arXiv subjects

Roman Vainshtein

Publications and source records attributed to Roman Vainshtein.

18 recordsLinked to original sources

arXiv2026

An Actionable Diagnosis of Multilingual, Multi-Agent Planning Failures

Multilingual multi-agent systems exhibit substantial degradation beyond English, yet prior work rarely identifies how task-critical information is lost when user requests are converted into executable plans. We study the planner in a multi-agent system as the request-to-action interface and derive an actionable taxonomy of planning-grounding failures from failed real-world task executions. LLM-based analysis shows that these failures constitute an increasing share of unsuccessful executions as language-resource availability declines, with the strongest effects in low-resource languages. To test whether the taxonomy supports mitigation, we introduce TART, Taxonomy-Guided Actionable Representation, that makes the taxonomy's key aspects explicit to the planner and downstream sub-agents. Across multiple languages, three LLM backbones, two datasets, and two agentic configurations, TART consistently improves performance. On multilingual GAIA, it raises a state-of-the-art system's accuracy by 5.6 percentage points averaged across eleven languages spanning low- to high-resource settings.

cs.MA↗

arXiv2026

RIFT-Bench: Dynamic Red-teaming For Agentic AI Systems

Agentic AI systems powered by large language models (LLMs) are rapidly evolving into autonomous decision-making systems, exposing attack vectors beyond those of traditional LLM vulnerabilities. Existing security evaluations are often tied to specific implementations or domains, limiting unified comparison across heterogeneous systems. To address this gap, we introduce RIFT-Bench, a representation-driven methodology for dynamic red-teaming that enables unified evaluations across diverse agentic architectures. Building on a novel hierarchical representation, RIFT-Bench operates in two automated phases: Discovery, which extracts system structure, and Scanning, which executes adaptive adversarial attacks. It directly evaluates the examined system using 105 adaptive adversarial probes spanning diverse attack vectors and objectives. We demonstrate the effectiveness of the proposed evaluation pipeline across 45 agentic systems spanning a diverse range of implementations, showing that the approach generalizes effectively to heterogeneous agentic architectures. Beyond systems and attacks, RIFT-Bench also supports direct evaluation of mitigation strategies. These key capabilities make RIFT-Bench a scalable foundation for security evaluation of agentic AI systems in practice. Infrastructure code and benchmark artifacts are available at https://tinyurl.com/RIFTBench.

cs.AI↗

arXiv2026

When Scanners Lie: Evaluator Instability in LLM Red-Teaming

Automated LLM vulnerability scanners are increasingly used to assess security risks by measuring different attack type success rates (ASR). Yet the validity of these measurements hinges on an often-overlooked component: the evaluator who determines whether an attack has succeeded. In this study, we demonstrate that commonly used open-source scanners exhibit measurement instability that depends on the evaluator component. Consequently, changing the evaluator while keeping the attacks and model outputs constant can significantly alter the reported ASR. To tackle this problem, we present a two-phase, reliability-aware evaluation framework. In the first phase, we quantify evaluator disagreement to identify attack categories where ASR reliability cannot be assumed. In the second phase, we propose a verification-based evaluation method where evaluators are validated by an independent verifier, enabling reliability assessment without relying on extensive human annotation. Applied to the widely used Garak scanner, we observe that 22 of 25 attack categories exhibit evaluator instability, reflected in high disagreement among evaluators. Our approach raises evaluator accuracy from 72% to 89% while enabling selective deployment to control cost and computational overhead. We further quantify evaluator uncertainty in ASR estimates, showing that reported vulnerability scores can vary by up to 33% depending on the evaluator. Our results indicate that the outputs of vulnerability scanners are highly sensitive to the choice of evaluators. Our framework offers a practical approach to quantify unreliable evaluations and enhance the reliability of measurements in automated LLM security assessments.

cs.CR↗

arXiv2026

Adversarial Intent is a Latent Variable: Stateful Trust Inference for Securing Multimodal Agentic RAG

Current stateless defences for multimodal agentic RAG fail to detect adversarial strategies that distribute malicious semantics across retrieval, planning, and generation components. We formulate this security challenge as a Partially Observable Markov Decision Process (POMDP), where adversarial intent is a latent variable inferred from noisy multi-stage observations. We introduce MMA-RAG^T, an inference-time control framework governed by a Modular Trust Agent (MTA) that maintains an approximate belief state via structured LLM reasoning. Operating as a model-agnostic overlay, MMA-RAGT mediates a configurable set of internal checkpoints to enforce stateful defence-in-depth. Extensive evaluation on 43,774 instances demonstrates a 6.50x average reduction factor in Attack Success Rate relative to undefended baselines, with negligible utility cost. Crucially, a factorial ablation validates our theoretical bounds: while statefulness and spatial coverage are individually necessary (26.4 pp and 13.6 pp gains respectively), stateless multi-point intervention can yield zero marginal benefit under homogeneous stateless filtering when checkpoint detections are perfectly correlated.

cs.CR↗

arXiv2026

Inference-Time Backdoors via Chat Templates: From LLM Supply Chains to Agentic System Compromise

Open-weight language models are increasingly used in production settings, raising new security challenges. One prominent threat is backdoor attacks, in which adversaries embed hidden behaviors that activate under specific conditions. Previous work has assumed that adversaries have access to training pipelines or deployment infrastructure. We propose a novel attack surface requiring neither: the "chat template". Chat templates are executable programs invoked at every inference call, often implemented in Jinja2, that occupy a privileged position between user input and model processing. We show that an adversary who distributes a model with a maliciously modified template can implant an inference-time backdoor without modifying model weights, poisoning training data, or controlling runtime infrastructure. We evaluate this attack across three deployment tiers. At the LLM level, triggered backdoors reduce factual accuracy from 90% to 15% on average and induce attacker-controlled URL emission with success rates exceeding 80%, while benign inputs show no measurable degradation; these results hold across eighteen models. At the agent level, template backdoors hijack tool-use across two benchmarks spanning 3,868 episodes, bypassing every tested injection defense offered by the benchmarks while remaining fully dormant absent the trigger. At the multi-agent system level, we demonstrate how a single poisoned artifact compromises a real-world agentic deployment and propagates supply-chain code poisoning downstream. The poisoned artifacts evade all security scans on the largest open model distribution platform; and because the payload is rendered by the template before user input is processed, it is architecturally unreachable by input-level defenses such as prompt injection guardrails. These results establish chat templates as a reliable and undefended attack in the open-weight AI supply chain.

cs.CR↗

arXiv2026

AgenTRIM: Tool Risk Mitigation for Agentic AI

AI agents are autonomous systems that combine LLMs with external tools to solve complex tasks. While such tools extend capability, improper tool permissions introduce security risks such as indirect prompt injection and tool misuse. We characterize these failures as unbalanced tool-driven agency. Agents may retain unnecessary permissions (excessive agency) or fail to invoke required tools (insufficient agency), amplifying the attack surface and reducing performance. We introduce AgenTRIM, a framework for detecting and mitigating tool-driven agency risks without altering an agent's internal reasoning. AgenTRIM addresses these risks through complementary offline and online phases. Offline, AgenTRIM reconstructs and verifies the agent's tool interface from code and execution traces. At runtime, it enforces per-step least-privilege tool access through adaptive filtering and status-aware validation of tool calls. Evaluating on the AgentDojo benchmark, AgenTRIM substantially reduces attack success while maintaining high task performance. Additional experiments show robustness to description-based attacks and effective enforcement of explicit safety policies. Together, these results show that AgenTRIM provides a practical, capability-preserving approach to safer tool use in LLM-based agents.

cs.CR↗

arXiv2025

General and Domain-Specific Zero-shot Detection of Generated Images via Conditional Likelihood

The rapid advancement of generative models, particularly diffusion-based methods, has significantly improved the realism of synthetic images. As new generative models continuously emerge, detecting generated images remains a critical challenge. While fully supervised, and few-shot methods have been proposed, maintaining an updated dataset is time-consuming and challenging. Consequently, zero-shot methods have gained increasing attention in recent years. We find that existing zero-shot methods often struggle to adapt to specific image domains, such as artistic images, limiting their real-world applicability. In this work, we introduce CLIDE, a novel zero-shot detection method based on conditional likelihood approximation. Our approach computes likelihoods conditioned on real images, enabling adaptation across diverse image domains. We extensively evaluate CLIDE, demonstrating SOTA performance on a large-scale general dataset and significantly outperform existing methods in domain-specific cases. These results demonstrate the robustness of our method and underscore the need of broad, domain-aware generalization for the AI-generated image detection task. Code is available at https://tinyurl.com/clide-detector.

eess.IV↗

arXiv2025

Training-Free Policy Violation Detection via Activation-Space Whitening in LLMs

As organizations increasingly deploy LLMs in sensitive domains such as legal, financial, and medical settings, ensuring alignment with internal organizational policies has become a priority. Existing content moderation frameworks remain largely confined to the safety domain and lack the robustness to capture nuanced organizational policies. LLM-as-a-judge and fine-tuning approaches, though flexible, introduce significant latency and training cost. To address these limitations, we frame policy violation detection as an out-of-distribution (OOD) problem in the model's activation space. We propose a training-free method that operates directly on the LLM internal representations, leveraging prior evidence that decision-relevant information is encoded within them. Inspired by whitening techniques, we derive policy-violation scores directly from normalized representations of LLM hidden activations. Our method requires only the policy text and a small number of illustrative samples, making it lightweight and easily deployable. We extensively evaluate our method across multiple LLMs and challenging policy benchmarks. It achieves up to 86.0% F1, outperforming both fine-tuned and LLM-as-a-judge baselines while requiring substantially less computation. Our code is publicly available at: https://github.com/FujitsuResearch/LLM-policy-violation-detection

cs.LG↗

arXiv2025

Counterfactual-based Agent Influence Ranker for Agentic AI Workflows

An Agentic AI Workflow (AAW), also known as an LLM-based multi-agent system, is an autonomous system that assembles several LLM-based agents to work collaboratively towards a shared goal. The high autonomy, widespread adoption, and growing interest in such AAWs highlight the need for a deeper understanding of their operations, from both quality and security aspects. To this day, there are no existing methods to assess the influence of each agent on the AAW's final output. Adopting techniques from related fields is not feasible since existing methods perform only static structural analysis, which is unsuitable for inference time execution. We present Counterfactual-based Agent Influence Ranker (CAIR) - the first method for assessing the influence level of each agent on the AAW's output and determining which agents are the most influential. By performing counterfactual analysis, CAIR provides a task-agnostic analysis that can be used both offline and at inference time. We evaluate CAIR using an AAWs dataset of our creation, containing 30 different use cases with 230 different functionalities. Our evaluation showed that CAIR produces consistent rankings, outperforms baseline methods, and can easily enhance the effectiveness and relevancy of downstream tasks.

cs.AI↗

arXiv2025

MAPS: A Multilingual Benchmark for Agent Performance and Security

Agentic AI systems, which build on Large Language Models (LLMs) and interact with tools and memory, have rapidly advanced in capability and scope. Yet, since LLMs have been shown to struggle in multilingual settings, typically resulting in lower performance and reduced safety, agentic systems risk inheriting these limitations. This raises concerns about the accessibility of such systems, as users interacting in languages other than English may encounter unreliable or security-critical agent behavior. Despite growing interest in evaluating agentic AI and recent initial efforts toward multilingual interaction, existing benchmarks do not yet provide a comprehensive, multi-domain, security-aware evaluation of multilingual agentic systems. To address this gap, we propose MAPS, a multilingual benchmark suite designed to evaluate agentic AI systems across diverse languages and tasks. MAPS builds on four widely used agentic benchmarks - GAIA (real-world tasks), SWE-Bench (code generation), MATH (mathematical reasoning), and the Agent Security Benchmark (security). We translate each dataset into eleven diverse languages, resulting in 805 unique tasks and 9,660 total language-specific instances - enabling a systematic analysis of the Multilingual Effect on AI agents' performance and robustness. Empirically, we observe a degradation in both performance and security when transitioning from English to other languages, with severity varying by task and correlating with the amount of translated input. This work establishes the first standardized evaluation framework for multilingual agentic AI, encouraging future research towards equitable, reliable, and accessible agentic AI. MAPS benchmark suite is publicly available at https://huggingface.co/datasets/Fujitsu-FRE/MAPS

cs.DB↗

arXiv2025

Identifying Memorization of Diffusion Models through $p$-Laplace Analysis: Estimators, Bounds and Applications

Diffusion models, today's leading image generative models, estimate the score function, i.e. the gradient of the log probability of (perturbed) data samples, without direct access to the underlying probability distribution. This work investigates whether the estimated score function can be leveraged to compute higher-order differentials, namely the p-Laplace operators. We show that these operators can be employed to identify memorized training data. We propose a numerical p-Laplace approximation based on the learned score functions, showing its effectiveness in identifying key features of the probability landscape. Furthermore, theoretical error-bounds to these estimators are proven and demonstrated numerically. We analyze the structured case of Gaussian mixture models, and demonstrate that the results carry-over to text-conditioned image generative models (text-to-image), where memorization identification based on the p-Laplace operator is performed for the first time, showing its advantage on 500 memorized prompts ($\sim$3000 generated images) in a post-generation regime, especially when the conditioning text is unavailable.

cs.CV↗

arXiv2025

Manifold Induced Biases for Zero-shot and Few-shot Detection of Generated Images

Distinguishing between real and AI-generated images, commonly referred to as 'image detection', presents a timely and significant challenge. Despite extensive research in the (semi-)supervised regime, zero-shot and few-shot solutions have only recently emerged as promising alternatives. Their main advantage is in alleviating the ongoing data maintenance, which quickly becomes outdated due to advances in generative technologies. We identify two main gaps: (1) a lack of theoretical grounding for the methods, and (2) significant room for performance improvements in zero-shot and few-shot regimes. Our approach is founded on understanding and quantifying the biases inherent in generated content, where we use these quantities as criteria for characterizing generated images. Specifically, we explore the biases of the implicit probability manifold, captured by a pre-trained diffusion model. Through score-function analysis, we approximate the curvature, gradient, and bias towards points on the probability manifold, establishing criteria for detection in the zero-shot regime. We further extend our contribution to the few-shot setting by employing a mixture-of-experts methodology. Empirical results across 20 generative models demonstrate that our method outperforms current approaches in both zero-shot and few-shot settings. This work advances the theoretical understanding and practical usage of generated content biases through the lens of manifold analysis.

cs.CV↗

arXiv2024

Insights and Current Gaps in Open-Source LLM Vulnerability Scanners: A Comparative Analysis

This report presents a comparative analysis of open-source vulnerability scanners for conversational large language models (LLMs). As LLMs become integral to various applications, they also present potential attack surfaces, exposed to security risks such as information leakage and jailbreak attacks. Our study evaluates prominent scanners - Garak, Giskard, PyRIT, and CyberSecEval - that adapt red-teaming practices to expose these vulnerabilities. We detail the distinctive features and practical use of these scanners, outline unifying principles of their design and perform quantitative evaluations to compare them. These evaluations uncover significant reliability issues in detecting successful attacks, highlighting a fundamental gap for future development. Additionally, we contribute a preliminary labelled dataset, which serves as an initial step to bridge this gap. Based on the above, we provide strategic recommendations to assist organizations choose the most suitable scanner for their red-teaming needs, accounting for customizability, test suite comprehensiveness, and industry-specific use cases.

cs.CR↗

arXiv2024

The Branch Not Taken: Predicting Branching in Online Conversations

Multi-participant discussions tend to unfold in a tree structure rather than a chain structure. Branching may occur for multiple reasons -- from the asynchronous nature of online platforms to a conscious decision by an interlocutor to disengage with part of the conversation. Predicting branching and understanding the reasons for creating new branches is important for many downstream tasks such as summarization and thread disentanglement and may help develop online spaces that encourage users to engage in online discussions in more meaningful ways. In this work, we define the novel task of branch prediction and propose GLOBS (Global Branching Score) -- a deep neural network model for predicting branching. GLOBS is evaluated on three large discussion forums from Reddit, achieving significant improvements over an array of competitive baselines and demonstrating better transferability. We affirm that structural, temporal, and linguistic features contribute to GLOBS success and find that branching is associated with a greater number of conversation participants and tends to occur in earlier levels of the conversation tree. We publicly release GLOBS and our implementation of all baseline models to allow reproducibility and promote further research on this important task.

cs.CL↗

arXiv2020

Automatic Machine Learning Derived from Scholarly Big Data

One of the challenging aspects of applying machine learning is the need to identify the algorithms that will perform best for a given dataset. This process can be difficult, time consuming and often requires a great deal of domain knowledge. We present Sommelier, an expert system for recommending the machine learning algorithms that should be applied on a previously unseen dataset. Sommelier is based on word embedding representations of the domain knowledge extracted from a large corpus of academic publications. When presented with a new dataset and its problem description, Sommelier leverages a recommendation model trained on the word embedding representation to provide a ranked list of the most relevant algorithms to be used on the dataset. We demonstrate Sommelier's effectiveness by conducting an extensive evaluation on 121 publicly available datasets and 53 classification algorithms. The top algorithms recommended for each dataset by Sommelier were able to achieve on average 97.7% of the optimal accuracy of all surveyed algorithms.

cs.LG↗

arXiv2020

Sequence Preserving Network Traffic Generation

We present the Network Traffic Generator (NTG), a framework for perturbing recorded network traffic with the purpose of generating diverse but realistic background traffic for network simulation and what-if analysis in enterprise environments. The framework preserves many characteristics of the original traffic recorded in an enterprise, as well as sequences of network activities. Using the proposed framework, the original traffic flows are profiled using 200 cross-protocol features. The traffic is aggregated into flows of packets between IP pairs and clustered into groups of similar network activities. Sequences of network activities are then extracted. We examined two methods for extracting sequences of activities: a Markov model and a neural language model. Finally, new traffic is generated using the extracted model. We developed a prototype of the framework and conducted extensive experiments based on two real network traffic collections. Hypothesis testing was used to examine the difference between the distribution of original and generated features, showing that 30-100\% of the extracted features were preserved. Small differences between n-gram perplexities in sequences of network activities in the original and generated traffic, indicate that sequences of network activities were well preserved.

cs.NI↗

arXiv2019

RankML: a Meta Learning-Based Approach for Pre-Ranking Machine Learning Pipelines

The explosion of digital data has created multiple opportunities for organizations and individuals to leverage machine learning (ML) to transform the way they operate. However, the shortage of experts in the field of machine learning -- data scientists -- is often a setback to the use of ML. In an attempt to alleviate this shortage, multiple approaches for the automation of machine learning have been proposed in recent years. While these approaches are effective, they often require a great deal of time and computing resources. In this study, we propose RankML, a meta-learning based approach for predicting the performance of whole machine learning pipelines. Given a previously-unseen dataset, a performance metric, and a set of candidate pipelines, RankML immediately produces a ranked list of all pipelines based on their predicted performance. Extensive evaluation on 244 datasets, both in regression and classification tasks, shows that our approach either outperforms or is comparable to state-of-the-art, computationally heavy approaches while requiring a fraction of the time and computational cost.

cs.LG↗

arXiv2019

Assessing the Quality of Scientific Papers

A multitude of factors are responsible for the overall quality of scientific papers, including readability, linguistic quality, fluency,semantic complexity, and of course domain-specific technical factors. These factors vary from one field of study to another. In this paper, we propose a measure and method for assessing the overall quality of the scientific papers in a particular field of study. We evaluate our method in the computer science domain, but it can be applied to other technical and scientific fields.Our method is based on the corpus linguistics technique. This technique enables the extraction of required information and knowledge associated with a specific domain. For this purpose, we have created a large corpus, consisting of papers from very high impact conferences. First, we analyze this corpus in order to extract rich domain-specific terminology and knowledge. Then we use the acquired knowledge to estimate the quality of scientific papers by applying our proposed measure. We examine our measure on high and low scientific impact test corpora. Our results show a significant difference in the measure scores of the high and low impact test corpora. Second, we develop a classifier based on our proposed measure and compare it to the baseline classifier. Our results show that the classifier based on our measure over-performed the baseline classifier. Based on the presented results the proposed measure and the technique can be used for automated assessment of scientific papers.

cs.IR↗