# Research Papers — 2026-03 · bundle page 75 of 243 > Full-text markdown of papers #3701–3750 of 12123 published in 2026-03, newest first. AI-readable mirrors of the full 12123-paper 2026-03 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-03/page/75 [← Previous bundle](https://aipod.dev/papers/archive/2026-03/page/74.md) [Next bundle →](https://aipod.dev/papers/archive/2026-03/page/76.md) ### Training-Free Adaptive 360-degree Video Streaming via Semantic Potential Fields Authors: Aizierjiang Aiersilan, Zhangfei Yang Published: 2026-03-22 Categories: cs.NI, cs.CV, cs.MM, cs.RO, eess.IV arXiv: https://arxiv.org/abs/2603.20999v2 Adaptive 360° video streaming for teleoperation faces two coupled challenges: viewport prediction under uncertain gaze patterns and bitrate adaptation over fluctuating wireless channels. While Deep Reinforcement Learning (DRL) methods achieve high Quality of Experience (QoE), their lack of interpretability and dependence on offline training limit deployment in safety-critical systems. We propose OrbitStream, a training-free framework that formulates viewport prediction as a Gravitational Viewport Prediction (GVP) problem, where semantic objects generate potential fields that attract operator gaze, and employs a Saturation-Based Proportional-Derivative (PD) Controller for buffer regulation. On object-rich teleoperation traces, OrbitStream achieves 94.7% zero-shot viewport prediction accuracy without user-specific profiling, approaching trajectory-extrapolation baselines (~98.5%). Across 3,600 Monte Carlo simulations, it ranks second among 12 algorithms (QoE 2.71 vs. BOLA-E's 2.80), outperforming FastMPC (1.84), with 1.01 ms decision latency and minimal rebuffering. --- ### When Does Content-Based Routing Work? Representation Requirements for Selective Attention in Hybrid Sequence Models Authors: Abhinaba Basu Published: 2026-03-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.20997v2 We identify a routing paradox in hybrid sequence models: content-based routing - deciding which tokens deserve expensive attention - requires pairwise computation, and this requirement is inescapable. Through 20+ controlled experiments across three tasks, multiple scales (200K to 1.4B parameters), and 15+ routing mechanisms, we map the routing landscape exhaustively. Every system that achieves high routing precision does so through pairwise token comparison. Every mechanism that avoids pairwise computation fails: recurrent models (Mamba-1.4B: 29%), memory banks (12%), bandits (0.7-3.6%), contrastive pretraining (1.6%), and 12 other approaches all cluster at 1-29%. Routing needs two ingredients: (1) per-token representations with bidirectional context and (2) pairwise token comparison. Bidirectional Mamba (O(n)) + pairwise comparison achieves 99.5%; replacing the full pairwise router with rank-1 projection improves this to 99.7%. Adding one bidirectional layer to frozen Pythia-1B recovers 99.4% routing. Six different O(n) preprocessing mechanisms (bidirectional Mamba, Perceiver inducing points, causal attention with E2E training, sparse attention, bidirectional attention, rank-1 projection) all succeed; global mean pooling (1.9%) and Fourier mixing (0.9%) fail. The routing signal occupies a ~34-dimensional latent subspace, invisible to cosine similarity. Non-learned indices (Bloom filter: 90.9%; BM25: 82.7%) bypass the bottleneck for exact/keyword matching. Combining O(n) bidirectional Mamba with rank-1 pairwise projection yields 99.7% routing at linear inference cost. --- ### The Intelligent Disobedience Game: Formulating Disobedience in Stackelberg Games and Markov Decision Processes Authors: Benedikt Hornig, Reuth Mirsky Published: 2026-03-22 Categories: cs.AI, cs.GT, cs.LG arXiv: https://arxiv.org/abs/2603.20994v1 In shared autonomy, a critical tension arises when an automated assistant must choose between obeying a human's instruction and deliberately overriding it to prevent harm. This safety-critical behavior is known as intelligent disobedience. To formalize this dynamic, this paper introduces the Intelligent Disobedience Game (IDG), a sequential game-theoretic framework based on Stackelberg games that models the interaction between a human leader and an assistive follower operating under asymmetric information. It characterizes optimal strategies for both agents across multi-step scenarios, identifying strategic phenomena such as ``safety traps,'' where the system indefinitely avoids harm but fails to achieve the human's goal. The IDG provides a needed mathematical foundation that enables both the algorithmic development of agents that can learn safe non-compliance and the empirical study of how humans perceive and trust disobedient AI. The paper further translates the IDG into a shared control Multi-Agent Markov Decision Process representation, forming a compact computational testbed for training reinforcement learning agents. --- ### Long-Term Outlier Prediction Through Outlier Score Modeling Authors: Yuma Aoki, Joon Park, Koh Takeuchi, Hisashi Kashima, Shinya Akimoto, Ryuichi Hashimoto, Takahiro Adachi, Takeshi Kishikawa, Takamitsu Sasaki Published: 2026-03-22 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2603.20993v1 This study addresses an important gap in time series outlier detection by proposing a novel problem setting: long-term outlier prediction. Conventional methods primarily focus on immediate detection by identifying deviations from normal patterns. As a result, their applicability is limited when forecasting outlier events far into the future. To overcome this limitation, we propose a simple and unsupervised two-layer method that is independent of specific models. The first layer performs standard outlier detection, and the second layer predicts future outlier scores based on the temporal structure of previously observed outliers. This framework enables not only pointwise detection but also long-term forecasting of outlier likelihoods. Experiments on synthetic datasets show that the proposed method performs well in both detection and prediction tasks. These findings suggest that the method can serve as a strong baseline for future work in outlier detection and forecasting. --- ### Structural Sensitivity in Compressed Transformers: Relative Error Propagation and Layer Removal Authors: Abhinaba Basu, Kumkum Basu, Koushik Deb Published: 2026-03-22 Categories: cs.LG, cs.AI, cs.CL, cs.LO arXiv: https://arxiv.org/abs/2603.20991v2 Compressing transformer weights makes large language models cheaper to deploy. But each layer's compression introduces an error. These errors accumulate as the signal passes through later layers, and how they accumulate is not well understood. We measure this directly: at each layer, we take the ratio of output to input error, calling it rho. A value below one means the layer absorbs the error; above one means it grows. Computing rho on six transformers (117M to 8B parameters) yields three findings. (i) Errors at layer t scale downstream by the product of later rho values, predicting representation drift (Spearman r = -0.44, p < 10^-4). This explains why compressing early layers hurts more than late ones, and why depth-decreasing sparsity schedules outperform uniform ones. Across architecture families, however, model width and redundancy matter more than rho alone. (ii) Within a layer, naive pruning shows a ~600x spread in component sensitivity. Activation-aware pruning (Wanda) shrinks this to 3-7x; the ranking reverses across architectures, so fixed importance scores do not transfer. (iii) For depth pruning, ranking layers by how far rho is from one takes two forward passes. It beats ShortGPT's Block Influence with 1.6x lower perplexity at eight layers removed, and physical deletion delivers 1.22x wall-clock speed-up. A blend of the two criteria does best (perplexity 14.2, 60.0% downstream accuracy on LLaMA-2-7B). Twelve Lean 4 norm inequalities provide machine-checked per-matrix error bounds. The contraction profile thus gives a training-free instrument for two decisions: where to compress within layers, and which to remove. --- ### $\mathrm{ECI}_{\mathrm{sem}}$: Semantic Residual Effective Contrastive Information for Evaluating Hard Negatives Authors: Aarush Sinha, Rahul Seetharaman, Aman Bansal Published: 2026-03-22 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2603.20990v4 Hard-negative source selection for dense retrieval is usually decided only after fine-tuning and downstream evaluation. We propose ECIsem, a validity-weighted diagnostic that ranks candidate hard-negative sources using frozen target-encoder embeddings. ECIsem is training-free, not label-free: each scored example requires a query, a labeled positive, and an explicit candidate negative. Each negative is weighted by target consistency, semantic locality, and lexical residuality, and the source score aggregates weighted residual directions through a log-determinant. At the scale we study, the log-determinant operates in its linear regime, so the score reduces to the mean validity weight; we report this reduction as a finding, and gate ablations show that a single lexical scalar, mean lexical residuality, reproduces the transfer ordering in every block. On MS MARCO negative sources, in-family ECIsem recovers the aggregate BEIR transfer ordering across DistilBERT, E5-base, and Contriever. Constructed controls expose an identifiability limit: a plausible false negative is non-inverted, near the positive, and lexically novel, exactly the profile the gates reward, and across twelve weightings of the three gates and three backbones only one of thirty-six cells is safe, and it is not transfer aligned, a limit that extends to margin-style filters built from the same statistics. The same gate statistics separate the controls sharply when used as audit features rather than as rewards, so we pair the score with a reported audit profile instead of a reweighting. --- ### $\mathrm{ECI}_{\mathrm{sem}}$: Semantic Residual Effective Contrastive Information for Evaluating Hard Negatives Authors: Aarush Sinha, Rahul Seetharaman, Aman Bansal Published: 2026-03-22 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2603.20990v3 Hard-negative source selection for dense retrieval is usually decided only after fine-tuning and downstream evaluation. We propose $\mathrm{ECI}_{\mathrm{sem}}$, a semantic residual variant of Effective Contrastive Information (ECI) that ranks candidate negative sources using frozen target-encoder embeddings. $\mathrm{ECI}_{\mathrm{sem}}$ is training-free, not label-free: each scored example requires a query, a labeled positive, and an explicit candidate negative. $\mathrm{ECI}_{\mathrm{sem}}$ builds a weighted residual information matrix from target consistency, semantic locality, lexical residuality, and a log-determinant diversity objective. On MS MARCO negative sources, in-family $\mathrm{ECI}_{\mathrm{sem}}$ ranks LLM negatives highest among non-hybrid sources and Dense+LLM highest among hybrid sources, matching the strongest aggregate BEIR transfer results across DistilBERT, E5-base, and Contriever. Controlled ablations show that this alignment depends on using the target encoder family, while additional ablations show stability under sample-size, temperature, tokenizer, and IDF-corpus perturbations. The theory gives a local linearized link to loss reduction, while the empirical study treats downstream evaluation as the final test. --- ### Toward an automated science of the mind Authors: Akshay K. Jagadish, Milena Rmus, Kristin Witte, Marvin Mathony, Marcel Binz, Eric Schulz Published: 2026-03-22 Categories: cs.AI, q-bio.NC arXiv: https://arxiv.org/abs/2603.20988v2 Artificial Intelligence is starting to automate scientific discovery, raising the prospect of an automated science of the mind. We chart this field around four challenges: representing experiments, generating synthetic behavior, synthesizing models, and closing the loop to discover psychological theories. --- ### Can we automatize scientific discovery in the cognitive sciences? Authors: Akshay K. Jagadish, Milena Rmus, Kristin Witte, Marvin Mathony, Marcel Binz, Eric Schulz Published: 2026-03-22 Categories: cs.AI, q-bio.NC arXiv: https://arxiv.org/abs/2603.20988v1 The cognitive sciences aim to understand intelligence by formalizing underlying operations as computational models. Traditionally, this follows a cycle of discovery where researchers develop paradigms, collect data, and test predefined model classes. However, this manual pipeline is fundamentally constrained by the slow pace of human intervention and a search space limited by researchers' background and intuition. Here, we propose a paradigm shift toward a fully automated, in silico science of the mind that implements every stage of the discovery cycle using Large Language Models (LLMs). In this framework, experimental paradigms exploring conceptually meaningful task structures are directly sampled from an LLM. High-fidelity behavioral data are then simulated using foundation models of cognition. The tedious step of handcrafting cognitive models is replaced by LLM-based program synthesis, which performs a high-throughput search over a vast landscape of algorithmic hypotheses. Finally, the discovery loop is closed by optimizing for ''interestingness'', a metric of conceptual yield evaluated by an LLM-critic. By enabling a fast and scalable approach to theory development, this automated loop functions as a high-throughput in-silico discovery engine, surfacing informative experiments and mechanisms for subsequent validation in real human populations. --- ### Interpreting the Synchronization Gap: The Hidden Mechanism Inside Diffusion Transformers Authors: Emil Albrychiewicz, Andrés Franco Valiente, Li-Ching Chen, Viola Zixin Zhao Published: 2026-03-22 Categories: cs.LG, cond-mat.dis-nn, cond-mat.stat-mech arXiv: https://arxiv.org/abs/2603.20987v1 Recent theoretical models of diffusion processes, conceptualized as coupled Ornstein-Uhlenbeck systems, predict a hierarchy of interaction timescales, and consequently, the existence of a synchronization gap between modes that commit at different stages of the reverse process. However, because these predictions rely on continuous time and analytically tractable score functions, it remains unclear how this phenomenology manifests in the deep, discrete architectures deployed in practice. In this work, we investigate how the synchronization gap is mechanistically realized within pretrained Diffusion Transformers (DiTs). We construct an explicit architectural realization of replica coupling by embedding two generative trajectories into a joint token sequence, modulated by a symmetric cross attention gate with variable coupling strength g. Through a linearized analysis of the attention difference, we show that the replica interaction decomposes mechanistically. We empirically validate our theoretical framework on a pretrained DiT-XL/2 model by tracking commitment and per layer internal mode energies. Our results reveal that: (1) the synchronization gap is an intrinsic architectural property of DiTs that persists even when external coupling is turned off; (2) as predicted by our spatial routing bounds, the gap completely collapses under strong coupling; (3) the gap is strictly depth localized, emerging sharply only within the final layers of the Transformer; and (4) global, low frequency structures consistently commit before local, high frequency details. Ultimately, our findings provide a mechanistic interpretation of how Diffusion Transformers resolve generative ambiguity, isolating speciation transitions to the terminal layers of the network. --- ### AutoMOOSE: An Agentic AI for Autonomous Phase-Field Simulation Authors: Sukriti Manna, Henry Chan, Subramanian K. R. S. Sankaranarayanan Published: 2026-03-22 Categories: cs.AI, cond-mat.mes-hall arXiv: https://arxiv.org/abs/2603.20986v2 Phase-field modeling links thermodynamics and kinetics to microstructural evolution, but multiphysics frameworks such as MOOSE require expertise to construct inputs, manage campaigns, diagnose failures, and validate results. We introduce AutoMOOSE, an open-source multi-agent framework that orchestrates the simulation lifecycle from a single natural-language prompt. Six specialized agents--Architect, Input Writer, Runner, Reviewer, Visualization, and a physics-grounded Skeptic--generate, execute, analyze, and adversarially test simulations against conservation laws, asymptotic limits, and scaling relations. We validate AutoMOOSE in two domains: non-conserved copper grain growth governed by Allen-Cahn dynamics and conserved Fe-Cr spinodal decomposition governed by Cahn-Hilliard dynamics. On a prospectively specified 25-task grain-growth benchmark spanning temperature, grain count, resolution, and model formulation, AutoMOOSE generates valid inputs for all 25 tasks, completes 19 simulations with measurable coarsening, and delivers 15 results that satisfy Burke-Turnbull kinetics ($R^2 \geq 0.90$) and survive Skeptic falsification. The failures trace to two identifiable generator defects. An ensemble of 1000 simulations recovers the prescribed activation energy to within 1% after finite-size extrapolation ($Q_\infty = 0.228$ eV versus $0.230$ eV). In the conserved domain, an agent-generated CALPHAD-based Fe-Cr simulation limits relative mass drift to $6.3 \times 10^{-6}$, dissipates free energy to a stable plateau, and evolves toward the equilibrium tie-line. Controlled ablations show that the full pipeline converts stochastic raw generation--3-7 successful tasks out of 8 across repeated trials--into consistent success on all 8 tasks. AutoMOOSE therefore provides a practical route from natural-language specification to reproducible, physics-validated phase-field simulation campaigns. --- ### AutoMOOSE: An Agentic AI for Autonomous Phase-Field Simulation Authors: Sukriti Manna, Henry Chan, Subramanian K. R. S. Sankaranarayanan Published: 2026-03-22 Categories: cs.AI, cond-mat.mes-hall arXiv: https://arxiv.org/abs/2603.20986v1 Multiphysics simulation frameworks such as MOOSE provide rigorous engines for phase-field materials modeling, yet adoption is constrained by the expertise required to construct valid input files, coordinate parameter sweeps, diagnose failures, and extract quantitative results. We introduce AutoMOOSE, an open-source agentic framework that orchestrates the full simulation lifecycle from a single natural-language prompt. AutoMOOSE deploys a five-agent pipeline in which the Input Writer coordinates six sub-agents and the Reviewer autonomously corrects runtime failures without user intervention. A modular plugin architecture enables new phase-field formulations without modifying the core framework, and a Model Context Protocol (MCP) server exposes the workflow as ten structured tools for interoperability with any MCP-compatible client. Validated on a four-temperature copper grain growth benchmark, AutoMOOSE generates MOOSE input files with 6 of 12 structural blocks matching a human expert reference exactly and 4 functionally equivalent, executes all runs in parallel with a 1.8x speedup, and performs an end-to-end physical consistency check spanning intent, finite-element execution, and Arrhenius kinetics with no human verification. Grain coarsening kinetics are recovered with R^2 = 0.90-0.95 at T >= 600 K; the recovered activation energy Q_fit = 0.296 eV is consistent with a human-written reference (Q_fit = 0.267 eV) under identical parameters. Three runtime failure classes were diagnosed and resolved autonomously within a single correction cycle, and every run produces a provenance record satisfying FAIR data principles. These results show that the gap between knowing the physics and executing a validated simulation campaign can be bridged by a lightweight multi-agent orchestration layer, providing a pathway toward AI-driven materials discovery and self-driving laboratories. --- ### Consistent but Dangerous: Per-Sample Safety Classification Reveals False Reliability in Medical Vision-Language Models Authors: Binesh Sadanandan, Vahid Behzadan Published: 2026-03-22 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.20985v1 Consistency under paraphrase, the property that semantically equivalent prompts yield identical predictions, is increasingly used as a proxy for reliability when deploying medical vision-language models (VLMs). We show this proxy is fundamentally flawed: a model can achieve perfect consistency by relying on text patterns rather than the input image. We introduce a four-quadrant per-sample safety taxonomy that jointly evaluates consistency (stable predictions across paraphrased prompts) and image reliance (predictions that change when the image is removed). Samples are classified as Ideal (consistent and image-reliant), Fragile (inconsistent but image-reliant), Dangerous (consistent but not image-reliant), or Worst (inconsistent and not image-reliant). Evaluating five medical VLM configurations across two chest X-ray datasets (MIMIC-CXR, PadChest), we find that LoRA fine-tuning dramatically reduces flip rates but shifts a majority of samples into the Dangerous quadrant: LLaVA-Rad Base achieves a 1.5% flip rate on PadChest while 98.5% of its samples are Dangerous. Critically, Dangerous samples exhibit high accuracy (up to 99.6%) and low entropy, making them invisible to standard confidence-based screening. We observe a negative correlation between flip rate and Dangerous fraction (r = -0.89, n=10) and recommend that deployment evaluations always pair consistency checks with a text-only baseline: a single additional forward pass that exposes the false reliability trap. --- ### Joint Surrogate Learning of Objectives, Constraints, and Sensitivities for Efficient Multi-objective Optimization of Neural Dynamical Systems Authors: Frithjof Gressmann, Ivan Georgiev Raikov, Seung Hyun Kim, Mattia Gazzola, Lawrence Rauchwerger, Ivan Soltesz Published: 2026-03-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.20984v2 Gaussian process surrogates dominate constrained multi-objective optimization because they are effective in data-scarce regimes, but their cubic scaling in training samples limits their ability to capture shared structure between objectives and constraints as problems grow in dimensionality. We show that deterministic neural network surrogates, equipped with feature tokenization and adaptive output normalization, match or exceed Gaussian process accuracy, while scaling to high-dimensional output spaces and training on all data including infeasible samples. Jointly training a single Feature Tokenizer Transformer to predict objectives, constraint satisfaction, and parameter sensitivities yields a unified gradient that simultaneously improves objective values, steers toward feasibility, and identifies the most influential parameters: a coherent search signal that disjoint per-output models cannot provide. We validate this on biophysical neural optimization problems of increasing complexity. In the hardest regime, with wide, uninformed parameter bounds where random sampling finds zero feasible solutions, descending the surrogate's learned constraint gradient steers the search into the feasible region and recovers near-optimal solutions where standard surrogate optimization and constrained Bayesian optimization find none. --- ### Joint Surrogate Learning of Objectives, Constraints, and Sensitivities for Efficient Multi-objective Optimization of Neural Dynamical Systems Authors: Frithjof Gressmann, Ivan Georgiev Raikov, Seung Hyun Kim, Mattia Gazzola, Lawrence Rauchwerger, Ivan Soltesz Published: 2026-03-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.20984v1 Biophysical neural system simulations are among the most computationally demanding scientific applications, and their optimization requires navigating high-dimensional parameter spaces under numerous constraints that impose a binary feasible/infeasible partition with no gradient signal to guide the search. Here, we introduce DMOSOPT, a scalable optimization framework that leverages a unified, jointly learned surrogate model to capture the interplay between objectives, constraints, and parameter sensitivities. By learning a smooth approximation of both the objective landscape and the feasibility boundary, the joint surrogate provides a unified gradient that simultaneously steers the search toward improved objective values and greater constraint satisfaction, while its partial derivatives yield per-parameter sensitivity estimates that enable more targeted exploration. We validate the framework from single-cell dynamics to population-level network activity, spanning incremental stages of a neural circuit modeling workflow, and demonstrate efficient, effective optimization of highly constrained problems at supercomputing scale with substantially fewer problem evaluations. While motivated by and demonstrated in the context of computational neuroscience, the framework is general and applicable to constrained multi-objective optimization problems across scientific and engineering domains. --- ### GraphDiffMed: Knowledge-Constrained Differential Attention with Pharmacological Graph Priors for Medication Recommendation Authors: Krati Saxena, Tomohiro Shibata Published: 2026-03-21 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.20188v1 Recommending safe and effective medication combinations from electronic health records (EHRs) is a core clinical AI problem, yet it remains difficult because patient trajectories are long, noisy, and clinically heterogeneous. Existing methods typically excel at either temporal modeling across visits or pharmacological knowledge integration (e.g., drug-drug interactions, DDIs), but rarely achieve both while robustly suppressing noise. We present GraphDiffMed, a knowledge-constrained medication recommendation framework built on dual-scale Differential Attention v2. Differential attention is applied at both intra-visit and inter-visit levels to filter spurious signals within encounters and across longitudinal history, while pharmacological constraints are incorporated during learning. Experiments on MIMIC-III and ablation studies show that this design consistently improves recommendation quality and ranking over strong baselines while achieving a more favorable safety performance balance. We further find that the strongest-performing configuration uses only demographic auxiliary features under our experimental setting. Overall, GraphDiffMed demonstrates that combining noise-aware attention with pharmacological constraints yields more reliable and clinically meaningful medication recommendation. We open-source our code at https://github.com/saxenakrati09/GraphDiffMed. --- ### Exploring Data Augmentation and Resampling Strategies for Transformer-Based Models to Address Class Imbalance in AI Scoring of Scientific Explanations in NGSS Classroom Authors: Prudence Djagba, Kevin Haudek, Clare G. C. Franovic, Leonora Kaldaras Published: 2026-03-21 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.19754v1 Automated scoring of students' scientific explanations offers the potential for immediate, accurate feedback, yet class imbalance in rubric categories particularly those capturing advanced reasoning remains a challenge. This study investigates augmentation strategies to improve transformer-based text classification of student responses to a physical science assessment based on an NGSS-aligned learning progression. The dataset consists of 1,466 high school responses scored on 11 binary-coded analytic categories. This rubric identifies six important components including scientific ideas needed for a complete explanation along with five common incomplete or inaccurate ideas. Using SciBERT as a baseline, we applied fine-tuning and test these augmentation strategies: (1) GPT-4--generated synthetic responses, (2) EASE, a word-level extraction and filtering approach, and (3) ALP (Augmentation using Lexicalized Probabilistic context-free grammar) phrase-level extraction. While fine-tuning SciBERT improved recall over baseline, augmentation substantially enhanced performance, with GPT data boosting both precision and recall, and ALP achieving perfect precision, recall, and F1 scores across most severe imbalanced categories (5,6,7 and 9). Across all rubric categories EASE augmentation substantially increased alignment with human scoring for both scientific ideas (Categories 1--6) and inaccurate ideas (Categories 7--11). We compared different augmentation strategies to a traditional oversampling method (SMOTE) in an effort to avoid overfitting and retain novice-level data critical for learning progression alignment. Findings demonstrate that targeted augmentation can address severe imbalance while preserving conceptual coverage, offering a scalable solution for automated learning progression-aligned scoring in science education. --- ### Clinical Note Bloat Reduction for Efficient LLM Use Authors: Jordan L. Cahoon, Chloe Stanwyck, Asad Aali, Rachel Madding, Emma Sun, Yixing Jiang, Renumathy Dhanasekaran, Emily Alsentzer Published: 2026-03-21 Categories: cs.CY, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2604.16364v1 Health systems are rapidly deploying large language models (LLMs) that use clinical notes for clinical decision support applications. However, modern documentation practices rely heavily on templates, copy--paste shortcuts, and auto-populated fields, producing extensive duplicated text (``note bloat'') that dilutes clinically meaningful signal and substantially increases the computational cost of LLM use. We introduce TRACE, a scalable preprocessing pipeline that removes note bloat by leveraging EHR attribution metadata to identify templated and copied content and applying frequency-based deduplication when metadata are unavailable. We evaluated TRACE across four real--world clinical cohorts spanning liver transplantation, obstetrics, and inpatient care (5.3 million notes) using blinded physician review and downstream modeling tasks. TRACE removed 47.3% of chart text while preserving performance for information extraction and clinical outcome prediction. At a large academic medical center, this reduction corresponds to an estimated $9.5 million annual decrease in LLM inference costs assuming one query per encounter. These findings show how underutilized EHR metadata can enable more scalable and cost-efficient deployment of LLM-based clinical systems. --- ### IWLV-Ramayana: A Sarga-Aligned Parallel Corpus of Valmiki's Ramayana Across Indian Languages Authors: Sumesh VP Published: 2026-03-21 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.13078v1 The Ramayana is among the most influential literary traditions of South and Southeast Asia, transmitted across numerous linguistic and cultural contexts over two millennia. Despite extensive scholarship on regional Ramayana traditions, computational resources enabling systematic cross-linguistic analysis remain limited. This paper introduces the IWLV Ramayana Corpus, a structured parallel corpus aligning Valmiki's Ramayana across multiple Indian languages at the level of the sarga (chapter). The corpus currently includes complete English and Malayalam layers, with Hindi, Tamil, Kannada, and Telugu layers in active production. The dataset is distributed in structured JSONL format with explicit provenance metadata, enabling applications in comparative literature, corpus linguistics, digital humanities, and multilingual natural language processing. To our knowledge, this is the first sarga-aligned multilingual parallel corpus of the Valmiki Ramayana with explicit provenance metadata and machine-readable format. --- ### Can Large Language Models Reliably Extract Physiology Index Values from Coronary Angiography Reports? Authors: Sofia Morgado, Filipa Valdeira, Niklas Sander, Diogo Ferreira, Marta Vilela, Miguel Menezes, Cláudia Soares Published: 2026-03-21 Categories: cs.CL arXiv: https://arxiv.org/abs/2604.13077v1 Coronary angiography (CAG) reports contain clinically relevant physiological measurements, yet this information is typically in the form of unstructured natural language, limiting its use in research. We investigate the use of Large Language Models (LLMs) to automatically extract these values, along with their anatomical locations, from Portuguese CAG reports. To our knowledge, this study is the first addressing physiology indexes extraction from a large (1342 reports) corpus of CAG reports, and one of the few focusing on CAG or Portuguese clinical text. We explore local privacy-preserving general-purpose and medical LLMs under different settings. Prompting strategies included zero-shot, few-shot, and few-shot prompting with implausible examples. In addition, we apply constrained generation and introduce a post-processing step based on RegEx. Given the sparsity of measurements, we propose a multi-stage evaluation framework separating format validity, value detection, and value correctness, while accounting for asymmetric clinical error costs. This study demonstrates the potential of LLMs in for extracting physiological indices from Portuguese CAG reports. Non-medical models performed similarly, the best results were obtained with Llama with a zero-shot prompting, while GPT-OSS demonstrated the highest robustness to changes in the prompts. While MedGemma demonstrated similar results to non-medical models, MedLlama's results were out-of-format in the unconstrained setting, and had a significant lower performance in the constrained one. Changes in the prompt techinique and adding a RegEx layer showed no significant improvement across models, while using constrained generation decreased performance, although having the benefit of allowing the usage of specific models that are not able to conform with the templates. --- ### Alignment midtraining for animals Authors: Jasmine Brazilek, Miles Tidmarsh Published: 2026-03-21 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.13076v4 We investigate the robustness of value alignment via midtraining with synthetic documents, using animal compassion as a value that is both important in its own right and orthogonal to existing alignment efforts. To evaluate compassionate reasoning, we develop and publicly release Animal Norms In Moral Assessment (ANIMA), a 26-question evaluation spanning 13 ethical dimensions, publicly available as a dataset and Inspect evaluation. On ANIMA, training with 3000 documents achieves 77% compared to 40% for instruction-tuning approaches, with generalization to human compassion and no degradation in standard safety benchmarks or capabilities. However, subsequent unrelated instruction-tuning degrades the intervention, with the advantage disappearing after 5000 samples. Our exploratory results suggest document-based value interventions may require explicit preservation strategies to remain effective through typical training pipelines. --- ### Alignment midtraining for animals Authors: Jasmine Brazilek, Miles Tidmarsh Published: 2026-03-21 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2604.13076v3 We investigate the robustness of value alignment via midtraining with synthetic documents, using animal compassion as a value that is both important in its own right and orthogonal to existing alignment efforts. To evaluate compassionate reasoning, we develop and publicly release Animal Norms In Moral Assessment (ANIMA), a 26-question evaluation spanning 13 ethical dimensions, publicly available as a dataset and Inspect evaluation. On ANIMA, training with 3000 documents achieves 77% compared to 40% for instruction-tuning approaches, with generalization to human compassion and no degradation in standard safety benchmarks or capabilities. However, subsequent unrelated instruction-tuning degrades the intervention, with the advantage disappearing after 5000 samples. Our exploratory results suggest document-based value interventions may require explicit preservation strategies to remain effective through typical training pipelines. --- ### TRACE: Traceroute-based Internet Route change Analysis with Ensemble Learning Authors: Raul Suzuki, Rodrigo Moreira, Pedro Henrique A. Damaso de Melo, Larissa F. Rodrigues Moreira, Flávio de Oliveira Silva Published: 2026-03-21 Categories: cs.NI, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2604.02361v1 Detecting Internet routing instability is a critical yet challenging task, particularly when relying solely on endpoint active measurements. This study introduces TRACE, a MachineLearning (ML)pipeline designed to identify route changes using only traceroute latency data, thereby ensuring independence from control plane information. We propose a robust feature engineering strategy that captures temporal dynamics using rolling statistics and aggregated context patterns. The architecture leverages a stacked ensemble of Gradient Boosted Decision Trees refined by a hyperparameter-optimized meta-learner. By strictly calibrating decision thresholds to address the inherent class imbalance of rare routing events, TRACE achieves a superior F1-score performance, significantly outperforming traditional baseline models and demonstrating strong effective ness in detecting routing changes on the Internet. --- ### Experimental Design for Missing Physics Authors: Arno Strouwen, Sebastián Micluţa-Câmpeanu Published: 2026-03-21 Categories: stat.ML, cs.LG, physics.comp-ph arXiv: https://arxiv.org/abs/2604.01231v1 For most process systems, knowledge of the model structure is incomplete. This missing physics must then be learned from experimental data. Recently, a combination of universal differential equations and symbolic regression has become a popular tool to discover these missing physics. Universal differential equations employ neural networks to represent missing parts of the model structure, and symbolic regression aims to make these neural networks interpretable. These machine learning techniques require high-quality data to successfully recover the true model structure. To gather such informative data, a sequential experimental design technique is developed which is based on optimally discriminating between the plausible model structures suggested by symbolic regression. This technique is then applied to discovering the missing physics of a bioreactor. --- ### Interpretable Battery Aging without Extra Tests via Neural-Assisted Physics-based Modelling Authors: Yuan Qiu, Wei Li, Wei Zhang, Yi Zhou, Fang Liu, Jianbiao Wang, Zhi Wei Seh Published: 2026-03-21 Categories: eess.SP, cs.LG arXiv: https://arxiv.org/abs/2604.01229v1 State of health (SoH) is widely used for battery management, but it is a single scalar and offers limited interpretability. Two batteries with similar SoH can exhibit very different degradation behaviors and the lack of interpretability hinders optimal battery operation. In this paper, we propose IBAM for interpretable battery aging modelling with a neural-assisted physics-based framework. IBAM outputs a 2-D aging fingerprint without extra diagnostic tests and uses only routine logs from the battery management system. The fingerprint offers great interpretability by capturing a battery's curve-wide polarization voltage loss and the tail loss near the end-of-discharge. IBAM first creates a physics-based battery model based on a fractional-order equivalent circuit model, and then extracts per-cycle fingerprints from the model using a two-stage least-squares method. IBAM further anchors fingerprints on the SoH axis with physics-guided regression, where the per-cycle SoH is estimated via a bidirectional gated recurrent unit with customized multi-channel voltage features. Across batteries with short-, medium-, and long-lifespans, IBAM consistently yields the best physics model fidelity at different aging stages, and provides clear interpretations of degradation mechanisms and fingerprint patterns about batteries of different lifespans. The resulting fingerprints support interpretable battery health assessment and can inform battery control choices. --- ### Learning to Remember: Attentive Reinforcement Learning for Edge Serverless Autoscaling Authors: Faraz Shaikh, Gianluca Reali, Mauro Femminella Published: 2026-03-21 Categories: cs.DC, cs.LG arXiv: https://arxiv.org/abs/2603.28790v2 In edge computing, the stochastic and bursty nature of serverless workloads challenges autonomous resource orchestration. Traditional reactive controllers, such as the Kubernetes Horizontal Pod Autoscaler (HPA), suffer from reaction latency, leading to Service Level Objective (SLO) violations during traffic spikes and resource flapping during ramp-downs. While Deep Reinforcement Learning (DRL) offers a pathway toward proactive management, standard agents suffer from \textit{temporal blindness}, an inability to exploit the recent temporal context in non-Markovian edge environments. To bridge this gap, we propose a stability-aware autoscaling framework unifying short-horizon temporal context and control via an Attention-Enhanced Double-Stacked LSTM architecture integrated within a Proximal Policy Optimization (PPO) agent. Unlike shallow recurrent models, our approach employs a learned attention mechanism that weights recent historical states non-uniformly, suppressing high-frequency jitter while preserving the trend that precedes demand shifts. We validate the framework on two independent Kubernetes clusters using real-world Azure Functions traces. Against the single-layer LSTM ablation and the static HPA baseline, our approach reduces P90 latency by $\approx$67\%, and holds average latency within the 50ms hard SLO for 98.8\% of the run against 49.6\% and 43.5\% respectively. Against Kubernetes Event-Driven Autoscaling (KEDA), it matches latency performance at 75\% fewer replica-steps and 59\% less churn, with P90 hard-SLO violation bursts of at most 5 consecutive intervals against up to 24 for KEDA. These results indicate that mitigating temporal blindness through deep attentive memory improves the reliability and stability of Kubernetes autoscaling under bursty edge workloads. --- ### Mitigating Temporal Blindness in Kubernetes Autoscaling: An Attention-Double-LSTM Framework Authors: Faraz Shaikh, Gianluca Reali, Mauro Femminella Published: 2026-03-21 Categories: cs.DC, cs.LG arXiv: https://arxiv.org/abs/2603.28790v1 In the emerging landscape of edge computing, the stochastic and bursty nature of serverless workloads presents a critical challenge for autonomous resource orchestration. Traditional reactive controllers, such as the Kubernetes Horizontal Pod Autoscaler (HPA), suffer from inherent reaction latency, leading to Service Level Objective (SLO) violations during traffic spikes and resource flapping during ramp-downs. While Deep Reinforcement Learning (DRL) offers a pathway toward proactive management, standard agents suffer from temporal blindness, an inability to effectively capture long-term dependencies in non-Markovian edge environments. To bridge this gap, we propose a novel stability-aware autoscaling framework unifying workload forecasting and control via an Attention-Enhanced Double-Stacked LSTM architecture integrated within a Proximal Policy Optimization (PPO) agent. Unlike shallow recurrent models, our approach employs a deep temporal attention mechanism to selectively weight historical states, effectively filtering high-frequency noise while retaining critical precursors of demand shifts. We validate the framework on a heterogeneous cluster using real-world Azure Functions traces. Comparative analysis against industry-standard HPA, stateless Double DQN, and a single-layer LSTM ablation demonstrates that our approach reduces 90th percentile latency by approximately 29% while simultaneously decreasing replica churn by 39%, relative to the single-layer LSTM baseline. These results confirm that mitigating temporal blindness through deep attentive memory is a prerequisite for reliable, low-jitter autoscaling in production edge environments. --- ### Beyond Static Visual Tokens: Structured Sequential Visual Chain-of-Thought Reasoning Authors: Guangfu Guo, Xiaoqian Lu, Yue Feng, Mingming Sun Published: 2026-03-21 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.26737v2 Current multimodal LLMs encode images as static visual prefixes and rely on text-based reasoning, lacking goal-driven and adaptive visual access. Inspired by human visual perception-where attention is selectively and sequentially shifted from the most informative regions to secondary cues-we propose Structural Sequential Visual CoT SSV-CoT. First, a question-relevant saliency map identifies and organizes key visual regions, explicitly modeling the spatial distribution of visual importance. Second, reasoning is performed following this discriminative order, inducing a curriculum-like semantic progression from primary to secondary cues. This method is trained end-to-end, using text cot and answer supervision, without relying on region-level annotations or specialized external tools. Experiments on diverse visual reasoning benchmarks show gains, validating structured and sequential visual cognition. --- ### Beyond Static Visual Tokens: Structured Sequential Visual Chain-of-Thought Reasoning Authors: Guangfu Guo, Xiaoqian Lu, Yue Feng, Mingming Sun Published: 2026-03-21 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.26737v1 Current multimodal LLMs encode images as static visual prefixes and rely on text-based reasoning, lacking goal-driven and adaptive visual access. Inspired by human visual perception-where attention is selectively and sequentially shifted from the most informative regions to secondary cues-we propose Structural Sequential Visual CoT SSV-CoT. First, a question-relevant saliency map identifies and organizes key visual regions, explicitly modeling the spatial distribution of visual importance. Second, reasoning is performed following this discriminative order, inducing a curriculum-like semantic progression from primary to secondary cues. This method is trained end-to-end, using text cot and answer supervision, without relying on region-level annotations or specialized external tools. Experiments on diverse visual reasoning benchmarks show gains, validating structured and sequential visual cognition. --- ### Ordinal Semantic Segmentation Applied to Medical and Odontological Images Authors: Mariana Dória Prata Lima, Gilson Antonio Giraldi, Jaime S. Cardoso Published: 2026-03-21 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.26736v1 Semantic segmentation consists of assigning a semantic label to each pixel according to predefined classes. This process facilitates the understanding of object appearance and spatial relationships, playing an important role in the global interpretation of image content. Although modern deep learning approaches achieve high accuracy, they often ignore ordinal relationships among classes, which may encode important domain knowledge for scene interpretation. In this work, loss functions that incorporate ordinal relationships into deep neural networks are investigated to promote greater semantic consistency in semantic segmentation tasks. These loss functions are categorized as unimodal, quasi-unimodal, and spatial. Unimodal losses constrain the predicted probability distribution according to the class ordering, while quasi-unimodal losses relax this constraint by allowing small variations while preserving ordinal coherence. Spatial losses penalize semantic inconsistencies between neighboring pixels, encouraging smoother transitions in the image space. In particular, this study adapts loss functions originally proposed for ordinal classification to ordinal semantic segmentation. Among them, the Expanded Mean Squared Error (EXP_MSE), the Quasi-Unimodal Loss (QUL), and the spatial Contact Surface Loss using Signal Distance Function (CSSDF) are investigated. These approaches have shown promising results in medical imaging, improving robustness, generalization, and anatomical consistency. --- ### Distilled Large Language Model-Driven Dynamic Sparse Expert Activation Mechanism Authors: Qinghui Chen, Zekai Zhang, Zaigui Zhang, Kai Zhang, Dagang Li, Wenmin Wang, Jinglin Zhang, Cong Liu Published: 2026-03-21 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2603.26735v1 High inter-class similarity, extreme scale variation, and limited computational budgets hinder reliable visual recognition across diverse real-world data. Existing vision-centric and cross-modal approaches often rely on rigid fusion mechanisms and heavy annotation pipelines, leading to sub-optimal generalization. We propose the Distilled Large Language Model (LLM)-Driven Sparse Mixture-of-Experts (DS-MoE) framework, which integrates text-guided dynamic routing and lightweight multi-scale comprehension. The DS-MoE framework dynamically aligns textual semantics with defect-specific visual patterns through a sparse MoE architecture, where task-relevant experts are adaptively activated based on semantic relevance, resolving inter-class ambiguity. A lightweight MobileSAM encoder enables real-time inference while preserving multi-scale defect details. Extensive experiments on PCB, aluminum foil, and mold defect datasets demonstrate that our framework achieves superior performance compared to existing pure vision models. \textbf{DS-MoE} surpasses YOLOv8/YOLOX with gains of +13.9, +1.4, and +2.0 pp mAP@ 0.5:0.95 on BBMP, aluminum, and PCB, respectively, while also improving precision and recall. --- ### Implicit Turn-Wise Policy Optimization for Proactive User-LLM Interaction Authors: Haoyu Wang, Yuxin Chen, Liang Luo, Buyun Zhang, Ellie Dingqiao Wen, Pan Li Published: 2026-03-21 Categories: cs.LG arXiv: https://arxiv.org/abs/2603.23550v1 Multi-turn human-AI collaboration is fundamental to deploying interactive services such as adaptive tutoring, conversational recommendation, and professional consultation. However, optimizing these interactions via reinforcement learning is hindered by the sparsity of verifiable intermediate rewards and the high stochasticity of user responses. To address these challenges, we introduce Implicit Turn-wise Policy Optimization (ITPO). ITPO leverages an implicit process reward model to derive fine-grained, turn-wise process rewards from sparse outcome signals. Unlike volatile token-level rewards, these turn-level signals exhibit superior robustness and may utilize a normalization mechanism to further enhance training stability. We evaluate ITPO across three representative multi-turn collaborative tasks: math tutoring, document writing, and medical recommendation. Empirical results demonstrate that ITPO, when combined with PPO, GRPO, or RLOO, consistently achieves improved convergence than existing baselines. Elaborate trajectory analysis confirms that ITPO infers turn-wise preferences that are semantically aligned with human judgment. Code is publicly available at https://github.com/Graph-COM/ITPO. --- ### Errors in AI-Assisted Retrieval of Medical Literature: A Comparative Study Authors: Jenny Gao, Yongfeng Zhang, Mary L Disis, Lanjing Zhang Published: 2026-03-21 Categories: cs.IR, cs.LG, stat.AP, stat.ME arXiv: https://arxiv.org/abs/2603.22344v1 Large language models (LLMs) assisted literature retrieval may lead to erroneous references, but these errors have not been rigorously quantified. Therefore, we quantitatively assess errors in reference retrieval of widely used free-version LLM platforms and identify the factors associated with retrieval errors. We evaluated 2,000 references retrieved by 5 LLMs (Grok-2, ChatGPT GPT-4.1, Google Gemini Flash 2.5, Perplexity AI, and DeepSeek GPT-4) for 40 randomly-selected original articles (10 per journal) published Jan. 2024 to July 2025 from British Medical Journal (BMJ), Journal of the American Medical Association, and The New England Journal of Medicine (NEJM). Primary outcomes were a multimetric score ratio combining validity of digital object identifier, PubMed ID, Google-Scholar link, and relevance; and complete miss rate (proportion of references failing all applicable metrics). Multivariable regression was used to examine independent associations. LLM platforms completely failed to retrieve correct reference data 47.8% of the time. The average score ratio of the 5 LLM platforms was 0.29 (standard deviation, 0.35; range, 0-1.25), with a higher score ratio indicating a higher accuracy in retrieving relevant references and correct bibliographic data. The highest and lowest accuracies were achieved by Grok (0.57) and Genimi (0.11), respectively. Compared with BMJ, NEJM articles had lower score ratios and higher complete miss rates. Multivariable analysis shows LLM platforms and journals were independently associated with score ratios and complete miss rate, respectively. We show modest overall performance of LLMs and significant variability in retrieval accuracy across platforms and journals. LLM platforms and journals are associated with LLM's performance in retrieving medical literature. Bibliographic data should be carefully reviewed when using LLM-assisted literature retrieval. --- ### Cloud-Edge Collaborative Large Models for Robust Photovoltaic Power Forecasting Authors: Nan Qiao, Shuning Wang, Sijing Duan, Wenpeng Cui, Yuzhe Chen, Qingchen Yang, Xingyuan Hua, Ju Ren Published: 2026-03-21 Categories: cs.LG, cs.NI arXiv: https://arxiv.org/abs/2603.22343v2 Photovoltaic (PV) power forecasting in edge-enabled grids requires balancing forecasting accuracy, robustness under weather-driven distribution shifts, and strict latency constraints. Existing models work well under normal conditions but often struggle with rare ramp events and unexpected weather changes. Relying solely on cloud-based large models often leads to significant communication delays, which can hinder timely and efficient forecasting in practical grid environments. To address these issues, we propose a condition-adaptive cloud-edge collaborative framework *CAPE* for PV forecasting. *CAPE* consists of three main modules: a site-specific expert model for routine predictions, a lightweight edge-side model for enhanced local inference, and a cloud-based large retrieval model that provides relevant historical cases when needed. These modules are coordinated by a screening module that evaluates uncertainty, out-of-distribution risk, weather mutations, and model disagreement. Furthermore, we employ a Lyapunov-guided routing strategy to dynamically determine when to escalate inference to more powerful models under long-term system constraints. The final forecast is produced through adaptive fusion of the selected model outputs. Experiments on two real-world PV datasets demonstrate that *CAPE* achieves superior performance in terms of forecasting accuracy, robustness, routing quality, and system efficiency. --- ### Neutrino Oscillation Parameter Estimation Using Structured Hierarchical Transformers Authors: Giorgio Morales, Gregory Lehaut, Antonin Vacheret, Frederic Jurie, Jalal Fadili Published: 2026-03-21 Categories: hep-ph, cs.LG arXiv: https://arxiv.org/abs/2603.22342v1 Neutrino oscillations encode fundamental information about neutrino masses and mixing parameters, offering a unique window into physics beyond the Standard Model. Estimating these parameters from oscillation probability maps is, however, computationally challenging due to the maps' high dimensionality and nonlinear dependence on the underlying physics. Traditional inference methods, such as likelihood-based or Monte Carlo sampling approaches, require extensive simulations to explore the parameter space, creating major bottlenecks for large-scale analyses. In this work, we introduce a data-driven framework that reformulates atmospheric neutrino oscillation parameter inference as a supervised regression task over structured oscillation maps. We propose a hierarchical transformer architecture that explicitly models the two-dimensional structure of these maps, capturing angular dependencies at fixed energies and global correlations across the energy spectrum. To improve physical consistency, the model is trained using a surrogate simulation constraint that enforces agreement between the predicted parameters and the reconstructed oscillation patterns. Furthermore, we introduce a neural network-based uncertainty quantification mechanism that produces distribution-free prediction intervals with formal coverage guarantees. Experiments on simulated oscillation maps under Earth-matter conditions demonstrate that the proposed method is comparable to a Markov Chain Monte Carlo baseline in estimation accuracy, with substantial improvements in computational cost (around 240$\times$ fewer FLOPs and 33$\times$ faster in average processing time). Moreover, the conformally calibrated prediction intervals remain narrow while achieving the target nominal coverage of 90%, confirming both the reliability and efficiency of our approach. --- ### T-MAP: Red-Teaming LLM Agents with Trajectory-aware Evolutionary Search Authors: Hyomin Lee, Sangwoo Park, Yumin Choi, Sohyun An, Seanie Lee, Sung Ju Hwang Published: 2026-03-21 Categories: cs.CR, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2603.22341v2 While prior red-teaming efforts have focused on eliciting harmful text outputs from large language models (LLMs), such approaches fail to capture agent-specific vulnerabilities that emerge through multi-step tool execution, particularly in rapidly growing ecosystems such as the Model Context Protocol (MCP). To address this gap, we propose a trajectory-aware evolutionary search method, T-MAP, which leverages execution trajectories to guide the discovery of adversarial prompts. Our approach enables the automatic generation of attacks that not only bypass safety guardrails but also reliably realize harmful objectives through actual tool interactions. Empirical evaluations across diverse MCP environments demonstrate that T-MAP substantially outperforms baselines in attack realization rate (ARR) and remains effective against frontier models, including GPT-5.2, Gemini-3-Pro, Qwen3.5, and GLM-5, thereby revealing previously underexplored vulnerabilities in autonomous LLM agents. Code is available at https://github.com/pwnhyo/T-MAP. --- ### T-MAP: Red-Teaming LLM Agents with Trajectory-aware Evolutionary Search Authors: Hyomin Lee, Sangwoo Park, Yumin Choi, Sohyun An, Seanie Lee, Sung Ju Hwang Published: 2026-03-21 Categories: cs.CR, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2603.22341v1 While prior red-teaming efforts have focused on eliciting harmful text outputs from large language models (LLMs), such approaches fail to capture agent-specific vulnerabilities that emerge through multi-step tool execution, particularly in rapidly growing ecosystems such as the Model Context Protocol (MCP). To address this gap, we propose a trajectory-aware evolutionary search method, T-MAP, which leverages execution trajectories to guide the discovery of adversarial prompts. Our approach enables the automatic generation of attacks that not only bypass safety guardrails but also reliably realize harmful objectives through actual tool interactions. Empirical evaluations across diverse MCP environments demonstrate that T-MAP substantially outperforms baselines in attack realization rate (ARR) and remains effective against frontier models, including GPT-5.2, Gemini-3-Pro, Qwen3.5, and GLM-5, thereby revealing previously underexplored vulnerabilities in autonomous LLM agents. --- ### Graphs RAG at Scale: Beyond Retrieval-Augmented Generation With Labeled Property Graphs and Resource Description Framework for Complex and Unknown Search Spaces Authors: Manie Tadayon, Mayank Gupta Published: 2026-03-21 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2603.22340v1 Recent advances in Retrieval-Augmented Generation (RAG) have revolutionized knowledge-intensive tasks, yet traditional RAG methods struggle when the search space is unknown or when documents are semi-structured or structured. We introduce a novel end-to-end Graph RAG framework that leverages both Labeled Property Graph (LPG) and Resource Description Framework (RDF) architectures to overcome these limitations. Our approach enables dynamic document retrieval without the need to pre-specify the number of documents and eliminates inefficient reranking. We propose an innovative method for converting documents into RDF triplets using JSON key-value pairs, facilitating seamless integration of semi-structured data. Additionally, we present a text to Cypher framework for LPG, achieving over 90% accuracy in real-time translation of text queries to Cypher, enabling fast and reliable query generation suitable for online applications. Our empirical evaluation demonstrates that Graph RAG significantly outperforms traditional embedding-based RAG in accuracy, response quality, and reasoning, especially for complex, semi-structured tasks. These findings establish Graph RAG as a transformative solution for next-generation retrieval-augmented systems. --- ### Problems with Chinchilla Approach 2: Systematic Biases in IsoFLOP Parabola Fits Authors: Eric Czech, Zhiwei Xu, Yael Elmatad, Yixin Wang, William Held Published: 2026-03-21 Categories: cs.LG, cs.CL, stat.ML arXiv: https://arxiv.org/abs/2603.22339v3 Chinchilla Approach 2 is among the most widely used methods for fitting neural scaling laws. Its parabolic approximation introduces systematic biases in compute-optimal allocation estimates, even on noise-free synthetic data. Applied to published Llama 3 IsoFLOP data at open frontier compute scales, these biases imply a parameter underallocation corresponding to 6.5% of the $3.8\times10^{25}$ FLOP training budget and \$1.4M (90% CI: \$412K-\$2.9M) in unnecessary compute at 50% H100 MFU. Simulated multimodal model misallocations show even greater opportunity costs due to higher loss surface asymmetry. Three sources of this error are examined: IsoFLOP sampling grid width (Taylor approximation accuracy), uncentered IsoFLOP sampling, and loss surface asymmetry ($α\neq β$). Chinchilla Approach 3 largely eliminates these biases but is often regarded as less data-efficient, numerically unstable, prone to local minima, and harder to implement. Each concern is shown to be unfounded or addressable, especially when the partially linear structure of the objective is exploited via Variable Projection, enabling unbiased inference on all five loss surface parameters through a two-dimensional optimization that is well-conditioned, analytically differentiable, and amenable to dense, or even exhaustive, grid search. It may serve as a more convenient replacement for Approach 2 or a more scalable alternative for adaptations of Approach 3 to richer scaling law formulations. See https://github.com/Open-Athena/vpnls for details and https://openathena.ai/scaling-law-analysis for other results from this study. --- ### Causal Direct Preference Optimization for Distributionally Robust Generative Recommendation Authors: Chu Zhao, Enneng Yang, Jianzhe Zhao, Guibing Guo Published: 2026-03-21 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2603.22335v2 Direct Preference Optimization (DPO) guides large language models (LLMs) to generate recommendations aligned with user historical behavior distributions by minimizing preference alignment loss. However, our systematic empirical research and theoretical analysis reveal that DPO tends to amplify spurious correlations caused by environmental confounders during the alignment process, significantly undermining the generalization capability of LLM-based generative recommendation methods in out of distribution (OOD) scenarios. To mitigate this issue, we propose CausalDPO, an extension of DPO that incorporates a causal invariance learning mechanism. This method introduces a backdoor adjustment strategy during the preference alignment phase to eliminate interference from environmental confounders, explicitly models the latent environmental distribution using a soft clustering approach, and enhances robust consistency across diverse environments through invariance constraints. Theoretical analysis demonstrates that CausalDPO can effectively capture users stable preference structures across multiple environments, thereby improving the OOD generalization performance of LLM-based recommendation models. We conduct extensive experiments under four representative distribution shift settings to validate the effectiveness of CausalDPO, achieving an average performance improvement of 17.17% across four evaluation metrics. --- ### Cyber Deception for Mission Surveillance via Hypergame-Theoretic Deep Reinforcement Learning Authors: Zelin Wan, Jin-Hee Cho, Mu Zhu, Ahmed H. Anwar, Charles Kamhoua, Munindar P. Singh Published: 2026-03-21 Categories: cs.CR, cs.AI, cs.GT, cs.MA arXiv: https://arxiv.org/abs/2603.20981v1 Unmanned Aerial Vehicles (UAVs) are valuable for mission-critical systems like surveillance, rescue, or delivery. Not surprisingly, such systems attract cyberattacks, including Denial-of-Service (DoS) attacks to overwhelm the resources of mission drones (MDs). How can we defend UAV mission systems against DoS attacks? We adopt cyber deception as a defense strategy, in which honey drones (HDs) are proposed to bait and divert attacks. The attack and deceptive defense hinge upon radio signal strength: The attacker selects victim MDs based on their signals, and HDs attract the attacker from afar by emitting stronger signals, despite this reducing battery life. We formulate an optimization problem for the attacker and defender to identify their respective strategies for maximizing mission performance while minimizing energy consumption. To address this problem, we propose a novel approach, called HT-DRL. HT-DRL identifies optimal solutions without a long learning convergence time by taking the solutions of hypergame theory into the neural network of deep reinforcement learning. This achieves a systematic way to intelligently deceive attackers. We analyze the performance of diverse defense mechanisms under different attack strategies. Further, the HT-DRL-based HD approach outperforms existing non-HD counterparts up to two times better in mission performance while incurring low energy consumption. --- ### From Causal Discovery to Dynamic Causal Inference in Neural Time Series Authors: Dmitry Zaytsev, Valentina Kuskova, Michael Coppedge Published: 2026-03-21 Categories: cs.LG, cs.AI, stat.AP, stat.ML arXiv: https://arxiv.org/abs/2603.20980v3 Time-varying causal models provide a powerful framework for studying dynamic scientific systems, yet most existing approaches assume that the underlying causal network is known a priori - an assumption rarely satisfied in real-world domains where causal structure is uncertain, evolving, or only indirectly observable. This limits the applicability of dynamic causal inference in many scientific settings. We propose Dynamic Causal Network Autoregression (DCNAR), a two-stage neural causal modeling framework that integrates data-driven causal discovery with time-varying causal inference. In the first stage, a neural autoregressive causal discovery model learns a sparse directed causal network from multivariate time series. In the second stage, this learned structure is used as a structural prior for a time-varying neural network autoregression, enabling dynamic estimation of causal influence without requiring pre-specified network structure. We evaluate the scientific validity of DCNAR using behavioral diagnostics that assess causal necessity, temporal stability, and sensitivity to structural change, rather than predictive accuracy alone. Experiments on multi-country panel time-series data demonstrate that learned causal networks yield more stable and behaviorally meaningful dynamic causal inferences than coefficient-based or structure-free alternatives, even when forecasting performance is comparable. These results position DCNAR as a general framework for using AI as a scientific instrument for dynamic causal reasoning under structural uncertainty. --- ### Detection of adversarial intent in Human-AI teams using LLMs Authors: Abed K. Musaffar, Ambuj Singh, Francesco Bullo Published: 2026-03-21 Categories: cs.LG, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2603.20976v1 Large language models (LLMs) are increasingly deployed in human-AI teams as support agents for complex tasks such as information retrieval, programming, and decision-making assistance. While these agents' autonomy and contextual knowledge enables them to be useful, it also exposes them to a broad range of attacks, including data poisoning, prompt injection, and even prompt engineering. Through these attack vectors, malicious actors can manipulate an LLM agent to provide harmful information, potentially manipulating human agents to make harmful decisions. While prior work has focused on LLMs as attack targets or adversarial actors, this paper studies their potential role as defensive supervisors within mixed human-AI teams. Using a dataset consisting of multi-party conversations and decisions for a real human-AI team over a 25 round horizon, we formulate the problem of malicious behavior detection from interaction traces. We find that LLMs are capable of identifying malicious behavior in real-time, and without task-specific information, indicating the potential for task-agnostic defense. Moreover, we find that the malicious behavior of interest is not easily identified using simple heuristics, further suggesting the introduction of LLM defenders could render human teams more robust to certain classes of attack. --- ### DiscoUQ: Structured Disagreement Analysis for Uncertainty Quantification in LLM Agent Ensembles Authors: Bo Jiang Published: 2026-03-21 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2603.20975v1 Multi-agent LLM systems, where multiple prompted instances of a language model independently answer questions, are increasingly used for complex reasoning tasks. However, existing methods for quantifying the uncertainty of their collective outputs rely on shallow voting statistics that discard the rich semantic information in agents' reasoning. We introduce DiscoUQ, a framework that extracts and leverages the structure of inter-agent disagreement -- both linguistic properties (evidence overlap, argument strength, divergence depth) and embedding geometry (cluster distances, dispersion, cohesion) -- to produce well-calibrated confidence estimates. We propose three methods of increasing complexity: DiscoUQ-LLM (logistic regression on LLM-extracted structure features), DiscoUQ-Embed (logistic regression on embedding geometry), and DiscoUQ-Learn (a neural network combining all features). Evaluated on four diverse benchmarks (StrategyQA, MMLU, TruthfulQA, ARC-Challenge) with a 5-agent system using Qwen3.5-27B, DiscoUQ-LLM achieves an average AUROC of 0.802, outperforming the best baseline (LLM Aggregator, 0.791) while being substantially better calibrated (ECE 0.036 vs. 0.098). The learned features generalize across benchmarks with near-zero performance degradation and provide the largest improvements where they are most needed: in the ambiguous "weak disagreement" tier where simple vote counting fails. --- ### GraPHFormer: A Multimodal Graph Persistent Homology Transformer for the Analysis of Neuroscience Morphologies Authors: Uzair Shah, Marco Agus, Mahmoud Gamal, Mahmood Alzubaidi, Corrado Cali, Pierre J. Magistretti, Abdesselam Bouzerdoum, Mowafa Househ Published: 2026-03-21 Categories: cs.CV arXiv: https://arxiv.org/abs/2603.20970v1 Neuronal morphology encodes critical information about circuit function, development, and disease, yet current methods analyze topology or graph structure in isolation. We introduce GraPHFormer, a multimodal architecture that unifies these complementary views through CLIP-style contrastive learning. Our vision branch processes a novel three-channel persistence image encoding unweighted, persistence-weighted, and radius-weighted topological densities via DINOv2-ViT-S. In parallel, a TreeLSTM encoder captures geometric and radial attributes from skeleton graphs. Both project to a shared embedding space trained with symmetric InfoNCE loss, augmented by persistence-space transformations that preserve topological semantics. Evaluated on six benchmarks (BIL-6, ACT-4, JML-4, N7, M1-Cell, M1-REG) spanning self-supervised and supervised settings, GraPHFormer achieves state-of-the-art performance on five benchmarks, significantly outperforming topology-only, graph-only, and morphometrics baselines. We demonstrate practical utility by discriminating glial morphologies across cortical regions and species, and detecting signatures of developmental and degenerative processes. Code: https://github.com/Uzshah/GraPHFormer --- ### Understanding Contextual Recall in Transformers: How Finetuning Enables In-Context Reasoning over Pretraining Knowledge Authors: Bhavya Vasudeva, Puneesh Deora, Alberto Bietti, Vatsal Sharan, Christos Thrampoulidis Published: 2026-03-21 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2603.20969v1 Transformer-based language models excel at in-context learning (ICL), where they can adapt to new tasks based on contextual examples, without parameter updates. In a specific form of ICL, which we refer to as \textit{contextual recall}, models pretrained on open-ended text leverage pairwise examples to recall specific facts in novel prompt formats. We investigate whether contextual recall emerges from pretraining alone, what finetuning is required, and what mechanisms drive the necessary representations. For this, we introduce a controlled synthetic framework where pretraining sequences consist of subject-grammar-attribute tuples, with attribute types tied to grammar statistics. We demonstrate that while such pretraining successfully yields factual knowledge, it is insufficient for contextual recall: models fail to implicitly infer attribute types when the grammar statistics are removed in ICL prompts. However, we show that finetuning on tasks requiring implicit inference, distinct from the ICL evaluation, using a subset of subjects, triggers the emergence of contextual recall across all subjects. This transition is accompanied by the formation of low-dimensional latent encodings of the shared attribute type. For mechanistic insight, we derive a construction for an attention-only transformer that replicates the transition from factual to contextual recall, corroborated by empirical validation. --- ### Hard labels sampled from sparse targets mislead rotation invariant algorithms Authors: Avrajit Ghosh, Bin Yu, Manfred Warmuth, Peter Bartlett Published: 2026-03-21 Categories: stat.ML, cs.LG, math.ST arXiv: https://arxiv.org/abs/2603.20967v2 One of the most common machine learning setups is logistic regression. In many classification models, including neural networks, the final prediction is obtained by applying a logistic link function to a linear score. In binary logistic regression, the feedback can be either soft labels, corresponding to the true conditional probability of the data (as in distillation), or sampled hard labels (taking values $\pm 1$). We point out a fundamental problem that arises even in a particularly favorable setting, where the goal is to learn a noise-free soft target of the form $σ(\mathbf{x}^{\top}\mathbf{w}^{\star})$. In the over-constrained case (i.e. the number of samples $n$ exceeds the input dimension $d$) with examples $(\mathbf{x}_i,σ(\mathbf{x}_i^{\top}\mathbf{w}^{\star}))$, it is sufficient to recover $\mathbf{w}^{\star}$ and hence achieve the Bayes risk. However, we prove that when the examples are labeled by hard labels $y_i$ sampled from the same conditional distribution $σ(\mathbf{x}_i^{\top}\mathbf{w}^{\star})$ and $\mathbf{w}^{\star}$ is $s$-sparse, then rotation-invariant algorithms are provably suboptimal: they incur an excess risk $Ω\!\left(\frac{d-1}{n}\right)$, while there are simple non-rotation invariant algorithms with excess risk $O(\frac{s\log d}{n})$. The simplest rotation invariant algorithm is gradient descent on the logistic loss (with early stopping). A simple non-rotation-invariant algorithm for sparse targets that achieves the above upper bounds uses gradient descent on the weights $u_i,v_i$, where now the linear weight $w_i$ is reparameterized as $u_iv_i$. --- ### Learning to Aggregate Zero-Shot LLM Agents for Corporate Disclosure Classification Authors: Kemal Kirtac Published: 2026-03-21 Categories: q-fin.TR, cs.AI, cs.MA, q-fin.CP, q-fin.ST arXiv: https://arxiv.org/abs/2603.20965v2 This paper studies whether a lightweight supervised aggregator can combine diverse zero-shot large language model outputs into a stronger downstream signal for corporate disclosure classification. Zero-shot LLMs can read disclosures without task-specific fine-tuning, but their predictions often vary across prompt perspectives, model families, and confidence levels. I examine this problem with a multi-prompt framework in which three fixed zero-shot LLM classifiers read each disclosure from different financial perspectives and output a sentiment label, a confidence score, and a short rationale. A logistic meta-classifier then aggregates these outputs to predict next-day stock return direction. To reduce pretrained-model contamination, I restrict evaluation to a post-release sample of 9{,}860 U.S.\ corporate disclosures issued by large publicly traded firms between January 2025 and March 2026, after the release of the frozen base LLMs used in the experiment. Results show that the trained aggregator outperforms single classifiers, majority vote, confidence-weighted voting, a zero-shot LLM judge, and a FinBERT baseline. Balanced accuracy rises from 0.566 for the best single classifier to 0.606 for the trained aggregator. The gain is largest in mixed-signal disclosures where classifiers disagree. The results suggest that zero-shot LLM outputs contain complementary financial signals, while also showing that the strongest gains come from supervised aggregation rather than from zero-shot voting alone. --- ### Elite Lanes: Evolutionary Generation of Realistic Small-Scale Road Networks Authors: Artur Morys-Magiera, Marek Długosz, Paweł Skruch Published: 2026-03-21 Categories: cs.NE arXiv: https://arxiv.org/abs/2603.20964v1 We present a comparative study of methods for generating realistic, constrained small- to medium-scale road networks with built-in redundancy. In this research, we evaluate the proposed Evolutionary Algorithm (EA) with connectivity and redundancy constraints against the Wave Function Collapse (WFC) method - commonly used in procedural terrain generation for games - and swarm algorithms: Particle Swarm (PSO) and Gray Wolf (GWO). Our focus is on producing realistic, redundant road networks suitable for vision, localization and navigation problems. We evaluate metrics: connectivity, cycles, intersections, dead ends, graph cut-edges while enforcing physical plausibility. We propose an EA and its extended version with elitism via MAP-Elites method. We detail the implementation, constraints, metrics and provide both visual and quantitative comparisons with baselines. Results highlight how fitness function design choices affect the structural characteristics of generated networks and highlight the impact of specific constraints in practical applications. Our contribution is a method for creating realistic synthetic datasets from sparse tile definitions derived from real-world data. We demonstrate a practical application by generating realistic maps using a laboratory-collected tileset from a Duckietown city model. Our approach performs coherent geometric transformations on metadata, in this work exemplified by semantic segmentation masks of the generated road networks. --- ### Integrative Learning of Dynamically Evolving Multiplex Graphs and Nodal Attributes Using Neural Network Gaussian Processes with an Application to Dynamic Terrorism Graphs Authors: Jose Rodriguez-Acosta, Sharmistha Guha, Lekha Patel, Kurtis Shuler Published: 2026-03-21 Categories: stat.AP, stat.ME, stat.ML arXiv: https://arxiv.org/abs/2603.20962v1 Exploring the dynamic co-evolution of multiplex graphs and nodal attributes is a compelling question in criminal and terrorism networks. This article is motivated by the study of dynamically evolving interactions among prominent terrorist organizations, considering various organizational attributes like size, ideology, leadership, and operational capacity. Statistically principled integration of multiplex graphs with nodal attributes is significantly challenging due to the need to leverage shared information within and across layers, account for uncertainty in predicting unobserved links, and capture temporal evolution of node attributes. These difficulties increase when layers are partially observed, as in terrorism networks where connections are deliberately hidden to obscure key relationships. To address these challenges, we present a principled methodological framework to integrate the multiplex graph layers and nodal attributes. The approach employs time-varying stochastic latent factor models, leveraging shared latent factors to capture graph structure and its co-evolution with node attributes. Latent factors are modeled using Gaussian processes with an infinitely wide deep neural network-based covariance function, termed neural network Gaussian processes (NN-GP). The NN-GP framework on latent factors exploits the predictive power of Bayesian deep neural network architecture while propagating uncertainty for reliability. Simulation studies highlight superior performance of the proposed approach in achieving inferential objectives. The approach, termed as dynamic joint learner, enables predictive inference (with uncertainty) of diverse unobserved dynamic relationships among prominent terrorist organizations and their organization-specific attributes, as well as clustering behavior in terms of friend-and-foe relationships, which could be informative in counter-terrorism research. ---