# Research Papers — 2026-04 · bundle page 94 of 233 > Full-text markdown of papers #4651–4700 of 11612 published in 2026-04, newest first. AI-readable mirrors of the full 11612-paper 2026-04 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-04/page/94 [← Previous bundle](https://aipod.dev/papers/archive/2026-04/page/93.md) [Next bundle →](https://aipod.dev/papers/archive/2026-04/page/95.md) ### Scalable Quantum Error Mitigation with Physically Informed Graph Neural Networks Authors: Huaxin Wang, Xinge Wu, Jiajun Liu, Ruiqing He, Jiandong Shang, Hengliang Guo, Qiang Chen Published: 2026-04-18 Categories: quant-ph, cs.LG arXiv: https://arxiv.org/abs/2604.16815v1 Quantum error mitigation (QEM) provides a practical route for estimating reliable observables on noisy intermediate-scale quantum (NISQ) devices. Traditional QEM strategies, including zero-noise extrapolation (ZNE) and Clifford data regression (CDR), rely on noise scaling or global regression, and their performance is constrained by the exponential growth of the system degrees of freedom. We construct a graph-enhanced mitigation (GEM) framework, which incorporates physical information into the model representation. In this work, quantum circuits are encoded as attributed graphs. Hardware-level physical information is mapped to node and edge features: local noise parameters such as calibration parameters $T_1$, $T_2$, and readout errors are encoded at nodes, while coupling-related information such as two-qubit gate errors is encoded as edge features. Graph neural networks are used to model how errors propagate along the physical coupling structure and build up into non-local correlations. This allows the model to capture local interactions and part of the resulting non-local correlations across qubits. A dual-branch affine correction is applied to maintain consistency with physical constraints. Experiments on 10-qubit and 16-qubit random circuits executed on superconducting quantum processors show that GEM provides a level of accuracy comparable to CDR at small scales, while yielding lower mean absolute error and improved stability in zero-shot transfer to larger systems. Results of the traditional QEM strategy indicate that global regression methods remain effective in low-dimensional settings but become less reliable as system degrees of freedom grow. In contrast, GEM makes use of local physical structures to show better scalability and generalization, while preserving the overall error propagation patterns. This work provides a practical scalable approach to QEM for NISQ devices. --- ### PersonalHomeBench: Evaluating Agents in Personalized Smart Homes Authors: Manasa Bharadwaj, Yolanda Liu, InJung Yang, Sungil Kim, Nikhil Verma, KoKeun Kim, Kevin Ferreira, YoungJoon Kim Published: 2026-04-18 Categories: cs.AI, cs.CL, cs.DB arXiv: https://arxiv.org/abs/2604.16813v3 Agentic AI systems are rapidly advancing toward real-world applications, yet their readiness in complex and personalized environments remains insufficiently characterized. To address this gap, we introduce PersonalHomeBench, a benchmark for evaluating foundation models as agentic assistants in personalized smart home environments. The benchmark is constructed through an iterative process that progressively builds rich household states, which are then used to generate personalized, context-dependent tasks. To support realistic agent-environment interaction, we provide PersonalHomeTools, a comprehensive toolbox enabling household information retrieval, appliance control, and situational understanding. PersonalHomeBench evaluates both reactive and proactive agentic abilities under unimodal and multimodal observations. Thorough experimentation reveals a systematic performance reduction as task complexity increases, with pronounced failures in counterfactual reasoning and under partial observability, where effective tool-based information gathering is required. These results position PersonalHomeBench as a rigorous evaluation platform for analyzing the robustness and limitations of personalized agentic reasoning and planning. --- ### Introspection Adapters: Training LLMs to Report Their Learned Behaviors Authors: Keshav Shenoy, Li Yang, Abhay Sheshadri, Sören Mindermann, Jack Lindsey, Sam Marks, Rowan Wang Published: 2026-04-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.16812v2 When model developers or users fine-tune an LLM, this can induce behaviors that are unexpected, deliberately harmful, or hard to detect. It would be far easier to audit LLMs if they could simply describe their behaviors in natural language. Here, we study a scalable approach to rapidly identify learned behaviors of many LLMs derived from a shared base LLM. Given a model $M$, our method works by finetuning models $M_i$ from $M$ with implanted behaviors $b_i$; the $(M_i, b_i)$ pairs serve as labeled training data. We then train an introspection adapter (IA): a single LoRA adapter jointly trained across the finetunes $M_i$ to cause them to verbalize their implanted behaviors. We find that this IA induces self-description of learned behaviors even in finetunes of $M$ that were trained in very different ways from the $M_i$. For example, IAs generalize to AuditBench, achieving state-of-the-art at identifying explicitly hidden concerning behaviors. IAs can also be used to detect encrypted finetuning API attacks. They scale favorably with model size and training data diversity. Overall, our results suggest that IAs are a scalable, effective, and practically useful approach to auditing fine-tuned LLMs. --- ### A Mechanism Study of Delayed Loss Spikes in Batch-Normalized Linear Models Authors: Peifeng Gao, Wenyi Fang, Yang Zheng, Difan Zou Published: 2026-04-18 Categories: stat.ML, cs.LG, math.OC arXiv: https://arxiv.org/abs/2604.16809v1 Delayed loss spikes have been reported in neural-network training, but existing theory mainly explains earlier non-monotone behavior caused by overly large fixed learning rates. We study one stylized hypothesis: normalization can postpone instability by gradually increasing the effective learning rate during otherwise stable descent. To test this hypothesis at theorem level, we analyze batch-normalized linear models. Our flagship result concerns whitened square-loss linear regression, where we derive explicit no-rising-edge and delayed-onset conditions, bound the waiting time to directional onset, and show that the rising edge self-stabilizes within finitely many iterations. Combined with a square-loss decomposition, this yields a concrete delayed-spike mechanism in the whitened regime. For logistic regression, under highly restrictive active-margin assumptions, we prove only a supporting finite-horizon directional precursor in a knife-edge regime, with an optional appendix-only loss lower bound under an extra non-degeneracy condition. The paper should therefore be read as a stylized mechanism study rather than a general explanation of neural-network loss spikes. Within that scope, the results isolate one concrete delayed-instability pathway induced by batch normalization. --- ### BioLip: Language-Generalizable Lip-Sync Deepfake Detection via Biomechanical Constraint Violation Modeling Authors: Hao Chen, Junnan Xu Published: 2026-04-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.16808v4 Existing lip-sync deepfake detectors rely on pixel artifacts or audio-visual correspondence, and both fail under generator or language shift because the features they learn are tied to the training distribution. We take a different approach. Authentic lip motion is constrained by tissue mechanics and neuromuscular bandwidth; current generators typically do not impose these constraints, producing trajectories with elevated variance in velocity, acceleration, and jerk that real speech does not exhibit. We exploit this signal, which we term temporal lip jitter, by computing kinematic statistics from 64 perioral landmarks over short sliding windows and feeding them into a lightweight three-branch network. The model uses only landmark coordinates: no pixels, no audio, and no voiceprint data. We train only on English data and test in a zero-shot setting on five unseen generators and seven languages. --- ### Channel Attention-Guided Cross-Modal Knowledge Distillation for Referring Image Segmentation Authors: Chen Yang Published: 2026-04-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.16806v1 Referring image segmentation (RIS) requires accurate segmentation of target regions in images according to language descriptions, which is a cross-modal task integrating vision and language. Existing RIS methods typically employ large-scale vision and language encoding models to improve performance, but their enormous parameter size severely restricts deployment in scenarios with limited computing resources. To solve this problem, this paper proposes a channel attention-guided cross-modal knowledge distillation method, which transfers the high-order fine-grained correlations between vision and language learned by the teacher network, as well as the correlations between semantic components represented by each channel, to the student network. Compared with the traditional pixel-wise relational distillation, this method not only enables the student to learn the knowledge of the teacher, but also retains part of its independent learning ability, alleviating the transfer of learning bias. Experimental results on two public datasets show that the proposed distillation method does not introduce additional parameters during inference and can achieve significant performance improvement for the student model. --- ### AutoOR: Scalably Post-training LLMs to Autoformulate Operations Research Problems Authors: Sumeet Ramesh Motwani, Chuan Du, Aleksander Petrov, Christopher Davis, Philip Torr, Antonio Papania-Davis, Weishi Yan Published: 2026-04-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.16804v4 Optimization problems are central to decision-making in manufacturing, logistics, scheduling, and other industrial settings. Translating complicated descriptions of these problems into solver-ready formulations requires specialized operations research (OR) expertise, making it hard to scale. We present AutoOR, a scalable synthetic data generation and reinforcement learning pipeline that trains LLMs to autoformulate optimization problems specified in natural language across linear, mixed-integer, and non-linear categories. AutoOR generates verified training data from standard optimization forms and uses solver execution feedback as the reward signal for RL post-training. AutoOR applied to an 8B model achieves state-of-the-art or competitive results across six established OR benchmarks, matching significantly larger frontier models. For a non-linear problem class involving physical dynamics, where frontier models score near 0%, we introduce a curriculum RL strategy that bootstraps from limited initial training data to make this class tractable for post-training. We believe that methods such as AutoOR can significantly accelerate industrial decision-making with AI. --- ### AutoOR: Scalably Post-training LLMs to Autoformalize Operations Research Problems Authors: Sumeet Ramesh Motwani, Chuan Du, Aleksander Petrov, Christopher Davis, Philip Torr, Antonio Papania-Davis, Weishi Yan Published: 2026-04-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.16804v2 Optimization problems are central to decision-making in manufacturing, logistics, scheduling, and other industrial settings. Translating complicated descriptions of these problems into solver-ready formulations requires specialized operations research (OR) expertise, making it hard to scale. We present AutoOR, a scalable synthetic data generation and reinforcement learning pipeline that trains LLMs to autoformalize optimization problems specified in natural language across linear, mixed-integer, and non-linear categories. AutoOR generates verified training data from standard optimization forms and uses solver execution feedback as the reward signal for RL post-training. AutoOR applied to an 8B model achieves state-of-the-art or competitive results across six established OR benchmarks, matching significantly larger frontier models. For a non-linear problem class involving physical dynamics, where frontier models score near 0%, we introduce a curriculum RL strategy that bootstraps from limited initial training data to make this class tractable for post-training. We believe that methods such as AutoOR can significantly accelerate industrial decision-making with AI. --- ### Continuous Limits of Coupled Flows in Representation Learning Authors: Zilin Li, Weiwei Xu, Xuchun Tong, Xuanbo Lu, Xuanqi Zhao Published: 2026-04-18 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.16801v1 While modern representation learning relies heavily on global error signals, decentralized algorithms driven by local interactions offer a fundamental distributed alternative. However, the macroscopic convergence properties of these discrete dynamics on continuous data manifolds remain theoretically unresolved, notoriously suffering from parameter explosion. We bridge this gap by formalizing decentralized learning as a coupled slow-fast dynamical system on Riemannian manifolds. First, using measure-theoretic limits, we prove that the discrete spatial transitions converge uniformly to an overdamped Langevin stochastic differential equation. Second, via the Itô-Poisson resolvent and a stochastic extension of LaSalle's Invariance Principle, we establish that the representation weights unconditionally avoid divergence and align strictly with the principal eigenspace of the spatial measure. Finally, we construct a joint Lyapunov functional for the fully coupled spatial-parametric flow. This proves global dissipativity and demonstrates that orthogonally disentangled, linearly separable features emerge spontaneously at the stationary limit. Our framework bridges discrete algorithms with continuous stochastic analysis, providing a formal theoretical baseline for decentralized representation learning. --- ### Frequency-Decomposed INR for NIR-Assisted Low-Light RGB Image Denoising Authors: Ligen Shi, Zengyu Pang, Chang Liu, Shuchen Sun, Jun Qiu Published: 2026-04-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.16800v1 Addressing the issues of severe noise and high frequency structural degradation in visible images under low-light conditions, this paper proposes a Near Infrared (NIR) aided low light image restoration method based on Frequency Decoupled Implicit Neural Representation (FDINR). Based on the statistical prior of RGB-NIR cross-modal frequency correlations, specifically that low-frequency RGB signals are more reliable, whereas high frequency NIR signals exhibit higher correlation, we explicitly decompose images into distinct frequency components via multi-scale wavelet transforms and construct a dual-branch implicit neural representation framework. Within this framework, we design a cross modal differentiated frequency supervision mechanism, leveraging low light RGB to guide the reconstruction of low frequency luminance and color, and utilizing high-SNR NIR signals to constrain the generation of high frequency texture details, thereby achieving complementary advantages in the frequency domain. Furthermore, an uncertainty-based adaptive weighting loss function is introduced to automatically balance the contributions of different frequency tasks, solving the problems of color distortion and artifacts caused by rigid fusion in the spatial domain common in traditional methods. Experimental results demonstrate that FD-INR not only effectively restores image luminance consistency and structural details but also, benefitting from its implicit continuous representation, outperforms existing methods in arbitrary-resolution reconstruction tasks, significantly enhancing the reliability of low light perception. --- ### Generative Semantic Communication via Alternating Dual-Domain Posterior Sampling Authors: Shunpu Tang, Qianqian Yang Published: 2026-04-18 Categories: cs.CV, cs.IT, eess.SP arXiv: https://arxiv.org/abs/2604.16796v1 Generative semantic communication (SemCom) harnesses pretrained generative priors to improve the perceptual quality of wireless image transmission. Existing generative SemCom receivers, however, rely on maximum a posteriori (MAP) estimation, which fundamentally cannot preserve the data distribution and thus limits achievable perceptual quality. Moreover, current diffusion-based approaches using single-domain guidance face significant limitations: latent-domain guidance is sensitive to channel noise, while image-domain guidance inherits decoder bias. Simply combining both domains simultaneously yields an overconfident pseudo-posterior. In this paper, we formulate semantic decoding as a Bayesian inverse problem and prove that posterior sampling achieves optimal perceptual quality by preserving the data distribution. Building on this insight, we propose alternating dual-domain posterior sampling (ADDPS), a diffusion-based SemCom receiver that alternately enforces latent-domain and image-domain consistency during the sampling process. This alternating strategy decomposes joint posterior sampling into simpler subproblems, avoiding gradient conflicts while retaining the complementary strengths of both domains. Experiments on FFHQ demonstrate that the proposed ADDPS achieves superior perceptual quality compared with existing methods. --- ### Improving Radio Interferometry Imaging by Explicitly Modeling Cross-Domain Consistency in Reconstruction Authors: Kai Cheng, Ruoqi Wang, Qiong Luo Published: 2026-04-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.16794v1 Radio astronomy plays a crucial role in understanding the universe, particularly within the realm of non-thermal astrophysics. Images of celestial objects are derived from the signals (called visibility) measured by radio telescopes. Such imaging results, called dirty images, contain artifacts due to factors such as sparsity and therefore require reconstruction to improve imaging quality. Existing methods typically restrict reconstruction to a unimodal domain, either to the dirty image after imaging or to the sparse visibility prior to imaging. Focusing solely on each unimodal reconstruction results in the loss of complementary in-context information in either the visibility or image domain, leading to an incomplete modeling of mutual dependency and consistency. To address these challenges, we propose CDCRec, a multimodal radio interferometric data reconstruction method that explicitly models cross-domain consistency. We design a hierarchical multi-task and multi-stage framework to enhance the exploration of interplays between domains during reconstruction. Our experimental results demonstrate that CDCRec improves imaging performance through enhanced cross-domain correlation extraction. In particular, our self-supervised complementary modeling strategy is better than current methods at interferometric domain translations that rely heavily on recovering dense information from constrained source-domain data. --- ### AstroSURE: Learning to Remove Noise from Astronomical Images Without Ground Truth Data Authors: Omid Vaheb, Sebastien Fabbro, Stark Draper Published: 2026-04-18 Categories: astro-ph.IM, cs.CV arXiv: https://arxiv.org/abs/2604.16793v1 In astronomical imaging, the low photon count of exposures necessitates extensive post-processing steps, including contamination removal and denoising. This paper evaluates deep-learning denoising methods that can be trained without clean ground-truth images and assesses their utility for detection11 oriented analysis of astronomical data. We adapt and compare Noise2Noise, Stein's Unbiased Risk Estimator, and blind-spot-based methods using synthetic data and real observations from the Hubble Space Telescope (HST) and the Canada-France-Hawaii Telescope (CFHT). Performance is evaluated using object-detection metrics, including correct detection rate and false alarm rate, together with image-based metrics and pixel-distribution diagnostics. The results show that these methods can improve faint-source detectability relative to the original noisy images, with encouraging gains on HST data after domain-consistent initialization, while transfer to CFHT data is more limited, highlighting the importance of instrument/domain similarity for unsupervised adaptation. --- ### Bias in the Loop: Auditing LLM-as-a-Judge for Software Engineering Authors: Zixiao Zhao, Amirreza Esmaeili, Fatemeh Fard Published: 2026-04-18 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.16790v1 Large Language Models are increasingly used as judges to evaluate code artifacts when exhaustive human review or executable test coverage is unavailable. LLM-judge is increasingly relevant in agentic software engineering workflows, where it can help rank candidate solutions and guide patch selection. While attractive for scale, current practice lacks a principled account of reliability and bias: repeated evaluations of the same case can disagree; small prompt edits can swing outcomes; and seemingly semantics-preserving, human-equivalent perturbations may elicit divergent verdicts. This paper studies LLM-as-a-Judge for code through a measurement-first lens. We analyze two pointwise judging regimes across code generation, code repair task, and test generation, and we systematically probe prompt-induced biases. Our study considers difficulty levels for repeated runs and controlled prompt interventions that isolate one presentation cue at a time, and it evaluates judges using consistency and sensitivity to bias. We find that judge decisions are highly sensitive to prompt biases even when the underlying code snippet is unchanged. Across all three tasks, several biases systematically shift preferences toward the option favored by the prompt, improving accuracy when that option aligns with the gold answer but substantially reducing it otherwise. In some settings, these effects are large enough to change task-level conclusions and alter relative model rankings. These findings show that reported judge performance may reflect prompt artifacts rather than stable assessment ability, posing a direct threat to the validity and reproducibility of code evaluation. We therefore argue that LLM-as-a-Judge studies should report bias sensitivity alongside accuracy and incorporate explicit controls to support more trustworthy model comparison in software engineering. --- ### When Informal Text Breaks NLI: Tokenization Failure, Distribution Shift, and Targeted Mitigations Authors: Avinash Goutham Aluguvelly Published: 2026-04-18 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.16787v1 We study how informal surface forms degrade NLI accuracy in ELECTRA-small (14M) and RoBERTa-large (355M) across four transforms applied to SNLI and MultiNLI: slang substitution, emoji replacement, Gen-Z filler tokens, and their combination. Slang substitution (replacing formal words with informal equivalents, e.g., "going to" -> "gonna", "friend" -> "homie") causes minimal degradation (at most 1.1pp): slang vocabulary falls largely within WordPiece coverage, so the tokenizer handles it without signal loss. Emoji replaces content words with Unicode characters that ELECTRA's WordPiece tokenizer maps to (UNK), destroying the input signal before any learned parameters see it (93.6% of emoji examples contain at least one (UNK), mean 2.91 per example). Noise tokens (no cap, deadass, tbh) are fully in-vocabulary but absent from NLI training data, consistent with the model assigning them inferential weight they do not carry. The two failure modes respond to different interventions: preprocessing recovers emoji accuracy by normalizing text before tokenization; augmentation handles noise by exposing the model to noise-bearing examples during training. A hybrid of both achieves 88.93% on the combined variant for ELECTRA on SNLI (up from 75.88%), with no statistically significant drop on clean text. Against GPT-4o-mini zero-shot, unmitigated ELECTRA is significantly worse on transformed variants (p < 0.0001); hybrid ELECTRA surpasses it across all SNLI variants and reaches statistical parity on MultiNLI. --- ### Bridging Coarse and Fine Recognition: A Hybrid Approach for Open-Ended Multi-Granularity Object Recognition in Interactive Educational Games Authors: Hanling Yi, Feng Lin, Mao Luo, Yifan Yang, Xiaotian Yu, Rong Xiao Published: 2026-04-18 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2604.16785v1 Recent advances in Multimodal Large Language Models (MLLMs) have enabled open-ended object recognition, yet they struggle with fine-grained tasks. In contrast, CLIP-style models excel at fine-grained recognition but lack broad coverage of general object categories. To bridge this gap, we propose \textbf{HyMOR}, a \textbf{Hy}brid \textbf{M}ulti-granularity open-ended \textbf{O}bject \textbf{R}ecognition framework that integrates an MLLM with a CLIP model. In HyMOR, the MLLM performs open-ended and coarse-grained object recognition, while the CLIP model specializes in fine-grained identification of domain-specific objects such as animals and plants. This hybrid design enables accurate object understanding across multiple semantic granularities, serving as a robust perceptual foundation for downstream multi-modal content generation and interactive gameplay. To support evaluation in content-rich and educational scenarios, we introduce TBO (TextBook Objects), a dataset containing 20,942 images annotated with 8,816 object categories extracted from textbooks. Extensive experiments demonstrate that HyMOR narrows the fine-grained recognition gap with CLIP to 0.2\% while improving general object recognition by 2.5\% over a baseline MLLM, measured by average Sentence-BERT (SBert) similarity. Overall, HyMOR achieves a 23.2\% improvement in average SBert across all evaluated datasets, highlighting its effectiveness in enabling accurate perception for multi-modal game content generation and interactive learning applications. --- ### EdgeVTP: Exploration of Latency-efficient Trajectory Prediction for Edge-based Embedded Vision Applications Authors: Seungjin Kim, Reza Jafarpourmarzouni, Christopher Neff, Hamed Tabkhi, Vinit Katariya Published: 2026-04-18 Categories: cs.CV arXiv: https://arxiv.org/abs/2604.16783v1 Vehicle trajectory prediction is central to highway perception, but deployment on roadside edge devices necessitates bounded, deterministic end-to-end latency. We present EdgeVTP, an embedded-first trajectory predictor that combines interaction-aware graph modeling with a lightweight transformer backbone and a one-shot curve decoder. By predicting future motion as compact curve parameters (anchored at the last observed position) rather than horizon-scaled autoregressive waypoints, EdgeVTP reduces decoding overhead while producing smooth trajectories. To keep runtime predictable in crowded scenes, we explicitly bound interaction complexity via a locality graph with a hard neighbor cap. Across three highway benchmarks and two Jetson-class platforms, EdgeVTP achieves the lowest measured end-to-end latency under a protocol that includes graph construction and post-processing, while attaining state-of-the-art (SotA) prediction accuracy on two of the three datasets and competitive error on other benchmarks. Our code is available at https://github.com/SeungjinStevenKim/EdgeVTP. --- ### FairNVT: Fair Classification via Noise Injection in Vision Transformers Authors: Qiaoyue Tang, Sepidehsadat Hosseini, Mengyao Zhai, Thibaut Durand, Greg Mori Published: 2026-04-18 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.16780v2 This paper presents FairNVT, a lightweight debiasing framework for pretrained transformer-based encoders that improves prediction fairness while preserving task performance. FairNVT is motivated by the intuition that reducing sensitive-attribute information in the representation used by the downstream classifier can facilitate fairer predictions. Our approach learns task-relevant and sensitive embeddings via lightweight adapters, applies calibrated Gaussian noise to the sensitive embedding, and fuses it with the task representation. Together with orthogonality constraints and fairness regularization, these components jointly reduce sensitive-attribute leakage in the learned embeddings and encourage fairer downstream predictions. Across three datasets spanning vision and language, FairNVT reduces sensitive-attribute attacker accuracy, improves fairness metrics such as demographic parity difference and equalized odds, and maintains competitive task performance. --- ### FairNVT: Improving Fairness via Noise Injection in Vision Transformers Authors: Qiaoyue Tang, Sepidehsadat Hosseini, Mengyao Zhai, Thibaut Durand, Greg Mori Published: 2026-04-18 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.16780v1 This paper presents FairNVT, a lightweight debiasing framework for pretrained transformer-based encoders that improves both representation and prediction level fairness while preserving task accuracy. Unlike many existing debiasing approaches that address these notions separately, we argue they are inherently connected: suppressing sensitive information at the representation level can facilitate fairer predictions. Our approach learns task-relevant and sensitive embeddings via lightweight adapters, applies calibrated Gaussian noise to the sensitive embedding, and fuses it with the task representation. Together with orthogonality constraints and fairness regularization, these components jointly reduce sensitive-attribute leakage in the learned embeddings and encourage fairer downstream predictions. The framework is compatible with a wide range of pretrained transformer encoders. Across three datasets spanning vision and language, FairNVT reduces sensitive-attribute attacker accuracy, improves demographic-parity and equalized-odds metrics, and maintains high task performance. --- ### Q-SINDy: Quantum-Kernel Sparse Identification of Nonlinear Dynamics with Provable Coefficient Debiasing Authors: Samrendra Roy, Syed Bahauddin Alam Published: 2026-04-18 Categories: quant-ph, cs.LG arXiv: https://arxiv.org/abs/2604.16779v3 Quantum feature maps offer expressive embeddings for classical learning tasks, and augmenting sparse identification of nonlinear dynamics (SINDy) with such features is a natural but unexplored direction. We introduce \textbf{Q-SINDy}, a quantum-kernel-augmented SINDy framework, and identify a specific failure mode that arises: \emph{coefficient cannibalization}, in which quantum features absorb coefficient mass that rightfully belongs to the polynomial basis, corrupting equation recovery. We derive the exact cannibalization-bias formula $Δξ_P = (P^\top P)^{-1}P^\top Q\,\hatξ_Q$ and prove that orthogonalizing quantum features against the polynomial column space at fit time eliminates this bias exactly. The claim is verified numerically to machine precision ($<10^{-12}$) on multiple systems. Empirically, across six canonical dynamical systems (Duffing, Van der Pol, Lorenz, Lotka-Volterra, cubic oscillator, Rössler) and three quantum feature map architectures (ZZ-angle encoding, IQP, data re-uploading), orthogonalized Q-SINDy consistently matches vanilla SINDy's structural recovery while uncorrected augmentation degrades true-positive rates by up to 100\%. A refined dynamics-aware diagnostic, $R^2_Q$ for $\dot X$, predicts cannibalization severity with statistical significance (Pearson $r=0.70$, $p=0.023$). An RBF classical-kernel control across 20 hyperparameter configurations fails more severely than any quantum variant, ruling out feature count as the cause. Orthogonalization remains robust under depolarizing hardware noise up to 2\% per gate, and the framework extends without modification to Burgers' equation. --- ### Federation over Text: Insight Sharing for Multi-Agent Reasoning Authors: Dixi Yao, Tahseen Rabbani, Manzil Zaheer, Tian Li Published: 2026-04-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.16778v2 We propose a federated learning-like framework, Federation over Text (FoT), that enables multiple clients solving different tasks to collectively generate a shared library of metacognitive insights by iteratively federating their local reasoning processes without sharing actual problem instances or task instructions. Instead of federation over gradients (e.g., as in distributed training), FoT operates at the semantic level without any gradient optimization or supervision signal. Iteratively, each client runs an LLM agent that does local thinking and self-improvement on their specific tasks independently, and shares reasoning traces with a central server, which aggregates and distills them into a cross-task (and cross-domain) insight library that existing and future agents can leverage to improve performance on related tasks. Experiments show that FoT improves reasoning effectiveness and efficiency across a wide range of challenging applications, including mathematical problem solving, cross-domain collaboration, real-world daily tasks, and machine learning research insight discovery. Specifically, it improves average performance scores by 25% while reducing the reasoning tokens by 4% across the first three applications. In the research insight discovery application, FoT is able to generate insights that cover over 80% of the major contributions in the subsequent papers. --- ### SAVE: A Generalizable Framework for Multi-Condition Single-Cell Generation with Gene Block Attention Authors: Jiahao Li, Jiayi Dong, Peng Ye, Xiaochi Zhou, Haohai Lu, Fei Wang Published: 2026-04-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.16776v1 Modeling single-cell gene expression across diverse biological and technical conditions is crucial for characterizing cellular states and simulating unseen scenarios. Existing methods often treat genes as independent tokens, overlooking their high-level biological relationships and leading to poor performance. We introduce SAVE, a unified generative framework based on conditional Transformers for multi-condition single-cell modeling. SAVE leverages a coarse-grained representation by grouping semantically related genes into blocks, capturing higher-order dependencies among gene modules. A Flow Matching mechanism and condition-masking strategy further enhance flexible simulation and enable generalization to unseen condition combinations. We evaluate SAVE on a range of benchmarks, including conditional generation, batch effect correction, and perturbation prediction. SAVE consistently outperforms state-of-the-art methods in generation fidelity and extrapolative generalization, especially in low-resource or combinatorially held-out settings. Overall, SAVE offers a scalable and generalizable solution for modeling complex single-cell data, with broad utility in virtual cell synthesis and biological interpretation. Our code is publicly available at https://github.com/fdu-wangfeilab/sc-save --- ### Representation Before Training: A Practical Benchmark for Generative Medical Event Model Tokenization Authors: Inhyeok Lee, Luke Solo, Michael C. Burkhart, Bashar Ramadan, Sahil Sethi, Sarah Jabbour, William F. Parker, Brett K. Beaulieu-Jones Published: 2026-04-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.16775v3 Generative medical event models use tokenized sequences of patient timelines as input, but practical guidance on the many decisions around tokenization is limited. We benchmark quantization granularity, reference-range anchoring, code--value fusion, numeric and temporal encodings, and native versus harmonized event representations from an expert-mapped common data model. Using both Llama and Qwen architectures, 156 models were trained on full hospitalizations from three initialization seeds, with each configuration following a shared training recipe for up to five epochs. We evaluated learned representations from the first 24 hours of hospitalization with linear probes to predict binary and continuous outcomes during hours 24-48. Fused tokens pairing codes with value deciles increased performance across all eight outcome families relative to the equivalent unfused tokenized input with area under the receiver operating characteristic curve (AUROC) gains of $+0.002$ to $+0.033$ and Spearman correlation gains of $+0.025$ to $+0.114$. Neither anchoring value bins to reference ranges nor increasing quantization granularity consistently improved performance, while xVal variants underperformed both discrete and soft encodings. Alternatives to explicit time tokens, such as event-order and admission-relative rotary position embeddings (RoPE), yielded higher family-mean point estimates across all eight families while reducing input length. When evaluating native input against input mapped to the Common Longitudinal Intensive Care Unit Data Format (CLIF), the CLIF full-hospitalization training sequences contained 28.6% as many tokens as the native sequences and improved performance across six of eight outcome families. These findings show that tokenization and event encoding are consequential design choices when learning patient representations for downstream classification and regression tasks. --- ### Representation Before Training: A Fixed-Budget Benchmark for Generative Medical Event Models Authors: Inhyeok Lee, Luke Solo, Michael C. Burkhart, Bashar Ramadan, William F. Parker, Brett K. Beaulieu-Jones Published: 2026-04-18 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2604.16775v1 Every prediction from a generative medical event model is bounded by how clinical events are tokenized, yet input representation is rarely isolated from other system and architectural choices. We evaluate how representation decisions affect downstream prediction after a shared one-epoch pretraining budget. We train 28 matched transformers on MIMIC-IV and evaluate them on 30 clinical outcomes in three experiments: (1) quantization granularity, reference-range anchoring, and code-value fusion; (2) value encoding (hard bins, soft discretization, code-normalized xVal) crossed with temporal encoding (event order, time tokens, admission-relative RoPE); and (3) native MIMIC laboratory/vital codes versus the Common Longitudinal ICU Format (CLIF)-remapped laboratory/vital codes with compression-preserving perturbation arms. In Experiment 1, fused code-value tokenization improves mortality AUROC from 0.891 to 0.915 (BH-adjusted p < 0.001), hospital length-of-stay AUROC from 0.763 to 0.788 (BH-adjusted p < 0.001), and, for the decile fused-vs-unfused comparison, mean regression Spearman rho across the 13 regression outcomes from 0.414 to 0.494. Across the three temporal encodings, event order only and admission-relative RoPE match or exceed inserting time tokens on average while shortening sequences by 11%. CLIF remapping preserves downstream performance in our single-site setting while yielding a smaller, clinically interpretable token set compatible with multi-site use. Finer-than-decile quantization, reference-range anchoring, and soft discretization help in selective outcomes, while code-normalized xVal remains well below the discrete and soft families, consistent with near-median suppression that persists after the affine variant. --- ### Retention Consequence in Lifecycle Memory Control Authors: Jiarui Han Published: 2026-04-18 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.16774v2 Persistent memory can fail after successful admission: a premise is written, then becomes a silent assumption, and later maintenance treats it as ordinary residue to be compressed, demoted, or evicted. We study this post-admission failure as a lifecycle-control problem. Existing memory systems already perform admission, update, compression, retrieval, and eviction. Our claim is not that such systems lack maintenance, but that retention consequence is often operationalized only indirectly through validity, similarity, recency, frequency, importance, or summarization signals rather than exposed as a separate lifecycle state. We therefore treat confidence as carried-forward validity/support evidence, and introduce strength as an explicit lifecycle state for retention consequence. We operationalize this distinction in StageMem, a small staged controller whose transient, working, and durable stores expose promotion, compression, and eviction pressure points. Across controlled premise-realization, compression, pressure, and implicit-heuristic diagnostics, the experiments separate writing too little, retaining the wrong high-cue content, forgetting costly premises, and preserving everything by saturation. Explicit retention consequence, used through lifecycle settlement, provides a control surface between omission and hoarding. For the targeted post-admission failure mode, the results support a lifecycle view of persistent memory: reliability depends not only on what enters memory, but on whether admission validity and retention consequence remain available during maintenance. --- ### The Reliance Negotiation Framework: A Dynamic Process Model of Student LLM Engagement in Academic Writing Authors: Shahin Hossain Published: 2026-04-18 Categories: cs.CY, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2604.16772v1 Student engagement with large language models (LLMs) in academic writing is not a stable trait, an adoption decision, or a competency level; it is a continuously negotiated process that existing frameworks cannot adequately theorize. Typological models provide categories without mechanisms; technology acceptance models explain adoption but not post-adoption quality; AI literacy frameworks treat competency as a static predictor rather than a live input. None accounts for within-student variability across tasks, the developmental paradox whereby experience produces habituation rather than sophistication, or principled non-use as a form of ethical reasoning. This article introduces the Reliance Negotiation Framework (RNF), developed from a sequential explanatory mixed-methods study of 382 undergraduates at a public minority-serving institution in the United States (survey, N = 382; 14 semi-structured interviews; three qualitative survey strands; 1,435 coded instances). The RNF reconceptualizes LLM reliance as an ongoing negotiation among four concurrent inputs (perceived benefits, perceived risks, ethical commitments, and situational demands) with outputs that recursively modify subsequent decisions. A Two-Model Architecture accommodates the 13.0% of participants whose categorical ethical commitments foreclose negotiation entirely. The framework generates four falsifiable predictions with implications for AI literacy pedagogy, academic integrity policy, and equity-centered practice at minority-serving institutions. --- ### When Misinformation Speaks and Converses: Rethinking Fact-Checking in Audio Platforms Authors: Chaewan Chun, Delvin Ce Zhang, Dongwon Lee Published: 2026-04-18 Categories: cs.CL, cs.CY arXiv: https://arxiv.org/abs/2604.16767v1 Audio platforms have evolved beyond entertainment. They have become central to public discourse, from podcasts and radio to WhatsApp voice notes and live streams. With millions of shows and hundreds of millions of listeners, audio platforms are now a major channel for misinformation. Yet existing fact-checking pipelines are mostly designed for written claims, overlooking the unique properties of spoken media. We argue that audio misinformation is not merely textual content with transcripts: it is structurally different because it is both spoken - carrying persuasive force through prosody, pacing, and emotion - and conversational - unfolding across turns, speakers, and episodes. These dual properties introduce verification difficulties that traditional methods rarely face. This position paper synthesizes evidence across modalities and platforms, examines datasets and methods, and highlights why existing pipelines fail on audio. We argue that advancing fact-checking requires rethinking verification pipelines around the spoken and conversational realities of audio. --- ### Mapping Election Toxicity on Social Media across Issue, Ideology, and Psychosocial Dimensions Authors: Lei Cao, Wen Zeng, Xinyue Wu, Eun Cheol Choi, Emilio Ferrara Published: 2026-04-18 Categories: cs.SI, cs.CL, cs.CY arXiv: https://arxiv.org/abs/2604.16765v1 Online political hostility is pervasive, yet it remains unclear how toxicity varies across campaign issues and political ideology, and what psychosocial signals and framing accompany toxic expression online. In this work, we present a large-scale analysis of discourse on X (Twitter) during the five weeks surrounding the 2024 U.S. presidential election. We categorize posts into 10 major campaign issues, estimate the ideology of posts using a human-in-the-loop LLM-assisted annotation process, detect harmful content with an LLM-based toxicity detection model, and then examine the psychological drivers of toxic content. We use these annotated data to examine how harmful content varies across campaign issues and ideologies, as well as how emotional tone and moral framing shape toxicity in election discussions. Our results show issue heterogeneity in both the prevalence and intensity of toxicity. Identity-related issues displayed the highest toxicity intensity. As for specific harm categories, harassment was most prevalent and intense across most of the issues, while hate concentrated in identity-centered debates. Partisan posts contained more harmful content than neutral posts, and ideological asymmetries in toxicity varied by issue. In terms of psycholinguistic dimensions, we found that toxic discourse is dominated by high-arousal negative emotions. Left- and right-leaning posts often exhibit similar emotional profiles within the same issue domain, suggesting emotional mirroring. Partisan groups frequently rely on overlapping moral foundations, while issue context strongly shapes which moral foundations become most salient. These findings provide a fine-grained account of toxic political discourse on social media and highlight that online political toxicity is highly context-dependent, underscoring the need for issue-sensitive approaches to measuring and mitigating it. --- ### LLM-Extracted Covariates for Clinical Causal Inference: Rethinking Integration Strategies Authors: Lei Liu, Jialin Chen, Kathy Macropol Published: 2026-04-18 Categories: cs.LG arXiv: https://arxiv.org/abs/2604.16763v3 Causal inference from electronic health records (EHR) is fundamentally limited by unmeasured confounding: critical clinical states such as frailty, goals of care, and mental status are documented in free-text notes but absent from structured data. Large language models can extract these latent confounders as interpretable, structured covariates, yet how to effectively integrate them into causal estimation pipelines has not been systematically studied. Using the MIMIC-IV database with 21,859 sepsis patients, we compare seven covariate-integration strategies for estimating the effect of early vasopressor initiation on 28-day mortality, spanning tabular-only baselines, traditional NLP representations, and three LLM-augmented approaches. A central finding is that not all integration strategies are equally effective: directly augmenting the propensity score model with LLM covariates achieves the best performance, while dual-caliper matching on text-derived categorical distances restricts the donor pool and degrades estimation. In semi-synthetic experiments with known ground-truth effects, LLM-augmented propensity scores reduce estimation bias from 0.0143 to 0.0003 relative to tabular-only methods, and this advantage persists under substantial simulated extraction error. On real data, incorporating LLM-extracted covariates reduces the estimated treatment effect from 0.055 to 0.027, directionally consistent with the CLOVERS randomized trial, and a doubly robust estimator yielding 0.031 confirms the robustness of this finding. Our results offer practical guidance on when and how text-derived covariates improve causal estimation in critical care. Our code is available at https://github.com/fpxlei/LLM-Covariates-Causal. --- ### CapSeal: Capability-Sealed Secret Mediation for Secure Agent Execution Authors: Shutong Jin, Ruiyi Guo, Ray C. C. Cheung Published: 2026-04-18 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2604.16762v1 Modern AI agents routinely depend on secrets such as API keys and SSH credentials, yet the dominant deployment model still exposes those secrets directly to the agent process through environment variables, local files, or forwarding sockets. This design fails against prompt injection, tool misuse, and model-controlled exfiltration because the agent can both use and reveal the same bearer credential. We present CapSeal, a capability-sealed secret mediation architecture that replaces direct secret access with constrained invocations through a local trusted broker. CapSeal combines capability issuance, schema-constrained HTTP execution, broker-executed SSH actions, anti-replay session binding, policy evaluation, and tamper-evident audit trails. We describe a Rust prototype integrated with an MCP-facing adapter, formulate conditional security goals for non-disclosure, constrained use, replay resistance, and auditability, and define an evaluation plan spanning prompt injection, tool misuse, and SSH abuse. The resulting system reframes secret handling for agentic systems from handing the model a key to granting the model a narrowly scoped, non-exportable action capability. --- ### Frozen Vision Transformers for Dense Prediction on Small Datasets: A Case Study in Arrow Localization Authors: Maxwell Shepherd Published: 2026-04-18 Categories: cs.CV, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.16758v1 We present a system for automated detection, localization, and scoring of arrow punctures on 40\,cm indoor archery target faces, trained on only 48 annotated photographs (5{,}084 punctures). Our pipeline combines three components: a color-based canonical rectification stage that maps perspective-distorted photographs into a standardized coordinate system where pixel distances correspond to known physical measurements; a frozen self-supervised vision transformer (DINOv3 ViT-L/16) paired with AnyUp guided feature upsampling to recover sub-millimeter spatial precision from $32 \times 32$ patch tokens; and lightweight CenterNet-style detection heads for arrow-center heatmap prediction. Only 3.8\,M of 308\,M total parameters are trainable. Across three cross-validation folds, we achieve a mean F1 score of $0.893 \pm 0.011$ and a mean localization error of $1.41 \pm 0.06$\,mm, comparable to or better than prior fully-supervised approaches that require substantially more training data. An ablation study shows that the CenterNet offset regression head, typically essential for sub-pixel refinement, provides negligible detection improvement while degrading localization in our setting. This suggests that guided feature upsampling already resolves the spatial precision lost through patch tokenization. On downstream archery metrics, the system recovers per-image average arrow scores with a median error of 1.8\% and group centroid positions to within a median of 4.00\,mm. These results demonstrate that frozen foundation models with minimal task-specific adaptation offer a practical paradigm for dense prediction in small-data regimes. --- ### Expressing Social Emotions: Misalignment Between LLMs and Human Cultural Emotion Norms Authors: Sree Bhattacharyya, Manas Mehta, Leona Chen, Cristina Salvador, Agata Lapedriza, Shiran Dudy, James Z. Wang Published: 2026-04-18 Categories: cs.CL, cs.CY arXiv: https://arxiv.org/abs/2604.16757v1 The expression of emotions that serve social purposes, such as asserting independence or fostering interdependence, is central to human interactions and varies systematically across cultures. As LLMs are increasingly used to simulate human behavior in culturally nuanced interactions, it is important to understand whether they faithfully capture human patterns of social emotion expression. When LLM responses are not culturally aligned, their utility is compromised -- particularly when users assume they are interacting with a culturally attuned interlocutor, and may act on advice that proves inappropriate in their cultural context. We present a psychologically informed evaluation framework of cross-cultural social emotion expression in LLMs. Using a human study comparing European American and Latin American participants' expression of engaging and disengaging emotions, we evaluate six frontier LLMs on their ability to reflect culturally differentiated patterns for expressing social emotions. We find systematic misalignment between model and human behavior: all models express engaging emotions more than disengaging ones, with particularly stark differences observed for the generally well-represented European American persona. We further highlight that LLM responses are highly concentrated and deterministic, failing to capture the diversity of human responses in expressing social emotions. Our ablation analyses reveal that these patterns are robust to sampling temperatures, partially sensitive to prompt language, and dependent on the response elicitation format. Together, our findings highlight limitations in how current LLMs represent the interaction of cultural and emotional axes, particularly when expressing social emotions, with direct implications for their deployment in cross-cultural affective contexts. --- ### Mitigating Prompt-Induced Cognitive Biases in General-Purpose AI for Software Engineering Authors: Francesco Sovrano, Gabriele Dominici, Alberto Bacchelli Published: 2026-04-18 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2604.16756v2 Prompt-induced cognitive biases are changes in a general-purpose AI (GPAI) system's decisions caused solely by biased wording in the input (e.g., framing, anchors), not task logic. In software engineering (SE) decision support (where problem statements and requirements are natural language) small phrasing shifts (e.g., popularity hints or outcome reveals) can push GPAI models toward suboptimal decisions. We study this with PROBE-SWE, a dynamic benchmark for SE that pairs biased and unbiased versions of the same SE dilemmas, controls for logic and difficulty, and targets eight SE-relevant biases (anchoring, availability, bandwagon, confirmation, framing, hindsight, hyperbolic discounting, overconfidence). We ask whether prompt engineering mitigates bias sensitivity in practice, focusing on actionable techniques that practitioners can apply off-the-shelf in real environments. Testing common strategies (e.g., chain-of-thought, self-debiasing) on cost-effective GPAI systems, we find no statistically significant reductions in bias sensitivity on a per-bias basis. We then adopt a Prolog-style view of the reasoning process: solving SE dilemmas requires making explicit any background axioms and inference assumptions (i.e., SE best practices) that are usually implicit in the prompt. So, we hypothesize that bias-inducing features short-circuit assumption elicitation, pushing GPAI models toward biased shortcuts. Building on this, we introduce an end-to-end method that elicits best practices and injects axiomatic reasoning cues into the prompt before answering, reducing overall bias sensitivity by 51% on average (p < .001). Finally, we report a thematic analysis that surfaces linguistic patterns associated with heightened bias sensitivity, clarifying when GPAI use is less advisable for SE decision support and where to focus future countermeasures. --- ### Machine individuality: Separating genuine idiosyncrasy from response bias in large language models Authors: Valentin Kriegmair, Dirk U. Wulff Published: 2026-04-18 Categories: cs.AI arXiv: https://arxiv.org/abs/2604.16755v2 As large language models (LLMs) are increasingly integrated into daily life, in roles ranging from high-stakes decision support to companionship, understanding their behavioral dispositions becomes critical. A growing literature uses psychometric inventories and cognitive paradigms to profile LLM dispositions. However, these approaches cannot determine whether behavioral differences reflect stable, stimulus-specific individuality or global response biases and stochastic noise. Here, we apply crossed random-effects models -- widely used in psychometrics to separate systematic effects -- to 74.9 million ratings provided by 10 open-weight LLMs for over 100,000 words across 14 psycholinguistic norms. On average, 16.9% of variance is attributable to stimulus-specific individuality, robustly exceeding a statistical null model. Cross-norm prediction analyses reveal this individuality as a coherent fingerprint, unique to each model. These results identify individual differences among LLMs that cannot be attributed to response biases or stochastic noise. We term these differences machine individuality. --- ### FineServe: A Fine-Grained Dataset and Characterization of Global LLM Serving Workloads Authors: Tiancheng Zhang, Shaoyuan Huang, Mingyuan Wang, Yunfeng Zhao, Xiaofei Wang, Wenyu Wang Published: 2026-04-17 Categories: cs.AI arXiv: https://arxiv.org/abs/2607.19349v1 Large language models (LLMs) are increasingly deployed as always-on online services, making efficient LLM serving a critical systems challenge. Achieving low latency and high throughput under volatile demand requires deep understanding of real-world serving workloads, yet existing studies often rely on proxy traces or coarse-grained characterizations that fail to capture the heterogeneity of modern multi-model LLM platforms. We present FineServe, an in-the-wild, multi-model LLM serving workload dataset collected from a global commercial marketplace, enabling fine-grained characterization of real-world serving dynamics across heterogeneous models and tasks. Leveraging FineServe, we conduct a comprehensive analysis of arrival dynamics and token behavior, revealing fundamentally different fluctuation regimes across model architectures, scales and task intents. Building on these insights, we develop the FineServe workload generator, which composes fine-grained model-aware workloads into configurable mixtures tailored for benchmarking multi-model serving platforms. By exposing these fine-grained workload dynamics, FineServe provides a realistic foundation for evaluating routing, scheduling, and capacity-planning strategies in LLM serving systems. FineServe is available at https://github.com/hihiztc1/FineServe. --- ### BatchDAG: LLM-Planned Execution Graphs for Scalable Ad-Hoc Analysis Over Enterprise Data Authors: Anupreet Walia Published: 2026-04-17 Categories: cs.AI arXiv: https://arxiv.org/abs/2607.18241v1 Large language models (LLMs) excel at analyzing individual documents but break down on exhaustive, cross-entity analytical questions over enterprise-scale datasets due to context overflow, loss of per-entity attribution, and linear latency from sequential tool calls. We present BatchDAG, a system in which an LLM generates a typed directed acyclic graph (DAG) of operations -- SQL queries, semantic searches, in-memory transforms, parallel fan-outs, and single-shot analyses -- which a deterministic engine evaluates with topological-wave parallelism and structured JSON data flow. A key optimization, entity-aware batching, groups rows by logical entity before fan-out, reducing LLM calls by up to 47x. BatchDAG is not primarily an accuracy improvement over hand-optimized pipelines; rather, it is a general-purpose orchestration layer that replaces multiple hand-engineered workflows with a single system that generates the appropriate execution strategy from natural language. In controlled experiments on 12 transcript-heavy queries, BatchDAG (3.74/5) achieves quality comparable to an expert-designed pipeline (3.25/5) and significantly outperforms a ReAct agent (3.09/5, p<0.01), with superior provenance (77% transcript evidence rate vs. 46-60% for baselines). A controlled ablation shows structured JSON intermediates reduce hallucinations by 27% versus prose summaries (paired t-test, p=0.107, n=12). The planner achieves 98.8% valid-DAG rate across 300 planning calls. In production at Brevian.ai, BatchDAG processes queries over 50,000+ meetings in under 60 seconds, with measured per-query costs of $0.02-$0.24 at published GPT-5.1 pricing. --- ### From Consumption to Reflection: Designing Human-AI Relations for Stable Reasoning Authors: Rikard Rosenbacke, Carl Rosenbacke, Victor Rosenbacke, Martin McKee Published: 2026-04-17 Categories: cs.CY, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2606.11195v1 Large language models (LLMs) have transformed how humans access information, but not how we reason with it. Their fluency accelerates consumption while bypassing the slow, reflective processes that underpin sound judgment. This paper introduces Relational Reflective Intelligence (RRI), an inference-time governance layer that operationalizes reflection through auditable reasoning loops. RRI operates not inside the model but around it, providing a practical structure for stable, auditable reasoning between humans and LLMs. The core premise is that LLMs inherit cognitive vulnerabilities similar to those that shape human thought: reliance on intuitive shortcuts, confusion between representation and reality, and a preference for coherence over falsification. When humans and models share these tendencies, their errors compound. We refer to this as relational drift, a failure that arises from interaction rather than from the model alone. Addressing this requires a shift from modeling relations between words to structuring relations between model outputs and human reasoning. RRI provides this missing layer through three components: the Rose-Frame, which identifies likely breakdowns in reasoning; the Architect's Pen, which introduces targeted reflection steps at critical moments; and an inference-time workflow that embeds these steps without retraining the model. Together, these elements transform human-AI interaction into a joint reasoning system with explicit checkpoints, conflict surfacing, and an auditable trail of assumptions. Rather than making machines think like humans or forcing humans to reason like machines, RRI creates a structured interaction in which both compensate for each other's limitations. It reframes AI safety as a cognitive architecture problem, where reliable decisions depend on embedding reflection directly into the interaction process. --- ### The Virtual Roundtable: Multi-Agent Personas Simulating the Dynamics of Human Brainstorming Authors: Tim Dorn, Saara A. Khan, Julie Mumford Published: 2026-04-17 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2606.05178v1 As AI-driven product development accelerates, the bottleneck is shifting from how we build to what we build. Traditional human brainstorming faces challenges including groupthink, echo chambers, and limited diversity. To address this, we present a multi-agentic architecture that simulates roundtable brainstorming through two phases: divergent thinking to generate diverse ideas, and convergent thinking to evaluate and rank the most promising ones. The system employs diverse AI personas that engage in roundtable discussions, guided by an agentic facilitator that steers the discussion toward productive outcomes. Personas maintain private thoughts while commenting publicly, with ideas emerging organically throughout the discussion. Per-persona quotas on idea submissions and votes promote balanced participation while producing natural rankings. Throughout the session, the system tracks each idea's lineage, capturing how concepts originate and cross-pollinate over time. We demonstrate this approach through a case study generating consumer ideas for AI smart glasses, showing (i) it produces diverse, relevant ideas with insights into their evolution; (ii) the cumulative exchange of perspectives across personas cultivates a shared context that progressively deepens the quality of discussion and the ideas produced. --- ### MCBench: A Multicontext Safety Assessment Benchmark for Omni Large Language Models Authors: Manh Luong, Tamas Abraham, Junae Kim, Amar Kaur, Rollin Omari, Gholamreza Haffari, Trang Vu, Lizhen Qu, Dinh Phung Published: 2026-04-17 Categories: cs.CL, cs.AI, eess.AS arXiv: https://arxiv.org/abs/2606.05177v2 Existing multimodal safety benchmarks focus solely on visual inputs and cannot assess Omni Large Language Models (LLMs) that process vision, audio, and text. We introduce MCBench, a benchmark with 1196 scenarios spanning four safety categories that require integrating multiple modalities for accurate safety assessment. Each unsafe scenario is paired with a minimally different safe counterpart to assess model sensitivity. Our evaluations of state-of-the-art models reveal significant challenges. Omni LLMs struggle with subtle or non-physical risks but perform better when salient visual or acoustic cues are present. Analysis of reasoning traces shows that, although models can extract modality-specific information, they often fail to integrate these cues effectively for safety judgments. Our findings reveal that current Omni LLMs lack robust cross-modal reasoning in safety-critical settings, underscoring the need for improved architectures and training strategies for multimodal safety. --- ### MCBench: A Multicontext Safety Assessment Benchmark for Omni Large Language Models Authors: Manh Luong, Tamas Abraham, Junae Kim, Amar Kaur, Rollin Omari, Gholamreza Haffari, Trang Vu, Lizhen Qu, Dinh Phung Published: 2026-04-17 Categories: cs.CL, cs.AI, eess.AS arXiv: https://arxiv.org/abs/2606.05177v1 Existing multimodal safety benchmarks focus solely on visual inputs and cannot assess Omni Large Language Models (LLMs) that process vision, audio, and text. We introduce MCBench, a benchmark with 1196 scenarios spanning four safety categories that require integrating multiple modalities for accurate safety assessment. Each unsafe scenario is paired with a minimally different safe counterpart to assess model sensitivity. Our evaluations of state-of-the-art models reveal significant challenges. Omni LLMs struggle with subtle or non-physical risks but perform better when salient visual or acoustic cues are present. Analysis of reasoning traces shows that, although models can extract modality-specific information, they often fail to integrate these cues effectively for safety judgments. Our findings reveal that current Omni LLMs lack robust cross-modal reasoning in safety-critical settings, underscoring the need for improved architectures and training strategies for multimodal safety. --- ### PEFT of SLM for Telecommunications Customer Support: A Comparative Study of LoRA Configurations with Energy Consumption Analysis Authors: Lucas Tamic, Ilan Jaffeux-Cheniout, Xavier Marjou Published: 2026-04-17 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2606.05176v2 While large language models (LLMs) show strong performance in natural language understanding and generation, their evaluation and adaptation to domain-specific constraints in telecommunications customer support remain limited. In addition, data sovereignty, regulatory constraints, and the handling of sensitive customer and network information complicate the use of externally hosted foundation models in this domain. We present a systematic study of parameter-efficient fine-tuning (PEFT) using Low-Rank Adaptation (LoRA) applied to Qwen2.5-3B to build a domain-specific conversational assistant. We introduce a combinatorial synthetic data generation approach based on a glossary of 52 industry-specific terms, producing approximately 30,000 training examples across 1,560 distinct problem scenarios via a generative pipeline powered by Gemini 2.0 Flash. We evaluate 16 LoRA configurations by varying hyperparameters and target modules. Our evaluation extends beyond standard metrics by incorporating energy consumption analysis and qualitative assessment using an LLM-as-a-judge framework with GPT-5.2 and Claude 4.5 Sonnet. Results show a clear divergence between quantitative and qualitative performance: models achieving the lowest validation loss do not necessarily obtain the best human-aligned rankings. The best validation loss (0.5024) ranks only 6th-7th in qualitative evaluation, while the worst loss (0.6807) ranks first according to both judges. This work contributes (1) a combinatorial method for synthetic dataset construction, (2) insights into the impact of target module selection for LoRA injection, (3) evidence that validation loss alone is insufficient for selecting fine-tuning configurations in conversational AI, and (4) an energy-performance trade-off analysis for sustainable LLM deployment. --- ### Generic Triple-Latent Compression with Gated Associative Retrieval Authors: Liu Xiao Published: 2026-04-17 Categories: cs.CL arXiv: https://arxiv.org/abs/2606.05175v1 We study generic triple-latent sequence models that maintain a running token state and compressed pair-memory pathway to capture higher-order token interactions without benchmark-specific parsing. The triple-latent family improves a small Transformer baseline on byte-level WikiText-2 and on a tokenizer-based MiniMind language-model benchmark, while a recall-focused gated key-value retrieval extension improves associative recall but remains seed-sensitive and much slower in the current reference implementation. --- ### Improving Heart-Focused Medical Question Answering in LLMs via Variance-Aware Rubric Rewards with GRPO Authors: Arash Ahmadi, Parisa Masnadi Khiabani, Sarah Sharif, Charles Nicholson, David Ebert, Mike Banad Published: 2026-04-17 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2606.05174v1 Large Language Models (LLMs) have shown strong promise in healthcare applications. Yet deploying general-purpose models in real-world settings remains difficult due to data privacy constraints, inference costs, and limited suitability for edge or on-device use. These challenges motivate the development of smaller, more efficient models that require robust post-training strategies to ensure reliable medical reasoning. In this work, we investigate Group Relative Policy Optimization (GRPO) for post-training LLMs on heart-focused medical question answering with rubric-based supervision derived from RaR-Medicine. We propose a Variance-Aware Reward Framework that extends the Explicit Aggregation and Implicit Aggregation strategies of Rubrics as Rewards by replacing weighted binary criterion aggregation and single overall Likert-style scoring with continuous analytical reward functions derived from criterion-level rubric outcomes. This formulation provides richer optimization signals for feedback that is sparse, multi-criteria, and difficult to verify automatically, and enables more stable on-policy reinforcement learning. On a held-out heart-related subset of HealthBench, our best GRPO variant improves accuracy from 0.362 to 0.502 and F1 from 0.532 to 0.668 relative to the Qwen3-14B base model, while remaining competitive with GPT-OSS-120B (0.508 accuracy, 0.674 F1). Our findings show that carefully designed rubric-based rewards provide a practical strategy for improving heart-focused medical question answering in LLMs, with potential to extend to other rubric-based tasks. --- ### Learning after COVID-19 and the ICT career aspirations: Are students entering the AI era with weaker skills? Authors: Diana Maria Popa, Simona-Vasilica Oprea, Adela Bâra Published: 2026-04-17 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2605.27391v1 This paper examines whether students are entering the generative AI era with sufficiently strong educational foundations, focusing on the relationship between learning environments and changes in ICT related career aspirations across countries. The analysis uses country-level data from PISA 2018 and 2022, combining indicators of student autonomy, digital skills and teacher support. A mixed-method approach is applied, including descriptive statistics, regression analysis, clustering, latent representation learning (using Variational Autoencoder-VAE), discriminant analysis and probabilistic modeling to capture both observable and latent dimensions of educational readiness. Unlike prior research that treats learning loss, digital skills and career expectations separately, our analysis integrates them within a comparative longitudinal framework. It shifts the focus from short-term post-pandemic effects to the structural capacity of education systems to prepare students for digital and AI-driven labor markets. Results show a global but uneven increase in ICT career aspirations. Digital skills emerge as the strongest and most consistent predictor, while teacher support plays a complementary role. Autonomy shows weaker, context-dependent effects. Educational readiness is multidimensional, and ICT aspirations evolve relatively independently from other career domains. --- ### EvoSpec: Evolving Speculative Decoding via Real-Time Vocabulary and Parameter Adaptation Authors: Shuyu Zhang, Lingfeng Pan, Qicheng Wang, Yaqi Shi, Yueyang Tan, Ruyu Yan, Jiaqi Chen, Lixing Du, Lu Wang Published: 2026-04-17 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.27390v3 Speculative decoding accelerates Large Language Model inference through draft-then-verify generation, yet lightweight draft models face coupled efficiency and quality limitations: large-vocabulary output projection is costly, while limited draft capacity and static parameters reduce acceptance under specialized or shifting inputs. Vocabulary pruning lowers projection cost, but static variants miss locally important long-tail tokens, while dynamic variants remain sensitive to preset selection policies and budgets. Moreover, limited draft capacity can leave the draft distribution misaligned even when the target token is covered. Online alignment improves draft quality, but full-parameter updates introduce substantial memory and latency overhead. We introduce EvoSpec, which jointly adapts the active vocabulary and lightweight draft parameters from verification feedback. EvoSpec asynchronously retrieves semantic and statistical token neighbors and performs curriculum-weighted online LoRA alignment while preserving exact target-model verification. On Qwen3-8B/EAGLE-2, EvoSpec reaches a $2.18\times$ speedup over vanilla decoding and a $1.20\times$ gain over EAGLE-2, while improving specialized-domain coverage and using $27\%$ less auxiliary GPU adaptation memory than full-parameter online adaptation. --- ### High-Risk AI Systems and the Problem of Identity in the European AI Act Authors: Andrea Ferrario Published: 2026-04-17 Categories: cs.CY, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.23922v3 The EU Artificial Intelligence Act (AIA) establishes a lifecycle governance regime for high-risk AI systems built around ex-ante conformity assessment, post-market monitoring, and re-assessment upon "substantial modification." These obligations presuppose AI identity judgments: regulators and providers must decide when an updated system remains the same system over time. In this work, we show how this logic is clarified by the function+ framework of artifact identity, which individuates AI systems by their intended function together with context-sensitive criteria of appropriate functioning, captured as "AI trustworthiness." We further argue that the AIA does not provide an internal, auditable criterion for synchronic identity--when two AI systems at a given time should count as the same for regulatory purposes--and instead largely defers such sameness determinations to sectoral or harmonization instruments. function+ supplies a synchronic identity test anchored in intended function and trustworthiness profiles and levels, making synchronic identity decisions inspectable in governance settings such as procurement, liability, and market surveillance. Our contribution is a conceptual and auditing lens: we provide a correspondence map between AIA lifecycle obligations and function+ identity components, and we make the synchronic case operationally legible via a minimal decision flow for audit and dispute contexts. We conclude with two implementation-facing recommendations: (1) more precise, testable reporting of intended purpose, and (2) standardized, auditable trustworthiness reporting that supports comparability over time and across deployments. --- ### Authority Signals in Claude AI Health Citations: A Descriptive Analysis Using the Authority Signals Framework Authors: Erin T. Jacques, Erela Datuowei, Elizabeth Quaye, Corey H. Basch, Arijit Chatterjee, Juanita Davis Published: 2026-04-17 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2605.23921v1 This study seeks to determine the authority signals used by Anthropic's Claude AI in its presentation of sources when answering consumer health questions. While there exists a great deal of discourse around the quality of health citations that LLMs produce, there is limited information on the integrity of the sources the citations originate from, and to what extent the sources are, from what health professionals would consider, credible sources. This descriptive cross-sectional study used data from HealthSearchQA, which contains 3,172 consumer health questions curated by Google Research. After exclusions, a final dataset of 3,075 questions yielding 10,038 citations was analyzed. The Authority Signals Framework (Jacques et al., 2026) was applied to examine 10 authority signals across four domains for a disproportionate stratified sample of 542 sources. Established institutional sources accounted for 97.8% of all citations (n = 9,818). Medical Institutions were the most frequently cited organization type (36.5%), followed by Government Resources (31.6%) and Professional Associations (28.4%). Commercial Health Information comprised 2.2% (n = 220). The top 10 organizations accounted for 57.8% of all citations, with Mayo Clinic alone representing 24.7%. Among commercial sources in the focused sample, 86.4% displayed medical review statements, 82.5% used schema markup, and 71.8% had comprehensive content, while traditional institutional sources appeared in Claude's citations with or without these same markers. As Anthropic positions Claude for HIPAA-ready healthcare applications, these findings establish a baseline for Claude's citation behavior and demonstrate the utility of the Authority Signals Framework as a tool for ongoing, cross-platform evaluation of AI-mediated health information. --- ### Artificial Effort Authors: Federico Belotti, Stefano Coniglio, Antonio Cosma, Francesco Fallucchi Published: 2026-04-17 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2605.23920v1 Real-effort tasks, in which participants perform cognitively costly activities whose outcomes depend on actual performance, are widely used in experimental economics. Their validity, however, rests on the assumption that a human performs them. We study whether this assumption still holds in the era of Artificial Intelligence (AI) and Large Language Models (LLMs). Using 8 canonical real-effort tasks and 23 LLMs from three major providers, we show that most tasks can now be solved accurately and at a negligible cost, while only a few resist automation. Performance improves with each model generation, and midtier models are rapidly closing the gap with frontier ones, broadening the set of widely accessible models that can automate these tasks. Additionally, we show that verbally offering monetary incentives has no effect on LLM performance. Our findings establish a boundary condition for the use of real-effort tasks in unsupervised settings: when participants can cheaply outsource task completion to an LLM, observed performance may no longer reflect genuine human effort. --- ### Governance by Design: Architecting Agentic AI for Organizational Learning and Scalable Autonomy Authors: Nelly Dux, Cristina Alaimo, Philippe Roussiere, Abhishek Kumar Mishra Published: 2026-04-17 Categories: cs.CY, cs.AI, cs.MA arXiv: https://arxiv.org/abs/2605.20210v1 Agentic AI systems - systems that can pursue goals through multi-step planning and tool-mediated action with limited direct supervision - are moving from experimental prototypes to enterprise deployments. This transition introduces tensions in implementation, scaling, and governance: organizations seek scalable autonomy for knowledge and coordination work, yet must preserve accountability, safety, cost control, and responsibility as systems initiate actions, access enterprise data, and evolve through iterative updates. Building on an in-depth qualitative case of a large IT services company's 2025 development and staged rollout of an agentic system integrated with enterprise tools; we show that governance is implemented through concrete architectural and working arrangements that determine what the system is allowed to do, which tools and data it can use, how memory is handled, and how performance improvements are introduced over time. We then distill seven lessons that explain how to build effective governance into agentic AI during operationalization and scaling. --- ### M3DocDep: Multi-modal, Multi-page, Multi-document Dependency Chunking with Large Vision-Language Models Authors: Joongmin Shin, Jeongbae Park, Jaehyung Seo, Heuiseok Lim Published: 2026-04-17 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2605.18774v1 In long, multi-page industrial documents, retrieval-augmented generation (RAG) depends heavily on whether chunk boundaries follow the document's true structure. Existing text-centric chunkers and generative hierarchy parsers often miss cross-page parent-child relations, figure/table-caption bindings, and boundary cues, which leads to fragmented or redundant chunks and degrades both retrieval and answer quality. We propose M3DocDep, an LVLM-based pipeline that first recovers block-level dependencies and then constructs chunks along the recovered document tree. The pipeline uses SharedDet as a common DP+OCR preprocessing layer, extracts multimodal block embeddings with boundary-aware SoftROI pooling, scores candidate parent-child edges with a biaffine head, decodes a globally valid dependency tree with MST constraints, and builds tree-guided chunks annotated with section paths and page ranges. Under a shared-block evaluation protocol, M3DocDep improves STEDS by +28.5 to +39.6 percent on DHP benchmarks, retrieval nDCG by +1.1 to +15.3 percent, and QA ANLS by +4.5 to +15.3 percent on corpus-level RAG benchmarks. These results show that recovering document dependencies before chunking yields more coherent retrieval units for long, multi-page multimodal documents. ---