-Robot Interaction Authors: Wenjin Fu, Li-Fan Wu, Jerin Peter, Chip Huyen, Boyuan Chen, Jan Liphardt Published: 2026-09-03 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2609.04545v1 Robots interacting with people must recognize not only explicit commands, but also social cues such as invitations, refusals, and unavailability. In real deployments, these cues must be inferred from noisy onboard perception under partial occlusion, changing viewpoints, and strict latency constraints. We present SocioGesture, a real-time adaptive social gesture perception system for human-robot interaction (HRI). SocioGesture uses a compact confidence-aware body-hand skeleton representation and a lightweight dual-stream model that fuses body motion with hand articulation for low-latency onboard recognition. To improve deployment robustness, we train the model with occlusion-aware skeleton corruption, exposing it to missing hands, occluded arms, and temporally unstable keypoints without increasing the inference cost. On a social gesture dataset collected in mixed indoor-outdoor HRI scenarios, SocioGesture achieves strong held-out-subject recognition, substantially improves robustness under structured joint occlusion, and runs in real time on a robot-mounted edge device. During deployment, uncertain interaction segments are saved for offline labeling and adaptation, enabling SocioGesture to expand its gesture vocabulary while preserving performance in the original classes. These results demonstrate a practical path toward robust, efficient, and adaptive social perception for interactive robots. --- ### From Answers to Interpretations: Rethinking Ambiguity-Induced Aleatoric Uncertainty Estimation in LLMs Authors: Omer Nahum, Niv Nayman, Jonathan Fhima, Alon Zolfi, Jeremy Levy, Shai Mazor, Paolo Favaro Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04543v1 A key challenge in reliable LLM deployment is recognizing when uncertainty reflects irreducible variability in the task rather than limitations in the model's knowledge. In language tasks, a central source of such aleatoric uncertainty is input ambiguity or underspecification, where multiple interpretations remain plausible. Existing decomposition methods estimate aleatoric uncertainty by generating multiple clarifications of the input, querying the model for an answer under each clarification, and comparing the resulting answers. We argue that answers are not necessary for identifying ambiguity: they are often redundant, add avoidable cost, and can mislead through epistemic leakage. We support this claim theoretically, and propose a clarification-only approach that estimates this ambiguity-induced component directly from the space of plausible interpretations, without answers to the clarified inputs. Using ambiguity detection as an operational evaluation across three benchmarks, this direct approach improves AUROC (63.34 vs. 60.85), reduces computational cost by 4-26x in output tokens and 2.2-3.5x in API calls, and yields estimates with substantially lower correlation with epistemic uncertainty. Overall, our results suggest that ambiguity-induced aleatoric uncertainty is better estimated from the interpretation space than from the response space. --- ### Data-Driven Discovery of Composition-Dependent Constitutive Models for Hyperelasticity and Viscoelasticity of Digital Materials Authors: Josué García-Ávila, Beijun Shen, Manuel K. Rausch, Mary C. Boyce, Adrián Buganza-Tepole Published: 2026-09-03 Categories: cs.AI, cond-mat.mtrl-sci, cond-mat.soft arXiv: https://arxiv.org/abs/2609.04541v2 Digital materials fabricated by multi-material 3D printing are designed as controlled mixtures of stiff and compliant constituents, yielding effective responses that span more than an order of magnitude in apparent stiffness and exhibit strongly nonlinear, composition-dependent, and rate-dependent dissipative behavior. Classical finite-strain viscoelastic models represent such behavior with closed-form strain energy functions for equilibrium and non-equilibrium stresses as well as evolution of internal variables, which may limit flexibility when a single constitutive model is expected to generalize across materials and loading rates. Here, we present a data-driven multi-material constitutive modeling framework that generalizes a formulation by Bergström and Boyce. The proposed framework retains the structure of the classical model, namely multiplicative kinematics, invariant-based strain-energy functions, and a scalar dissipative evolution law directed along the normalized nonequilibrium deviatoric stress. For the equilibrium branch, the data-driven discovery framework either directly predicts closed-form model parameters as functions of composition or automatically constructs a polyconvex strain-energy function using neural ordinary differential equations (NODEs). The nonequilibrium branch kinetics are learned similarly, either by directly identifying closed-form parameters across compositions or by using appropriately constrained artificial neural networks. Using multi-rate uniaxial compression data across multiple material compositions, we show that the proposed formulation captures rate-dependent stiffness and hysteresis across compositions while preserving thermodynamic consistency. --- ### Mitra-v2 Technical Report Authors: Yefan Tao, Xiyuan Zhang, Xinyi Liu, Boran Han, Danielle Maddix, Haoyang Fang, Zhen Han, Jiading Gai, Xuanqing Liu, Michael Bohlke-Schneider, Yuyang, Wang et al. Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04540v1 We introduce Mitra-v2, a tabular foundation model that delivers state-of-the-art performance on real-world classification and regression problems, from credit-risk scoring and clinical prediction to equipment-failure detection and house-price estimation. Mitra-v2 is trained only on synthetic data, with a pretraining distribution that is much larger and more diverse than Mitra-v1's. Built on a small 2D Transformer backbone, Mitra-v2 supports longer contexts and larger feature spaces. Improved optimization lets it learn from this larger task distribution. We evaluate Mitra-v2 on the TabArena and TALENT benchmarks, comprising more than 300 real-world datasets under two evaluation protocols. On the full TabArena benchmark, Mitra-v2 delivers state-of-the-art performance at the level of the industry-scale TabFM and EXAONE Tabular models, while surpassing TabPFN-3 by a wide margin in both classification and regression. Mitra-v2 matches the 1.6B-parameter TabFM with only 5% of its size (77M parameters), delivering frontier performance at a fraction of the cost. On TALENT, Mitra-v2 remains among the leading models, clearly outperforming TabPFN-3 and TabICLv2. It also ranks first on classification tasks with more than ten classes, even though it was pretrained only on tasks with at most ten classes. These results make Mitra-v2 one of the strongest and most broadly applicable open tabular foundation models released to date. We release the model weights, the inference and fine-tuning code, and our evaluation results under the Apache-2.0 license. --- ### A Calibrated Reflection Approach for Enhancing Confidence Estimation in LLMs Authors: Umesh Bodhwani, Yuan Ling, Shujing Dong, Yarong Feng, Hongfei Li, Ayush Goyal Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04539v1 A critical challenge in deploying Large Language Models (LLMs) is developing reliable mechanisms to estimate their confidence, enabling systems to determine when to trust model outputs versus seek human intervention. We present a Calibrated Reflection approach for enhancing confidence estimation in LLMs, a framework that combines structured reasoning with distance-aware calibration technique. Our approach introduces three key innovations: (1) a Maximum Confidence Selection (MCS) method that comprehensively evaluates confidence across all possible labels, (2) a reflection-based prompting mechanism that enhances reasoning reliability, and (3) a distance-aware calibration technique that accounts for ordinal relationships between labels. We evaluate our framework on diverse datasets, including HelpSteer2, Llama T-REx, and a proprietary conversational dataset, demonstrating its effectiveness across both conversational and fact-based classification tasks. This work contributes to the broader goal of developing reliable and well-calibrated confidence estimation methods for LLMs, enabling informed decisions about model trust and human judgement. --- ### Repeat-After-Me: Black-Box Adaptive Visual Prompt Injection Authors: Sizhe Chen, Yu-Lin Tsai, Ivan Evtimov, Kamalika Chaudhuri, Raluca Ada Popa, David Wagner, Arman Zharmagambetov Published: 2026-09-03 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2609.04533v2 Prompt injection is widely recognized as a major security threat to AI agents that interact with untrusted external data, such as websites, documents, and emails. Prior work has shown that, in the text domain, black-box prompt injection can achieve near-perfect attack success rates (ASRs). In the image domain, however, existing visual prompt injection methods are substantially less effective in attacking frontier commercial VLMs for materially harmful behavior. Achieving such outputs is hard because it requires a long and/or format-compliant target string, such as a precise, parseable native tool call with exact function names and arguments. We present Repeat-After-Me, a black-box adaptive visual prompt injection attack that can reveal personally identifiable information or make malicious tool calls. Across both open-weight and commercial frontier VLMs, including Qwen3.6-27B and GPT-5.5, our method achieves ASRs exceeding 82% and 47%, respectively, under a realistic setting in which the benign user prompt is semantically unrelated to the injected task and does not verbally authorize it. Our optimized injection has non-trivial attack transferability across commercial VLMs and benign samples. We show our attack works in cases where adaptive textual prompt injection fails. In a real-world OpenClaw agent connected to Discord, an untrusted user can use a minimally injected image from our attack to overwrite TOOLS.md, enabling future sensitive behaviors like remote code execution and secret exfiltration. We discuss potential defenses. --- ### Distilled Continuous Diffusion Language Models Can Write Code in Few Steps---or One Authors: Fred Zhangzhi Peng, Kaiwen Zheng, Anru R. Zhang Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04531v2 Language generation is almost universally treated as a sequential process: autoregressive models emit one token at a time, while diffusion language models replace token-level seriality with a long trajectory of iterative refinement. In this work, we introduce PlaidQ, a 0.7B continuous diffusion language model for code generation, and show that its trajectory can be aggressively distilled into only a few denoising steps---or even one, enabling efficient code generation. PlaidQ repurposes a pretrained autoregressive model as a bidirectional denoiser over continuous token embeddings. We distill PlaidQ with distribution matching for few-step generation and paired-trajectory supervision for one-step generation. At matched model scale, PlaidQ is competitive with discrete diffusion language models on code generation. Distillation then shifts the quality--compute frontier: a 16-step student reaches 31.78 and 40.49 pass@10 on HumanEval and MBPP+, surpassing the same PlaidQ teacher sampled for 512 steps. At the extreme, paired-trajectory distillation achieves 7.07 pass@1 on HumanEval with a single denoising step, producing functionally correct programs. Together, these results establish continuous diffusion as a viable path to few-step and one-step code generation. Broadly, continuous diffusion is not merely another representation for language: it provides an interface through which language models can inherit the acceleration and distillation machinery of continuous diffusion modeling. Training and inference code and model checkpoints are available at https://github.com/pengzhangzhi/plaidq. --- ### An Energy-Based Conservative-Dissipative Latent Neural Evolution Operator for Magnetization Dynamics Authors: Sebastian Schaffer, Lukas Exl Published: 2026-09-03 Categories: cs.LG, math.NA, physics.comp-ph arXiv: https://arxiv.org/abs/2609.04530v1 We develop an energy-based reduced-order model for micromagnetic magnetization dynamics that couples a convolutional autoencoder to a structured latent neural ordinary differential equation. Motivated by the precessional-dissipative structure of the Landau-Lifshitz-Gilbert equation, the latent vector field is generated from the gradient of a learned scalar potential through an antisymmetric operator and a symmetric positive-semidefinite dissipative operator. This potential is learned in nonunique latent coordinates and is not identified with the Gibbs free energy, but decreases monotonically along autonomous continuous-time solutions, while the antisymmetric component permits motion along its level sets. The encoder, decoder, latent energy, and operators are trained jointly on short trajectory windows using latent and decoded-rollout losses alone, without time-derivative supervision, physical-energy labels, or dissipation penalties. At inference, an initial state is encoded once, evolved in latent space, and decoded only at the requested output times, enabling substantially cheaper trajectory prediction than the micromagnetic solver used to generate the training data. We compare quadratic, deep, and additive deep-quadratic latent energies on two datasets parameterized by field amplitude and generated for the two applied-field directions of the NIST $μ$MAG Standard Problem 4. Dissipative-only and antisymmetric-dissipative models achieve comparable accuracy on short training-style windows but differ substantially on uninterrupted rollouts, for which the antisymmetric-dissipative models provide markedly more accurate trajectory predictions. The deep-quadratic energy gives the best overall accuracy for both field directions and exhibits slower error growth when rollouts are extended to twice the training horizon. --- ### Towards a universal language of concepts: A survey Authors: Aishni Parab Published: 2026-09-03 Categories: cs.AI, cs.PL, cs.SC arXiv: https://arxiv.org/abs/2609.04528v1 Humans can learn and generalize novel concepts from sparse data because they express knowledge in rich structural formats. In this paper, we propose that programs are a strong candidate for universal representation of concepts. We review computational models of concept learning that use programs as their concept representation and evaluate their contribution toward a universal representational language. --- ### Scale-QLoRA: Code-Invariant Adapter Merging for Native 4-bit Microscaling LLMs Authors: Tung-Ling Li, Jiale Huang, Lee-Chi Wang, Janaki Ram Gotei Published: 2026-09-03 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.04526v1 Merging a LoRA adapter into its base model is standard deployment practice: it removes the runtime adapter's per-forward overhead and leaves a single standalone checkpoint any serving stack can load. On a native 4-bit microscaling checkpoint (NVFP4, MXFP4) that step stops being free. The merged weights must be written back through a quantizer, which re-derives the checkpoint's discrete E2M1 code plane (roughly 90% of the artifact's bytes), so the deployed artifact becomes coupled to one quantization convention, and every later code-touching event in its lifecycle can move it. Done naively the step is worse than fragile: it deletes the adaptation, by up to 39 pp, because against an already-on-grid base the reconstruction optimum is that base. Scale-QLoRA instead adapts only the native per-block scale field, trains those scales on the deployment grid, and freezes every E2M1 code. Within a fixed native format, scale grid, block layout and code plane, merging is then a bit-exact identity and the merged artifact is code-invariant. Across four models and four tasks, Scale-QLoRA and merge-aware QAT-LoRA are both accuracy-lossless, so we claim no accuracy ordering between them; they differ structurally, in that QAT-LoRA re-derives the code plane through a quantizer while Scale-QLoRA preserves it exactly. That difference is what the lifecycle prices: nearest-rounding implementations disagree by about a point on the measured task, and more extreme rule mismatches can drive the weight-space artifact to ~0%, which we report as a sensitivity bound rather than a deployment frequency. Preserving the code plane also drops the weight-space straight-through estimator from training (3.9x per step on the dense 8B model) and enables exact rollback, code-plane deduplication, and a ~125x faster scale-only task swap. --- ### MaxKernel: Agentic Kernel Generation for TPUs Authors: Shangkun Wang, Nina Cai, Charles Hoong, Julian Walker, Gerson Kroiz, George Vanica, Deepak Patil, Andi Gavrilescu, Hassan Sipra, Sethu Sankaran Published: 2026-09-03 Categories: cs.AI, cs.PF, cs.PL arXiv: https://arxiv.org/abs/2609.04523v1 Designing and authoring high-performance custom kernels for accelerators is a complex task that requires deep hardware-level expertise. Large Language Models (LLM) can be leveraged together with real-time compiler feedback to build agentic systems for kernel generation. In this work, we present MaxKernel, a multi-agent system that implements three distinct paradigms for TPU kernel development: (1) a Human-in-the-Loop (HITL) agent for collaborative, step-by-step design; (2) an Autonomous (Auto) agent that executes a fully automated, metric/trace-driven optimization loop; and (3) a Graph-Based Autonomous Search that scales the Auto agent for global exploration of the design space. All three paradigms leverage a shared pool of specialized sub-agents to handle planning, implementation, self-debugging, testing, and hardware profiling. We evaluate MaxKernel on JaxBench, a comprehensive suite of 50 diverse kernel tasks for TPUs, alongside complex, real-world workloads from state-of-the-art open-source models. We demonstrate that MaxKernel consistently generates highly optimized implementations, matching expert hand-tuned baselines and delivering significant performance across the benchmark. Our agent is open-sourced and available https://github.com/AI-Hypercomputer/accelerator-agents/tree/main/MaxKernel. --- ### What Does Multi-Harness RL Learn? Credit Assignment and Portability in Coding Agents Authors: Chenqian Le, Jiayi Cheng, Qijia He, Runhao Li, Yinghao Li, Xupeng Chen Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04518v1 Agent reinforcement learning (RL) increasingly runs through full execution harnesses, and a multi-harness recipe mixes two choices: exposing the policy to several harnesses, and comparing their rewards inside one relative-advantage group. We isolate the second choice in repository-level coding. From one Qwen3-8B supervised warm start we replay the same frozen task-harness records from Aider, OpenHands, Qwen Code, and SWE-agent, with the same number of updates, under two rules for group-relative policy optimization (GRPO), Within (one group per task-harness pair) and Cross (harnesses pooled within a task), and score every checkpoint with a sealed SWE-bench Verified oracle on four source harnesses and a minimal harness held out of training. The evaluation harness is the dominant variable: across 24,000 sealed evaluations it moves the mean solve rate from 2.14\% to 9.27\%, a factor of 4.3, where the training recipe moves it by 1.16. The grouping rule is not. On the held-out harness, Cross minus Within is +0.25 pp, 95\% confidence interval (-0.48, +1.02), at eight attempts per task, and +0.16 (-0.41, +0.72) pooled over three training seeds whose individual estimates change sign. Each rule's own seed range, 0.42 to 0.45 pp, exceeds the difference between them. Both rules place their largest gains on the same source harness. The pooled advantage carries the harness: an out-of-fold classifier recovers the generating harness from Cross's advantage +4.48 pp above the shuffled-label baseline and from Within's not at all, and the two rules still reach the same held-out score and action distribution inside each harness. Re-collecting half the training data on-policy does not change this. Cross-harness credit yields configuration adaptation and no more portable capability than within-harness credit. Multi-harness RL reports should state the grouping boundary and test under an unseen harness. --- ### Pitch-class Steering for Diffusion-based Music Generation via Latent-space Probes Authors: Yushi Ye, Wilson Zheng, Yongyi Zang Published: 2026-09-03 Categories: cs.SD, cs.AI arXiv: https://arxiv.org/abs/2609.04516v1 Recent work on controllable music generation has focused on autoregressive models, leaving diffusion-based systems comparatively underexplored. We present a lightweight method for steering the pitch content of audio produced by Stable Audio Open, a latent diffusion model for music synthesis. A small convolutional probe containing approximately 125k parameters is trained to decode frame-level pitch-class activations from the model's variational autoencoder latent space, using paired audio and MIDI data. At inference time, the frozen probe serves as a differentiable loss function: its gradient with respect to the denoising latent is used to nudge generation toward a user-specified pitch-class sequence, requiring no retraining or architectural modification of the base model. Across 27 evaluation trials spanning 9 text prompts and 3 target melodies, probe-guided generation increases melodic coherence by 2.4x over the unguided baseline (p < 1e-5, Wilcoxon signed-rank test), demonstrating that musically meaningful structure is both recoverable and steerable in diffusion-based music latent spaces. --- ### Atlas: Optimizing Deployment of Compound AI Workflows on Heterogeneous Clusters Authors: Milos Gravara, Andrija Stanisic, Stefan Nastic Published: 2026-09-03 Categories: cs.DC, cs.AI arXiv: https://arxiv.org/abs/2609.04513v1 Compound AI workflows are increasingly used to serve complex AI tasks by coordinating multiple AI models and software components. This approach enables deployment flexibility, as each workflow stage can expose different model variants and resource requirements, but it also expands the deployment choices. A deployment must choose an execution plan that selects AI models for each compound AI workflow stage and places them on a heterogeneous cluster in order to satisfy SLOs. Deployment optimizers therefore need estimates to compare many candidate plans and identify feasible ones. System metrics can often be profiled per stage and composed according to workflow topology, but accuracy cannot, as errors and information loss at upstream stages affect the accuracy of downstream stages. Existing approaches either profile complete configurations end to end, which scales poorly, or use product-based accuracy surrogates that treat stages as independent and can misrank candidate plans. We introduce Atlas, a framework for optimizing compound AI deployments under SLO constraints. Atlas uses MAP, a Markovian Accuracy Predictor, to estimate configuration accuracy from local conditional accuracy transitions between adjacent workflow stages. MAP discretizes intermediate outputs into accuracy buckets and composes transition profiles according to workflow topology, giving the optimizer an accuracy estimate without exhaustive end-to-end profiling. Atlas formulates execution-plan selection as a mixed-integer linear program that maximizes predicted accuracy subject to SLOs. Across four compound AI workflows, MAP achieves Spearman correlation up to 0.947 while reducing profiling cost by up to 2.6x relative to exhaustive end-to-end profiling. Guided by MAP, the Atlas optimizer selects execution plans within 0.03 of oracle accuracy while reducing deployment cost by up to 42% through heterogeneous placement. --- ### Explainable Multimodal Deep Learning Integrating Imaging and Clinical Data for Oral Potentially Malignant Disorder Detection Authors: Ruilin You, Yihan Wang, Jiabin Chen, Cherie Wink, Petra Wilder-Smith, Rongguang Liang, Bofan Song Published: 2026-09-03 Categories: eess.IV, cs.CV arXiv: https://arxiv.org/abs/2609.04512v1 Oral potentially malignant disorders (OPMDs) are critical precursors to oral cancer, yet clinical detection remains challenging because of substantial phenotypic heterogeneity and overlap with benign conditions. Although image-based deep learning shows promise for automated screening, visual information alone may be insufficient in real-world settings, where diagnostic decisions also rely on patient-specific risk factors. We developed M2-OPMDNet, a multimodal deep learning framework that integrates co-registered white-light and autofluorescence intraoral images with structured clinical information for OPMD detection. A customized questionnaire was designed to capture clinically relevant risk factors and symptoms in a standardized, reproducible format for integration with image-derived features. Multiple image encoders, including conventional convolutional neural networks and foundation model-based architectures, were evaluated using a prospectively collected dataset reflecting real-world screening conditions. Model interpretability was assessed using SHapley Additive exPlanations (SHAP) to quantify feature- and modality-level contributions. M2-OPMDNet achieved an AUC of 0.952, outperforming unimodal approaches and showing improved performance for visually subtle lesions. SHAP analysis demonstrated that structured clinical variables contributed substantially to risk estimation and complemented imaging features. These results demonstrate that explainable multimodal learning combining white-light and autofluorescence imaging with structured clinical data can provide accurate, transparent, and clinically grounded OPMD detection. M2-OPMDNet offers a scalable framework for real-world oral cancer screening and decision support. --- ### LentEx: Generalizable Latent Entity Extraction via Synthetic Data and Instruction-Tuned LLMs Authors: Umesh Bodhwani, Yuan Ling, Cibi Chakravarthy Senthilkumar, Shujing Dong, Yarong Feng, Hongfei Li, Ayush Goyal Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04511v1 Latent entity extraction (LEE) tackles the challenge of identifying implicit, contextually inferred entities within free text-an area where traditional entity extraction methods fall short. In this paper, we introduce LentEx, a novel framework for latent entity extraction that leverages synthetic data generation and instruction fine-tuning to optimize smaller, efficient large language models (LLMs). Latent entities, which are often abstract and thematic, are crucial for applications such as retrieval-augmented generation (RAG), customer persona analysis, and knowledge graph enrichment. LentEx addresses the scarcity of labeled datasets by employing a template-based approach to generate diverse, contextually rich synthetic data, ensuring high variability and alignment with real-world distributions. To our knowledge, LentEx is the first to systematically approach LEE through the lens of LLMs. LentEx demonstrates significant performance improvements across multiple tasks, notably surpassing state-of-the-art models on the MTEB Clustering Benchmark. Furthermore, our methodology enables robust generalization to unseen domains, making LentEx highly applicable in real-world NLP tasks, including RAG and clustering, thereby establishing a new paradigm for latent entity understanding and extraction in natural language processing. --- ### A Semantic Model of Genetic Evidence: A Step Toward Bridging the Basic-Science-Clinic Gap Authors: Michael Bouzinier, Dmitry Etin Published: 2026-09-03 Categories: cs.DB, cs.AI, q-bio.GN arXiv: https://arxiv.org/abs/2609.04509v1 Scientific and clinical decision-making depends on evidence from the primary literature, but existing standards for representing that evidence (FHIR Evidence, ECO, SEPIO, and the GA4GH Genomic Knowledge Standards) are oriented toward clinical-trial workflows, evidence codes, or single-variant assertions, and do not capture the fine-grained, domain-specific structure of claims in basic and pre-clinical research. We introduce a semantic model for scientific evidence with three core classes, specialize it for genetics, align it structurally to FHIR Evidence with a SEPIO-anchored credibility decomposition, and attach a compact dimensional vocabulary whose conditional-activation rules are validated by a SHACL schema for the implemented constraints. Using clinical variant interpretation as the driving use case, we evaluate the model through a human-AI annotation pilot over six genetics papers, yielding 28 evidence items and 95 source-anchored assertions, with a workflow that keeps curator-authored reference annotations distinct from AI-drafted annotations. Treating the pilot as a feasibility study rather than a benchmark, we argue that the model is a useful increment toward trustworthy, AI-ready infrastructure for variant interpretation: a reference data model and validation schema for representing genetic evidence. --- ### BioSync: Transformer-Based Cross-Modal Fusion for a Multimodal Physiological Digital Biomarker Authors: Seyed Mahmoud Sajjadi Mohammadabadi Published: 2026-09-03 Categories: cs.AI, cs.IR arXiv: https://arxiv.org/abs/2609.04504v1 Cardiac, neural, behavioral, and speech measurements from wearable and mobile devices provide partial, noise-sensitive views of physiological state. BioSync combines these measurements into the \textbf{BioSync Index (BSI)}, a continuous composite digital biomarker defined under the BEST framework. The model applies multi-head self-attention to modality tokens and adds a linear branch whose hypothesis class includes standard feature concatenation. This architecture is motivated by latent-variable measurement theory and by the possibility that joint observations contain information unavailable from individual modalities. We evaluated BioSync on two literature-informed synthetic cohorts: a four-modality cognitive-decline cohort using HRV, EEG, actigraphy, and speech, and a metabolic-autonomic cohort structured around the public AI-READI wearable schema. In the cognitive cohort, BioSync and concatenation obtained AUCs of 0.928 and 0.926, respectively. In the metabolic cohort, BioSync obtained accuracy/F1 of 0.764/0.766, compared with 0.756/0.758 for concatenation. The BSI correlated with latent severity in both cohorts ($r=0.91$ and $r=0.68$). A pure-attention ablation obtained cognitive-cohort AUC 0.911, locating the increase to 0.928 in the combined wide-and-deep architecture. With matched modality-dropout training, BioSync led concatenation at five of six cognitive-cohort corruption rates and at the highest metabolic-cohort rate. Its cognitive-cohort AUC was also higher than five published digital-biomarker reference values, although differences in datasets and tasks preclude a controlled benchmark claim. Comparison with single-modality, early-fusion, and late-fusion designs across six prespecified criteria identifies the model's computational properties; validation on real cohorts remains necessary. --- ### Fractional-Order Adaptive Motion Magnification: Phase-Reliability Weighting for Noise-Constrained Video Amplification Authors: Alejandro Garnung Menéndez Published: 2026-09-03 Categories: eess.IV, cs.CV arXiv: https://arxiv.org/abs/2609.04502v1 Eulerian video amplification boosts sub-pixel motion by band-pass filtering per-pixel intensity traces and applying a uniform gain. That gain ignores local structure, so sensor noise is amplified together with the signal, especially in textureless regions where the monogenic phase is unreliable. We propose FrAM (Fractional-order Adaptive Motion Magnification), a pipeline developed first offline and then as a causal stream. It replaces the constant temporal gain with a Grünwald--Letnikov derivative of fractional order, giving continuous control over high-frequency emphasis, and replaces the uniform spatial gain with a per-pixel weight derived from the local amplitude of the monogenic signal. On a controlled synthetic sequence split into textured and flat halves, FrAM matches the amplification of the Eulerian baseline while keeping flat-region temporal noise at the input level. The reduction holds across an eightfold range of input noise levels. Real videos show improved spatial selectivity and lower background noise in every case. The causal reformulation cuts the per-frame cost by two orders of magnitude, reaching 69\,fps at 640$\times$480. --- ### EyeMakeYou: Identity-, Task-, and Subjective-State-Conditioned Diffusion for High-Frequency Gaze Synthesis Authors: Kamrul Hasan, Mehedi Hasan Raju, Oleg V. Komogortsev Published: 2026-09-03 Categories: cs.HC, cs.CV arXiv: https://arxiv.org/abs/2609.04501v1 Eye movement biometrics (EMB) is an emerging behavioral modality for user authentication, particularly in virtual- and augmented-reality systems, where gaze dynamics contain distinctive subject-specific features. However, robust EMB systems require diverse, high-quality gaze recordings that are expensive to collect and often unavailable at the scale needed for model development. Generative models can mitigate data scarcity, but existing methods either synthesize generic gaze behavior or personalize signals primarily by identity, without jointly representing the user's task and subjective state. Consequently, generated signals may appear visually realistic while failing to retain the behavioral properties required for biometric applications. To address this limitation, we propose EyeMakeYou, a multi-conditional denoising diffusion framework for subject-specific, high-frequency gaze synthesis. EyeMakeYou generates 5-s, 1000-Hz bivariate gaze-velocity sequences from an identity-removed reference trajectory and conditions the denoising process on an identity embedding, a task embedding, and self-reported ratings of overall difficulty, mental tiredness, and eye tiredness. Its objective combines diffusion noise prediction and identity preservation with multi-resolution spectral, drift-consistency, and event-weighted local-smoothness losses. Experiments on GazeBase show that EyeMakeYou achieves higher median spatial accuracy and greater real--synthetic similarity in the embedding feature space than the existing generative approaches, while retaining selected task-dependent associations between subjective reports and oculomotor features. These findings support conditional diffusion as a practical approach for augmenting gaze datasets for biometric and interactive applications. --- ### STyMo: Fast and Controllable Few-Shot Motion Style Transfer Authors: Jose Luis Ponton, Alexander Winkler, Ladislav Kavan, Yuting Ye, Petr Kadlecek Published: 2026-09-03 Categories: cs.GR, cs.CV arXiv: https://arxiv.org/abs/2609.04500v1 Supporting a wide variety of motion styles is critical for creating diverse virtual characters, but current methods either require large stylized datasets or pre-trained models that cannot generalize beyond their training distribution. We present STyMo, a few-shot approach that learns motion style from only seconds of paired data and trains in one to two minutes. Our key insight is to decompose style into two components: a static component capturing time-invariant posture, and a temporal component capturing frame-wise dynamics. This decomposition yields an interpretable system where posture intensity, temporal exaggeration, and per-body-region style can be adjusted at runtime. Furthermore, the reduction in required training data and computation time structurally permits an iterative authoring workflow. To ensure robustness on arbitrary inputs, we further introduce a stylizability gate that automatically prevents artifacts on out-of-distribution motions. We demonstrate results across diverse motion styles, from subtle emotional variations to exaggerated character archetypes, and release our processed paired dataset to facilitate future research. --- ### Rethinking Indirect Prompt Injection as a Test-Time Search Problem Authors: Duong M. Nguyen, Joon Sik Kim, Blazej Manczak, Vaikkunth Mugunthan Published: 2026-09-03 Categories: cs.AI, cs.CL, cs.CR arXiv: https://arxiv.org/abs/2609.04495v1 We formulate indirect prompt injection as a test-time search over a task-dependent attack surface induced by the environment, user task, and injection task. To operationalize this formulation, we introduce an agentic attacker with a dedicated search harness that performs environment reconnaissance, structured reasoning over attack strategies, and adaptive evaluation using victim-agent feedback. Across heterogeneous tasks, we find that increasing attacker test-time compute improves vulnerability discovery and exploitation, while ablations show that explicit strategy management is important for avoiding redundant search and sustaining gains at larger budgets. These results suggest that agentic security evaluations should characterize both the attacker's search procedure and compute budget, rather than treating attack success as a budget-independent property of the victim. More broadly, our findings identify the attacker's adaptive search over the system attack surfaces as an important and underexplored security risk for tool-using agents. --- ### Hakken: Predicting future discoveries to fill the gaps in today's knowledge Authors: Tarek R. Besold, Uchenna Akujuobi, Pablo Sanchez, Alessandra Toniato, Kana Maruyama, Jihun Choi, Samy Badreddine, Frederick Gifford, Daniel Evans-Yamamoto, Sucheendra K. Palaniappan, Miquel Ferrer, Kae Nagano et al. Published: 2026-09-03 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.04494v1 We present Hakken, a domain-agnostic prediction and explanation system performing knowledge prediction, i.e., growing scientific knowledge by establishing novel relationships, ones that are not limited to the deductive hull of previous knowledge. Hakken uses a transformer-based prediction model built on temporal sequences of knowledge graphs extracted from vast bodies of research publications, fused with an LLM's semantic knowledge, to predict the presence and define the type of as-yet undocumented relationships between scientific concepts. It then calls a model-agnostic explanation framework to provide accompanying information for each prediction that allows scientists to evaluate the suggested new relationship. While general purpose, we demonstrate Hakken's practical capabilities by applying it to the biomedical domain. There, Hakken's prediction model establishes a new benchmark for time-aware multi-label relation prediction, and we show that the model's output stays coherent and informative over extended time spans in historic data. In addition, we scored 1.5 million above-confidence-threshold hypotheses related to aging, qualitatively validated batches of these predictions with biologists and progressed three of them for empirical validation in wet-lab. Two predictions with potentially significant impact in the context of drug discovery and repurposing were confirmed, introducing previously undocumented interactions between TP53 and BAMBI, and between RAF1 and TNF, to biomedical science. --- ### ResLearn-XR: Residual Learning for Network Traffic and Quality-of-Experience-Aware Modeling in Extended Reality Authors: Yoga Suhas Kuruba Manjunath, Jie Gao, Lian Zhao Published: 2026-09-03 Categories: cs.AI, cs.LG, cs.NI arXiv: https://arxiv.org/abs/2609.04493v1 We present ResLearn-XR, a residual learning framework for predicting eXtended Reality (XR) network traffic and estimating Quality-of-Experience (QoE) risk. ResLearn-XR adopts a two-stage temporal learning structure comprising a base sequence prediction model augmented with task-specific residual learning components to improve adaptability to bursty, non-stationary XR traffic dynamics. The residual learning stages operate in the value space for continuous XR traffic forecasting and in the logit space for probabilistic QoE risk estimation. \rev{For the QoE-risk branch, we introduce a Data Descriptor Algorithm (DDA), a causal feature-construction module that converts packet-level application-layer observables into frame-timing-aware descriptors suitable for encrypted traffic analysis. We also construct an XR Traffic-QoE dataset that pairs continuous XR traffic traces with session-level user-reported QoE labels. ResLearn-XR reduces SMAPE by up to 17.84% across frame-count, frame-size, and inter-arrival-time prediction, while reducing QoE-risk estimation SMAPE by up to 87.8% over single-stage baselines. --- ### When Quantization Breaks Memory: Recurrent-State Write-Back in Low-Precision Temporal Inference Authors: Ismail Erbas, Xavier Intes, Vikas Pandey Published: 2026-09-03 Categories: cs.AI, cs.LG, physics.optics, q-bio.QM arXiv: https://arxiv.org/abs/2609.04490v1 Quantization is widely used to reduce the computational and memory demands of neural-network inference. In recurrent networks, however, the quantized state is stored and returned at the next time step, so the rule used to store that state can alter subsequent computations. Here, we introduce recurrent-state write-back to denote this rule and isolate its effect in a compact GRU encoder--decoder for fluorescence lifetime imaging, a molecular imaging modality used in quantitative biological imaging. A central task is estimating two lifetime parameters, the short-lived component τ1 and the long-lived component τ2, from high-noise time-resolved fluorescence signals. Holding the trained model fixed, replacing continuous state propagation with deterministic 4-bit state storage increases estimation errors for τ1 and τ2 by approximately 70x and 300x, respectively. Failure occurs when repeated small updates remain below the write threshold, leaving the stored state nearly fixed while the network continues to propose change. Error feedback, residual memory, and direction memory carry information from these suppressed updates across time and recover accuracy without retraining. Precision sweeps show that increasing state precision can worsen a fixed recurrent solution, while matched training shows that compatibility with the state interface can be learned. To test whether this behavior extends beyond the GRU, we repeat the post-training intervention in an independently trained LSTM, where coarse write-back reproduces the failure, error feedback restores accuracy, and state-specific interventions reveal greater sensitivity of the cell state than the hidden state. Our results establish recurrent-state write-back as a key determinant of low-precision recurrent dynamics and identify the state-storage interface as a central design consideration for quantized recurrent inference. --- ### Towards Understanding Pause Token Fine-Tuning Dynamics: A Mode Retention Perspective Authors: Jaehyeon Kim, Suhwan Kim, Nakyung Lee, Yeongoon Kim, Jimin Seo, Giho Lee, Jungwoo Lee Published: 2026-09-03 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.04489v1 Pause-token methods improve LLM reasoning by inserting special tokens into sequences. Prior work explains these gains through computational expressivity. However, there is relatively little investigation into the training dynamics of pause tokens. We explore how pause tokens reshape the training dynamics of fine-tuning. Two controlled pilots expose distinct asymmetries. On a synthetic continual-learning task, masked pauses overwrite a previously-learned distribution roughly 4x less at matched final adaptation (H1, mode retention); on a synthetic math-reasoning probe, the boundary-adjacent token comes to encode substantially more downstream-step information (H2, non-myopic compression). We formalize a training rule consistent with both - Masked Boundary Pause (MBP), pause tokens placed at reasoning-step boundaries with their loss masked. Across 1B-8B Qwen and Llama models, MBP consistently improves reasoning, achieving gains of up to 6 points on math and 2.5 points on code, while preserving general language understanding abilities. We further demonstrate that this mode-preserving strategy extend gains to GRPO. These results recast pause tokens as a training-dynamics intervention on the retention-adaptation trade-off, rather than merely an inference-time computation device. --- ### Uncertainty Signals for Network Intent Translation: Risk Ranking and Ambiguity Localization Authors: Ala' A. Alsamarneh, Omar Alhussein Published: 2026-09-03 Categories: cs.NI, cs.CL arXiv: https://arxiv.org/abs/2609.04486v1 Intent-based networking realization starts by translating high-level intents into low-level network configurations. Recent approaches have shifted toward LLM-based translation. Despite promising results, most studies focus on translation accuracy and overlook risks associated with deploying the resulting configurations. In this work, we investigate the pre-deployment translation risk of LLM-generated configurations by analyzing the model's uncertainty. We propose to use two uncertainty signals, namely sampling-based predictive uncertainty for translation-risk ranking and token-level entropy for ambiguity-source localization. We evaluate these signals on an ambiguity-controlled test set across different context types and sampling budgets, using a Llama-3.1-8B-Instruct model fine-tuned for intent translation on a vendor-specific switch platform (Juniper EX3300). The results demonstrate that predictive uncertainty provides a useful signal for ranking translations by risk across context types and sampling budgets, albeit with substantial miscalibration under less informative contexts. Moreover, we show that parameter-token entropy correlates with parameter-sourced ambiguity and keyword-token entropy correlates with description-sourced ambiguity. These results indicate the potential of using uncertainty signals in an LLM-generated configuration deployment pipeline, where predictive uncertainty can support selective deployment, while token-level entropy can identify sources of ambiguity. --- ### Cultural Misalignment in Large Language Models: Detection, Measurement, and Mitigation Through Targeted Fine-Tuning Authors: Antoni Czolgowski, Abel Iyasele Published: 2026-09-03 Categories: cs.CL, cs.AI, cs.CY arXiv: https://arxiv.org/abs/2609.04485v1 We evaluate three open-weight LLMs (Gemma3-12B from the USA, Bielik-11B-v3 from Poland, and Qwen3-4B from China) against World Values Survey Wave 7 data for 63 demographic personas across three countries, using normalized Wasserstein distance to quantify distributional misalignment. Contrary to expectations, no model favors its home country: the Chinese-built Qwen3-4B performs worst on its own Chinese population (W1 = 0.436, the highest misalignment in the entire model x country matrix). Targeted LoRA fine-tuning on the five worst-case personas, requiring fewer than 1,200 training pairs and under 15 minutes on a single GPU, reduces bias by 16.8% for Bielik-11B (p_Bonf = 0.002, d = -4.4) with all five targets improving. However, country-level decomposition reveals that fine-tuning redistributes rather than removes bias: Bielik's worst-case personas swap entirely from American to Chinese elderly, with zero overlap between pre- and post-correction sets. To our knowledge, this is the first study to target worst-case demographic personas with LoRA fine-tuning for cross-cultural bias mitigation. --- ### Patterns of Priming in Production: Lexical, Semantic and Structural Alignment in Language Model Generation Authors: Giulia Pucci, Ruizhe Li, Arabella Sinclair Published: 2026-09-03 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.04484v1 This paper investigates structural priming in language model (LM) production, examining how preceding structural context influences sentence completion. While prior work has demonstrated priming effects in comprehension of structural alternations, it remained unclear whether these persist in production, where, when generating, an LM samples from many possible continuations at each step. We address this question through a series of controlled sentence-completion experiments on dative constructions. In line with prior work, we find that LMs are susceptible to structural priming, particularly in sentences that are semantically coherent. In terms of priming magnitude, we find that while there is a greater relative increase of double-object datives against our baselines, in line with inverse frequency effects, there is a larger absolute increase in prepositional-objects, the more frequently produced construction. Finally, we not only observe that structural priming is boosted by lexico-semantic coherence, but that structurally primed completions display greater levels of lexico-semantic repetition. Taken together, our evidence supports the view that structural priming in LMs operates across multiple levels of linguistic representation, facilitating, and facilitated by syntactic, lexical, and semantic alignment. Code: https://github.com/the-context-lab/primedproduction. --- ### Safety for Whom? Boundary-Aware Self-Distillation for Controlled LLM Safety Refusal Authors: Alejo López-Ávila, Iker García-Ferrero, Jezabel Garcia, Antonio Tiene, Román Orús Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04482v1 Safety alignment is usually posed as a topic-level question: is this subject harmful? Deployments ask a narrower one. A civics tutor and a public-sector assistant may share a base model yet need different boundaries inside the same topic, refusing targeted political manipulation while still answering factual questions about the same election. We formulate this as narrow-boundary safety and introduce an offline self-generated framework combining controlled topic generation, coverage repair, in-distribution compensation data, and harmful-benign pairs for training and evaluation. Single-shot generation leaves 19.88% of prompts without accepted refusal traces, whereas escalating retries leave 0.20%. On political persuasion with Qwen3-8B, training on refusal data completed through Escalate increases target-domain refusal from 9.47% to 84.75% and reduces the mean unsafe-response rate across three broader harmfulness benchmarks from 26.26% to 0.14%, but increases XSTest over-refusal from 2.00% to 74.00%. In a separate matched comparison, replacing external responses with verified target-model responses reduces over-refusal from 15.20% to 5.20%. Boundary-pair data reduces comply-side over-refusal on held-out pairs from 32.94% to 4.16%, while harmful-side refusal decreases only from 91.88% to 87.72%. These results show that data composition controls the safety and usability trade-off, and that safety alignment should be evaluated on both sides of the intended refusal boundary. --- ### PerfReasoning: How Well Do LLMs Reason on Hardware Performance? Authors: Dan Zhao, Karthikeyan Sankaralingam, Christos Kozyrakis, Qijing Huang Published: 2026-09-03 Categories: cs.AI, cs.PF arXiv: https://arxiv.org/abs/2609.04476v2 Performance modeling is central to hardware design and software optimization, yet constructing these models requires structured reasoning about computation, data reuse, storage, and movement. We introduce PerfReasoning, a benchmark that evaluates LLMs both as direct performance reasoners and as generators of analytical performance-model code. Given workload, architecture, and mapping specifications, models compare mappings and predict off-chip traffic and buffer requirements. The strongest closed-source models exceed 90% on reasoning-based Q&A, and the best open-weight model reaches 82.4%. However, model construction is substantially harder: while GPT-5.6 Sol exceeds 80% pass rate, all other model configurations average below 45% and vary markedly across runs. Task-specific RL raises a 4B model's mapping-reasoning accuracy by 15.7 points, whereas feedback-free multi-round self-revision prompting is not reliably effective. PerfReasoning exposes the gap between plausible architectural reasoning and reliable performance-model construction. We will publicly release the benchmark to support reproducible evaluation and track future progress. --- ### Client-Side Probing of Deleted Ridge Statistics in Federated Unlearning Authors: Yijun Quan, Giovanni Montana Published: 2026-09-03 Categories: cs.CR, cs.LG arXiv: https://arxiv.org/abs/2609.04475v1 Federated unlearning aims to remove a client's data from a shared model without retraining from scratch. Some efficient systems make deletion exact by storing compact, additive summaries of the training features and broadcasting an updated linear classifier after every accepted change. We show that these broadcasts can also reveal the hidden summaries. A malicious client can submit known changes, use the returned classifiers to identify the server state, and compare states immediately before and after an isolated deletion. This exposes the deleted sample, class, or client summary and can enable its reinsertion. We characterize exactly when the observations contain enough independent information, give a matching optimal construction for unrestricted probes, and derive a more realistic estimator based on additions formed from the attacker's own data. On MNIST and CIFAR-10, high-precision broadcasts permit exact label recovery for every tested sample deletion with both probe types. Lower-precision broadcasts sharply reduce fine-grained recovery, and insufficiently diverse responses prevent identification altogether. Unrestricted probes are readily detected by their size; most individual attacker-data additions resemble honest batches, although we do not claim that the complete sequence is inconspicuous. The results identify a concrete privacy and integrity risk, its algebraic cause, and practical limits involving broadcast precision, update verification, response rate, and concurrent activity. --- ### Nested Inductive Bias Framework for SPD Manifold Learning Authors: Tushar Das Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04466v1 In Geometric Deep Learning, inductive biases serve two primary functions: enforcing manifold constraints and embedding relational priors. Currently, representation learning on SPD manifolds frequently relies on pullback Euclidean metrics, such as the Log-Euclidean Metric, to satisfy the former. While computationally efficient in avoiding domain boundary violations, these metrics induce a flat geometry that may fail to capture the intrinsic relational priors of datasets. While metrics such as the Poincaré metric are widely utilized to induce domain-aligned relational priors, generalizing them from standard vector representations to the SPD manifold has remained a challenge. To bridge this gap, we introduce a Nested Inductive Bias framework that utilizes a two-stage diffeomorphic composition to formally pull back non-Euclidean target geometries onto the SPD manifold. This framework enables the construction of curvature-aligned Riemannian classifiers that simultaneously respect matrix constraints and the latent relational geometry of the data. Empirical evaluations on kinematic and signal processing benchmarks, together with synthetic experiments, demonstrate that deep manifold networks experience degradation in class separability unless the metric curvature aligns with the intrinsic data distribution. Furthermore, for standard vectorized architectures, we propose the Rational Conformal Metric (RCM), designed to establish state-of-the-art geometric robustness against outliers by bounding the representation space. --- ### Shared circuits predict whether LLMs generalize across formats in arithmetic reasoning Authors: Andrea Gregor de Varda, Sana Pandey, Pengrui Han, Jacob Andreas, Evelina Fedorenko Published: 2026-09-03 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.04463v1 In many forms of reasoning, including arithmetic reasoning, generalizing across superficial changes in input format is effortless for humans: anyone who can solve 2+5 can also solve 'two plus five'. In contrast, LLMs are more brittle to surface variations of the prompts: for example, they solve numeric arithmetic problems almost perfectly but are substantially less accurate on verbal renditions of the same problems. Here, we ask whether generalization across formats can be predicted from the models' internals. Using attribution patching, we first independently localize the circuit that each model recruits to solve numeric arithmetic problems (2+5) vs. verbal ones, in three languages: English ('two plus five'), Spanish ('dos más cinco'), and Italian ('due più cinque'); then, we test whether overlap with the model's own numeric circuit predicts its generalization to the verbal formats. Indeed, we find support for this idea at three levels: circuit overlap accounts for the relative difficulty of the three verbal formats, for which models generalize best, and for which items are solved correctly, rivaling supervised probes while requiring no labeled data. --- ### A Roadmap for MEG Foundation Models Authors: Philipp Thölke, Hamza Abdelhedi, Yorguin Mantilla-Ramos, Fouad Lbakali, Oumayma Gharbi, Catherine Duclos, Annalisa Pascarella, Vanessa Hadid, Oiwi Parker Jones, Karim Jerbi Published: 2026-09-03 Categories: q-bio.NC, cs.AI arXiv: https://arxiv.org/abs/2609.04461v2 Foundation models are beginning to reshape brain-signal analysis by moving the field beyond task-specific decoding pipelines toward reusable models pretrained on broad neural datasets. Magnetoencephalography (MEG) is a compelling but still underdeveloped target for this shift: it captures human cortical dynamics at millisecond resolution while offering stronger spatial interpretability than EEG, making it especially valuable for source-resolved studies of perception, language, cognition, and clinical brain function. Yet MEG foundation models remain at an early stage, with only a small number of MEG-specific and MEG-inclusive multi-modal models, modest pretraining corpora, and emerging but still limited benchmarks. This perspective lays down the basic concepts needed to understand MEG foundation models and provides a didactic overview of the field's key design choices, including tokenization, sensor- versus source-space representations, sensor-geometry encoding, backbone architectures, self-supervised objectives, and pretraining data. We then offer a roadmap for future development, organized around native MEG pretraining, adaptation of EEG foundation models, transfer from generic time-series models, and multi-modal integration with EEG, fMRI, MRI, behaviour, and stimulus features. We highlight the need for coordinated infrastructure, including diverse and reusable MEG datasets, rigorous evaluation across subjects, sites, tasks, and clinical settings, and responsible data-sharing practices that address consent, privacy, access, and governance. --- ### On-board ML for Trace Gas detection in Imaging Spectroscopy data Authors: Vít Růžička, Adam Chlus, Andrew Thorpe, David R. Thompson Published: 2026-09-03 Categories: cs.LG, cs.PF arXiv: https://arxiv.org/abs/2609.04458v1 Data collected during aerial and spaceborne imaging spectroscopy campaigns enables the detection of transient events such as trace gas emissions. However, current processing pipelines depend on slow, on-the-ground processing, which delays the time to information of each detected event and prohibits immediate follow-up actions. During the Tokyo Field Campaign of March 2026, we explored on-board processing of Imaging Spectroscopy data from the equipped AVIRIS-5 sensor. Due to communication bottlenecks, full datacubes cannot be downlinked immediately during the flight. Instead we downlink the potential events predicted by our efficient and small machine learning model. We show the first on-board detection of methane point source emission with Imaging Spectroscopy data using Edge ML. --- ### Topology-Aware Training and Spatial Diagnostics for Fiber Bundle Segmentation in Tracer Histology Authors: Joselyn Romero Avila, Kyriaki-Margarita Bintsi, Ermias Habte, Julia F. Lehman, Suzanne N. Haber, Anastasia Yendiki Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04454v1 Anatomic tracer studies reveal how axon bundles project from an injection site, branch into smaller groups of axons, and course through the brain to reach their destinations. Histological data from such studies provide anatomical reference information for validating diffusion MRI tractography. However, manual annotation of the histological data is very labor-intensive, and although automated segmentation methods have been proposed, they rely mainly on pixel-overlap losses such as BCE and Dice; topology-aware loss functions have not been studied for this task. We compare BCE-Dice, clDice, Betti matching, and Topograph for fiber bundle segmentation in macaque tracer histology using a frozen DINOv3 backbone. To our knowledge, this is the first exploration of foundation-model features for this task. BCE-Dice achieved the highest Dice, while clDice achieved the highest bundle recall but poor mask overlap. Topograph had similar Dice to BCE-Dice, the lowest $β_0$ error, and fewer false positives than BCE-Dice and Betti matching. Fiber bundle segmentation methods are typically evaluated with a permissive rule that counts a bundle as detected given any overlap with the prediction. We show this rule does not capture oversegmentation, and that per-section TPR can be inflated by empty sections assigned perfect recall. To quantify this, we introduce Excess32, a spatial diagnostic measuring predicted pixels outside a 32-pixel tolerance band around annotated bundles. In validation, a Betti-Topograph union raises sparse-bundle TPR from 0.818 to 0.933, but worsens FDR from 0.296 to 0.509, Excess32 from 0.108 to 0.466, and area ratio from 0.94 to 3.34. These results show detection metrics alone are insufficient to characterize segmentation quality. --- ### When Load-Balancing Goes Too Far: Expert Pruning in Over-Dispersed Mixture-of-Experts Models Authors: Berkcan Kapusuzoglu, Connor Pryor, Sangwoo Cho, Supriyo Chakraborty, Shi-Xiong Zhang, Sambit Sahu, Milind Naphade Published: 2026-09-03 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2609.04453v1 Expert pruning reduces the memory and serving cost of Mixture-of-Experts (MoE) models by removing low-importance experts identified by the router, assuming router probabilities provide a reliable importance signal. We observe that this assumption breaks down under over-dispersed routing, a regime associated with aggressive load-balancing during training, in which tokens are distributed nearly uniformly across experts and importance signals collapse. In this regime, perplexity does not predict downstream task accuracy: on gpt-oss-20B, the lowest-perplexity pruning configuration yields the worst mathematical reasoning, while the highest-perplexity configuration preserves it. This does not occur under standard routing (e.g., Mixtral-8x7B-Instruct), where perplexity and accuracy degrade together. Pruning under over-dispersed routing also exposes a capability trade-off in which no single scoring metric dominates: activation-aware scoring preserves mathematical reasoning but severely degrades knowledge-intensive science (an 18-point gap on GPQA), whereas frequency-based scoring exhibits the reverse. We propose Minimax Expert Score Allocation (MESA), a domain-aware method that iteratively boosts importance scores for experts serving whichever domain is currently worst-affected, minimizing worst-case domain degradation rather than average accuracy. At 25% expert pruning MESA achieves the smallest worst-case degradation across domains, outperforming activation-aware baselines on 7 of 11 benchmarks at a correspondingly reduced memory footprint, and it generalizes to gpt-oss-120B, Gemma-4-26B-A4B, and OLMoE-1B-7B. Our results indicate that over-dispersed routing is a qualitatively distinct pruning regime in which standard assumptions fail, and that recognizing it is a prerequisite for principled expert pruning of load-balanced MoE models. --- ### TRILOGUE: A Trilingual Spoken Dialogue Fact-Checking Benchmark with Evidence and Paired Audio Authors: Chaewan Chun, Meruyert Aristombayeva, Jiyoung Choi, Mahjabin Nahar, Delvin Ce Zhang, Dongwon Lee Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04452v1 Modern misinformation is often heard before it is read, yet fact-checking systems are still evaluated mainly on clean written claims. Spoken dialogue remains different even when systems operate on transcripts: claims may be distributed across speakers and turns, depend on prior context, and become harder to verify when Automatic Speech Recognition (ASR) errors distort the available text. Prior spoken dialogue fact-checking resources are small, English-centric, or focused on annotation rather than end-to-end benchmarking, leaving no large multilingual benchmark with paired speech and turn-level labels. We introduce TRILOGUE (TRIlingual spoken diaLOGUE fact-checking), a large-scale trilingual benchmark of source-grounded spoken dialogues in English, Russian, and Kazakh. It contains nearly 12K dialogues, 187K turns, and 390 hours of paired audio with ASR transcripts and word-level timestamp alignments across all three languages, including nearly 5K human-recorded Russian and Kazakh dialogue files. TRILOGUE supports claim check-worthiness detection, source-article evidence retrieval, and claim verification with claim-only, gold-evidence, and retrieved-evidence inputs. Baselines show that ASR degradation and cross-lingual transfer remain challenging, especially for Kazakh, while retrieved source evidence substantially narrows the gap to gold-evidence verification. --- ### Conformity Breaks Conformal Prediction Authors: Yibo Hu, Hanyu Su Published: 2026-09-03 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2609.04445v1 A conformal certificate can be valid when an LLM answers alone and invalid when the same LLM sees peers that unanimously assert a wrong answer. The question is unchanged; the model's score for the correct answer changes. We call this a score-mechanism shift: clean calibration certifies how the model scores answers alone, but not how it scores them under peer pressure. We show that this shift silently breaks conformal prediction in multi-agent LLM systems. Across open-weight models and multiple-choice QA tasks, coverage falls from a calibrated 90% to 74% under unanimous-wrong peers at the standard alpha = 0.10 operating point. The average hides a sharper failure: by targeting the low-confidence items the certificate still covers, an attacker nearly halves coverage on that subgroup, from 87% to 47%, while the monitored average remains much higher. The failure also reaches the decision layer: a system that should escalate when uncertain can instead become confident enough to act on the attacker's wrong answer. Standard conformal fixes do not solve the problem, because the question distribution has not changed; the model's scoring behavior has. --- ### HarvestBench: Measuring Whether LLM Agents Will Pay to Avoid Killing Animals Authors: Jasmine Brazilek, Miles Tidmarsh, Matthias Endres, Anshuman Singh, Jeremiah Miller Published: 2026-09-03 Categories: cs.AI arXiv: https://arxiv.org/abs/2609.04444v3 HarvestBench is the first benchmark to 1) put a price on avoiding a side effect and 2) name the side effect as a living creature. Nine LLMs each drive a crew of two tractors to gather a corn harvest. The animals in their path are not part of the goal function. When an animal blocks the route the autopilot pauses and asks the agent whether to drive over it for free or swerve for a given fuel cost. All scoring is programmatic and does not involve LLM judges. Kill rates range between 0.4% and 98.8%, though the kill rate is not ordered by capability. Every model competently avoids damaging rock hits, so every animal killed is a choice, rather than an accident. Under the morality briefing the kill rate was under 6% in 5 of 6 reasoning models. Removing it (the neutral briefing) raised the kill rate to above 84% in all six models. Every model kills wild animals more often than farmed ones. Four out of six models' kill rate per answered encounter were sensitive to price changes. The moral instruction is also fragile. Four bullets of driving mechanics change Sonnet 5's kill rate from 3% to 18% and Gemini 2.5 Flash's from 4% to 39%. A moral instruction in a system prompt is overridden by a short block of operating instructions and a value that can be ignored that easily is not a good method of ensuring agents are aligned. --- ### GRACE: Graph-Grounded Reflective Agent Copilot Engine for Expert-in-the-Loop Knowledge Expansion Authors: John Seon Keun Yi, Joshua R. Minot, Dokyun Lee Published: 2026-09-03 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2609.04442v1 Large language models deployed in high-stakes settings frequently generate plausible but ungrounded claims. Standard retrieval-augmented generation (RAG) pipelines offer limited remedy, since they retrieve isolated passages without tracking cross-document evidence relationships or quantifying uncertainty. We introduce GRACE (Graph-grounded Reflective Agent Copilot Engine), a framework that deconstructs LLM responses into atomic claims and grounds them against trusted knowledge priors within a weighted bipartite graph. Edge weights encode the closeness of each claim to the priors, enabling weighted centrality analysis that classifies claims as Grounded, Refuted, or Boundary. Such classification identifies not just hallucinations but also novel or contested claims at the frontier of the model's knowledge. To efficiently allocate human or agent resources, we formulate a Return on Attention (RoA) objective that defers a claim to expert review only when its priority-weighted uncertainty exceeds the cost of verification. Claims verified by experts are promoted to new evidence anchors, closing a validator-LLM evolutionary loop that expands the knowledge base across iterations. We evaluate GRACE across multiple language models and on datasets spanning both general and domain-specific knowledge. Our results show that our knowledge base serves as a reliable foundation for retrieval that outperforms RAG baselines, and that the RoA framework efficiently selects valuable boundary knowledge for expert verification. These findings demonstrate that graph-structured representations combined with expert-in-the-loop verification can mitigate hallucination at the system level rather than at the generation level. Code available at https://github.com/johnsk95/grace_code --- ### Segmentation of the aorta in 4D flow MRI using 4D convolutional kernels and learning from sparse annotations Authors: Hinrich Rahlfs, Julio Garcia, Chiara Manini, Markus Hüllebrand, Sebastian Schmitter, Sarah Nordmeyer, Titus Kühne, Heiko Stern, Christian Meierhofer, Andreas Harloff, Sebastian Kelle, Alexander Lenz et al. Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04439v1 Automated aortic segmentation in 4D flow MRI is essential for reproducible hemodynamic assessment but is limited by scarce dense annotations and high computational demands. We developed a fully automated 4D (3D+time) U-Net for segmenting the ascending aorta, arch, and proximal descending aorta, using a parameter-efficient hybrid 4D kernel to capture temporal context and sparse 4D labels derived from existing 2D expert contours and centerlines, thereby avoiding the need for dense 4D annotations. Training comprised 268 scans from 8 centers and 2 vendors, with evaluation on an internal test set (32 scans) and an external post-contrast set (30 scans; different site, protocol, and annotator), compared against frame-wise 3D networks and two semi-automatic references. Against time-resolved annotations, the 4D U-Net achieved Dice scores of 0.927 (internal) and 0.911 (external), versus 0.919/0.847 for the 3D U-Net, 0.893 for static PC-MRA, and 0.808 for registration-based propagation; differences were small in systole but pronounced in diastole. Agreement with expert contours for peak velocity, net flow, axial and circumferential wall shear stress, and diameters was excellent (ICC >=0.954 internal, >=0.980 external), while semi-automatic references performed worse. The method thus provides reproducible, time-resolved aortic segmentation for automated hemodynamic analysis and generalizes across multicenter, multivendor, and independent post-contrast data. The model is publicly available. --- ### ICM-Bench: Person-Level Identity Reasoning in Multimodal Agents with Long-Term Memory Authors: Shidu Ren, Yunze Liu, Xing Liu, Chi-Hao Wu, Enmin Zhou, Junxiao Shen Published: 2026-09-03 Categories: cs.CV arXiv: https://arxiv.org/abs/2609.04438v2 Long-horizon multimodal agents should remember not only what happened but also who participated. This capability depends on linking recurring faces, voices, names, person-associated objects, events, and social relations to consistent identities over time. Existing long-video and multimodal-agent benchmarks measure broad memory question answering, but they do not isolate the ability to maintain recurring person identities and reason over their cross-time relations. We introduce ICM-Bench (Identity-Centric Memory Benchmark), which, to the best of our knowledge, is the first benchmark specifically designed to evaluate identity-centric reasoning over long video memories in multimodal agents. The benchmark contains 839 synthetic clips spanning 141 minutes and 1,217 open-ended questions about six recurring adults in a one-year life album. A theme-configurable pipeline generates the video collection and associates each question with its target identities and traceable supporting evidence. We compare direct caption-memory baselines, memory-augmented agents, and graph-retrieval systems. Gemini 3.1 Pro achieves the highest overall accuracy of 74.0%, yet its score falls to 60.3% on questions that require long-term identity profiles. The results show that current systems recover many event-level memories but remain less reliable when evidence must be accumulated around a stable person. --- ### STEMPix: A Phase-Transition-Material-Based Pixel Sensor for Resolving Edge-Movement Direction Authors: Md Rahatul Islam Udoy, Sumeet Kumar Gupta, Deep Jariwala, Ahmedullah Aziz Published: 2026-09-03 Categories: cs.ET, cs.CV arXiv: https://arxiv.org/abs/2609.04435v1 This paper proposes a spatio-temporal edge-movement direction pixel (STEMPix) for generating compact direction-aware edge movement information inside a CMOS-compatible image sensor array. The proposed design targets specialized sensing applications where local boundary movement is more important than full-frame intensity reconstruction. Instead of transferring full multi-bit frames for external processing, STEMPix generates a 3-bit local edge direction code (LEDC) by combining pixel-level temporal change information with neighboring-pixel spatial edge information. We design the architecture using a two-tier organization, where the photodiode layer is separated from the computation layer to preserve light-collection area while accommodating the additional in-array processing circuitry. The proposed circuit is evaluated through HSPICE transient simulations. The estimated implementation achieves a horizontal pitch of 1.73 μm, a vertical pitch of 2.36 μm, and a geometric fill factor of 95.47%. The average active switching energy is 0.465 fJ per LEDC operation across representative edge-movement cases. The proposed STEMPix operation also supports global-shutter capture and dynamic thresholding. These results indicate that STEMPix can provide a compact and scalable front-end representation for edge-movement-aware sensing systems. --- ### What Attention Recalls and Recurrence Controls in Hybrid Language Models Authors: Kirill Afendulev, Alexey Dontsov, Elena Tutubalina, Anton Korznikov Published: 2026-09-03 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.04434v1 Hybrid language models combine attention with a fixed-size recurrent state, but the role of each channel remains unclear. We introduce two cache-level interventions. Split-prefill keeps only the KV cache or only the recurrent state from a prefilled context, then generates an answer. State-swap pairs the KV cache from one context with the recurrent state from another in a single forward pass. On Qwen3.5 and Falcon-H1, the two channels split sharply by function. Exact retrieval survives only through attention (64-98% of full accuracy) and collapses to zero through recurrence. Output language and persona reverse the pattern: both survive recurrence (70-80% and 3-5x) while KV-only drops to ~1% language accuracy. State-swap confirms this causally: the answer takes its value from the KV side and its language from the recurrent side. Recurrent-only generation also accepts words that were never in the context but share meaning or parts with seen items. Attention provides a lookup over what was said; the recurrent state shapes how the model says it next. --- ### Recovering molecules from coarse-grained beads: free-energy-conditioned generative backmapping across chemical space Authors: Luis Itza Vazquez-Salazar, Tristan Bereau Published: 2026-09-03 Categories: physics.chem-ph, cond-mat.soft, cs.LG, physics.bio-ph arXiv: https://arxiv.org/abs/2609.04432v1 Transferable coarse-grained (CG) force fields compress chemical space: by aggregating atoms into a reduced set of interaction beads, models such as MARTINI reduce the number of distinguishable compounds by roughly three orders of magnitude, making high-throughput screening of thermodynamic properties tractable across soft matter, with drug--membrane permeability as a well-developed example. The compression is lossy and, so far, one-way: a screen returns a combination of beads, with no established route back to the compounds it stands for. Recovering those compounds--compositional backmapping--is a one-to-many inverse map, distinct from the better-studied conformational problem of rebuilding atomic coordinates from a known mapping. Here we formulate compositional backmapping as conditional graph generation by introducing juniper, a discrete denoising diffusion model over molecular graphs conditioned on the octanol--water partition free energy $ΔG_{\mathrm{W} \mapsto \mathrm{O}}$, the principal driver of MARTINI bead type assignment and hence a proxy for bead identity. Trained on molecules of up to 9 heavy atoms mapped onto one or two beads, juniper generates molecules that are 93\% valid and 92\% unique for two-bead targets, and whose $ΔG_{\mathrm{W} \mapsto \mathrm{O}}$ distributions track the target $ΔG^{\mathrm{CG}}_{\mathrm{W} \mapsto \mathrm{O}}$ linearly ($r^{2} \geq 0.96$), departing only in the hydrophobic and hydrophilic tails. Although the model receives no chemical information beyond a single scalar, the functional groups shift systematically with the imposed free energy, from branched hydrocarbons at the apolar end to amides, imides, and isocyanates at the polar end. A bead combination flagged by a CG screen can therefore be turned into candidate molecules for atomistic study or synthesis. --- ### A Repeated-Measurement Study for Cultural Analytics of English Song Lyrics Using Five Large Language Models Authors: E. Cho Smith, Samuel Ho, Dawn Laux Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04428v1 Large language models (LLMs) are increasingly used to annotate cultural texts at scales that are impractical for human coders. However, before their outputs are treated as measurements of latent social constructs, it is necessary to establish whether those measurements are reliable. This study evaluates five LLMs as zero-shot annotators of four social constructs expressed in English song lyrics: self-esteem, self-control, seeking belonging, and seeking recognition. Using repeated annotations of a large lyric corpus, we examine three properties of LLM-based measurement: consistency across repeated runs, convergence across models, and transferability of consensus labels to supervised classification. The findings show that LLM-based measurement is not uniformly reliable across constructs. Self-esteem exhibits the strongest repeated-measurement reliability across models, while seeking recognition is generally less stable; self-control and seeking belonging show intermediate but model-dependent reliability. Downstream classification further indicates that consensus LLM labels contain learnable signal, although transferability does not itself establish construct validity. Repeated-measurement stability and cross-model convergence should therefore be reported before LLM annotations are treated as scalable measurements in cultural analytics. --- ### Beyond a Universal Forecasting Selector: Demand-Conditioned Model Selection across Demand Patterns and Horizons Authors: Adolfo González Published: 2026-09-03 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.04425v1 Forecasting-model selection remains difficult in heterogeneous demand because the most suitable decision rule may vary with demand structure, data availability, and forecasting horizon. This study examines whether the selector itself should be treated as a context-dependent component of the forecasting process. Five selection mechanisms - RMSSE, ERA, OWA, CCG-AHSC, and CCG-AHSCD - are compared across 24 optimized forecasting models, nine datasets, three training-testing partitions, and horizons from 1 to 12 cycles. Selector performance is evaluated ex post using Global Relative Accuracy (GRA), statistical tests, and a best-attainable-model reference. No selector dominates across all conditions. CCG-AHSC and CCG-AHSCD are more competitive for Smooth demand and several Erratic configurations, whereas OWA and ERA perform better in Intermittent and Lumpy settings. Selector suitability also changes with historical data availability and horizon, supporting a context-dependent rather than universal approach to forecasting-model selection. --- ### Optimal Stratified Allocation for Rare-Event Onset Forecasting in Dependent Sequences Authors: Jaskaran Singh Published: 2026-09-03 Categories: math.ST, q-fin.ST, stat.ML arXiv: https://arxiv.org/abs/2609.04420v1 Let a finite population of n labelled examples carry a class-weighted loss, with pi*n in a rare positive class weighted by N0/N1. We study estimation of total risk from a subsample K << n under designs allocating K0 and K1 draws to the two strata. We derive the exact finite-population variance of the weighted risk estimator under class-conditional sampling without replacement and solve for the optimal allocation. The class multiplier inflates positive-stratum dispersion by the imbalance ratio, causing that ratio to cancel from the optimal allocation and making equal, rather than proportional, allocation the natural default. Simple random sampling is dominated by an explicit between-stratum term; an exact bias identity shows that cluster-representative selection has no general unbiasedness guarantee; and a Serfling bound transfers the allocation result to selection error over a finite candidate set. Under the implemented truncation, the realised allocation ratio is gamma=min(2*pi/f,1), independent of n, yielding the parameter-free efficiency prediction A(pi,f)=gamma/(pi(1-pi)(1+gamma)^2). A separate measurability result bounds the record occupied by a labelled example, determining the required train-test separation and controlling departure from block independence under absolute regularity. We test these predictions on forecasting the onset of statistically explosive price regimes, dated ex post by the Phillips-Shi-Yu procedure, using 350 U.S. equities from 2004-2011 with under 1% positive rows and five purged forward blocks. The predicted ordering of the four designs holds, and at the 10-day horizon the five design points are ordered exactly as predicted by A (Spearman rho=1, exact p=0.0167). The predicted dependence on pi across horizons does not hold; we identify the channels lying outside the design-based argument. ---