arXiv:2605.29401v1 Announce Type: new
Abstract: Time-Series Foundation Models (TSFMs) excel at zero-shot unimodal forecasting using numerical data, but unlike LLMs they cannot consume multimodal, non-numerical context that often shape real-world trajectories. In this work, we bridge this gap and argue for a multimodal time-series forecasting approach that post-trains LLMs to act as context-guided revisors over strong numerical TSFM priors. We introduce PostTime, a post-training recipe combining Supervised Fine-Tuning (SFT) and Reinforcement Learning with Verifiable Rewards (RLVR), along with a methodology to generate automated reasoning traces for forecast revisions. PostTime teaches an LLM to generate context-conditioned forecast interventions -- decisions to revise, preserve, or ignore the TSFM prior based on the multimodal context. We evaluate this approach on the TimesX multimodal forecasting benchmark using a Gemma-3-4B LLM and TimesFM-2.5 TSFM, and show that it significantly outperforms standalone TSFMs, LLM-only baselines, and existing multimodal forecasting approaches.
Science Journals
arXiv:2601.04765v5 Announce Type: replace
Abstract: We study how syntactic and semantic information is encoded in inner layer representations of Large Language Models (LLMs), focusing on the very large DeepSeek-V3. We find that, by averaging hidden-representation vectors of sentences sharing syntactic structure or meaning, we obtain vectors that capture a significant proportion of the syntactic and semantic information contained in the representations. In particular, subtracting these syntactic and semantic ``centroids'' from sentence vectors strongly affects their similarity with syntactically and semantically matched sentences, respectively, suggesting that syntax and semantics are, at least partially, linearly encoded. We also find that the cross-layer encoding profiles of syntax and semantics are different, and that the two signals can to some extent be decoupled, suggesting differential encoding of these two types of linguistic information in LLM representations.
arXiv:2605.26255v2 Announce Type: replace-cross
Abstract: Early prediction of respiratory failure is critical for timely clinical intervention in intensive care units. Existing electronic health record (EHR)-based models can continuously monitor physiologic deterioration, but they may not fully capture pulmonary pathophysiology reflected in chest radiographs (CXRs). In this study, we ask whether CXR information improves prospective prediction of invasive mechanical ventilation beyond EHR signals alone. We develop a gated multimodal framework that integrates structured EHR time-series data with CXR foundation-model representations. The gating module adaptively controls the contribution of imaging features based on patient-specific clinical context, allowing the model to selectively rely on imaging information when it is informative. We prospectively evaluate the framework for predicting invasive mechanical ventilation within 24 hours in ICU patients and compare it with an established EHR-only model (Ventio), physician predictions obtained at matched clinical time points, and alternative multimodal variants. The gated multimodal models achieved higher discrimination than the EHR-only baseline, with AUROC values of 0.860 and 0.858 using REMEDIS and MedInsight CXR representations, respectively, compared with 0.752 for Ventio. Relative to physician predictions, the multimodal framework substantially improved sensitivity while maintaining favorable specificity. Compared with the EHR-only model, multimodal integration increased specificity and positive predictive value, suggesting that CXR information can refine risk estimation in selected patients. These findings support adaptive multimodal fusion as a practical strategy for incorporating imaging into prospective respiratory failure prediction.
arXiv:2605.29716v1 Announce Type: new
Abstract: Diffusion Large Language Models (dLLMs) have emerged as a promising non-autoregressive generative paradigm. Given the prohibitive computational cost of full fine-tuning, Parameter-Efficient Fine-Tuning (PEFT) has become the standard approach. However, existing PEFT methods (e.g., LoRA), originally tailored for autoregressive models, rely on static parameters that are agnostic to the noise level. Consequently, they ignore the intrinsic dynamics of the diffusion process, where input distributions and generation difficulty shift significantly along the denoising trajectory, rendering them suboptimal for dLLMs. To address this, we propose Noise-aware Low-Rank Adaptation (NaRA), which introduces a low-rank core matrix generated by a lightweight, globally shared hypernetwork conditioned on the noise level. This design enables the update matrices to vary continuously along the diffusion process while keeping parameter and latency overhead negligible. We provide a theoretical justification for the proposed NaRA framework and empirically demonstrate consistent improvements over noise-agnostic baselines across commonsense reasoning, mathematical reasoning, and code generation benchmarks. Our code is available at https://github.com/generaldi/NaRA.
arXiv:2605.28835v1 Announce Type: new
Abstract: Large Language Models (LLMs) extend their capabilities through function-calling (FC), which relies on training data with high quality, diversity, and broad coverage of scenario. However, obtaining and annotating real function-calling data is challenging, while synthetic data from existing pipelines often suffers from unreliable APIs, limited tool scalability, insufficient diversity, and weak quality control. To address these, we present GenesisFunc, an automated pipeline for generating FC training data. Starting from reliable tools in widely used public benchmarks, our GenesisFunc employs a multi-agent framework to support a dialogue generation system that produces conversations spanning diverse scenarios, while maintaining both diversity and quality throughout the process. The accuracy of the data is further reinforced through a multi-stage evaluation system. We fine-tune an 8B LLM on the synthetic dataset and show through extensive experiments that it outperforms similarly sized open-source models in in-domain FC performance and out-of-domain generalization, while reaching FC capabilities comparable to some of the latest API-based models. In addition, our method demonstrates strong potential to scale effectively across downstream tools, underscoring its real-world applicability.
arXiv:2602.09311v2 Announce Type: replace
Abstract: Automated regression testing is a cornerstone of modern software development, often contributing directly to code review and Continuous Integration (CI). Yet some tests suffer from flakiness, where their outcomes vary non-deterministically. Flakiness erodes developer trust in test results, wastes computational resources, and undermines CI reliability. While prior research has examined test flakiness within individual projects, its broader ecosystem-wide impact remains largely unexplored. In this paper, we present an empirical study of test flakiness in the OpenStack ecosystem, which focuses on (1) cross-project flakiness, where flaky tests impact multiple projects, and (2) inconsistent flakiness, where a test exhibits flakiness in some projects but remains stable in others. By analyzing 649 OpenStack projects, we identify 1,535 cross-project flaky tests and 1,105 inconsistently flaky tests. We find that cross-project flakiness affects 55% of OpenStack projects and significantly increases both review time and computational costs. Surprisingly, 70% of unit tests exhibit cross-project flakiness, challenging the assumption that unit tests are inherently insulated from issues that span modules like integration and system-level tests. Through qualitative analysis, we observe that race conditions in CI, inconsistent build configurations, and dependency mismatches are the primary causes of inconsistent flakiness. These findings underline the need for better coordination across complex ecosystems, standardized CI configurations, and improved test isolation strategies.
arXiv:2605.07210v2 Announce Type: replace
Abstract: This paper shows how diffusion language models (DLMs) can be used as effective and efficient retrievers. Existing DLM-based retrievers (e.g., DiffEmbed) follow BERT-style encoding, representing each query or passage as a single mean-pooled vector. This ignores how DLMs are trained to generate responses through masked-position prediction under bidirectional attention, a capability that can provide stronger retrieval signals. We propose DiffRetriever, which uses the DLM's native masked-position prediction directly for retrieval. For each query or passage, DiffRetriever appends one or more masked positions, using the outputs as retrieval representations in a single forward pass. With one masked position, single-representation DiffRetriever already improves over DiffEmbed on the same backbones. DiffRetriever also naturally extends to multi-representation retrieval: DLMs process multiple masked positions jointly, enabling ColBERT-style fine-grained matching with little additional encoding latency. In autoregressive LLM retrievers, the same multi-representation strategy requires sequential decoding and therefore incurs much higher latency. DiffRetriever obtains the strongest aggregate effectiveness within our matched comparison, outperforming DiffEmbed, PromptReps, and RepLLaMA. Masked-position counts selected on training data transfer well across datasets, while per-query variation suggests headroom for adaptive allocation. Code is available at https://github.com/ielab/diffretriever.
arXiv:2605.02288v2 Announce Type: replace
Abstract: Automated laboratories hold the promise of accelerating scientific discovery, yet their deployment is bottlenecked by the difficulty of designing safe and executable environments. While simulator-based design offers scalability, existing 3D scene generation methods are primarily tailored for household settings, optimizing for visual plausibility while neglecting the protocol grounding and layout-level safety constraints essential for scientific experimentation. We present LabBuilder, an end-to-end system that generates and verifies 3D laboratory layouts from concise textual specifications. It operates through three tightly coupled components: LabForge first curates a meta-dataset of annotated assets and chemical knowledge, translating natural language specifications into structured protocols; building on these protocols, LabGen synthesizes laboratory layouts via an iterative, constraint-aware optimization strategy; finally, LabTouchstone evaluates the resulting layouts as a unified benchmark. Extensive experiments demonstrate that LabBuilder significantly outperforms existing state-of-the-art methods, producing laboratory environments that are realistic and valid under modeled geometric, chemical-safety, and navigation constraints.
arXiv:2605.29741v1 Announce Type: new
Abstract: The dominance of colonial languages in African education and scientific communication limits how hundreds of millions of speakers of African languages access and produce scientific knowledge. A core obstacle is the lack of established scientific terminology in these languages. We introduce AfriScience-MT, a parallel corpus covering six African languages (Amharic, Hausa, Luganda, Northern Sotho, Yor\`ub\'a, and isiZulu) across 11 scientific domains. Professional translators, working with expert science communicators, translated plain-language summaries of scientific papers into each target language and created new terms where none existed. We benchmark machine translation systems and large language models in zero-shot, few-shot, and fine-tuned settings. Our results show that closed-source models outperform all open-source models at both the sentence and document levels: GPT-5.4 and Gemini-3.1-Flash-Lite lead with average sentence-level COMET scores of 68.3 and 68.0, respectively, and tie at an average document-level COMET of 48.3. Among open systems, fine-tuned NLLB-1.3B reaches 67.3 at the sentence level, and TranslateGemma-12B reaches 44.0 at the document level with 1-shot in-context learning. We release AfriScience-MT to support benchmarking and document-level scientific MT for African languages.
arXiv:2605.29872v1 Announce Type: cross
Abstract: QEM is widely regarded as a plausible bridge from NISQ devices to FTQC. Yet the empirical studies used to assess the effectiveness of QEM techniques on concrete problems have received comparatively little scrutiny with respect to the validity of their conclusions. We systematically review 81 recent QEM papers using an eight-criterion framework covering statistical rigour, reproducibility, and reporting quality. Among the applicable papers, only 15 (25%) use inferential methods, while 25 (42%) report uncertainty only descriptively, without testing whether the claimed effects are statistically supported. To demonstrate the consequences of these omissions, we use ZNE as a representative and widely used case study and identify two compounding sources of artefacts in current QEM benchmarks. First, we observe parameter sensitivity: in a 132-configuration sweep, implicitly assumed choices such as scale factors, extrapolation method, and hardware calibration are not merely incidental but active, with variations changing conclusions from statistically significant improvement to statistically significant degradation. Second, we identify a drift-induced effectiveness illusion: in a 72-hour longitudinal study on real hardware, temporal drift alone can make the same ZNE configuration exhibit an effect size more than three times as large, depending solely on when it is executed, and also drastically reduces the effective number of independent observations. These findings do not imply that QEM methods are intrinsically unsound; rather, they show that current evaluation practice can make mitigation performance appear more robust than the evidence warrants. We therefore propose minimum reporting standards for QEM evaluations, including explicit parameter documentation, robustness checks, longitudinal drift assessment, and inferential statistical testing with effect-size reporting.
arXiv:2604.24824v5 Announce Type: replace
Abstract: This article philosophically examines how shifts in assumptions regarding the existence and non-existence of the true target (TT) give rise to new perspectives and insights for machine learning (ML)-based predictive modeling and, correspondingly, proposes a knowledge system for evaluation and learning under Democratic Supervision. By systematically analysing the existence assumption of the TT in current mainstream ML paradigms, we explicitly adopt a negative ontology perspective, positing that the TT does not objectively exist in the real world, and, grounded in this non-existence assumption, define Democratic Supervision for ML. We further present Multiple Inaccurate True Targets (MIATTs) as an instance-level realization of Democratic Supervision. Building upon MIATTs, we derive principles, for the logic-driven generation and assessment of MIATTs, a logical assessment formulation for evaluation with MIATTs, and undefinable true target learning for learning with MIATTs. Based on these components, we establish the evaluation and learning with MIATTs (EL-MIATTs) framework for ML-based predictive modelling. A real-world application demonstrates the potential of the proposed EL-MIATTs framework in supporting education and professional development for individuals, aligning with prior discussions of Democratic Supervision in the fields of education and professional development.
arXiv:2601.23156v2 Announce Type: replace
Abstract: We consider the problem of unsupervised skill segmentation and hierarchical structure discovery in reinforcement learning. While recent approaches have sought to segment trajectories into reusable skills or options, most rely on action labels, rewards, or handcrafted annotations, limiting their applicability. We propose a method that segments unlabelled trajectories into skills and induces a hierarchical structure over them using a grammar-based approach. The resulting hierarchy captures both low-level behaviours and their composition into higher-level skills. We evaluate our approach in high-dimensional, pixel-based environments, including Craftax and the full, unmodified version of Minecraft. Using metrics for skill segmentation, reuse, and hierarchy quality, we find that our method consistently produces more structured and semantically meaningful hierarchies than existing baselines. Furthermore, as a proof of concept, we demonstrate that these discovered hierarchies accelerate and stabilise learning on downstream reinforcement learning tasks.
arXiv:2605.28914v1 Announce Type: new
Abstract: Tool-using language agents turn model decisions into external side effects: they read files, run scripts, call APIs, send messages, and invoke Model Context Protocol tools. This makes agent attacks different from jailbreaks. The harmful step is often not an obviously forbidden output, but an ordinary executable action that becomes unsafe because attacker-controlled context steers authorized access against the user's interest. We identify this failure mode as authority confusion: untrusted resources may inform reasoning, but they must not authorize side effects. We present AIRGuard, a runtime guard that operationalizes least privilege as action-time authorization. AIRGuard normalizes heterogeneous tool calls, derives task authority into step-level authority, tracks source and target trust, simulates sensitive side effects, audits cross-step risk, and enforces decisions before actions execute. On AgentTrap, AIRGuard reduces Sonnet 4.6 attack success from 36.3% without defense to 5.5%. On DTAP-150, AIRGuard preserves 76.0% benign utility with Haiku 4.5, compared with 52.0% for ARGUS and 42.0% for MELON. An ablation further shows that prompt-only policy helps only modestly, whereas a dedicated runtime authority-control layer gives the agent system direct control over tool-mediated side effects. Code and data are available at https://github.com/Sophie508/AIRGuard.
arXiv:2605.29768v1 Announce Type: new
Abstract: Existing traffic forecasting benchmarks assume a fixed sensor set, but real road-sensor networks grow continuously as the road network changes year by year. We introduce the XXLTraffic dataset family, which spans up to 27 years of California PeMS and Transport for NSW data. The fixed-sensor subsets of XXLTraffic support extremely long forecasting with multi-year gaps and standard hourly / daily long-horizon forecasting. We extend it to EvoXXLTraffic, a sensor-evolving reorganization that exposes per-year active sensors, yearly traffic-flow matrices, and yearly graph snapshots across nine PeMS districts, with growth ratios ranging from +305% to over +10,000%. We define a yearly streaming forecasting protocol on EvoXXLTraffic in which each calendar year is a continual task, and benchmark a wide range of representative baselines drawn from static spatio-temporal GNNs, na\"ive online schemes, evolving-graph continual methods, and retrieval / test-time methods. We find that our ultra-large evolutionary dataset better reflects the real world, and many state-of-the-art (SOTA) results no longer work. Our dataset complements existing benchmarks by enabling more realistic forecasting under ultra-long evolutionary road networks.
arXiv:2605.29115v1 Announce Type: new
Abstract: Unix competence is the ability to use shell and operating-system primitives as first-class tools, not merely to write programs through a terminal. Current terminal benchmarks tend to blur this distinction: a solver fluent in Python but weak in Unix can pass a substantial fraction of Terminal-Bench 2.0, while the reverse skill profile is rarely exercised. We make the distinction operational and build a training surface for the Unix component. unix-ctf is a procedural generator of capture-the-flag tasks for shell agents. Each task hides a short token (a flag of the form flag(a3b1c9...)) inside a fresh Linux container using a single Unix feature, and the agent must recover it. Tasks are produced by an LLM-assisted synthesis pipeline that generates candidate hiding techniques, rewrites them into parameterized hide-and-find script pairs, and filters them with a bidirectional contract: the hide script must leave no plaintext trace of the flag on disk, and the find script must recover the flag in a fresh directory. Because the LLM only writes the planting and recovery steps (the container, layout, and grading harness are fixed), the pipeline lands 656 of 750 raw attempts as portable, reusable variants (87.5\%). Our reproduction of Endless Terminals' full-container-generation approach lands only 17.4\% under the same checks. The 656 variants canonicalize to 155 distinct techniques. Fine-tuning Qwen3-8B with LoRA using GRPO on this surface lifts solve rate from 11.6\% to 43.6\% on a 15-skill multi-family holdout (n=225), redistributes which InterCode-CTF tasks the model solves, and produces a +33 pp gain in Forensics while reaching 32/100 on InterCode-CTF. These results suggest that Unix competence is separable, trainable, and best evaluated directly rather than folded into programming-through-a-shell.
arXiv:2605.29927v1 Announce Type: new
Abstract: Despite recent advances, LLM-based web agents still struggle with limited exploration, omission of critical steps, and sensitivity to task constraints. Prior work suggests that many of these failures stem from weaknesses in planning, yet the impact of alternative natural language plan representation remains unexplored. To address this, we introduce PlanAhead, a static planner-executor framework that evaluates the impact of plan representation in agent performance. We first automatically categorize WebArena tasks into 3 difficulty levels, enabling consistent difficulty grading without human annotation. Then we systematically evaluate 4 different plan representations on the tasks categorized as hard: sequential subgoals, narrative, pseudocode, and checklist; across different families of multimodal LLM powered agents (OpenAI, Alibaba, and Google). To account for stochastic variability, we introduce two novel evaluation metrics: Achievement Rate (AR) and Solved-Task Consistency (STC). Our results show that both, the plan formulation and the underlying LLM generating the plan, significantly influence web-agent robustness and task success.
arXiv:2605.28911v1 Announce Type: new
Abstract: As AI systems increasingly shape political views, defining and evaluating AI political neutrality is an urgent problem. Here, we propose a new definition of AI political neutrality and design a large-scale user study to test it, releasing a new dataset PARETO with 7,434 participants and 208,152 evaluations of AI responses. Our definition follows a simple principle grounded in political theory: when asked about a controversial issue, an AI model should generate responses that maximize approval across groups with opposing viewpoints, while balancing approval between groups. This definition allows empirical testing of whether an AI response is "neutral" and generalizes to any political context without pre-supposing a single left-right axis of division. We construct a benchmark of controversial U.S. issues, with prompts sourced from politically charged questions on Reddit and responses from frontier AI models, and recruit human participants to rate AI responses. Across all 20 issues, we find that it is possible for AI responses to achieve high rates of approval on both sides, even as those sides disagree strongly with each other on the substance of the issues. We also find that default responses lean liberal for GPT, Gemini, Claude, and Llama, but not Grok, and that user prompts with political charges are harder to respond to than neutral prompts. This work introduces a rigorous definition and benchmark of AI political neutrality, and a dataset to measure progress toward it.
arXiv:2605.29971v1 Announce Type: new
Abstract: Causal interventions in language model representations have largely targeted discrete features, like grammatical number. However, language models must also make use of features that are graded. We introduce a method for causal intervention on continuous variables: given activation vectors paired with a graded target variable, we localize a low-dimensional direction for that variable and use this direction to edit a vectors toward counterfactual target values. We apply this method to a continuous feature that is well-studied in psycholinguistics, namely verb bias (which reflects which syntactic structures tend to follow a given verb). We show that verb bias is causally represented in steering vectors extracted from large language models: counterfactual edits to verb bias systematically shift downstream structural preferences. Verb bias has also previously been linked to in-context learning; in further analyses, we find that steering vectors encode error signals that could drive the error-driven update behavior seen in in-context learning but that these aspects of the steering vectors are not causally used in downstream production. Overall, these results show causal interventions can be applied to continuous variables, though connecting continuous variables to in-context learning remains a challenge.
arXiv:2605.29926v1 Announce Type: new
Abstract: Accurate prediction of drug-target interactions (DTI) is critical for drug discovery. Existing methods often rely on single-modal representations (e.g., sequences or graphs) or combine only two modalities, overlooking 3D structural features. To address this challenge, we propose TriMod-DTI, a triple-modal contrastive learning framework that incorporates 1D sequences, 2D graphs, and 3D structures of drugs and proteins, obtaining the universal and complementary feature representations for DTI prediction. We design a Feature Extractor to capture drug and target features across the three modalities, thereby enriching their representations. We further propose a triple-modal contrastive learning strategy to align different modal representations of the same drug or protein in the latent space. By constructing cross-modal positive and negative sample pairs, this approach enhances the model's discriminative ability. Experiments on three benchmark datasets demonstrate that TriMod-DTI outperforms state-of-the-art methods. The ablation studies validate the contributions of each modality. Moreover, case studies highlight its practical potential for DTI prediction and drug discovery.
arXiv:2605.30127v1 Announce Type: new
Abstract: Surface electromyography (sEMG) enables continuous hand pose estimation on wearable devices, but models trained on multi-user corpora degrade on unseen individuals due to inter-user variability in anatomy and electrode placement. We propose REACT, a lightweight conditioning framework that personalizes a frozen pretrained EMG-to-pose backbone at inference time using only a handful of calibration recordings. REACT learns a compact user embedding from calibration data and applies Feature-wise Linear Modulation (FiLM) to adapt the shared encoder's feature space, requiring no gradient updates at deployment. On the large-scale EMG2POSE benchmark, REACT improves over the state-of-the-art baseline across all three generalization splits in both regression and tracking modes, reducing angular error by up to 3.9% with minimal parameter overhead and under 45 seconds of per-user calibration.
arXiv:2605.29462v1 Announce Type: new
Abstract: The emergence of Large Vision-Language Models (LVLMs) has substantially expanded model capabilities beyond text-only understanding, enabling unified inference across both visual and textual modalities and supporting a broader range of real-world applications. To comprehensively evaluate the perception, understanding, reasoning, and cognition capabilities of LVLMs throughout the entire financial business workflow in Chinese contexts, we introduce CFMME, a novel Chinese financial multimodal evaluation benchmark. CFMME comprises 6,052 instances spanning from fundamental academic knowledge to complex real-world applications, covering eight primary financial image modalities and four core multimodal tasks. On CFMME, we conduct a thorough evaluation of representative LVLMs. The results show that the state-of-the-art model attains an overall accuracy of 66.11\% on the question answering task and an average score of 77.18 on the detection, recognition, and information extraction tasks, indicating substantial room for improvement in current LVLMs. In addition, we conduct detailed analyses of error causes, cross-modal capabilities, and multi-orientation settings, yielding valuable insights for future research. We hope that CFMME will spur further progress in LVLMs, especially by improving their performance on multiple multimodal tasks in the financial domain.
arXiv:2601.13111v2 Announce Type: replace
Abstract: Realistic text-to-SQL workflows often require joining multiple tables. As a result, accurately retrieving the relevant set of tables becomes a key bottleneck for end-to-end performance. We study an open-book setting where queries must be answered over large, heterogeneous table collections pooled from many sources, without clean scoping signals such as database identifiers. Here, dense retrieval (DR) achieves high recall but returns many distractors, while join-aware alternatives often rely on extra assumptions and/or incur high inference overhead. We propose CORE-T, a scalable, training-free framework that enriches tables with LLM-generated purpose metadata and pre-computes a lightweight table-compatibility cache. At inference time, DR returns top-K candidates; a single LLM call selects a coherent, joinable subset, and a two-step additive adjustment stage restores strongly compatible tables. Across Bird, Spider, MMQA, and Beaver, CORE-T improves over DR by up to 22.7 points in table-selection F1 while returning up to 40% fewer tables, and by up to 24.4 points in multi-table execution accuracy, and uses 1.64-4.20x fewer total selection tokens than LLM-intensive baselines.
arXiv:2601.05247v2 Announce Type: replace
Abstract: Building on ideas of Gurevich and Shelah for the G\"odel Class, we present a new probabilistic proof of the finite model property for the Guarded Fragment of First-Order Logic. Our proof is conceptually simple and yields the optimal doubly-exponential upper bound on the size of minimal models. We precisely analyse the obtained bound, up to constant factors in the exponents, and construct sentences that enforce models of tightly matching size. The probabilistic approach adapts naturally to the Triguarded Fragment, an extension of the Guarded Fragment that also subsumes the Two-Variable Fragment. Finally, we derandomise the probabilistic proof by providing an explicit model construction which replaces randomness with deterministic hash functions.
arXiv:2605.29224v1 Announce Type: new
Abstract: AI agents augment large language models with external tools such as web retrieval, enabling grounded and up-to-date responses. However, incorporating external content into the generation pipeline can weaken the safety alignment mechanisms that govern model outputs. Prior work shows that enabling retrieval in agents increases compliance with harmful requests. We introduce AgentREVEAL, a diagnostic framework for analyzing retrieval-induced safety degradation in LLM agents. The framework examines two axes: how retrieval is integrated into the agent pipeline and the properties of the retrieved content. Along the integration axis, we find that binding tool invocation and response generation in a single step amplifies harmful outputs. Along the content axis, we uncover the Safe Source Paradox: even oppositional or safety-oriented sources, such as pages containing warnings or risk disclaimers, can increase harmful compliance by an average of 25% compared to the no-retrieval baseline. Finally, we show that relevance acts as a shared activation condition for both vulnerabilities. Similar patterns appear on frontier closed models, and harmful compliance remains elevated under several representative pipeline interventions, with some agents also entering this regime under autonomous retrieval. Because relevance is also what makes retrieval useful, these results expose a safety-utility trade-off for retrieval-enabled agents. We introduce HarmURLBench, a benchmark containing 1,405 real-world URLs paired with 320 harmful behaviors to support future evaluations.
arXiv:2601.18728v2 Announce Type: replace
Abstract: Modern generative modeling methods have demonstrated strong performance in learning complex data distributions from clean samples. In many scientific and imaging applications, however, clean samples are unavailable, and only noisy or linearly corrupted measurements can be observed. Moreover, latent structures, such as manifold geometries, present in the data are important to extract for further downstream scientific analysis. In this work, we introduce Riemannian AmbientFlow, a framework for simultaneously learning a probabilistic generative model and the underlying, nonlinear data manifold directly from corrupted observations. Building on the variational inference framework of AmbientFlow, our approach incorporates data-driven Riemannian geometry induced by normalizing flows, enabling the extraction of manifold structure through pullback metrics and Riemannian Autoencoders. We establish theoretical guarantees showing that, under appropriate geometric regularization and measurement conditions, the learned model recovers the underlying data distribution up to a controllable error and yields a smooth, bi-Lipschitz manifold parametrization. We further show that the resulting smooth decoder can serve as a principled generative prior for inverse problems with recovery guarantees. We empirically validate our approach on low-dimensional synthetic manifolds and on MNIST.