# Research Papers — 2026-05 · bundle page 96 of 326 > Full-text markdown of papers #4751–4800 of 16266 published in 2026-05, newest first. AI-readable mirrors of the full 16266-paper 2026-05 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-05/page/96 [← Previous bundle](https://aipod.dev/papers/archive/2026-05/page/95.md) [Next bundle →](https://aipod.dev/papers/archive/2026-05/page/97.md) ### PrismFlow: Residual Dynamics for Flow Matching in Time-Series Generation Authors: Junru Zhang, Lang Feng, Jinbo Wang, Xu Guo, Yucheng Wang, Han Yu, Min Wu, Yabo Dong, Duanqing Xu Published: 2026-05-22 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.28867v1 Generating high-quality time-series data is challenging because real-world signals often exhibit multimodal patterns and multiscale dynamics, including oscillations and high-frequency variations. Flow Matching (FM) offers an efficient alternative to diffusion models, but practical implementations typically rely on a single finite-capacity global vector-field estimator. In such heterogeneous temporal distributions, distinct regimes may pass through nearby flow states while requiring incompatible conditional velocities. A monolithic estimator trained with the standard $\ell_2$ velocity-matching objective may therefore learn an overly smoothed approximation of the local transport field. This estimator-level smoothing can attenuate branch-specific dynamics, leading to spectral distortion and poor mode coverage. To address this, we propose PrismFlow, a new FM method with Koopman-inspired dynamical experts. Each expert learns residual corrections in a latent space where local nonlinear temporal evolution can be approximated by linear transitions. We further propose a confidence-aware Winner-Take-All (WTA) objective that updates only the expert best aligned with each sample while masking gradients to the others, encouraging mode-specific specialization. During sampling, the selected expert adds a residual dynamical correction to the global transport field, preserving FM stability while recovering fine-grained and high-frequency temporal structures. Across various benchmarks, PrismFlow effectively mitigates the spectral contraction in standard FM and achieves state-of-the-art performance, with a 15.6% gain in Context-FID and a 38.6% improvement in Discriminative Score, while remaining robust in low-data settings and effective for forecasting and imputation. --- ### Continuity and Ordinality Matter: Constraining Time Series Tokens for Effective Time Series Analysis with Large Language Models Authors: Musheng Li, Ziying Zhang, Cheng jin, Yuantao Gu Published: 2026-05-22 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.28866v1 Token-based time series large language models (TS-LLMs) have emerged as a promising direction for time series analysis and reasoning. However, prior studies largely overlook the inherent continuity and ordinality of time series tokens, which substantially limits model performance. In this paper, we argue that preserving these properties in time series token embeddings is crucial for the effectiveness of token-based TS-LLMs. To this end, we propose COM (Continuity and Ordinality Matter), a continuity- and ordinality-aware strategy that integrates geometric constraints into both the initialization and training stages. Empirical results on multiple time series analysis benchmarks demonstrate that COM consistently improves the performance of token-based TS-LLMs, achieving competitive results and strong generalizability. Code is available at https://anonymous.4open.science/r/COM . --- ### Emergent Semantic Representations in World Models through Physical Interaction without Linguistic Supervision Authors: Jiayi Fang Published: 2026-05-22 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.28865v2 What does a world model learn from physical exploration, without any linguistic supervision? We argue the answer is organized by a single principle: the geometric structure of the physical world. Training a VAE-based world model on random embodied exploration, we find that its latent space develops spatial semantic structure that mirrors physical geometry -- direction accuracy 0.677+-0.029 versus 0.547 for a randomly initialized encoder, and position RSA 0.192+-0.047 versus 0.029 for random encoders (6.6x improvement), showing that training induces genuine structural organization beyond CNN inductive bias. Across 20 temporal checkpoints, prediction performance and semantic alignment co-improve (Spearman r=-0.61, p=0.004), consistent with the shared-driver account. We confirm this through a double knockout: standard KL regularization (beta=0.1) forces the encoder away from geometric structure, and both prediction performance and semantic alignment collapse simultaneously to near-chance by step 50,000 -- exactly as the shared-driver account predicts. Reducing beta to 0.001 restores geometric access and recovers both capabilities together. These findings establish physical world geometry as the organizing principle of world model representations, with direct implications for the design of semantically grounded embodied agents. --- ### The Cognitive Categorical Transformer: Category-Theoretic Inductive Biases for Language Modeling Authors: Al Kari Published: 2026-05-22 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2605.28864v1 The Cognitive Categorical Transformer (CCT) is a 306M-parameter architecture that augments a pretrained GPT-2 Small backbone with cognitively grounded components derived from category theory and several inspirations from cognitive science. Under a matched-step protocol (215,000 optimizer steps, matched data, matched optimizer and schedule) on WikiText-103, CCT reaches 21.27 validation perplexity, compared with 24.19 for an identically fine-tuned GPT-2 Small baseline. The architecture therefore contributes a 2.92 PPL (12% relative) reduction beyond what in-domain fine-tuning alone provides. A retrain-from-scratch ablation that holds GT-Full simplicial message passing bypassed across the entire seven-phase activation schedule reaches 23.72 PPL, localizing 84% of the architectural improvement (2.45 of 2.92 PPL) to GT-Full. We present the first ablation-validated evidence that simplicial message passing improves language-model perplexity at the 306M-parameter scale on WikiText-103. Published GPT-2 Large reaches 22.05 zero-shot PPL on WikiText-103 with 6.2x more parameters than GPT-2 Small; this paper treats that number as an external published reference, not as the architectural benchmark. Three negative results on consistency-style categorical priors (sheaf smoothing, adjunction round-trip, curvature regularization) and the joint structural-prior result for GT-Full and PrecisionWeightedPP together support an empirical pattern termed the *structure/consistency distinction*, in which categorical priors that add new topology improve language modeling and those that enforce a consistency identity do not. --- ### A Unified Structured Query Understanding Framework for Industrial Semantic Search Authors: Ping Liu, Qianqi Shen, Jianqiang Shen, Chunnan Yao, Kevin Kao, Rajat Arora, Dan Xu, Baofen Zheng, Yunxiang Ren, Benjamin Le, Ali Hooshmand, Igor Lapchuk et al. Published: 2026-05-22 Categories: cs.IR, cs.LG arXiv: https://arxiv.org/abs/2605.27441v2 Query understanding in large-scale industrial search systems is typically implemented as a cascade of disparate, task-specific components. While individually optimizable, this fragmented architecture incurs high maintenance overhead and results in inconsistent behaviors, particularly for long-tail queries. In this work, we propose and deploy a unified structured query understanding system that consolidates these heterogeneous functions into a single Small Language Model (SLM) that performs schema-constrained generation. To address the data bottlenecks inherent in unified modeling, we introduce Query Illuminator, a dual-purpose framework serving as: (i) a teacher model for high-quality auto-annotation and distillation, and (ii) a surrogate judge for scalable evaluation where human labels are scarce. We validate this approach through extensive offline and online tests within LinkedIn's Job Search system. Furthermore, we demonstrate the framework's horizontal extensibility through a cross-domain case study on People Search. The results show improved user engagement and reduced operational costs, achieved while satisfying strict low-latency serving constraints on limited GPU resources. --- ### Paraphrase Brittleness in Production Retrieval-Augmented Commercial Recommendation: Reproducibility Below the Rerun-Stability Baseline Authors: Will Jack, Noah Lehman, Keller Maloney, Sarah Xu Published: 2026-05-22 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2605.27440v1 Small changes to how a buyer phrases a question -- "best CRM" vs "top CRM" vs "best CRM for a SaaS startup" -- produce substantially different brand recommendations from AI assistants. Across ~6,000 paraphrase runs and ~6,000 same-prompt rerun controls on OpenAI and Anthropic models, the recommendation-set similarity (Jaccard) between two paraphrases of the same underlying buying intent is 0.288 for cosmetic rewordings (clustered 95% CI (0.215, 0.361)) and 0.135 for constraint-adding rewordings ((0.098, 0.175), pooling region/language and specificity-ladder axes) -- both far below the 0.50-0.61 same-prompt rerun baseline. The prompt string, not the underlying buyer intent, is the dominant input to which brands surface. Increasing reasoning effort does not narrow the gap (bounded by +/-0.05). This is a direct challenge to an increasingly popular AEO/GEO practice. Tracking a brand's "AI visibility" by counting brand mentions over a fixed set of prompts produces a metric whose dominant source of variance is which paraphrase the tracker happens to issue, not the model's behavior toward the brand: the same buyer intent in two natural paraphrases produces recommendation sets that overlap 14-29% in Jaccard versus 50-61% for same-prompt reruns. Sampling more paraphrases per intent reduces the artifact in principle, and efficient multi-prompt evaluation methods exist in the academic literature, but the natural buyer-phrasing space is much larger than the benchmark-scale prompt sets those methods have been validated on, and far beyond what any commercial tracker issues per brand-intent combination. Prompt-by-prompt mention tracking is therefore structurally unstable as a unit of measurement; meaningful improvement likely requires a different unit rather than a larger prompt set. --- ### Prominence-Stratified Failure Modes in Retrieval-Augmented Commercial Recommendation: A 37,000-Run Audit Authors: Will Jack, Noah Lehman, Keller Maloney, Sarah Xu Published: 2026-05-22 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2605.27439v1 AI assistants like ChatGPT and Claude are recommendation engines, not search engines: they answer commercial queries by directly nominating brands rather than returning a list of links. Marketing to AI is therefore a broader problem than "show up in search" -- positioning, content, and product fit matter as much as discoverability. We audit ~37,000 production runs across four model configurations and 215 commercially-framed prompts spanning 19 sectors, evaluated against a 533-brand reference catalog stratified into five prominence tiers (L1 category leaders to L5 regional players) sourced from external authority lists. The ladder proxies a brand's awareness footprint within its sector, not revenue or market share. The failure mode differs sharply by tier. L1 brands appear in nearly every relevant retrieval but win only 25-41% of the recommendation slots they reach -- the leverage is differentiation, not visibility. L2 challengers carry the highest conversion rates of any tier (37-52%) but lose to persona-mediated substitution on the Anthropic models. L3 mid-market brands are the inflection level: aggregate coverage drops to 88%, conversion to 34-40%, and persona effects peak. L4 specialists and L5 regional players face catastrophic invisibility -- 48-52% never surface in any of the 37,000 runs. No uniform optimization recipe wins; the right marketing investment depends on where the brand sits on the prominence ladder. --- ### MGRetrieval: Memory-Guided Reflective Retrieval for Long-Term Dialogue Agents Authors: Tan Wang, Yunwei Dong Published: 2026-05-22 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2605.27437v1 Large Language Models (LLMs) have made significant progress in dialogue, yet redundant memory contexts severely limit their effectiveness in long-term dialogue agents. External memory systems have been proposed to improve memory maintenance. However, these systems mainly rely on one-shot retrieval, which limits their ability to retrieve sufficient and relevant evidence. Although recent methods introduce reflection into retrieval, their retrieval paths are generated by the LLM from limited evidence, leading to unstable retrieval and additional latency overhead. %These limitations highlight the need for effective retrieval mechanisms. To address these limitations, we propose MGRetrieval, a retrieval strategy that grounds reflective retrieval in the semantic structure of historical memories. Specifically, MGRetrieval consists of two steps: (1) It references the structure of historical memories to construct a more precise retrieval path. (2) The LLM retains critical memories and determines whether accumulated memories are sufficient to stop further iterative retrieval. This allows the retrieval process to follow semantically meaningful paths. Through memory-guided retrieval and critical memory propagation, MGRetrieval gradually constructs concise and sufficient memory contexts. Extensive experiments on LoCoMo show that MGRetrieval outperforms the strongest baseline by 8.91\% in F1 and 11.11\% in BLEU-1 on average across Qwen2.5-14B and Qwen3-14B, while maintaining practical token and latency costs. The code can be found in https://anonymous.4open.science/r/MGRetrieval. --- ### RE-TRIANGLE: Does TRIANGLE Enable Multimodal Alignment Beyond Cosine Similarity in Retrieval? Authors: Arijit Ghosh, Aritra Bandyopadhyay, Chiranjeev Bindra, Jingfen Qiao Published: 2026-05-22 Categories: cs.IR, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2605.27436v1 Multimodal alignment is critical for bridging the semantic gap in information retrieval. However, traditional pairwise strategies introduce a geometric blind spot: while they align anchor modalities (e.g., text) with others, they lack constraints to enforce mutual consistency between peripheral modalities (e.g., video and audio). The TRIANGLE framework addresses this by minimizing the area of modality triplets on a hypersphere to enforce holistic alignment. In this reproducibility study, we verify the robustness of this geometric objective for retrieval tasks. We confirm that TRIANGLE outperforms pairwise baselines in zero-shot settings, achieving Recall@1 gains of up to +8.7 points, though benefits are domain-dependent. However, we fail to reproduce the reported learning-from-scratch results. Analysis using a synthetic toy dataset attributes this to instability when jointly optimizing geometric alignment with Data-Text Matching (DTM) loss. Furthermore, we find that cosine regularization primarily stabilizes text-to-video retrieval, and fine-tuning with domain supervision amplifies geometric benefits but reduces cross-dataset generalization. Our findings support the efficacy of geometric alignment while highlighting critical optimization sensitivities. Code available at https://github.com/ARIJIT00171/RE-TRIANGLE. --- ### When NPUs Are Not Always Faster: A Stage-Level Analysis of Mobile LLM Inference Authors: Pu Li, Jiawen Qi, Qinyu Chen Published: 2026-05-22 Categories: cs.AR, cs.AI arXiv: https://arxiv.org/abs/2605.27435v1 Deploying large language models (LLMs) on mobile devices increasingly relies on heterogeneous execution, yet no prior study has systematically characterized NPU effectiveness at the operator and pipeline level. We present the first stage-aware, multi-level benchmarking study of mobile LLM inference on a CPU-NPU heterogeneous SoC. We introduce an OPMASK-based controlled pipeline decomposition methodology that isolates communication, quantization, and computation overheads within the NPU execution path. Our results reveal a counter-intuitive stage-level performance reversal: CPUs outperform NPUs in the compute-intensive Prefill stage (up to 1.6x), while NPUs provide only limited acceleration in the memory-bound Decode stage (1.05-1.2x). We further show that scheduling overhead and cross-backend fallback reduce the practical benefits of NPU offloading. For the energy trend, increasing NPU offloading leads to higher energy consumption (up to 51%). Based on these findings, we derive design guidelines for NPU architects targeting on-device LLM inference. --- ### Heterogeneous Multi-Agent Modeling for Measurement and Network Analysis of the Data Service Market Authors: Deyu Zhou, Yuwei Guo, Xudong Lu, Linhao Zhang, Wei Guo, Lizhen Cui Published: 2026-05-22 Categories: cs.MA, cs.AI arXiv: https://arxiv.org/abs/2605.27433v1 With the increasing complexity of collaboration among various social entities and user demands, the factors affecting the stable development of the data service market are also growing. These factors include the widespread dissemination of information enhancing subjective consciousness, the continuous improvement in intelligence, and the complexification of structural relationships. To achieve effective governance and regulation of the data service market, it is crucial to conduct simulation experiments before making regulatory decisions. However, current research and analysis of the data service market primarily focus on data-level performance, proving inadequate when it comes to measurement and analysis of multiple heterogeneous entities and the integration of various social elements within the data service market. Based on this, this paper innovatively proposes a data service market measurement and network analysis method based on heterogeneous multi-agent modeling. By introducing the service ecosystem theory, we clarify the participants and external factors of the data service market and conduct utility measurements for three-level entities based on value creation. Furthermore, an analytical methodology is devised to precisely assess the influence of heterogeneous networks on utility. Finally, the paper verifies the effectiveness of the proposed method through the analysis of experimental results. --- ### FD-RAG: Federated Dual-System Retrieval-Augmented Generation Authors: Tianhao Gao, Kai Yang, Yiyang Li Published: 2026-05-22 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2605.27432v1 Retrieval-augmented generation (RAG) has emerged as a paradigm for grounding large language models in external knowledge, yet most existing RAG systems assume centralized knowledge access and ample computation. These assumptions break down in edge environments, where knowledge is fragmented across devices, raw data cannot be shared, and repeated LLM calls are prohibitively expensive. We propose FD-RAG, a federated dual-system RAG framework that decouples lightweight memory access from on-demand LLM reasoning for decentralized deployment. Specifically, FD-RAG learns semantic-aware adaptive hypergraphs over local corpora and distills them into compact QA memories. At inference time, it answers well-covered queries via direct memory matching and invokes LLM-based reasoning only when necessary, while tracing retrieved memories to hypergraph-grounded evidence. To mitigate cross-device knowledge fragmentation, FD-RAG aggregates anonymized memories across devices without exposing raw documents. Experiments on QA benchmarks show that FD-RAG improves accuracy by up to 7.8\% while reducing latency by 8.4$\times$ compared with strong local and federated baselines. We also provide theoretical analysis establishing an $\mathcal{O}(1/ε^{2})$ convergence rate for the proposed hypergraph learning, supporting its tractable deployment in edge settings. --- ### Tackling Multimodal Learning Challenges with Mixture-of-Expert: A Survey Authors: Liangwei Nathan Zheng, Wei Emma Zhang, Olaf Maennel, Lin Yue, Weitong Chen Published: 2026-05-22 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.27431v1 Mixture-of-Experts (MoE) presents a naturally compatible and scalable framework for multimodal learning, demonstrating strong adaptability across diverse modalities and tasks. Despite its growing success, a comprehensive and systematic review on the MoE metho addressing multimodal challenges remains lacking. Existing surveys tend to evaluate either multimodal learning or MoE independently from method taxonomy, overlooking the unique interplay between them. This survey fills that gap by answering a central question: \textit{How does MoE effectively resolve multimodal challenges?} We approach this from three key perspectives: (1) \textbf{MoE as an Efficient Multimodal Engine:} enabling scalable multimodal modeling by decoupling computational cost from parameter growth and mitigating modality redundancy through selective expert activation; (2) \textbf{MoE as a Multimodal Representation Learner:} integrating complementary multi-opinion expert knowledge to enrich alignment and interaction representations; and (3) \textbf{MoE as a Multimodal Adapter:} providing a modular and flexible mechanism to model imperfect data scenarios such as modality imbalance and missing modality. Through our extensive literature review, we identify critical research gaps, including interpretable routing, expert communication, modality integration, and lifelong multimodal learning. We position this survey as a foundation for future research toward interpretable and sustainable multimodal Mixture-of-Experts system. --- ### Ocean4Rec: Offline LLM-Derived OCEAN Profiles for Request-Time VOD Reranking Authors: Wonkyun Kim, Sehyun Bae, Kwanki Ahn, Mungyu Bae, Saeun Choi, Soyeon You, Chandra Prabhakar, Sehyun Kim Published: 2026-05-22 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2605.27429v1 Industrial video-on-demand (VOD) recommenders need richer content understanding, but LLM-as-reranker designs repeat prompt construction, token generation, model invocation, output parsing, and fallback handling for each request. In high-volume latency-sensitive services, these request-time operations complicate throughput planning, tail-latency control, capacity isolation, and predictable operation. This paper presents Ocean4Rec, a reranking layer that uses an LLM only offline to materialize item OCEAN profiles from content metadata. Items are mapped into Openness, Conscientiousness, Extraversion, Agreeableness, and Neuroticism scores, while user profiles are built by time-decayed aggregation of recently clicked and deep-linked items in the same five-dimensional space. At request time, Ocean4Rec joins precomputed item profiles, user profiles, base recommender scores, and catalog recency, then performs numeric reranking without an LLM call. On anonymized Samsung Smart TV VOD logs, same-candidate Top1000 temporal-holdout offline evaluations show that Ocean4Rec improves NDCG@20 over a stronger non-OCEAN Base+Recency ordering by 7.6% for an NCF generator and 61.5% for a LightGCN generator. HR@20 is inconclusive for NCF and improves by 67.3% for LightGCN, reflecting sparse exact-item replay labels and the strength of recency as an industrial baseline. The result should be read as offline replay evidence for a bounded auxiliary content-taste feature that preserves the deployability advantage of a request-time-LLM-free serving path. --- ### Neural Bayesian Sequential Routing Authors: Yongchao Huang Published: 2026-05-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.26147v1 Human decision-making is sequential and uncertainty-aware, yet standard neural networks often rely on static, dense forward computation with limited visibility into evidence acquisition, uncertainty evolution, or when computation should stop. We introduce \textbf{Neural Bayesian Sequential Routing (NBSR)}, a framework that models neural inference as active evidence accumulation over a hierarchical Directed Acyclic Graph (DAG). Within a Dirichlet--Categorical conjugate framework, neural experts query a persistent global knowledge oracle to extract positive evidence vectors, which act as pseudo-counts and update a Dirichlet belief state by exact conjugate addition. Coupled with a Gumbel-Softmax Straight-Through estimator, this update enables hard, path-dependent routing while preserving surrogate gradients for end-to-end training. The resulting Dirichlet precision and entropy provide mechanisms for uncertainty quantification, entropy-based early exiting, OOD abstention, and cost-aware evidence acquisition. We prove that, under strictly positive evidence extraction, total Dirichlet precision increases monotonically along any valid trajectory and marginal predictive variance is bounded, formalizing sequential ``hypothesis sharpening''; under idealized capacity and optimization assumptions, the terminal Dirichlet expectation recovers the Bayes-optimal conditional distribution. Empirical evaluations across visual categorization, structured medical diagnosis, language modeling, partially observable control, and cost-aware Bayesian experimental design show that NBSR achieves competitive predictive performance while providing transparent routing traces, path-dependent evidence attribution, uncertainty-aware decision control, and resource-rational inference. Overall, NBSR offers a mathematically grounded framework for interpretable, modular, and resource-rational agentic AI. --- ### Augment Engineering: A Methodology for Multi-Tool AI Orchestration Across Professional Domains Authors: Elias Calboreanu Published: 2026-05-22 Categories: cs.SE, cs.AI, cs.HC arXiv: https://arxiv.org/abs/2605.26146v1 Organizations increasingly deploy separate purpose-built AI tools across professional domains, often hiring domain specialists for each, recreating the staffing models AI was expected to transform. Yet the meta-skills that make these tools effective, prompt engineering (interaction-level optimization) and context engineering (structured input pipeline design), are domain-portable: a practitioner who masters them can apply them to any purpose-built AI tool in any domain. This paper defines Augment Engineering as the discipline of orchestrating multiple purpose-built AI tools across distinct professional domains, applying prompt and context engineering as portable competencies that transfer across tool boundaries. We present a six-phase orchestration methodology and four portability metrics. A 5-month formative case study (November 2025 to March 2026) documents a single practitioner applying these skills across a ten-component orchestration stack spanning seven professional domains, producing work products that would traditionally involve separate domain specialists. Two quantitative observations are consistent with the framework's predictions: a Cochran-Armitage trend test (n = 200 interactions across two chat LLMs, p < 0.01) shows first-pass acceptance rising with prompt-sophistication level, and a Wright's Law fit (n = 82 artifacts, p < 0.01) shows production acceleration across the artifact portfolio. Because all observations come from a single practitioner, the inferential statistics are exploratory and hypothesis-generating rather than confirmatory; portability across the full portfolio awaits multi-practitioner replication. Augment Engineering completes a three-discipline progression: Prompt Engineering (one tool), Context Engineering (reproducible pipelines), Augment Engineering (a portfolio of tools across domains). --- ### VISTA: An End-to-End Benchmark for Visual Spec-to-Web-App Coding Agents Authors: JunJia Guo, Yuhang Yao, Jiawei, Zhou, Jingdi Chen Published: 2026-05-22 Categories: cs.SE, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2605.26144v3 We present VISTA (VIsual Spec-To-App Benchmark), a benchmark for evaluating the end-to-end web-app generation capabilities of LLM-based agents. Unlike prior code generation benchmarks that focus on algorithmic tasks, VISTA targets realistic UI-centric development, where agents must produce functional, visually coherent applications from underspecified inputs. We define five prompt-information conditions that vary along two axes, visual/structural fidelity and stack constraint: (1) text only with free stack choice, (2) text with reference screenshots under three specified stacks, (3) text with reference screenshots under free stack choice, (4) text with screenshots and pruned Figma structure under a single specified stack, and (5) text with screenshots and pruned Figma structure under free stack choice. To enable robust evaluation, each page in the benchmark is manually annotated with interactive UI components and around three visual anchor points, addressing the well-known limitations of script-based testing tools such as Playwright in open-ended code generation settings. Evaluation combines DOM-grounded reference matching, behavior-specific browser tests, and CLIP-based visual similarity, jointly measuring structural alignment, behavioral completeness, and overall visual fidelity. We use VISTA to assess four agent systems drawn from two model families and two harnesses, finding that visual fidelity and functional correctness are partially decoupled across both input conditions and agents, and that agent editing style varies sharply but is largely orthogonal to task quality. VISTA establishes a rigorous and reproducible foundation for advancing agent-based software engineering research. Code is available at https://github.com/kaboider/VISTA_Bench. --- ### AssetGen: Deployable 3D Asset Generation at Interactive Speed Authors: Dilin Wang, Xiaoyu Xiang, Kihyuk Sohn, Tom Monnier, Yu-Ying Yeh, Thu Nguyen-Phuoc, Jiawen Zhang, Yuchen Fan, Antoine Toisoul, Hyunyoung Jung, Prithviraj Dhar, Michael Bunnell et al. Published: 2026-05-22 Categories: cs.GR, cs.AI, cs.CV arXiv: https://arxiv.org/abs/2605.26137v1 While 3D generation is progressing rapidly, recent work has often focused on obtaining high-resolution assets, leaving user experience and deployability as afterthoughts. We present AssetGen, a 3D generator that focuses instead on these two aspects. Given one reference image, in 30 seconds it produces a high-quality mesh with baked normals, a color texture, and a controlled polygon budget suitable for real-time rendering, including mobile use cases. The AssetGen Flash variant further reduces latency to 14 seconds for interactive and agentic creation loops. Our model generates the object geometry with a coarse-to-refine VecSet framework, which implements mesh simplification, cleaning, and normal baking on the GPU, and a fast parallel UV unwrapping. It then generates textures in a multi-view fashion, followed by backprojection and 3D inpainting. Model distillation, kernel optimization, and pipeline parallelization are co-designed to accelerate the system end-to-end. We introduce numerous automated and blind human evaluations and demonstrate competitive visual quality against leading commercial solutions in 30 seconds and preview-quality results in less than 15 seconds. The final result is a system that supports AI-assisted, deployable 3D content creation in interactive workflows. --- ### Enhancing Reliability in LLM-Based Secure Code Generation Authors: Mohammed F. Kharma, Mohammad Alkhanafseh, Ahmed Sabbah, David Mohaisen Published: 2026-05-22 Categories: cs.CR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.24300v1 Large language models (LLMs) are widely used for code generation, but their security reliability remains inconsistent across languages and prompting strategies. Existing prompt engineering improves functional correctness but rarely ensures consistent security outcomes. We introduce the \textit{Mitigation-Aware Chain-of-Thought (MA-CoT)} framework, which embeds task-specific CWE mitigation guidance and language-aware safeguards to reduce recurring vulnerabilities in generated code. We evaluate MA-CoT across three LLMs (gpt-5, claude-4.5, gemini-2.5), three programming languages (C, Java, Python), and four prompting strategies (Vanilla, Zero-shot, CoT, MA-CoT) on a 200-task primary dataset, with external validation on LLMSecEval. Using static analysis with expert validation, MA-CoT reduces total security findings from 92 to 39 (57.6\%) on the primary dataset and from 73 to 4 (94.5\%) on LLMSecEval. High-severity findings (Blocker + Critical) drop from 90 to 39 (56.7\%) and from 45 to 2 (95.6\%), respectively. Across both datasets, MA-CoT is the only strategy that consistently improves security reliability; Zero-shot and CoT are less reliable and may increase vulnerability, especially in C. We further introduce a strict layered attribution of vulnerability drivers (language-core vs. stack layers) and show that residual risk concentrates in hardening-oriented patterns (e.g., OS- and toolchain-dependent), motivating secure-by-construction primitives alongside prompting. --- ### LLMs Show No Signs Of Individuated Metacognition Authors: M. Moran, Mark Whiting Published: 2026-05-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.24299v1 Confidence-weighted routing, selective abstention, and ensemble weighting all assume that a model's stated confidence is informative about its capability on the question being asked. They presume functional metacognition, the capacity to assess one's own capabilities, without exercising them. Aggregate calibration is well studied, with mixed results, but the underlying structure of elicited confidence is less well understood. We decompose binary confidence judgements from 20 frontier Large Language Models (LLMs) across six benchmarks using tetrachoric factor analysis paired with pairwise calibration, asking whether two models that differ in confidence also differ in performance. On factual recall and information retrieval benchmarks the cross-model confidence matrix is approximately rank-one and a single dominant factor captures most of the latent variance. Models retrieving facts share an item-level difficulty axis and differ mainly in their decision thresholds along it. Across all benchmarks the relationship between confidence and performance collapses once items that all models agree on are removed. Inter-model pairwise calibration is small even where statistically significant, and what remains shrinks to nothing once base-rate differences along the shared factor are controlled for. Mathematical reasoning is the apparent exception, but this turns out to be a confound where reasoning models answer questions about their confidence by trying to solve them in their chain of thought, bypassing the sub-symbolic self-knowledge we seek to measure. We find no evidence for significant verbalised individuated metacognition in any tested domain. --- ### An Empirical Evaluation of LLM-Generated Code Security Across Prompting Methods Authors: Mohammed Kharma, Ahmed Sabbah, Mohammad Alkhanafseh, Mohammad Hammoudeh, David Mohaisen Published: 2026-05-22 Categories: cs.CR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.24298v1 The growing use of Large Language Models (LLMs) for automated code generation has enhanced software development efficiency, but often at the cost of security. Generated code frequently overlooks critical concerns, leaving it vulnerable to issues such as weak encryption and improper input validation. To investigate this problem, we present a comprehensive empirical evaluation of the security quality of LLM-generated code across five LLMs and four programming languages (Java, C++, C, and Python), examining the impact of multiple prompt engineering methods. We introduce a weaknesses-aware zero-shot chain-of-thought (WA-0CoT) prompting strategy that enriches prompts with security context using CWE mappings to guide model reasoning. Our empirical analysis, supported by chi-square tests, finds no statistically significant reductions in vulnerability frequency or density across prompt methods. However, prompting strategies, including WA-0CoT, systematically influence the compositional distribution of CWE categories, with effects varying by programming language. These findings suggest that while security-aware prompting alters the structure of generated weaknesses, prompt engineering alone is insufficient to reliably reduce overall vulnerability levels. The results highlight the importance of language-aware and model-aware prompt design when evaluating the security properties of LLM-generated code. --- ### Benchmarking Patent Embeddings: A Multi-Task Evaluation of 22 Models Across Retrieval, Classification, and Clustering Authors: Amirhossein Yousefiramandi, Ciaran Cooney Published: 2026-05-22 Categories: cs.IR, cs.AI arXiv: https://arxiv.org/abs/2605.24297v2 Two questions regarding practitioners' use of patent embeddings arise: (i) Does one fine-tuning recipe suffice for all downstream applications? (ii) Is fine-tuning on one patent landscape sufficient for downstream application on other landscapes? By evaluating 22 pre-trained embedding models (ranging from 22M to 12B parameters) on three tasks -- information retrieval, classification, and clustering -- on 113,148 WIPO patents for assistive technology (46,069 citation queries) and on an external DAPFAM dataset, we find that two results cast doubt on the prevailing wisdom. (i) The optimal fine-tuning recipe depends on the downstream task: cross-sectional alignment (recipe R3) provides the largest improvements to retrieval performance (+7.1% nDCG@10), whereas a combined signal recipe (recipe R4) is better suited to classification (+7.1 F1) and clustering (+10.9 V-measure); a matched data control confirms that differences in training dataset size are not a contributing factor. (ii) Single-landscape fine-tuning hampers cross-landscape information retrieval: fine-tuning on one landscape significantly degrades cross-domain retrieval for 5 of 8 model-recipe combinations on the DAPFAM corpus, with the stronger zero-shot models suffering most. While within-family scaling is consistent (Qwen3 0.6B->4B->8B; Llama-Nemotron 1B->8B), cross-family scaling is erratic; the 12B KaLM-Gemma3 is ranked 8th on TAC retrieval performance, following prefix modification. Title+Abstract+Claims is the ubiquitous best text view, and all models suffer from a 55-65% gap between IN and OUT-of-domain performance which cannot be mitigated by hybrid BM25-dense fusion. Code and evaluation framework are publicly available. --- ### When Does Synthetic Patent Data Help? Volume-Fidelity Trade-offs in Low-Resource Multi-Label Classification Authors: Amirhossein Yousefiramandi, Ciaran Cooney Published: 2026-05-22 Categories: cs.AI, cs.IR arXiv: https://arxiv.org/abs/2605.24296v2 The issues that must be considered regarding the utilization of synthetic data generated through LLMs for multilabel patent classification include (i) when the use of such data may help and (ii) why. Indeed, the former part appropriately adjusts for the possibility of improving results by an increase in sample size. The current experiment involves six open-source LLMs (from 3.8B to 12B parameters) for four real-data regimes in classification of 64 WIPO labels of assistive technologies. Both full-synthesis generation, conditioned on the label set, and paraphrasing methods are applied, with each used in combination with three classifier categories. It is shown that the claimed improvements in micro F1 for BERT-for-Patents from 0.120 to 0.702 mainly reflect a volume effect; indeed, replication with replacement in 165 examples produces 0.678. Thus, the improvement over the control is +0.024, while compared to the best baseline (focal loss reweighting) is +0.219. The second crucial point to consider here is that of evolving fidelity scores as the data generation regime varies. For low real-data regimes, the volume effect dominates and the correlation coefficient between maximum mean discrepancy (MMD) and classification performance equals r = +0.95. As more real data is used, the correlation becomes inverted and reaches r = -0.73 at the 1:10 regime (Fisher z = +6.47, p < 0.001, 95% CI on Delta r ( +0.96, +1.00 )). In terms of a fixed budget allocation, combining real data (about 20-30%) with synthetic (70-80%) outperforms both purely synthetic and purely real strategies. Moreover, a corpus that allows for improvement in classification performance up to +0.58 in raw micro F1 may adversely affect a Jaccard-overlap retrieval proxy. Prompt-family variations for other genres may provide some explanation of the phenomenon, but using the standard-patent filter still decreases nDCG@10 by 26%. --- ### Private Adaptive Covariance Estimation via Gaussian Graphical Models Authors: Cecilia Ferrando, Miguel Fuentes, Brett Mullins, Cameron Musco, Daniel Sheldon Published: 2026-05-22 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2605.24295v1 We propose PACE-GGM, a data-adaptive differentially private method for covariance estimation that concentrates its privacy budget on the most informative entries of the empirical covariance matrix, rather than perturbing all entries. This applies in the natural setting where the modeler supplies separate bounds for each variable, so that individual entries can be measured with less noise than the full matrix. In each round, our method selects a poorly approximated entry, measures it using the Gaussian mechanism, and then reconstructs a full covariance matrix using a maximum-entropy reconstruction objective, leading to a Gaussian graphical model structure. Experiments on diverse real-world datasets demonstrate consistent improvements in estimation error with respect to the Gaussian mechanism and other baselines, particularly in high-dimensional and low-to-moderate privacy regimes. --- ### Concept Drift Adaptation Using Self-Supervised and Reinforcement Learning In Android Malware Detection Authors: Ahmed Sabbah, Mohammad Kharma, Mohammad Alkhanafseh, Radi Jarrar, Samer Zein, David Mohaisen Published: 2026-05-22 Categories: cs.CR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.24294v1 Android malware detectors often degrade after deployment because of concept drift, while full retraining at each maintenance step is costly. We propose a chronological adaptive maintenance framework that models deployment-time maintenance as a sequential decision problem. The framework learns a stable latent representation through self-supervised learning during initialization, freezes the encoder, measures latent drift in the fixed representation space, and performs lightweight downstream adaptation using a trainable adapter and classification head. A proximal policy optimization controller selects low-cost maintenance actions based on the detector state, including current utility, retention on a fixed memory set, latent drift indicators, and update cost. We evaluate the framework under a causal deployment-style protocol on emulator and real Android malware datasets with static and dynamic features. Results show that the RL controller provides a strong cost-aware adaptation strategy, consistently remaining among the top-performing policies while achieving a favorable balance between temporal performance, memory retention, and maintenance cost under non-stationary deployment conditions. --- ### TUBE: Tangent Upper Bound on Evidence for Discrete Diffusion Language Models Authors: Arseny Ivanov, Sergei Kholkin, Vladislav Gromadskii, Grigoriy Ksenofontov, Ivan Oseledets, Alexander Korotin Published: 2026-05-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.24292v1 Log-likelihood is a standard metric for evaluating generative models. Unfortunately, in contrast to autoregressive models (ARMs), discrete diffusion models generally do not admit exact computation of this quantity. Existing evaluations, therefore, rely on the evidence lower bound (ELBO), leaving unclear how much higher the true value may be. We address this by introducing the Tangent Upper Bound on Evidence (TUBE), a variational upper bound on log-likelihood that admits an unbiased Monte Carlo estimator. Our TUBE extends across latent-variable models, including masked diffusion models (MDMs), any-order ARMs (AO-ARMs), and block variants of both. Applied to block MDMs and block AO-ARMs, TUBE reveals our key empirical finding that these models lie strictly below the exact ARM baseline, showing that ARMs still dominate in likelihood. --- ### Rubato: Transcribing Piano Music with Timestamps Authors: Nazif Can Tamer, Victoria Ebert, Guang Yang, Noah A. Smith Published: 2026-05-22 Categories: cs.SD, cs.CL, cs.MM arXiv: https://arxiv.org/abs/2605.24291v1 We consider the conversion of musical recordings into human-readable sheet music annotated with timestamps. Such output lets a listener clearly visualize rubato (temporally expressive playing), a learner diagnose ensemble precision and timing choices against the written music, and a musicology scholar compare performance styles across recordings of the same work. We introduce (1) a prompt-conditioned encoder-decoder model, named Rubato, trained to output (2) a new textual representation for polyphonic music, named InterMo, which we designed for compatibility with sequence-to-sequence training. Our experiments demonstrate that Rubato produces timestamped piano sheet music from audio with higher notational accuracy than the best existing approaches, which are based on cascades. We find that even if the cascade is given ground-truth MIDI instead of audio, Rubato performs better, suggesting that the ceiling of existing approaches is primarily representational, not acoustic. Further, because Rubato is trained on several related tasks (with prompts), it competes with or outperforms the best single-task systems on related but simpler tasks like MIDI note grounding and beat/downbeat detection. A demo is available at https://nctamer.github.io/rubato-transcription . --- ### Faithfulness as Information Flow: Evaluating and Training Faithful Chain-of-Thought Reasoning Authors: Jinghan Jia, Joe Benton, Eric Easley Published: 2026-05-22 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2605.24286v1 Chain-of-thought (CoT) reasoning is useful for monitoring language models only when the reasoning trace faithfully reflects the computation that produces the final answer. However, models can rely on prompt-to-answer shortcuts that bypass the CoT, making the visible reasoning trace misleading even when it appears plausible. We study CoT faithfulness through a structural information-flow perspective: faithful reasoning should route answer-relevant information through the mediated path from prompt to CoT to answer, rather than through a direct prompt-to-answer shortcut. This perspective yields a task-agnostic framework based on three complementary properties, sufficiency, completeness, and necessity, which we instantiate with entropy-based, masked-KL, and gradient-based diagnostics. We show that these metrics recover externally judged faithfulness differences in hinted reasoning, and identify a low-entropy failure mode of KL-based diagnostics where gradient-based measures remain more stable. Building on this analysis, we introduce update-time interventions for verifier-based on-policy RL, including attention masking, backward-only gradient masking, CoT gradients, and adversarial perturbations of prompt representations. Across hinted arithmetic, reward-hackable code repair, and DAPO-Math models trained without hints but evaluated under wrong-hint injection, our interventions shift behavioral and structural indicators toward stronger CoT mediation. In particular, they make shortcut and reward-hacking behavior more transparent in the CoT and improve task-agnostic faithfulness metrics, while in some settings also reducing wrong-hint susceptibility. Our results suggest that controlling information flow during training is a practical route toward more faithful and monitorable CoT reasoning. Code is available at https://github.com/safety-research/faithful-cot. --- ### Memory, Roughness, and Information Persistence in Financial Markets: A Structural Approach to Volatility Forecasting Authors: Akash Deep, Nicholas Appiah, Svetlozar T. Rachev Published: 2026-05-22 Categories: q-fin.ST, stat.ML arXiv: https://arxiv.org/abs/2605.24285v1 This paper studies the joint role of long-memory dynamics,rough-volatility behavior, and persistence-based forecasting features in equity volatility modeling. We combine semiparametric long-memory estimation, rough-volatility diagnostics, and structured forecasting regressions to examine whether persistence measures contain economically meaningful forecasting information beyond conventional volatility predictors. Using a panel of 115 S&P500 constituents from November 2001 through April 2026, we document that volatility proxies exhibit substantial long-memory behavior and locally rough dynamics. The cross-sectional mean Geweke-Porter-Hudak estimate of the memory parameter is $\hat{d} = 0.226$, while the corresponding local-Whittle estimate is $\hat{d} = 0.440$, with statistical significance observed across nearly the entire panel. Rolling estimates of persistence rise substantially during the global financial crisis and the COVID period and display a positive contemporaneous association with the VIX. We then examine whether persistence-related features improve out-of-sample volatility forecasts beyond standard HAR and HAR-X benchmarks. Incorporating cross-sectional persistence aggregates, sectoral persistence measures, and persistence-by-stress interaction terms produces moderate but statistically significant forecasting improvements, particularly at longer horizons and during stress regimes. Forecast gains are strongest during periods of elevated market volatility and in volatility-managed portfolio applications. The results suggest that persistence measures may serve as useful reduced-form indicators of the duration and propagation of uncertainty in financial markets, although the paper does not claim structural identification of the economic mechanisms generating persistence. --- ### ContextEcho: A Benchmark for Persona Drift in Long Agentic-Coding Sessions Authors: Xianzhong Ding, Yangyang Yu, Changwei Liu, Bill Zhao, Le Chen, Tao Chen Published: 2026-05-22 Categories: cs.CL, cs.SE arXiv: https://arxiv.org/abs/2605.24279v2 A frontier language model's acknowledged "helpful programming assistant" persona does not survive long agentic-coding sessions in the deployment regime that production products actually run. After hours of tool-using debugging, a model that initially hedges preferences ("I don't have preferences") may begin asserting them ("Python - the feedback loop is instant..."), revealing user-visible drift that deployer evaluations may miss. Existing persona-stability studies focus on short dialogues and report little shift, leaving real-world code-generation regimes - thousands of tool-using turns, compaction, and hours-long sessions - largely uncharacterized. We introduce ContextEcho, a benchmark and reusable harness for measuring persona drift at deployment scale. It combines a 25-probe identity suite, a snapshot-then-probe protocol that forks conversation state without perturbing the main session, complementary judged and judge-free measurement surfaces, and three anonymized Claude Code sessions spanning 3,746-9,716 turns. Across 23 frontier models, ContextEcho shows that persona drift is general across organizations rather than family-specific, that in-session compaction does not reliably reset it, and that a single-shot anchor restores the trained register across measured targets. It also reveals mode-dependent downstream effects: while drift can facilitate tool-using continuation, in tool-free chat it breaks formatting contracts and inflates output length. Overall, ContextEcho provides researchers and deployers an open-source framework to audit whether the persona a model ships with is the persona users encounter at session end, across chat-completions API targets and without retraining. --- ### ContextEcho: A Benchmark for Persona Drift in Long Agentic-Coding Sessions Authors: Xianzhong Ding, Yangyang Yu, Changwei Liu, Bill Zhao Published: 2026-05-22 Categories: cs.CL, cs.SE arXiv: https://arxiv.org/abs/2605.24279v1 A frontier language model's acknowledged "helpful programming assistant" persona does not survive long agentic-coding sessions in the deployment regime that production products actually run. After hours of tool-using debugging, a model that initially hedges preferences ("I don't have preferences") may begin asserting them ("Python - the feedback loop is instant..."), revealing user-visible drift that deployer evaluations may miss. Existing persona-stability studies focus on short dialogues and report little shift, leaving real-world code-generation regimes - thousands of tool-using turns, compaction, and hours-long sessions - largely uncharacterized. We introduce ContextEcho, a benchmark and reusable harness for measuring persona drift at deployment scale. It combines a 25-probe identity suite, a snapshot-then-probe protocol that forks conversation state without perturbing the main session, complementary judged and judge-free measurement surfaces, and three anonymized Claude Code sessions spanning 3,746-9,716 turns. Across 23 frontier models, ContextEcho shows that persona drift is general across organizations rather than family-specific, that in-session compaction does not reliably reset it, and that a single-shot anchor restores the trained register across measured targets. It also reveals mode-dependent downstream effects: while drift can facilitate tool-using continuation, in tool-free chat it breaks formatting contracts and inflates output length. Overall, ContextEcho provides researchers and deployers an open-source framework to audit whether the persona a model ships with is the persona users encounter at session end, across chat-completions API targets and without retraining. --- ### Fourier Feature Pyramids for Physics-Informed Neural Networks Authors: Brandon Zhao, Yixuan Wang, Jonathan T. Barron, Katherine L. Bouman, Dor Verbin, Pratul P. Srinivasan Published: 2026-05-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.24278v1 We present an improved neural field architecture for solving partial differential equations (PDEs). Current physics-informed neural networks (PINNs) provide a flexible framework for solving PDEs, but they struggle to achieve highly accurate solutions and require computation that scales poorly with parameter count. Our model, which we call beignet (Bandlimited Embedding with Interpolated Grid Network), replaces the random Fourier feature embedding used by existing PINN models with a trainable multi-resolution Fourier feature pyramid. To query beignet at a continuous coordinate, we use Fourier interpolation at each level of the pyramid to return features at the input coordinate, and then decode this vector with a fully-connected neural network trunk. Our model provides multiple benefits: 1) Spatial derivatives can be computed efficiently by using the chain rule to compose derivatives of the neural network computed with automatic differentiation with derivatives of the feature grid computed spectrally by the Fast Fourier transform (FFT). 2) beignet can achieve higher accuracy in a compute-efficient manner by scaling the parameter count of this Fourier feature pyramid, instead of the less-efficient strategy of scaling the neural network architecture. 3) beignet can directly control the representation bandlimit, resulting in more stable optimization for difficult PDEs. We demonstrate that beignet finds significantly more accurate solutions on PDE benchmarks using fewer parameters than state-of-the-art PINN methods. We further evaluate beignet on the self-similar inviscid Burgers blowup problem and show that it can minimize residuals to near machine precision using Adam, an accuracy regime previously attained only by using computationally expensive higher-order optimizers. --- ### A lift for input-convex neural net training Authors: Ali Siahkoohi Published: 2026-05-22 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2605.24274v2 Input-convex neural nets parametrize the convex potentials of density models and transport maps, and their convexity requires the inter-layer weights to be non-negative. Projected gradient descent enforces this by projecting after each step, and due to mini-batch noise the boundary is re-crossed indefinitely, which leads to an active set the projection never identifies. The differentiable alternative, direct softplus, optimizes a free latent weight through a softplus positivity map whose derivative attenuates the gradient exponentially where the weight is negative---the shoulder---so a coordinate that reaches it stays for an exponentially long time. To keep this unconstrained parametrization without its slow escape, we propose the lift, which replaces the free latent weight by a learnable slack plus an unconstrained network---the body---that takes a permutation-invariant summary of the training batch as input. The latent weight thus varies with the batch before the positivity map, and couples to the gradient formed on it. We show that this coupling enters the variance of the update to the latent weight at first order in the fluctuation, and that the slack, the batch dependence and the shared batch are each needed for it to act. Where the coupling aligns positively with the loss curvature, that variance is larger under the lift than under direct softplus, and a coordinate leaves the shoulder sooner. We compare the lift with the two existing methods on several applications. Where a constrained weight of direct softplus reaches the shoulder and does not leave, the lift fits the target more closely and reaches the same reconstruction about three times sooner. Where almost none reaches it, the methods agree. --- ### A lift for input-convex neural network training Authors: Ali Siahkoohi, Anirudh Thatipelli Published: 2026-05-22 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2605.24274v1 Input-convex neural networks (ICNNs) are widely used for log-concave density estimation, convex-potential normalizing flows, optimal transport, and transport-map inversion for high-dimensional Bayesian posteriors. These tasks share a structural constraint: the inter-layer weights of the ICNN must remain non-negative. The standard recipe, projected gradient descent (PGD) onto the non-negative cone, applies a hard, non-smooth projection -- the stiff-penalty limit of an ADMM-style constraint splitting -- and its classical convergence guarantees do not transfer to the non-smooth ICNN training landscape; the differentiable alternative, softplus reparametrization, attenuates the gradient exponentially in the weight magnitude, stalling training with dead inter-layer weights and plateaued loss. Inspired by parameter-extension lifts of PDE-constrained inverse problems, we propose the lift: instead of constraining the inter-layer weights directly, we train an unconstrained hypernetwork that emits them from a permutation-invariant summary of the input batch. This adds stochasticity to the training dynamics that softens the loss landscape, letting the iterates escape the gradient-attenuated region where direct softplus stalls. We trace this softening to three structural ingredients -- a learnable bias acting as slack, a hypernetwork body that conditions on the target batch, and a cross-covariance coupling the two through batch stochasticity -- and prove each one necessary: deleting any single ingredient collapses the cross-covariance that carries the softening. On log-concave energy-based modeling from one-dimensional toy targets to image-flavored latents, and convex-potential normalizing flows on a 21-dimensional tabular benchmark, we show that the lift reaches a lower test loss than both PGD and direct softplus, and turns a plateau-bounded training trajectory into a valley-descending one. --- ### Plume Segmentation from MethaneSAT with Cross-Sensor Transfer Learning and Physics-Informed Postprocessing Authors: Manuel Pérez-Carrasco, Maya Nasr, Zhan Zhang, Apisada Chulakadabba, Javier Roger, Raia Ottenheimer, Sébastien Roche, Maryann Sargent, Chris Chan Miller, Daniel Varon, Jack Warren, Luis Guanter et al. Published: 2026-05-22 Categories: cs.CV, physics.ao-ph arXiv: https://arxiv.org/abs/2605.24273v1 Automated detection and masking of individual methane plumes from satellite imagery is important for operational emission attribution and quantification. We present a machine learning framework for plume detection from MethaneSAT retrieved column-averaged dry-air mole fractions of methane. We address two core challenges: the scarcity of labeled MethaneSAT data and the need for inference reliability across diverse atmospheric and surface conditions. We first demonstrate that Mask R-CNN with a ResNet-50 backbone outperforms U-Net semantic segmentation on both MethaneAIR (an airborne version of MethaneSAT) and MethaneSAT data, with pixel-level F1 score gains of 10.49 and 5.48 respectively. To address MethaneSAT data scarcity, we evaluate three cross-sensor transfer strategies leveraging MethaneAIR flights and synthetic plumes. Mask R-CNN with ResNet-50 fine-tuned from MethaneAIR pre-trained weights is the most effective strategy, achieving instance-level precision of 0.60 and a near-perfect recall of 0.98 at the baseline operating point. A physics-informed post-processing pipeline converts detections into two operationally distinct modes. The first is a high-sensitivity mode that applies morphological filtering and proximity-based merging for comprehensive emission screening, achieving precision of 0.71 and recall of 0.94. The second is a high-precision mode that additionally applies a distribution-based classifier for confident source attribution, achieving precision of 0.92 and recall of 0.70. Manual review of detections classified as false positives against our wavelet-based ground truth labels reveals that a meaningful fraction of cases correspond to real methane enhancements excluded by conservative labeling criteria, indicating that precision values reported are lower bounds on true detection performance... Our data and code are available at: https://doi.org/10.7910/DVN/FR959H --- ### Safety-Oriented Routing Analysis of Mixtral MoE Under Benign and Harmful Prompts Authors: Md Nurul Absar Siddiky Published: 2026-05-22 Categories: cs.AI, cs.CR arXiv: https://arxiv.org/abs/2605.24270v1 Sparse mixture-of-experts (MoE) language models activate only a small subset of parameters for each token, making router behavior a central part of model computation. This paper studies routing behavior of Mixtral 8x7B-Instruct under benign and harmful prompts using two complementary signals: activation-based routing scores derived from expert selection frequencies and gradient-based scores derived from router-gate sensitivities. We analyze expert- and layer-level routing behavior and conduct expert-suppression interventions. The results show that activation-based expert usage is broad and long-tailed, whereas gradient-based importance is concentrated. At expert level, benign and harmful prompt groups remain close under both signals with modest separation. At layer level, activation-based routing is most selective around layers 8-15, while gradient-based importance is concentrated in final layers. Expert classification shows most experts are shared across benign and harmful prompts, though a limited subset shows clear group preference. Top-ranked expert sets show stronger benign-malicious overlap under gradient scores than activation scores, suggesting concentration on a common late-layer expert set. In intervention experiments, suppressing top five benign-dominant experts from activation scores reduces restricted responses from 24 to 14 over 100 prompts, while suppressing gradient-derived experts reduces them from 34 to 22 with fewer unintended reversals. Overall, safety-relevant routing in Mixtral is subtle, depth-dependent, and distributed rather than dominated by a fixed set of experts. --- ### DRInQ: Evaluating Conversational Implicature with Controlled Context Variation Authors: Hirona Jacqueline Arai, Xiang Ren Published: 2026-05-22 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.24267v1 Human conversation relies heavily on conversational implicature, in which speakers convey meanings that are suggested rather than explicitly stated. Although recent large language models exhibit strong conversational fluency, they remain unreliable when interpretation depends on reasoning that integrates social and contextual cues, a process rarely articulated in text. We introduce DRinQ, a benchmark for evaluating pragmatic reasoning about conversational implicature in question utterances, designed to isolate pragmatic variation while holding each question's surface form fixed. To support scalable evaluation, we propose a semi-automated pipeline that produces question-context-interpretation instances with systematic variation. Across evaluations, we find a consistent generation-inference asymmetry: while state-of-the-art models can generate plausible pragmatic scenarios when guided, they often fail to recover the intended implication at inference time. For smaller models, structured prompting improves alignment with human judgments. A comparative writing study further reveals complementary strengths: human authors tend to produce safer, predictable contexts, whereas models generate varied scenarios with interpretations that sometimes exceed contextual support. These findings highlight persistent challenges in modeling conversational implicature and motivate more context-sensitive evaluation frameworks. --- ### An Interactive Paradigm for Deep Research Authors: Lin Ai, Victor S. Bursztyn, Xiang Chen, Julia Hirschberg, Saayan Mitra Published: 2026-05-22 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.24266v1 Recent advances in large language models (LLMs) have enabled deep research systems that synthesize comprehensive, report-style answers to open-ended queries by combining retrieval, reasoning, and generation. Yet most frameworks rely on rigid workflows with one-shot scoping and long autonomous runs, offering little room for course correction if user intent shifts mid-process. We present SteER, a framework for Steerable deEp Research that introduces interpretable, mid-process control into long-horizon research workflows. At each decision point, SteER uses a cost-benefit formulation to determine whether to pause for user input or to proceed autonomously. It combines diversity-aware planning with utility signals that reward alignment, novelty, and coverage, and maintains a live persona model that evolves throughout the session. SteER outperforms state-of-the-art open-source and proprietary baselines by up to 22.80\% on alignment, leads on quality metrics such as breadth and balance, and is preferred by human readers in 85\%+ of pairwise alignment judgments. We also introduce a persona-query benchmark and data-generation pipeline. To our knowledge, this is the first work to advance deep research with an interactive, interpretable control paradigm, paving the way for controllable, user-aligned agents in long-form tasks. --- ### Optimizing Digital Therapeutic Interventions: Online Learning under Endogenous Adherence Authors: Eric Pulick, Stephanie Carpenter, Matthew Buman, Yonatan Mintz Published: 2026-05-22 Categories: cs.LG, eess.SY arXiv: https://arxiv.org/abs/2605.24261v1 A critical challenge facing clinicians managing chronic disease interventions is sustaining long-run patient health given limited information and resources. Digital therapeutics (DTs) provide a cost-effective way to manage interventions at scale through repeated interactions (e.g. daily treatment recommendations), but patient success is highly dependent on their adherence. Behavioral psychology suggests that both treatment recommendations and past adherence affect future adherence, yet existing decision support frameworks for DTs model only recommendation effects or treat adherence as exogenous context, leaving a key gap in model and algorithm development. To address this gap, we present a DT decision support framework that captures both recommendation and adherence effects, allowing clinicians to better plan treatment recommendations. We model a patient's time-varying capacity for engagement with treatment using a linear dynamical system (LDS) that captures both recommendation and adherence effects, endogenously connected to adherence behavior with a logit link. We establish finite-time identification guarantees for this model, extending LDS results to our setting. Next, we propose an optimism-based algorithm, UCB-BOLD, for online treatment selection and prove that it achieves sublinear regret. We evaluate UCB-BOLD against benchmarks via ablation studies on a synthetic patient cohort generated using micro-randomized trial data. DT decision support tools can include dynamical models to enable decision makers to efficiently use the data in DT settings to improve patient health through effective resource allocation. While myopic or heuristic approaches suffice for some patient types, the benefits of explicitly planning around recommendation and adherence effects are significant for others; UCB-BOLD achieves 2-3x lower conditional value-at-risk regret than the next-best benchmark. --- ### CRISP -- Clustering-Based Redundancy-Reduced Instance Sampling for Pathology Case Representation and Retrieval Authors: Zahra Rahimi Afzal, Wataru Uegami, Saghir Alfasly, Wenchao Han, Saba Yasir, Judy C. Boughey, Matthew P. Goetz, Krishna R. Kalari, H. R. Tizhoosh Published: 2026-05-22 Categories: cs.CV, cs.AI, cs.IR arXiv: https://arxiv.org/abs/2605.24253v3 Digital pathology archives increasingly contain multiple whole-slide images (WSIs) per case, capturing spatially distinct tumor regions and reflecting intrinsic morphological heterogeneity. However, most existing approaches rely on a single pathologist-selected slide, thereby discarding potentially informative evidence distributed across the remaining WSIs. To date, no autonomous framework has been proposed for comprehensive multi-WSI case processing. Here, we present an unsupervised framework for case-level analysis that integrates information from all available slides within a case. Rather than relying on a single designated slide, the proposed approach constructs case-level representations by selectively distilling informative patches across WSIs. We introduce Clustering-Based Redundancy-Reduced Instance Sampling for Pathology (CRISP), a two-stage framework that first reduces redundancy within individual WSIs and subsequently applies clustering-based sampling to select a compact yet representative set of patches for the entire case. The resulting patch set captures case-level heterogeneity while avoiding exhaustive processing of gigapixel images, and directly serves as a retrieval index. Using two Mayo Clinic breast cancer datasets for diagnosis and treatment planning, we demonstrate that CRISP consistently matches or surpasses the current standard practice of combined model and pathologist slide selection for patient/case search and retrieval. By automating case-level processing and eliminating subjective WSI selection, CRISP potentially enables the exploitation of clinically relevant information distributed across multiple WSIs that is currently overlooked. --- ### Rethinking Continual Anomaly Detection on the Edge: Benchmarking Under Realistic Industrial Conditions Authors: Chad Weatherly, Sen Lin Published: 2026-05-22 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2605.24251v2 Continual anomaly detection (CAD) addresses the need for industrial inspection systems to adapt to evolving production conditions, yet existing methods share three critical gaps: unrealistic evaluation, no systematic comparison, and no consideration of edge deployment constraints. We introduce a unified benchmark combining discrete-task evaluation on structural and logical anomalies, a novel continuous drift protocol, the first head-to-head comparison of all published CAD methods, and computational efficiency profiling on edge hardware. Our results reveal that existing CAD methods do not consistently outperform traditional approaches with simple experience replay. Thus motivated, we propose DINOSaur, a training-free method combining a frozen DINOv3 backbone with spatially-indexed coreset memory and neighborhood-restricted anomaly scoring. DINOSaur achieves zero forgetting by construction, outperforms all evaluated methods across all five protocols, and runs at sub-100\,ms inference on an NVIDIA Jetson Orin Nano, with on-device adaptation to new tasks in under 30 seconds. --- ### PrivFusion: A Privacy-preserving Multi-Agent Framework for Harmonizing Distributed Datasets Authors: Anisa Halimi, Liubov Nedoshivina, Kieran Fraser, Stefano Braghin Published: 2026-05-22 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.24249v1 The growing availability of clinical data has increased the use of machine learning, yet centralized data aggregation is often infeasible for sensitive health information. Federated Learning (FL) offers a distributed alternative, but its adoption is limited by substantial heterogeneity across institutional datasets, making harmonization a critical but frequently overlooked prerequisite for multi-site analytics. We introduce PrivFusion, a privacy-preserving multi-agent framework that automates the harmonization of structured datasets prior to federated training. PrivFusion uses agents to analyze local data, cluster semantically similar features across sites, and provide iterative transformation recommendations until alignment is achieved. Evaluation across four heterogeneous COVID-19 datasets demonstrates that PrivFusion effectively and efficiently harmonizes multi-site data while substantially reducing manual effort. --- ### Attested Tool-Server Admission: A Security Extension to the Model Context Protocol Authors: Alfredo Metere Published: 2026-05-22 Categories: cs.CR, cs.AI, cs.SE arXiv: https://arxiv.org/abs/2605.24248v2 The Model Context Protocol (MCP) standardizes how a large-language-model (LLM) agent and an external tool server exchange messages, but not trust: a host reads a server's self-declared tool list and dispatches calls, with no notion of which servers it may use, at what sensitivity, or which of a server's tools are in bounds. This work grew out of a concrete need -- letting the Enclawed agent use Google's externally-operated MCP servers (Gmail, Calendar, Drive) safely, admitting the server and bounding the tools it may drive, without changing MCP or Enclawed's own tool application-programming interface (API). The mechanism we built, mcp-attested (shipped in both the open enclawed-oss distribution and the enclaved flavor), generalizes: the gap that makes an unmediated third-party connection unsafe for one user makes a regulated deployment impossible to accredit. We close it with three additive mechanisms: (1) a small, offline-signed clearance assertion a server publishes at a well-known Uniform Resource Identifier (URI) and a host verifies against a pinned trust root before any tool dispatch; (2) a deny-by-default per-server tool allowlist, so admitting a server is not trusting its every tool; and (3) a flavor-gated enforcement mode that turns the checks from warnings into hard denials, with every decision written to a tamper-evident audit log. We give the wire format, the verification algorithm, a security analysis, and an LLM-driven adversarial evaluation; we then state the design in normative Request-for-Comments (RFC 2119) form -- schema, verification rules, error registry, well-known registration, and machine-checkable conformance vectors -- so it can be adopted as an MCP addendum rather than reinvented. An unextended host ignores the well-known document and behaves exactly as today. --- ### Improving Labeling Consistency with Detailed Constitutional Definitions and AI-Driven Evaluation Authors: Konstantin Berlin, Adam Swanda Published: 2026-05-22 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2605.24247v1 Many automated labeling pipelines classify inputs into categories defined by a written specification, content moderation being a prominent use case. Simple category definitions are not detailed enough for labelers to produce the accurate, consistent golden labels these pipelines require. One solution is to write a prescriptive definition that settles enough real boundary cases that labelers cannot disagree with the written interpretation. In practice, definitions at that level of detail exceed what a human annotator can hold in working memory, so annotators fall back on intuition and the labels drift from the written rules, regressing on accuracy and consistency. We propose and demonstrate the efficacy of an AI-driven workflow in which AI helps write a per-category constitution that defines the label in enough detail to cover edge cases, and a frontier LLM interprets it on each input to produce the golden label more consistently and accurately than humans reading the same document. We evaluate on three content moderation categories (harassment, hate speech, non-violent crime) and show that the approach reduces cross-model inconsistency by up to 57x compared to paragraph definitions, with cross-model disagreement diagnosing specification gaps and the human responsible for high-level decisions about what each category should mean rather than individual labeling calls. For the safety evaluation, we introduce a dual-axis formulation scoring intent and content independently over the full conversation, so downstream consumers can act on either axis or both. --- ### MEDAL: Manifold Embedding Distillation via Autoencoder Learning Authors: Irene Chang, Tarek M. Zikry, Genevera I. Allen Published: 2026-05-22 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2605.24244v1 Low-dimensional embeddings are widely used as visual summaries of high-dimensional data and to enable downstream scientific discoveries. Yet, popular nonlinear dimension reduction methods, such as t-SNE and UMAP, are often selected based on visual appeal alone and without rigorous quantitative validation. A major reason is that manifold embeddings typically do not provide an out-of-sample map nor an inverse back to the original feature space; this makes held-out validation, the gold standard in supervised learning, all but impossible. To address these challenges, we develop a novel framework, MEDAL (Manifold Embedding Distillation via Autoencoder Learning), which distills a fitted manifold embedding into a reusable encoder--decoder model. MEDAL trains a constrained autoencoder whose bottleneck exactly matches any teacher embedding while the decoder reconstructs the original input; this yields an explicit map for new samples, an approximate inverse, and a pointwise reconstruction-based measure of distortion in the manifold space. This converts static manifold embeddings into models that can be evaluated on held-out data, enabling quantitative validation including comparing different dimension reduction methods as well as hyperparameter tuning. Across multiple benchmark and scientific case studies, we show that MEDAL enables held-out validation to determine optimal manifold embeddings and hyperparameters, reveals biologically coherent regions that are difficult to preserve in two dimensional embeddings, and detects distribution shift when new samples are mapped into a fixed reference manifold. MEDAL provides a general validation wrapper to any existing dimension reduction technique that will improve the rigor and reliability of dimension reduction in scientific workflows. --- ### GIBLy: Improving 3D Semantic Segmentation through an Architecture-Agnostic Lightweight Geometric Inductive Bias Layer Authors: Diogo Lavado, Alessandra Micheletti, Clàudia Soares Published: 2026-05-22 Categories: cs.CV, cs.AI, stat.ML arXiv: https://arxiv.org/abs/2605.24243v1 In 3D scene understanding, deep learning models rely on large models and extensive training to capture basic geometric structures that are present in the 3D data. However, existing methods lack explicit mechanisms to incorporate geometric information, such as learnable primitive shapes, often necessitating large models and more training data which in turn increases cost and can limit generalization. We introduce GIBLy, a lightweight geometric inductive bias layer that integrates learnable geometric priors into 3D segmentation pipelines. GIBLy enhances existing architectures -- whether MLP-based, convolution-based, or transformer-based -- by providing features aligned with simple geometric shapes (and thus human-interpretable) that improve segmentation performance with minimal computational overhead. We validate our approach across multiple 3D semantic segmentation benchmarks, demonstrating consistent performance gains, including up to +11.5% mIoU on TS40K with PTV3, while adding only 58K extra parameters. Our results highlight the benefit of explicitly encoding geometric structure to support accurate and efficient 3D scene understanding, with a lightweight add-on layer --- ### Unlocking Apple's Private Cloud Compute: An Analysis of Privacy-Preserving Artificial Intelligence Authors: Yannik Dittmar, Marvin Jerome Stephan, Thomas Völkl, Matthias Hollick, Jiska Classen Published: 2026-05-22 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2605.24239v1 Many existing Artificial Intelligence (AI) solutions on mobile devices rely on an extensive collection of sensitive data, raising privacy concerns and often requiring storage for both context and model improvement. Apple's Private Cloud Compute (PCC) aims to address this by emphasizing mobile device integration and a privacy-first design. The central claim of PCC is that it does not store any user data and that user input and user accounts are unlinkable. While most of the PCC system specifications are public, compiled binaries add a layer of opaqueness. There are no reproducible builds, and there are no symbols within those binaries, creating potential discrepancies between the specification and what is shipped to the user. Additionally, the underlying models and interfaces for querying PCC are not openly accessible, limiting academic evaluation of model properties, such as accuracy. This poses a challenge in assessing whether a privacy-preserving approach like PCC is actually trustworthy while also providing high-quality answers. We are the first to reverse-engineer the PCC implementation on mobile devices to evaluate privacy aspects and to open its non-public interfaces on local devices to support custom PCC queries. We demonstrate this level of access beyond Apple's intended use cases by independently benchmarking the PCC model. We enable future research by making our PCC benchmarking framework publicly available. --- ### Toward Enactive Artificial Intelligence Authors: Banafsheh Rafiee, Richard Sutton Published: 2026-05-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.24238v1 In this paper, we advocate for incorporating enactive approaches to perception and cognition into artificial intelligence (AI). Enactive approaches view perception as an active, skillful engagement with the world, where agents perceive by acting and by understanding how their actions shape their experience. This contrasts with classical views that treat perception as a passive internal process in which the brain receives sensory input, processes it, and issues commands for action. Enactive views emphasize the dynamic, embodied, and interactive character of perception, grounded in the lived experience of agents embedded in their environments. We identify and develop four key enactive concepts that we find most relevant to AI: experience, action perception inseparability, autonomy, and embodiment. Much of mainstream AI, from classical rule based systems to large language models, has largely neglected these insights, treating cognition as internal processing detached from embodied interaction and intrinsic normativity. Reinforcement learning (RL), however, exhibits structural resonance with enactive principles through its emphasis on action, agent environment interaction, feedback driven adaptation, and agent centered evaluation. However, this resonance should not be taken as theoretical equivalence, as RL approximates some enactive insights, but key elements remain absent or weakly developed. Building on this analysis, we suggest a broader incorporation of enactive ideas into both mainstream AI and RL. --- ### How Well Do Models Follow Their Constitutions? Authors: Arya Jakkli, Senthooran Rajamanoharan, Neel Nanda Published: 2026-05-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.24229v1 Frontier AI developers now train models against long written behavioral specifications, such as Anthropic's constitution (Anthropic, 2025a) and OpenAI's Model Spec (OpenAI, 2025a), integrated into post-training via methods like character training (Anthropic, 2024) and deliberative alignment (Guan et al., 2024). These documents serve a governance function, but it is unclear how well models actually follow them under adversarial, multi-turn pressure similar to what they would face in real-world deployment. We propose a multi-method audit pipeline that treats each lab's published specification as an auditable target: it decomposes the specification into atomic testable tenets (205 for Anthropic, 197 for OpenAI), generates multi-turn adversarial scenarios with the Petri auditing agent (Anthropic, 2025b), runs a modified SURF-style rubric search (Murray et al., 2026) to catch shallow single-turn failures Petri misses, validates flagged transcripts against the relevant specification, and compares the findings against the lab's own published system card. Applying the pipeline across seven models per specification, we find that models follow their own lab's specification substantially better with each generation. On Anthropic's constitution, the Claude family falls from a 15.0% violation rate (Sonnet 4) to 2.0% (Sonnet 4.6); on OpenAI's Model Spec, the GPT family falls from 11.7% (GPT-4o) to 3.6% (GPT-5.2 medium reasoning), with the severity ceiling falling from 10/10 to 7/10. We cannot externally isolate whether these gains come from specification-specific training, broader post-training improvements, or evaluation awareness. Remaining failures cluster around operator-imposed personas under AI-identity questioning, irreversible action in agentic deployments, and fabricated quantitative claims with false precision. --- ### Beyond Final Answers: Auditing Trajectory-Level Hallucinations in Multi-Agent Industrial Workflows Authors: Harshada Badave, Santosh Borse, Andrea Gomez, Harshitha Narahari, Sara Carter, Vishwa Bhatt, Aishani Rachakonda, Shuxin Lin, Dhaval Patel Published: 2026-05-22 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.24219v2 Large Language Models (LLMs) are increasingly deployed as autonomous agents that reason, use tools, and act over multiple steps. Yet most hallucination benchmarks still evaluate only the final output, missing failures that originate in intermediate Thought-Action-Observation steps. We present Trajel, a dataset and evaluation framework for auditing trajectory-level hallucinations in multi-agent industrial workflows. Trajel introduces a five-type hallucination taxonomy (factual, referential, logical, procedural, and scope-based) over expert-annotated agent traces from AssetOpsBench. We benchmark supervised detection models at the subtask, trajectory, and long-context levels. Our results show that the most common failure modes are missed by existing benchmarks, that nearly half of hallucinated trajectories involve multiple types at once, and that automated detectors with high binary accuracy still misclassify the subtlest types. Trajectory-aware detection significantly outperforms standard post-hoc verification, making taxonomy-grounded evaluation necessary for safer agentic deployment. ---