# Research Papers — 2026-07 · bundle page 70 of 222 > Full-text markdown of papers #3451–3500 of 11075 published in 2026-07, newest first. AI-readable mirrors of the full 11075-paper 2026-07 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-07/page/70 [← Previous bundle](https://aipod.dev/papers/archive/2026-07/page/69.md) [Next bundle →](https://aipod.dev/papers/archive/2026-07/page/71.md) ### SubSplat: High-Resolution Pixel-aligned 3DGS via Sub-pixel Gaussian Reparameterization Authors: Jiun Lee, Jaekwang Kim, Sangmin Lee Published: 2026-07-23 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.20813v1 Pixel-aligned Gaussian splatting enables efficient and generalizable novel-view synthesis. However, high-resolution rendering faces a critical trade-off where increasing input resolution improves detail at the expense of quadratically rising network computational cost. Conversely, maintaining low-resolution inputs stabilizes this cost but results in insufficient Gaussian density and artifacts. To address this, we propose SubSplat, which introduces Sub-pixel Gaussian Reparameterizer(SPGR) to subdivide primary Gaussians into fine-grained primitives, restoring structural density directly from low-resolution features. We further enhance the reparameterization quality through feature aggregation, which effectively captures high-frequency details across multiple views. Experiments on RealEstate10K and ACID demonstrate that SubSplat achieves high-fidelity rendering with superior efficiency. Our results validate that the proposed framework successfully resolves the trade-off between reparameterization fidelity and network computational cost inherent in pixel-aligned Gaussian Splatting. --- ### New Complexity-Theoretic Frontiers of Tractability for Neural Network Training Authors: Cornelius Brand, Robert Ganian, Mathis Rocton Published: 2026-07-23 Categories: cs.LG, cs.DS arXiv: https://arxiv.org/abs/2607.20811v1 In spite of the fundamental role of neural networks in contemporary machine learning research, our understanding of the computational complexity of optimally training neural networks remains incomplete even when dealing with the simplest kinds of activation functions. Indeed, while there has been a number of very recent results that establish ever-tighter lower bounds for the problem under linear and ReLU activation functions, less progress has been made towards the identification of novel polynomial-time tractable network architectures. In this article we obtain novel algorithmic upper bounds for training linear- and ReLU-activated neural networks to optimality which push the boundaries of tractability for these problems beyond the previous state of the art. In particular, for ReLU networks we establish the polynomial-time tractability of all architectures where hidden neurons have an out-degree of $1$, improving upon the previous algorithm of Arora, Basu, Mianjy and Mukherjee. On the other hand, for networks with linear activation functions we identify the first non-trivial polynomial-time solvable class of networks by obtaining an algorithm that can optimally train network architectures satisfying a novel data throughput condition. --- ### Profiling Lightweight Large Language Models Authors: Tomohiro Harada, Enrique Alba, Gabriel Luque Published: 2026-07-23 Categories: cs.AI arXiv: https://arxiv.org/abs/2607.20806v1 Lightweight large language models (LLMs) are increasingly being deployed locally on personal computers and are expected to play a growing role in resource-constrained edge and mobile environments. In such settings, energy consumption, execution time, and memory usage directly affect practical usability, yet existing evaluations of LLM efficiency largely rely on proxy descriptors such as parameter count or FLOPs, often decoupled from task precision. This paper introduces a PTME-based experimental framework for the precision-aware profiling of lightweight LLM inference, jointly measuring Precision, execution Time, peak Memory usage, and Energy consumption through direct hardware-level measurements. The methodology is applied to a representative set of lightweight LLMs executed locally under edge-class resource envelopes on a controlled desktop platform, using benchmarks spanning code generation, mathematical reasoning, and multi-task understanding. We find that static proxy descriptors approximate inference cost well but fail to predict precision. Tightening the resource envelope increases cost without affecting precision, amplifying execution time more strongly than energy and penalizing larger models the most. Moreover, no single model dominates across all PTME dimensions, and a Pareto analysis reveals non-dominated configurations that would be hidden by accuracy-only or efficiency-only assessments, providing practical guidance for selecting models under different resource envelopes. These results show that selecting lightweight LLMs by size, FLOPs, latency, or accuracy alone can select the wrong deployment candidate; PTME profiling exposes configurations that preserve useful accuracy at lower physical cost. --- ### The Geometry of Personality: Activation Steering with Jungian Cognitive Functions Authors: Liu Zai, Yumeng Wang, Junchen Fu, Joemon M. Jose Published: 2026-07-23 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2607.20803v2 Activation steering enables control and interpretation of LLMs, yet existing work primarily models personality through static trait frameworks such as the Big Five. We investigate whether personality can instead be represented and controlled as a set of cognitive processes using the eight Jungian Cognitive Functions. To this end, we introduce a framework comprising a Jungian evaluation protocol and a dataset of over 2,100 role-playing character narrations. Activation steering vector extraction and evaluation experiments on Llama-3.1-8B demonstrate effective monotonic control over all eight cognitive functions through activation steering. Beyond controllability, our analysis reveals that: 1. personality information is concentrated in middle transformer layers; 2. steering vectors exhibit structured geometric relationships consistent with distinctions between rational and irrational functions; 3. effective multi-dimensional steering directions cannot be recovered as linear combinations of single-function directions. These findings provide new insights into the representation of personality in LLM activation space and establish a framework for studying interpretable, effective, and multi-dimensional personality control. --- ### Observing the Conduct of Systematic Reviews with Generative AI Support: An Experience Report from a Graduate Software Engineering Course Authors: Danilo Monteiro Ribeiro, Gilberto Sussumu Hida Published: 2026-07-22 Categories: cs.CY, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2609.26057v1 Context: Secondary studies are fundamental practices in Evidence- Based Software Engineering, but teaching them requires activities that expose students to authentic methodological decisions. Objective: This paper reports an experience in a graduate course in which ten doctoral students in Software Engineering, organized into three groups, piloted secondary studies with and without support from generative AI. Method: A single-day classroom session was organized and observed, in which the groups conducted pilot systematic reviews with and without generative AI support. Classroom observations, produced artifacts, and interaction threads with assistants configured in ChatGPT were analyzed to reconstruct how each group appropriated the technology throughout the activity. Results: LLMs reduced initial barriers, accelerated the generation of alternatives, and made methodological problems more explicit, but they also favored excessive delegation, superficial validation, operational difficulties, and a shift in focus from conducting the SLR to using the tool. Conclusion: The experience offers a situated, observational account of how doctoral students engaged with generative AI during a systematic review activity, and the resulting insights also inform the design of a subsequent controlled study. The findings indicate that generative AI can support practical learning about SLRs, provided that its use is accompanied by human supervision, decision records, and critical reflection on its limitations. --- ### SAGE: Schema-Guided LLMs for Grant Review Authors: Erik Varapaev, Andrei Chetvergov, Stepan Ukolov, Timofei Sivoraksha, Alexander Evseev, Sergey Bolovtsov Published: 2026-07-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.20829v1 Grant reviewers must apply detailed criteria to application forms, budgets, and supporting documents while producing assessments that colleagues can inspect. We present SAGE, Schema-Guided Aspect-Based Grant Evaluation, a system that translates a grant rubric into structured checks and links its judgements to evidence from the application package. We evaluate SAGE in two stages on 35 nonprofit grant applications. A post-factum comparison with 105 reviews from the original competition shows fair ordinal agreement (kappa = 0.29). The foundation then conducted a criterion-level re-review after inspecting SAGE, producing 202 assessments. In this assisted round, SAGE reached kappa = 0.58 and outperformed a one-prompt-per-criterion baseline (kappa = 0.33 on the common subset), with higher rank correlation and lower error. A claim-level audit further identifies confirmed, disputed, and unaddressed parts of the structured draft. SAGE operationalizes the review methodology by producing a detailed, evidence-linked, and auditable draft for expert correction. --- ### Beyond WER: Entity and Disfluency Recall in Accented Conversational ASR Authors: Fiza Husain, Ankit Pandey, Yash Singh Published: 2026-07-22 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.20828v1 ASR systems optimised for Word Error Rate (WER) often miss named entities and filled pauses in accented conversational English, both critical for language-learning feedback. We present a three-stage pipeline for speakers from India, Indonesia, and Latin America: (1) heuristic SQL filters curating entity-rich training data at 2.8x the entity density of random sampling, (2) regional LoRA adapters fine-tuned on Qwen2.5-Omni-3B producing both verbatim and corrected transcripts in a single forward pass, and (3) a six-category error taxonomy validated by an LLM-based judge (83.8% agreement, 210 human-labelled samples). The pipeline achieves 80-85% entity recall (up from 53-55%), 76-86% filler recall (up from <5%), and 6-10% WER across 6k test utterances, outperforming Whisper and a commercial ASR on entity recall while matching a zero-shot 30B model with 10x fewer parameters. Paired bootstrap tests confirm that curation alone accounts for 2.8-4.2 pp of entity recall gain (p<0.0001). --- ### From Discharge Notes to Patient Understanding: Persona-Grounded, Open-Ended Simulation of LLMs as Discharge Educators Authors: Won Seok Jang, Zonghai Yao, Hong Yu Published: 2026-07-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.20827v1 Hospital discharge education is an interactive teaching task: a clinician adapts a discharge plan to a patient's literacy, recall, and personality. Existing LLM evaluations target static or artifact-generation tasks and do not measure patient understanding under open-ended dialogue. We introduce DischargeBench, a persona-grounded simulation in which a candidate LLM educator conducts a multi-turn session with a Virtual Patient, while an Education Monitor Agent regulates patient realism without modifying the educator, protecting the evaluation signal. We curate MIMIC-IV-Ext-DischargeBench, 477 cases over 24 ICD chapters with persona axes (personality, education level, health literacy, past-medical-history recall) for stratified analysis. Each simulation is scored on four axes -- Conversation Quality, Topic Checklist, Comprehension, and Factual Consistency -- by an LLM-as-a-Judge aligned against physician annotations. Across closed- and open-source LLMs, aggregate scores conceal clinically relevant variation across ICD chapters and patient personas; difficult personas expose coverage failures, comprehension gaps, and reduced source-answer agreement. LLM evaluation for discharge education should center patient understanding, not text quality or answer accuracy alone. --- ### TALON: A Temporally Aware Longitudinal Framework for Radiology Report Generation Authors: Nien-Tsyr Sun, Min-Chen Chen, Hui Nien Hung, Vincent S. Tseng Published: 2026-07-22 Categories: cs.CL, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.20826v1 Current radiology report generation (RRG) models usually produce descriptive reports based on a single examination or only the most recent prior examination, limiting their ability to perform accurate and meaningful longitudinal comparisons and detect subtle interval changes. Although recent approaches have begun to incorporate multiple prior examinations, they usually aggregate a fixed-length history without explicitly modeling the role-dependent relevance of each prior examination before fusion. To address this, we propose TALON, a Temporally Aware LONgitudinal RRG framework that adaptively integrates variable-length patient histories. The underlying Dual-Channel Temporal Fusion Module (DCTFM) compares the current examination with each prior examination through complementary similarity and change channels to capture persistent findings and interval changes, respectively. The specially designed channel-specific attention estimates the relevance of each prior examination, while a learned prior-specific gate adaptively integrates informative longitudinal evidence and suppresses redundancy. Experiments on MIMIC-CXR show that TALON outperforms the current state-of-the-art method on various clinical efficacy and graph-based metrics. When more prior examinations become available, TALON's performance on these metrics improves even further, emphasizing the strength of TALON's DCTFM in modeling longitudinal RRG across longer and more complex patient histories than existing approaches. --- ### HERMES: Contrast-Aware Knowledge Graph Reasoning from Clinical Notes for Patient Outcome Prediction Authors: Gia-Bach Nguyen, Hoang-Ha Nguyen, Tuan-Cuong Vuong, Trang Mai Xuan, Duy Quoc Ngo, Tien-Cuong Nguyen, Huan Vu, Thien Van Luong Published: 2026-07-22 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.20825v1 Clinical predictive models often rely on structured Electronic Health Record data, such as time-series and procedure codes. While recent approaches have begun leveraging unstructured clinical notes, they typically encode them as flat sequences, which may lose explicit relational and temporal structure present in clinical narratives. In response, we propose HERMES, a graph-based framework that operates exclusively on clinical text while preserving clinical relationships. This approach builds on two key ideas. First, personalized Knowledge Graphs (KGs) are constructed through Large-Language-Model-guided extraction from clinical notes with Contrastive Logic Modeling that explicitly captures temporal dynamics and treatment failures and changes in outcomes. Second, a Graph Attention Network synthesizes patient representations through graph-based learning over the KGs. Experiments on MIMIC-III and MIMIC-IV for in-hospital mortality and 30-day readmission prediction show that HERMES consistently outperforms strong text-only baselines. Our findings demonstrate that explicit relational modeling with Contrastive Logic Modeling significantly advances predictive performance. --- ### A Noise Optimum in Rehearsal-Free Continual Learning: Isolation, Mechanism, and Scope Authors: Gunner Levi Howe Published: 2026-07-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.20162v1 Injecting stochastic noise into a consolidation rule can improve a network's retention of earlier tasks up to an optimal level, then degrade it -- an inverted-U in retention vs. noise. This paper isolates what produces that optimum and maps where it holds, entirely in simulation. (1) Phenomenon: the retention inverted-U appears on several related-task continual-learning benchmarks (Split-MNIST, FashionMNIST, continual Yin-Yang). (2) Isolation: a magnitude-matched ladder shows the effect requires coherent restoring toward the consolidated weights -- a random-direction force of identical magnitude produces no optimum, and a coherent force toward the wrong target actively hurts. (3) Active ingredient: most of the optimum is recovered by coupling the anchor gain to the injected-noise variance sigma^2 -- a one-line rule that neither Ornstein-Uhlenbeck Adaptation (fixed gain) nor MESU (posterior-variance gain) implements. A forced Ornstein-Uhlenbeck calculation derives the rising flank and predicts that the optimal noise rises with per-task interference g -- confirmed out-of-sample in direction against pre-existing measurements (the exponent is unresolved at our grid). The barrier-conditioning of the originating Doob h-transform is a low-sigma safety net that bounds forgetting where the coupled gain is too weak. (4) Scope: the optimum requires shared task structure -- it is absent on permuted-MNIST, and a controlled rotated-vs-permuted comparison localizes the boundary to task structure; the precise governing quantity is left open. (5) Length: at matched severity the advantage persists but attenuates with task count, and we show no rotation family can attribute the trend (a compact-group identity). A single-seed BrainScaleS-2 demonstration of the originating rule is reported separately (Howe, arXiv:2607.06924); this paper makes no hardware claim. --- ### Robust Conformal Intrusion Detection via Traffic-Aware Calibration and Attack-Orbit Invariance Authors: Zhenpeng Li Published: 2026-07-22 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2609.19241v1 Large language models fine-tuned for network intrusion detection emit single-point predictions without statistical validity guarantees. Conformal prediction supplies a finite-sample coverage guarantee, but a threshold calibrated on clean traffic fails once an adversary perturbs controllable network features. We demonstrate this failure across three intrusion detection benchmarks and propose traffic-aware conformal prediction, which calibrates on traffic drawn from the perturbation mechanism an attacker is expected to use and provably restores coverage whenever that mechanism is known and can be sampled. A stronger, adaptive attacker that queries the target model's own score can still degrade this matched-calibration guarantee. We address this second threat model by excluding attacker-controllable features and their deterministic descendants from the scored representation, and prove that this yields an exact, pathwise coverage guarantee rather than a probabilistic bound. Across three independently fine-tuned language model architectures, this representation remains completely unchanged under every evaluated attack attempt, at a quantified seven-to-fourteen-point cost in clean accuracy relative to the unrestricted feature set. --- ### Towards Proactive Detection of User-Side Implicit Conflicts in Human-LLM Dialogue Authors: Jinqiang Wang, Tao Zhu, Huansheng Ning Published: 2026-07-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.19155v1 In Human-LLM dialogue, follow-up user utterances may implicitly conflict with earlier intents, leading the LLM to misinterpret user needs and generate inappropriate responses. A reliable dialogue system should proactively detect user-side conflicts before generating a response and seek clarification when necessary. However, prior work has largely focused on LLM-side conflicts, leaving user-side conflicts underexplored. To fill this gap, we construct UC-Bench, a human-annotated benchmark for evaluating user-side conflict detection. Preliminary experiments show that existing LLMs struggle with this task, especially when conflicts arise from implicit incompatibilities grounded in dialogue history. To improve lightweight LLMs with limited training data, we investigate data synthesis for user-side conflict detection. Existing synthesis methods do not explicitly model the implicit incompatibilities between historical and current user utterances, making it difficult to capture the evolution of conflicts and to generate reliably labeled implicit conflict samples. We propose SynUC, a constraint-guided synthesis method that represents user-side conflicts in a constraint space and uses the SPEAKING framework to guide traceable constraint transformations. Applying SynUC to WildChat, we construct UC-Data, a user-side conflict training set containing 2,487 samples. On UC-Bench, Qwen3.5-4B trained on UC-Data outperforms larger general-purpose LLMs such as Claude Opus 4.8, as well as the same backbone trained on data synthesized by existing methods. --- ### Neo-Classic: A Benchmark for Evaluating Linguistic-Aesthetic Reasoning in Classical Chinese Poetry Authors: Han Zhang, Zihan Gu, Zhiyuan Wang, Tianyi Ma, Jiacheng Lu, Xinyan Zhang, Yuhao Wei, Cheng Hua Published: 2026-07-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.19154v1 While Large Language Models (LLMs) achieve high accuracy on established Classical Chinese Poetry benchmarks, it remains challenging to distinguish transferable Linguistic-Aesthetic Reasoning from reliance on familiar pre-training patterns. To address this issue, we introduce Neo-Classic, an evaluation benchmark that combines a constructionist Out-of-Sample (OOS) dataset with a suite of reverse understanding probes. Unlike traditional benchmarks that rely on verification or generation over historical corpora, Neo-Classic comprises strictly metrical poetry authored by contemporary experts, reducing the possibility of direct retrieval. We evaluate state-of-the-art models, including Qwen3-Max, Gemini-3-Pro, and DeepSeek-V3.2, across five behavioral probes designed to test hierarchical constraint satisfaction. Our results reveal two primary limitations. First, a performance gap of 20 to 50 percent emerges when models transition from historical to contemporary texts. Second, models exhibit substantial difficulties in discourse-level ordering tasks, with standard accuracy remaining low (0 to 13 percent). Although expert-level guidance improves the performance of reasoning-enhanced models to 36 percent, a notable gap with human experts persists. These findings suggest that while current LLMs capture local formal patterns, they struggle with global hierarchical planning required for robust Linguistic-Aesthetic Reasoning. --- ### Reading Between the Lines: Can LLMs Discover the Question Behind the Text? Authors: Claudiu Creanga, Liviu P. Dinu Published: 2026-07-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2609.19070v1 This paper introduces ``question archaeology'', a specific evaluation task focused on inferring the single, authentic "genesis question" that motivated the creation of a complete text. Distinct from question generation, which targets any plausible question, or discourse frameworks that model utterance-level acts, our task assesses a model's grasp of authorial intent. We present a new dataset of commissioned texts paired with their original research questions and plausible distractors. Our evaluation of both proprietary models, like Gemini Flash and Pro, as well as open source models like Mistral and Qwen, reveals significant progress in this task, with the newer versions outperforming the earlier ones, while BERT-based models performed poorly. Notably, our findings indicate that current LLMs surpass human performance on this task, suggesting advanced understanding of authorial intent. This capability has important implications for AI's role in tasks requiring nuanced interpretation of human communication. Our work thus provides a new framework and a challenging benchmark for future models. --- ### LightSleepX: A Lightweight, Inception-Based Dual-Modal Network for Sleep Staging Authors: Yi Wang Published: 2026-07-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.19062v1 Automatic sleep staging is fundamental to personal health monitoring, yet many existing approaches are ill-suited for real-world applications. Traditional pipelines often rely on hand-crafted features or shallow machine learning models that struggle to generalize, while state-of-the-art deep learning methods, though accurate, are computationally heavy and impractical for resource-constrained environments. This paper introduces LightSleepX, a lightweight framework designed to deliver robust sleep analysis in resource-constrained environments. LightSleepX combines an Inception-style architecture with depthwise separable convolutions and Multi-scale Enhanced Attention for efficient multi-modal EEG/EOG feature extraction, and a Mamba encoder for rule-free long-range temporal modeling. On public benchmark datasets, LightSleepX achieves 85.9% accuracy and a 0.803 macro-F1 score on Sleep-EDF-20, and 81.8% accuracy and a 0.796 macro-F1 score on the cross-subject ISRUC-S3 dataset. With 0.049M parameters and 195.9 MFLOPs, the framework targets practical local deployment where computational cost and privacy are central constraints. --- ### FedGuide: Diffusion Prior Alignment and Value Baseline Guidance for Heterogeneous Federated Reinforcement Learning Authors: Zhilin He, Gauri Joshi Published: 2026-07-22 Categories: cs.LG, cs.RO arXiv: https://arxiv.org/abs/2609.18964v1 Federated Reinforcement Learning (FRL) enables collaborative policy learning across distributed agents with heterogeneous environments. While recent methods based on variance reduction, divergence penalization, and momentum optimization improve FRL under heterogeneous settings, they still primarily synchronize policy or value-network parameters and do not explicitly address distributional mismatch among heterogeneous clients. Therefore, we propose \textbf{FedGuide}, a FRL framework that uses diffusion priors as behavior models to provide personalized data supported distributions for heterogeneous local policy learning. Instead of directly averaging local policies, FedGuide aggregates those diffusion priors through Optimal-Transport Mixture-of-Experts (OT-MoE), preserving heterogeneous behavior modes in distribution space. It further develops a Distribution Correction Estimation (DICE) value baseline to provide low-variance, return-aware guidance for local policy improvement. Experiments across heterogeneous environments show that FedGuide outperforms representative FRL methods in client-average returns, final-round performance, and worst-round robustness, while maintaining stable learning under stronger heterogeneity. --- ### BLADE: ReliaBle Dynamic Hardware-Aware SNN-ANN Boundary SeLection for Event-BAseD Object DEtection Authors: Mahdi Taheri, Alwin Paul Published: 2026-07-22 Categories: cs.NE, cs.AI, cs.AR, cs.DC, cs.ET, cs.LG arXiv: https://arxiv.org/abs/2609.17562v1 Hybrid Spiking Neural Network (SNN)-Artificial Neural Network (ANN) architectures combine the energy efficiency of SNNs with the superior detection accuracy of ANNs for event-based object detection. Existing hybrid SNN--ANN networks, however, employ static inference and select the SNN-ANN boundary primarily according to accuracy and energy consumption, without considering dynamic inference or reliability. This paper presents BLADE, the first reliability-aware boundary selection methodology for dynamic hybrid SNN-ANN networks with ANN early exit. The proposed framework jointly optimizes the SNN-ANN boundary and ANN early-exit configuration according to reliability, detection accuracy, execution time, and energy consumption, while incorporating reliability through hierarchical statistical fault injection during design-space exploration. Experimental evaluation on an event-based object detector achieves an mAP 0.5 of 0.691 while reducing the inference compute energy to 15.82~mJ when the ANN early exit fires. Reliability analysis identifies the most significant floating-point exponent bit as the dominant source of catastrophic failures, producing significant-or-worse accuracy degradation in 58.8% of its fault injections. Protecting this single bit with approximately 3% storage overhead eliminates catastrophic failures across the evaluated realistic technology fault rates. Furthermore, increasing the proportion of SNN computation improves fault tolerance, with the fully SNN configuration achieving a reliability retention of 0.965 under aggressive fault conditions. The results demonstrate that jointly optimizing reliability, accuracy, execution time, and energy consumption enables more dependable deployment of dynamic hybrid SNN--ANN systems for safety-critical edge AI applications. --- ### Pay Only for Disagreement: Certified No-Regression Verdicts for Model Updates with Matching Label-Complexity Bounds Authors: Vishnu Bindu Balachandran Published: 2026-07-22 Categories: cs.LG, cs.AI, cs.CV, stat.ML arXiv: https://arxiv.org/abs/2609.17560v1 Every production model is updated, by retraining, fine-tuning, quantization, or a silent vendor swap, and each update risks being worse than what it replaced. We formalize update promotion as certified paired risk-difference auditing. Our starting point is a support identity: the risk difference between two models lives on the inputs where they disagree, observable without labels. We build DISCERN, a sequential two-tier protocol. A zero-label tier certifies benign updates whose disagreement rate is below tolerance from unlabeled traffic alone. An audited tier labels only sampled disagreements through an anytime-valid confidence sequence, valid at every stopping time and under any label-routing rule, even an adversarial judge. We prove finite-sample validity and matching label-complexity bounds of order rho^2/eps^2 at the rate level, so exploiting free disagreement provably saves a factor 1/rho over any pairing-blind auditor, and the guarantee composes across an unbounded sequence of promotions from one error budget. Across 14,000+ replayed audit streams over 785 update pairs, including LoRA fine-tunes of language models up to 1.4B parameters, miscoverage is 0.0002 (nominal 5%), power 0.986 with zero false alarms, and 56% of benign updates certify with zero labels. Each audit emits a machine-checkable evidence record for post-market monitoring. --- ### A Cross Community Agenda for Speech AI Authors: Maria Teleki, Kimi Wenzel, Anna Seo Gyeong Choi, Tobias Weinberg, Shree Harsha Bokkahalli Satish, Stephanny Sanchez, Belu Ticona, Ariadna Sanchez, Yash Sonkar, Aarti Mathur, Christoph Minixhofer, Abraham Glasser et al. Published: 2026-07-22 Categories: cs.HC, cs.CL arXiv: https://arxiv.org/abs/2609.13168v1 Speech AI, any AI system that recognizes, transforms, or generates speech, is built and evaluated across two communities with only a small overlap: technical natural language processing (NLP) venues (e.g., ACL, ICASSP, Interspeech), and sociotechnical HCI venues (e.g., ASSETS, CHI, FAccT). In this position paper, we work toward a cross-community synthesis, organizing our critique around three problems: speech AI operates with an incomplete model of communication; it operates with an incomplete model of identity; and its metrics measure the wrong constructs. We draw on AAC as a setting where these failures are most visible and their stakes highest, alongside other underserved speakers - people who stutter, multilingual speakers, and non-binary and transgender users. For each problem we offer solution sketches oriented toward designing for human variability, nearly all of which require quantitative and qualitative methods in combination. We close on the venue structures that hold these methods apart, and on what program committees and individual authors can do to bring them together. --- ### Hyperion: An AI-powered HPC cluster for sciences and humanities research that utilizes ML for predicting job turnaround time Authors: Jun Zhou, Nathan Elgar, Tawnee Benedetto, John Richards, Ming Hu, Greg Wilsbacher, Lawrence Miao, Paul Sagona Published: 2026-07-22 Categories: cs.DC, cs.LG arXiv: https://arxiv.org/abs/2609.11946v1 Hyperion is an innovative high-performance computing (HPC) cluster developed for researchers in both science and humanities disciplines at the University of South Carolina (USC). Our approach involved constructing a HPC cluster designed to meet the current research needs while accommodating future expansion. Additionally, we developed and trained two machine learning (ML) models to predict turnaround time, including wait time and wall time, and seamlessly integrated them into the Slurm job submission. Finally, we showcase a variety of sample applications hosted on the Hyperion platform. --- ### Representation learning of human cortical folding to reveal long lasting neurodevelopmental signatures Authors: Julien Laval, Robin Guiavarch, Antoine Dufournet, Racim Menasria, Barthélémy Drabczuk, Cristobal Mendoza, Saeb Tounsi, Chikh Abdelghani Baroud, Merieme Bourenane, Vanessa Troiani, William Snyder, Marisa A Patti et al. Published: 2026-07-22 Categories: q-bio.QM, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.05438v2 The human brain folds in utero, primarily during late gestation. Shortly after birth, cortical folding patterns are established and remain stable thereafter, making them promising early neurodevelopmental markers. Yet it is unclear whether the representations given by current neuroimaging foundation models capture cortical folding variability. Here, we introduce Champollion, a self-supervised learning framework that learns interpretable local representations of cortical folding from structural MRI. Optimized on representative folding-related tasks, Champollion accurately captures known folding patterns across cortical regions and external datasets. In a comprehensive benchmark, it consistently outperforms neuroimaging and general-purpose foundation models. Furthermore, Champollion reveals richer genetic associations than conventional morphometric descriptors and identifies localized folding signatures associated with incomplete hippocampal inversion, prematurity, and maternal smoking. These results establish cortical folding as a rich and largely untapped source of neurodevelopmental information, and Champollion provides a unified framework for discovering, localizing and interpreting long lasting cortical folding signatures. --- ### Mitigating Fabrication in Multi-Stage LLM Pipelines for Hiring: An Empirical Evaluation of Prompt Guardrails and Human-in-the-Loop Checkpoints Authors: Hiroko Takano Published: 2026-07-22 Categories: cs.CY, cs.CL, cs.HC arXiv: https://arxiv.org/abs/2608.26171v2 Multi-stage LLM hiring pipelines (resume improvement, interview question generation, answer feedback) can fabricate credentials, inflate qualifiers, and invent experience. We evaluate two mitigations, prompt guardrails and human-in-the-loop (HITL) checkpoints, against a fully automated baseline. In a controlled experiment (10 synthetic resumes x 2 job descriptions x 3 repetitions x 3 conditions; 180 runs), the baseline (C1) produced at least one unsupported claim in 96.7% of outputs (mean 6.80 findings/output). Prompt guardrails (C2) reduced finding density by 86% (6.80 to 0.92/output), but 50.0% of outputs still contained a fabrication, showing prompt-level mitigation alone is insufficient. A human checkpoint after resume improvement (C3) eliminated all identity fabrications, reduced finding density by 59% (6.88 to 2.82/output), reduced item-level fabrication from 96.7% to 75.0% (p=.022), and cut capture of JD-embedded trap requirements from 47% to 2% (vs. 5% under the guardrail). An exploratory analysis of multi-specialty resumes shows contamination rising monotonically with domain distance between specialties, suggesting career changers are especially exposed. The reviewer in this study caught all flagrant fabrications, but subtle qualifier drops and plausible new claims survived review roughly half the time (54.5% removal). Neither mitigation degraded the deliverable: claim retention exceeded 99% under both. The interventions are complementary: the guardrail eliminates unprompted additions and qualifier inflation cheaply, while the checkpoint gives near-categorical guarantees against the most severe failures, invented identities and JD-baited claims. These results support a layered architecture combining guardrails with a human checkpoint. A supplementary run with a newer-generation model (90.0% baseline fabrication rate) suggests the problem is not resolved by model progress alone. --- ### From Plots to Words: Model-Aware Multimodal Explanations as a Foundation for Accessible, Non-Visual Interaction Authors: Nur Keleşoğlu, Łukasz Sobczak, Joanna Domańska Published: 2026-07-22 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2608.24910v1 Multimodal large language models are increasingly used in interactive systems, yet ensuring consistent, trustworthy reasoning across heterogeneous modalities remains challenging. We present a context-aware, multi-agent framework that integrates textual queries, numerical data, visual representations, and model-derived signals for explainable time-series forecasting. A distinctive feature is that it turns predominantly visual forecasting outputs (e.g., trend plots) into structured, model-aware textual explanations. We argue that this makes the approach a natural foundation for non-visual, accessible interaction of particular relevance to blind and visually impaired users, for whom plot-centric interfaces are largely inaccessible. The framework supports three progressively richer pipelines (baseline, interpretable, explainable), enabling systematic comparison of unimodal, perception-driven, and model-aware responses. In an exploratory evaluation using an LLM-based judge as an early-stage proxy for human assessment, the explainable configuration improves overall explanation quality by up to 32% over a numerical baseline, with notable gains in trustworthiness and model awareness. We position user-centered validation with target users, including screen-reader and speech-interface users, as the essential next step rather than a claim established here. --- ### Super Star: Towards Streaming Real-time Interactive Agents for Digital Humans Authors: Wentao Jiang, Youchen Xie, Haidi Fan, Yajing Chen, Xin Wang, Ye Shi, Jingya Wang Published: 2026-07-22 Categories: cs.HC, cs.CV, cs.SD arXiv: https://arxiv.org/abs/2608.24909v1 Existing co-speech gesture generation methods are predominantly studied in offline settings, where gestures are synthesized from complete speech segments. However, interactive digital humans in real-world scenarios are required to generate speech-synchronous gestures online, using only currently available response audio under strict latency constraints. As a result, prior methods are unsuitable for real-time interaction, as they either rely on future speech information or incur substantial inference delay. In this paper, we formulate online co-speech gesture generation for interactive digital humans and propose a real-time interactive framework that couples a streaming speech response module with an online gesture generation module. Specifically, the gesture generator is designed as a causal multimodal autoregressive model that predicts body motion from streaming response speech and motion history, enabling low-latency and speech-aligned gesture synthesis without access to future speech. To support this setting, we further propose an offline data synthesis pipeline tailored to virtual companion scenarios, which leverages topic- and emotion-aware subject corpora to construct diverse human-agent dialogues and then generates co-speech gestures conditioned on the agent responses. Moreover, to bridge the gap between offline data construction and online deployment, we establish a self-evolving training loop by incorporating user feedback collected during online interaction into the data generation process, enabling continual adaptation to user preferences. Extensive experiments demonstrate that our framework achieves superior better latency-quality trade-off, stronger speech-motion synchronization, and higher user preference than competitive existing baselines. Project Page: https://super-star-2026.github.io/ --- ### Can Conversational AI loosen Us-Versus-Them Boundaries? The Effects of Common, Dual, and Separate Identity Framings on Pro-Immigrant Intergroup Helping Authors: Oluwadamilola Jeboda, John F. Dovidio, Jonas R. Kunst Published: 2026-07-22 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2608.19220v1 Rising immigration has intensified intergroup tensions in many countries. Traditional bias-reduction programs remain difficult to scale and increasingly constrained by U.S. policy. This preregistered experiment tested whether conversational AI can shift how majority-group members categorize and relate to Latine immigrants. Drawing on the common ingroup identity model, a quota-representative national sample of 658 non-Latine White U.S. adults completed five rounds of dialogue with a LLM (GPT-4o). The model was instructed to frame Latine immigrants in terms of a common ingroup identity (a shared American identity), a dual identity (both Latine and American), or a separate identity (distinct cultural boundaries), or to discuss an unrelated topic in a control condition. The manipulations altered categorization: relative to control, common ingroup identity and dual identity conversations lowered separate categorization, and dual identity conversations raised dual categorization. Although direct effects on behavior and pro-diversity beliefs were nonsignificant, willingness to act was significantly higher in the conditions emphasizing a superordinate identity (common ingroup and dual identity). A path model further revealed indirect associations: both conditions reduced separate categorization, which in turn correlated with greater willingness to act. Semantic similarity analyses of the transcripts confirmed that conversations tracked their assigned narratives; participants' convergence with shared-identity language related positively, and with separate-identity language negatively, to willingness to act. These effects were largely consistent across moderators (need for closure, openness to experience, and political orientation). The findings show that brief AI conversations can loosen us-versus-them boundaries while underscoring the gap between cognitive recategorization and behavior. --- ### Are you Talking Logic to Me? Assessing Language Models Syllogistic Reasoning Capabilities Authors: Hanna Abi Akl, Fabien Gandon, Catherine Faron, Pierre Monnin Published: 2026-07-22 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2608.12374v1 Language models (LMs) struggle with logical tasks like reasoning on syllogisms. It has been shown that Knowledge Representation (KR) plays a crucial role in expressing input information to help models solve tasks. This observation motivates our study of the impact of different formal KR notations on syllogistic reasoning by extending the FOLIO and P-FOLIO datasets. Our experiments on Small Language Models (SLMs) in Supervised Fine-Tuning (SFT) and Zero-Shot (ZS) settings show that the choice of input notation can yield performances competitive with natural language while enabling faster inference. We also propose a syllogistic categorization method (SEF) and use it to enrich ZS prompts with logical definitions, which boost reasoning in small models. We open-source our framework, Common Logic Grammar Construction (CLGC), as the first Python library for automatically generating syllogisms in KR notations and defining their SEF categories. --- ### A Conceptual Framework for Refining Influence Knowledge from Simulation Evidence in Cyber-Physical Systems Authors: Barbara da Silva Oliveira, Julien Deantoni, Nicolas Ferry Published: 2026-07-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.11221v1 Cyber-physical systems (CPS) are typically developed by multiple stakeholders who produce artefacts tailored to their specific domains of expertise. The behaviour of these systems emerges from the interaction between those artefacts and their operational environment. Simulation and co-simulation have become essential approaches for analysing CPS behaviour and, through simulation campaigns, developers can explore system responses under changing conditions, including interactions with the environment. However, the lack of details and understanding of some environmentmediated interactions (typically the ones beyond direct sensing and actuation), which remain unmodelled due to their complexity, a lack of time, or a lack of domain experience, hinders the proper comprehension and exploitation of simulation results. To address these limitations, we propose a conceptual framework leveraging the novel concept of Influences to support the iterative and incremental refinement of simulation campaigns and deepen the understanding of the system behaviour. We demonstrate the proposed approach through a case study involving a mobile robot implemented using Simulink/Gazebo co-simulation. --- ### Google's AI & Economy ATLAS v1.0: Mapping Gemini Usage in the Economy Authors: Zanna Iscenko, Scott Strand, Yiyuan Chen, Guillaume Aimard, Mihai Codreanu, Vivek Sampathkumar, Alex Imas, Julian Jacobs, Evalyne Muiruri, Juan Mateos-Garcia, Jia Jen Ng, Samirah Javed et al. Published: 2026-07-22 Categories: econ.GN, cs.AI arXiv: https://arxiv.org/abs/2608.00038v1 This paper introduces the AI & Economy ATLAS (Activity, Task, Landscape, and Adoption Study), an ongoing economic research initiative using Google AI usage data. The first iteration of ATLAS is built on 15 million de-identified interactions across the Gemini App, Google AI Mode, and Gemini API. Using privacy-preserving algorithms as well as established and bespoke classification methods, we map AI usage to over 800 occupations, 4000 tasks, 300 household activities, 150 countries, and 140 languages. We then make a number of observations on what the data reveals about AI's diffusion, and its usage at work and in day-to-day life. In the workplace, we show that while AI adoption spans occupations covering just above 88% of US employment, penetration remains shallow and overwhelmingly collaborative in nature, with end-to-end task automation limited in scope. Outside of work, AI spans activities making up about 98% of Americans' non-sleep time, with disproportionately high use in high-friction tasks such as engaging with government and professional service providers, likely delivering economic value that standard national accounts may miss. Globally, adoption scales with national wealth and has broad linguistic distribution, with English queries representing only around a third of volume. As we build upon ATLAS and expand its scope and capabilities, we will continue to provide large-scale empirical evidence to inform the public, policy and academic questions about the ongoing AI transformation. --- ### An Ontology-Guided, Deduplication-Aware Extraction Layer for Knowledge Graph Construction from Heterogeneous Documents Authors: Vaibhav Dangaich, Kevin Lewis, Kundeshwar Pundalik Published: 2026-07-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2607.28662v1 Large language models extract entities and relationships from unstructured documents fluently but inconsistently: type vocabularies fracture across documents, the same person surfaces under several name variants, relationships duplicate, and distinct individuals who share a name risk silent conflation. This paper presents the design, implementation, and empirical refinement of a production extraction layer that converts a live document stream into a validated knowledge graph aligned to a formal ontology. The system consumes document metadata from Kafka, routes PDF, spreadsheet, Office, and image content through handlers built for each format, and extracts entities and relationships in two passes using a locally hosted Qwen3.5-9B model tuned on the ontology. Its distinguishing component is ontology-guided extraction: the relevant slice of a curated ontology is retrieved live from a graph database by embedding similarity and injected into the extraction prompt, reducing catalog overhead by about 94 percent relative to static domain slices. Extracted results then pass through a refinement pipeline of five stages: deterministic cleaning, merging across chunks, a second pass for relationships, six deduplication algorithms that require no model inference, and an embedding resolution subsystem whose conflict guard no similarity score can override. Evaluation on intelligence corpora improved search recall from roughly 70 to 95 percent with no false merges, and corrected seven classes of silent quality defect, ranging from a bug that truncated source text by a single character to the systematic duplication of entities that carried title prefixes. --- ### Are the Financial Reasoning from LLMs Credible? A Real World Test over Long-Horizon Statements Authors: Xinke Tong, Xuanming Zhang, Tianyi Tang, An Yang, Jiatu Hu, Guojie Lin, Zhenzhen Shi, Lingfeng Zeng, Boyu Yang, Bing Zhao, Hu Wei, Lin Qu et al. Published: 2026-07-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2607.28661v1 Do Large Language Models (LLMs) possess genuine structural reasoning, or merely rely on surface-level pattern matching? The financial domain, demanding numerical precision and multi-step logic over long contexts, is an ideal testbed. Existing benchmarks fail to capture real-world industrial complexity, predominantly relying on multiple-choice questions or single-hop QA over cropped tables while ignoring intricate cross-statement dynamics and temporal de-cumulation. To bridge this gap, we introduce FinIndices, a large-scale benchmark evaluating data-processing fidelity over uncropped financial statements (up to 32K tokens). Utilizing an automated synthesis pipeline with adversarial traps, FinIndices encompasses Single-Index computation and Table-Index tabulation to test complex domain, temporal, and caliber reasoning. Our evaluation reveals two severe LLM vulnerabilities. First, a "Knowledge Bottleneck": despite memorizing formulas during pre-training, models demonstrate fragile pattern matching. Removing explicit formula hints causes performance to collapse (e.g., Gemini-3.1-Pro drops from 70.70% to 38.22% on table tasks), exposing fatal flaws in temporal de-cumulation and stock-flow caliber mismatch. Second, a "Structural Bottleneck": the intense cognitive load of generating multi-metric, multi-period tables actively drains reasoning capacity. Under structural pressure, LLMs that flawlessly execute isolated derivations regress to shallow heuristics, such as fetching incorrect adjacent columns or substituting deep accounting adjustments with lazy literal arithmetic. Finally, Supervised Fine-Tuning (SFT) yields substantial zero-hint gains (+8.54% Single, +3.82% Table), validating that structured logic can be partially restored via data-centric alignment. --- ### Sympathetic Framing: Evaluating AI Alignment across Sociodemographic Groups Authors: Haran Shani-Narkiss, Michael Fire, Oren Tsur Published: 2026-07-22 Categories: cs.CL, cs.AI, cs.CY, cs.LG arXiv: https://arxiv.org/abs/2607.27232v2 Large Language Models (LLMs) are increasingly shaping how we consume information and form our worldview. This raises concerns beyond bias in AI: do LLMs grasp the emotional nuances conveyed via textual framing? In this work, we empirically evaluate how well an array of LLMs aligns with human emotional perception. Considering news headlines covering political and geopolitical conflicts, both human participants (n = 3011, a representative sample of the U.K. adult population, via a YouGov survey) and seven LLMs answered whether headlines evoked sympathy for a specified side in a conflict. We find that the correlation between AI and human evaluations varies across models, ranging from very high (0.789, GPT-5.2) to medium (0.4 ,Mistral Large 2512). Crucially, the leading models are broadly aligned with human judgments across all demographic subgroups, including age, gender, level of education, prior geopolitical knowledge, and participants' predispositions regarding the conflict, although there are statistically significant differences between groups. This research, with its robust design and large, demographically diverse dataset, offers the most comprehensive evaluation of LLMs' comprehension of news framing to date. Findings highlight an important, often-ignored aspect of differential alignment: even when aggregate performance is high, AI alignment is not universal -- it may correspond differently with demographic features and cultural norms. Considering or ignoring the need for differential alignment may therefore have significant implications for the development of ethical and useful AI systems. --- ### Sympathetic Framing: Evaluating AI Alignment across Sociodemographic Groups Authors: Haran Shani-Narkiss, Michael Fire, Oren Tsur Published: 2026-07-22 Categories: cs.CL, cs.AI, cs.CY, cs.LG arXiv: https://arxiv.org/abs/2607.27232v1 Large Language Models (LLMs) are increasingly shaping how we consume information and form our worldview. This raises concerns beyond bias in AI: do LLMs grasp the emotional nuances conveyed via textual framing? In this work, we empirically evaluate how well an array of LLMs aligns with human emotional perception. Considering news headlines covering political and geopolitical conflicts, both human participants (n = 3011, a representative sample of the U.K. adult population, via a YouGov survey) and seven LLMs answered whether headlines evoked sympathy for a specified side in a conflict. We find that the correlation between AI and human evaluations varies across models, ranging from very high (0.789, GPT-5.2) to medium (0.4 ,Mistral Large 2512). Crucially, the leading models are broadly aligned with human judgments across all demographic subgroups, including age, gender, level of education, prior geopolitical knowledge, and participants' predispositions regarding the conflict, although there are statistically significant differences between groups. This research, with its robust design and large, demographically diverse dataset, offers the most comprehensive evaluation of LLMs' comprehension of news framing to date. Findings highlight an important, often-ignored aspect of differential alignment: even when aggregate performance is high, AI alignment is not universal -- it may correspond differently with demographic features and cultural norms. Considering or ignoring the need for differential alignment may therefore have significant implications for the development of ethical and useful AI systems. --- ### KernelGenBench: Can LLMs and Agents Write Efficient Kernels Across Operator Sources and Hardware Platforms? Authors: Peiyu Zang, Jian Tao, Jialing Zhang, Yichen Yuan, Wentao Zhang, Guang Liu, Yonghua Lin Published: 2026-07-22 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2607.27231v3 Modern AI systems depend on specialized accelerator kernels, whose development is complicated by increasingly diverse operators and hardware. LLMs and agentic systems promise to automate this work, but existing evaluations do not show whether their performance transfers across operator sources and hardware platforms, or what such transfer costs. We present KernelGenBench, the first unified multi-source and multi-chip infrastructure for evaluating LLM- and agent-generated Triton kernels. With a common Triton target spanning six hardware platforms, it provides the broadest cross-vendor hardware coverage among existing kernel-generation benchmarks. We report two controlled analytical views: KernelGenBench-MS (Multi-Source) covers 210 operators from PyTorch ATen, production vLLM operators, and proprietary cuBLAS routines, while KernelGenBench-MC (Multi-Chip) evaluates a semantically stable 110-operator subset across six hardware platforms. Our evaluation consumed over 15 billion tokens. Agentic execution improved correctness, but no method dominated across sources and platforms: vLLM posed the strongest correctness challenge, cuBLAS set the highest performance ceiling, and AutoKernel accuracy fell from 87% on NVIDIA to 25% on Iluvatar CoreX. These improvements were costly: specialized agents averaged 4.99 million tokens per successful operator, rising to 6.25 million for CUDA Optimized Skill. The results establish operator source, hardware platform, and agentic scaffold as distinct dimensions of kernel-generation capability, and show that success in a familiar source-hardware setting is not a reliable proxy for deployment readiness. --- ### KernelGenBench: A Multi-Source and Multi-Chip Benchmark for LLM-based Kernel Generation Authors: Peiyu Zang, Jian Tao, Jialing Zhang, Yichen Yuan, Wentao Zhang, Guang Liu, Yonghua Lin Published: 2026-07-22 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2607.27231v1 Large language models (LLMs) have significantly increased the demand for efficient accelerator kernels, but kernel development remains a highly specialized and labor-intensive task. The recent rise of LLMs and agentic frameworks offers a promising pathway toward automatic kernel generation. However, despite rapid progress, there is still no comprehensive benchmark to rigorously evaluate LLM-generated kernels across diverse operator sources or heterogeneous hardware platforms. We present KernelGenBench, a unified benchmark for systematically evaluating LLM- and agent-generated Triton kernels across diverse operator sources and heterogeneous hardware platforms. It comprises two complementary sub-benchmarks: KernelGenBench-MS (Multi-Source), evaluating 210 operators from three sources beyond standard PyTorch-centric tasks, and KernelGenBench-MC (Multi-Chip), measuring performance portability across six heterogeneous hardware platforms using a 110-operator subset. Our large-scale evaluation, consuming over 15 billion tokens, shows: (1) agent-based methods consistently outperform pure LLM sampling methods, while cuBLAS operators are the most challenging across all methods; (2) generation performance varies significantly across hardware platforms, with even recent kernel-specialized agents experiencing severe cross-platform degradation (e.g., AutoKernel drops from 87% on NVIDIA to 25% on Platform E); (3) autonomous kernel generation remains highly cost-intensive, with specialized agent methods averaging 5.11 million tokens per successful operator (AKO4all reaches 5.19 million), orders of magnitude higher than simple LLM sampling approaches. --- ### Multi-Head Attention Residuals Authors: Cheng Luo, Zefan Cai, Junjie Hu Published: 2026-07-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2607.27230v2 Transformers propagate information across depth through a single additive residual stream: every sublayer reads only the most recent state. Attention residuals relax this by letting each sublayer attend, through a learned softmax. However, that read uses a single query shared across the entire width, so every feature subspace must read the depth history through one distribution. The cost of this forced compromise grows with how much the subspaces disagree about which layers to read, and disagreement grows with model width. We introduce Multi-Head Attention Residuals (MHAR): the routing query is reshaped into H per-subspace heads, each with its own softmax over the depth history. The read becomes block-diagonal, the reshape adds zero parameters and negligible compute, and H = 1 recovers attention residuals exactly. Trained from scratch on a deduplicated Nemotron-based anneal corpus that is quality-filtered and STEM- and code-heavy, MHAR improves validation loss over a standard Transformer at 100M, 350M, and 1B (-0.061, -0.149, and -0.140). It achieves the best result among four methods in every setting, with the gain increasing from 100M to the larger scales. The head count is a real design axis rather than a free knob: validation loss is U-shaped with respect to H, with a flat optimum at H = 4 or H = 8 across scales. We adopt H = 8 for large-scale models; over-splitting beyond this point (H = 16) consistently gives back part of the gain. A direct probe of the trained queries confirms that learned subspace disagreement is the underlying driver. Fused Triton routing kernels increase attention-residual training throughput from 0.2-0.5x to 0.55-0.88x of the baseline while maintaining near-baseline peak memory. An identity-preserving conversion using delta attention residuals supports 8B mid-training, yielding improvements of +3.2 on GSM8K and +3.1 on GPQA. --- ### A Diagnostic Gap Framework for Evaluating Reconstruction Fidelity in Weakly Supervised Mammography Authors: Vinceline Bertrand, Ionut Cardei Published: 2026-07-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.22740v1 Weakly supervised pipelines for medical imaging have become increasingly popular over the years. These systems often include multiple stages and components, such as reconstruction, generation, and localization, yet standard evaluation metrics provide limited insight into whether clinically relevant information is preserved across each stage. We present the diagnostic gap framework, a practical evaluation tool that measures decision preservation and explanation preservation as a function of measured reconstruction fidelity. To isolate the effect of reconstruction from localization, we evaluate on curated lesion ROI crops using a fidelity ladder of three class-conditional reconstructors---VQ-VAE-GAN, VAE-GAN, and diffusion (SDEdit)---spanning a twenty-fold range in perceptual distance (LPIPS 0.029--0.584). At autoencoder fidelity, both decision and explanation are preserved: AUC changes remain within $\pm$0.005 and attribution similarity (HiResCAM, Grad-CAM++) stays high. At diffusion fidelity, both collapse: pooled AUC drops by 0.253 and mass-pathology AUC falls below chance. The diagnostic gap is thus a measurable function of reconstruction fidelity rather than an intrinsic cost of reconstruction, and the framework provides an architecture-agnostic instrument for identifying when and where multi-stage pipelines lose diagnostic signal. --- ### Cortex: Compact Behavior Cloning for Quake with Frozen Visual Features Authors: Dzmitry Malyshau Published: 2026-07-22 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2607.22739v2 We study how far a deliberately simple behavioral-cloning policy can progress in a visually rich first-person game before adding reinforcement learning or explicit memory. Cortex is a compact Quake policy with 10.98 million trainable parameters in a six-layer transformer over a frozen DINOv3 encoder. It is trained on the Quake subset of the public Pixels2Play corpus: 6,849 recordings (about 474.7 hours), represented as 17.09 million cached decision frames with keyboard and mouse actions. One sampled training epoch uses 517,048 four-frame windows and takes 3.3 minutes of policy-head optimization on one RTX 5080, excluding one-time feature extraction. We evaluate two independent batches of 20 stochastic, 120-second episodes on Quake E1M1. Cortex does not complete the level, but every episode reaches the opening door, button room, and gate descent; 19 of 20 episodes in each batch record at least one kill. Under the same time-controlled harness, released P2P-150M and NitroGen checkpoints remain shallower in five matched-duration episodes each. These comparisons are limited by small reference samples and different native interfaces. Ablations show that denser visual tokens improve combat and survival, while longer optimization and naive action history improve offline metrics without consistently improving play. The remaining failures are consistent with covariate shift and motivate targeted corrective data. We release the policy implementation, checkpoint, and a representative rollout. --- ### Nova3D: Code-Native Generation of Programmable 3D Assets Authors: Nimra Noor, Muhammad Bilal, Abdullah Hussain, Hassan Baig Published: 2026-07-22 Categories: cs.GR, cs.CV arXiv: https://arxiv.org/abs/2607.22738v1 Current 3D generative models mostly produce a final surface: a visually strong but largely opaque mesh. Interactive 3D worlds need more than a surface. They need named parts, an assembly hierarchy, measurable constraints, local edit handles, and joints for articulation. We present Nova3D, a system that generates 3D assets as executable Blender source code; the compiled mesh, a binary glTF (GLB), is treated as the artifact, not the asset. Because the output is a program, semantic handles exist at generation time rather than being recovered afterward by segmentation or rigging. We evaluate on Nova3D-Bench, a frozen, spec-grounded benchmark of 54 items across six domains and three difficulty levels with text and image inputs, against eleven baselines in four families (mesh-native, part-structured, code-native, and CAD) plus a same-LLM ablation. Nova3D produces an executable program and a valid artifact for 54/54 items. Every asset exposes named parts organized in a parent-child assembly tree; no mesh-native, CAD, or segmentation baseline exposes either. It satisfies 51/52 prompt-stated numeric and count constraints (best baseline: 11/52), passes 14/18 blinded local edits with locality preserved in 18/18, and articulates 59 joints across 12 assets at 98.3% geometric validity, where every baseline exposes zero native joints. Its geometry is competitive: it wins the structured domains in a pairwise shape-quality tournament and is second only to the strongest mesh-native model, while conceding texture realism to baked-PBR systems. The central result is representational: code-native generation turns a generated 3D object from an opaque surface into a programmable asset that downstream systems can inspect, measure, edit, and animate. --- ### Benchmarking the Domain Gap: Model Selection Instability Under Domain Shift in Video Capsule Endoscopy Authors: Dan Hanson, Debesh Jha Published: 2026-07-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.22736v1 Video capsule endoscopy (VCE) classification is typically evaluated within a single dataset, yet clinical deployment demands robustness across acquisition sources, labeling policies, and patient populations. We examine this gap using Kvasir-Capsule, Capsule Vision 2024 (CV2024), and a shared-label subset of Galar. We fine-tune a suite of general-domain pretrained backbones on the official Kvasir-Capsule folds under a standardized protocol and evaluate the same checkpoints on two non-source targets within a documented shared-label decision space. We find that the predictive value of in-domain ranking is target-dependent: Kvasir-Capsule ranking aligns more closely with Galar than with CV2024, while the two non-source targets agree only weakly. Consequently, the strongest in-domain backbone leads on one target yet falls to mid-pack on the other, and no single evaluation target reliably predicts the others. A second CV2024-trained configuration set reproduces this target-dependent instability. We conclude that capsule endoscopy model selection should report cross-target ranking stability rather than peak single-dataset performance. --- ### Fast Fourier Convolutional GAN for 30 m Clear-Sky Land Surface Temperature Gap-Free Reconstruction Authors: Marwa Alfouly, Smajil Halilovic, Nils Bochow, Thomas Hamacher, Niklas Boers, Konrad Schindler Published: 2026-07-22 Categories: cs.CV, eess.IV, physics.geo-ph arXiv: https://arxiv.org/abs/2607.22734v1 Satellite-derived Land Surface Temperature (LST) provides spatially comprehensive data that ground stations cannot match. However, its utility is frequently limited by severe data gaps due to the presence of clouds. As LST is essential for understanding land-atmosphere interactions, numerous methods have been proposed to address this challenge. Yet, the development of a scalable and adaptable pipeline for generating gap-free LST datasets and reconstructing cloud-contaminated pixels remains challenging. Moreover, the reconstruction of extensive missing regions in fine-spatial-resolution observations is particularly difficult. To address this challenge, we propose a Multimodal Fast Fourier Convolutional GAN for reconstructing cloud-contaminated pixels in fine-resolution (30 m) Landsat imagery to generate gap-free clear-sky LST products. The method leverages Fast Fourier Convolution to enable a global receptive field across the image, and is guided by a stack of data consisting of satellite observations and Synthetic Aperture Radar (SAR) data. Across all LST quantiles, the interquartile range of scene-averaged RMSE (computed over reconstructed pixels) is consistently between 0.8 K and 1.8 K. The proposed approach enables the recovery of extensive missing regions, including scenes with more than 70% cloud-induced gaps, while relying on auxiliary data that are readily available at a near-global scale. --- ### Generative Augmentation for EEG Motor Imagery Classification: A Class-Conditional VAE with Cycle-Consistent Decoder Refinement Authors: Matei Moldoveanu, Alain Sirois, Claire Ben Ali, Fabien Lotte, Florian Yger Published: 2026-07-22 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2607.22733v1 We investigate whether a generative model can supply useful synthetic motor-imagery (MI) electroencephalography (EEG) trials that improve the accuracy of independent downstream classifiers. We train a class-conditional variational autoencoder (CVAE) with an integrated latent classifier on the Zhou motor-imagery dataset, using the learned per-class prior as a generator: sampling the prior for a given label and decoding it into a synthetic, label-consistent signal. A constraint on the covariance matrix of the generated data encourages preservation of covariance structure, and the model is trained with a schedule that alternates ordinary VAE training with a decoder-focused phase that sharpens the generative pathway used for augmentation. We measure the effect of adding synthetic trials to the training set under two evaluation protocols -- within-user (pooled 60/20/20 split across subjects) and cross-user (leave-one-subject-out, LOSO) -- across four representative EEG classification pipelines: Common Spatial Patterns with Linear Discriminant Analysis (CSP+LDA), tangent-space features with a Support Vector Machine (TGSP+SVM), Minimum Distance to Riemannian Mean (MDM), and a neural network based on EEGNetv4 (henceforth EEGNet). Results are aggregated across independent augmentation draws, random seeds (within-user), or leave-one-subject-out folds (cross-user), with uncertainty reported as 95\% confidence intervals (Student's $t$-distribution) computed over per-seed/per-fold averages. We find that synthetic EEG from the CVAE is most credible as a source of class-structured, covariance-like data rather than as a substitute for real raw EEG: it can raise the point estimate for MDM, but the broader augmentation claim remains conservative -- observed gains are small and classifier-dependent. --- ### Spatial Reasoning in LLM Game Agents: Impact of Causal Context and Multi-Step Planning Authors: Mohit Jiwatode, Ronja Fuchs, Robin Schmöcker, Bodo Rosenhahn, Alexander Dockhorn Published: 2026-07-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2607.22732v1 LLM-based game agents often perform poorly on more complex tasks. This work examines whether these failures are linked to limited spatial reasoning and evaluates whether causal prompt augmentation and multi-step planning can improve win-rates while managing response latency. Using the open-source Qwen3 model family, we conduct experiments across varying model scales, reasoning modes, and planning horizons. We further introduce a focused GVGAI benchmark consisting of three custom games with five difficulty levels to isolate spatial navigation. The evaluation follows two paradigms: an initial ``positioning experiment'' to test an agent's ability to find its exact coordinates, and a study of game-play success. Our results show that while larger models with an enabled thinking mode identify their positions more accurately, overall performance in coordinate matching remains limited for smaller models. Win rates decrease as game levels and layout complexity increase, validating the benchmark's difficulty scaling. Integrating causal context into the prompts tends to improve the agents' success rates, particularly for bigger models. While enabling thinking mode and longer planning horizons significantly improve performance, multi-step planning further reduces mean per-step response times, offering a practical trade-off between reasoning depth and execution speed. --- ### Calibration-Free 3D Multi-Camera People Tracking for Indoor Environment Authors: Ponleur Veng, Dominique Vaufreydaz, Phutphalla Kong Published: 2026-07-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.22731v1 Multi-Camera People Tracking (MCPT) traditionally relies on precise intrinsic and extrinsic camera calibration to project 2D detections into a unified 3D world coordinate system.However, manual calibration constitutes a major bottleneck in large-scale dataset generation from unconstrained video archives. This work proposes a unified calibration-free 3D MCPT framework that infers geometric structure directly from visual data using deep foundation models. The system integrates anchor-free detection (YOLOX), robust tracking (BoT-SORT), omni-scale appearance embedding (OsNet), pose estimation (HRNet via MMPose), and transformer-based geometric reconstruction using the Visual Geometry Grounded Transformer (VGGT). A pose-guided 3D lifting strategy projects head keypoints onto a reconstructed manifold, eliminating dependence on ground-plane homography. Global identity association is formulated as hierarchical agglomerative clustering under a joint appearance-geometry cost with strict velocity gating. Evaluation on the AI City Challenge 2024 demonstrates a HOTA score of 53.13% without access to ground-truth calibration matrices, establishing a strong baseline for purely vision-based 3D tracking. --- ### Open Your Model's Eyes: Video and Context-Aware Multimodal Backchannel Prediction Authors: Min-Jae Kim, Jun-Yeong Moon, Mujeen Sung, Gyeong-Moon Park Published: 2026-07-22 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2607.22729v1 Backchannels, which signal listener states like empathy and understanding, are fundamental to natural human interaction. However, current approaches rely solely on audio and text. This omits crucial visual cues, such as facial expressions and gestures, as well as broader conversational contexts, which are necessary for accurate prediction. In this paper, we introduce Context-Aware Multimodal Alignment for Backchannel Prediction (CAMA-BC), a novel framework that leverages visual information through Multi-Layer Multimodal Alignment (MMA). Our alignment process comprises two stages. First, Context Alignment (MMA-CA) utilizes unlabeled dialogues with videos to capture conversational contexts. Next, Backchannel Alignment (MMA-BA) fine-tunes the representations specifically for backchannel prediction. Experimental results show that CAMA-BC significantly outperforms both existing methods and simple multimodal baselines, with particular effectiveness in recognizing complex backchannels such as empathy. --- ### CrossSpine: Multi-scale Cross-sequence Attention with Anatomical Priors for Automated Pfirrmann Grading Authors: Hai Son Nguyen, Duong Ngoc Vu, Trong-Nghia Nguyen, Bien Tran Van, Van-Dem Pham, Trang Mai Xuan, Huan Vu, Thien Van Luong Published: 2026-07-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.22728v1 Automated grading of Lumbar Disc Degeneration is essential for the objective quantification of structural changes associated with low back pain. Observing that baseline models underperformed on our data, we propose a framework designed to overcome these limitations. First, we present the Cross-sequence Attention Spine (CrossSpine) framework, a novel architecture that employs a cross-sequence attention mechanism to adaptively fuse features from different MRI sequences at multiple spa- tial scales. Second, we contribute a meticulously curated dataset aimed at automated Pfirrmann grading. Finally, we introduce an IVD-aware classification technique that integrates anatomical disc-level information, enabling the model to learn level-specific degeneration priors. Our experi- ments demonstrate the superiority of this approach: CrossSpine achieved a relative improvement exceeding 125% in the Macro F1 score, while boosting the Mean AUPRC by 99% and the Mean AUROC by 36% com- pared to the baseline. --- ### Trustworthy Medical Segmentation: Uncertainty-Aware U-Net Evaluation Under Clinical Image Degradation Authors: Pranav Kaliaperumal, Manisha Kaliaperumal Published: 2026-07-22 Categories: cs.CV, cs.LG arXiv: https://arxiv.org/abs/2607.22727v1 Medical image segmentation models often report high benchmark accuracy under ideal imaging conditions, yet their failures under clinical degradation can be quiet: sensor noise, patient motion, low- resolution acquisition, and contrast variability may all alter model behavior without producing an obvious warning. We present a reproducible framework for evaluating uncertainty-aware segmentation under con- trolled clinical degradation. Our experiments use a synthetic multimodal brain tumor MRI cohort generated with a biophysical phantom simulator that follows the BraTS protocol. We train U-Net and Attention U-Net baselines for multi-class tumor sub-region segmentation and augment both models with Monte Carlo dropout to estimate per-voxel uncertainty. Across eight clinically motivated corruption types at five severity levels, we measure segmentation accuracy, calibration, failure detection, and selective prediction coverage. On clean data, Attention U-Net achieves a whole-tumor Dice of 0.990; under severe Gaussian noise, its performance falls to 0.089. Predictive uncertainty rises with degradation and tracks segmentation error (Pearson r = 0.53 under severity-3 Gaussian noise), allowing us to flag failures with an AUROC of 0.843. These results argue for uncertainty-aware inference as a practical safety layer in physician-in-the-loop radiology workflows. We release the code, trained models, and evaluation protocol to support direct reproduction. --- ### PCA: Persistence-Aware Compression and Aggregation for Fast Video Large Language Models Authors: Zihan Song, Shuo Ye, Bo Zhao, Ruixin Zhang, Jiayu Zhang, Shouhong Ding, Zitong Yu Published: 2026-07-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.22726v1 Despite advances in Video Large Language Models (VLLMs) that have displayed promising outcomes in video understanding, the redundancy in the long-duration frames remains a hindrance to efficient reasoning. This paper introduces a training-free $\mathbf{P}$ersistence-Aware $\mathbf{C}$ompression and $\mathbf{A}$ggregation (PCA) method designed to preserve high-fidelity raw visual information before the encoding stage. PCA can be built on arbitrary VLLMs and consists of two modules: 1) A Dynamic Downsampling (DD) module that adaptively removes redundant frames by analyzing frame-wise similarity. 2) A Persistence-Aware Motion Enhancement (PAME) module that enriches each selected keyframe by aggregating the temporal context of its neighbors, ensuring that essential information is preserved even after aggressive frame reduction. Our approach substantially reduces the computation of long-context modeling, while enhancing the performance of the baseline model. Extensive experiments demonstrate that PCA consistently outperforms existing state-of-the-art approaches in both efficiency and accuracy, achieving a speedup of 1.8$\times$ to 2.5$\times$ compared to the baseline VLLM. The code is open-sourced at https://github.com/Heisenberg10110/PCA. --- ### Structural Preservation Governs Data Augmentation in Deep Learning-Based Laser Speckle Material Classification Authors: Mohamed Abdallah Salem, Nourhan Zein Diab Published: 2026-07-22 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2607.22725v1 Data augmentation is routinely used to improve generalization in image classification, but the assumptions underlying standard policies are poorly matched to coherent imaging. Laser speckle patterns are not generic textures; they arise from coherent interference, and their discriminative content is carried by structured stochastic spatial and frequency statistics. This study examines how controlled augmentation perturbations influence speckle-based material classification on the SensiCut dataset. We train ResNet18 and EfficientNet-B0 under a parametric augmentation framework comprising rotation, Gaussian blur, independent Gaussian noise, spatially correlated speckle-aware noise, intensity jitter, and spatial masking, and evaluate test performance using macro F1-score averaged over three random seeds. Separate ordinary least squares models link augmentation parameters to performance for each architecture. Across both models, Gaussian blur exerts a strong negative effect (p < 0.001), indicating that low-pass filtering suppresses high-frequency structure that is informative for material discrimination. Independent pixel-wise noise is likewise harmful (p = 0.003 for EfficientNet-B0 and p = 0.001 for ResNet18), consistent with disruption of local spatial coherence. In contrast, spatially correlated perturbations yield significant positive coefficients (p = 0.004 for EfficientNet-B0 and p = 0.001 for ResNet18), showing that variability can improve robustness when it preserves speckle organization. The fitted models explain a substantial fraction of performance variation (R2 = 0.796 for EfficientNet-B0 and R2 = 0.879 for ResNet18). These results show that, in laser speckle imaging, augmentation effectiveness is determined primarily by structural preservation rather than perturbation magnitude. The findings motivate physics-aware augmentation design for coherent optical sensing. --- ### Progress-conditioned Group Policy Optimization for Long-Horizon Agentic Tasks Authors: Kaibing Yang, Guangfeng Cai, Shengtian Yang, Shuo He, Yu Li, Mengyi Liu, Pengwei Chen, Jun Xu, Lei Feng Published: 2026-07-22 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2607.22724v1 Group-based policy optimization has been increasingly used to train large language model (LLM) agents from sparse outcome rewards by comparing trajectories or steps within a group. However, on difficult long-horizon tasks, this comparison can suffer from a sampling imbalance: repeated or low-effect actions dominate the high-probability region of the policy while useful state-changing actions remain under-sampled. This imbalance produces many all-failed rollout groups, where outcome rewards provide no direction for correcting the policy. Together, these effects can form a self-reinforcing credit trap: failure-dominated sampling yields no outcome-based correction, allowing repeated low-effect actions to persist. To break this loop, we propose Progress-conditioned Group Policy Optimization (ProGPO), which uses first-visit observation coverage only when all samples in a group receive zero outcome reward. Specifically, within such groups, ProGPO assigns higher relative advantages to trajectories or steps that visit more new states since reaching new observations is a prerequisite for task success. Experiments on two challenging agentic benchmarks, ALFWorld and WebShop with Qwen2.5-1.5/7B-Instruct, show that ProGPO consistently improves over group-based baselines, with particularly large gains on hard tasks. ---