Structured Security Auditing and Robustness Enhancement for Untrusted Agent Skills

Agent Skills package SKILL.md files, scripts, reference documents, and repository context into reusable capability units, turning pre-load auditing from single-prompt filtering into cross-file security review. Existing guardrails often flag risk but recover malicious intent inconsistently under semantics-preserving rewrites. This paper formulates pre-load auditing for untrusted Agent Skills as a robust three-way classification task and introduces SkillGuard-Robust, which combines role-aware evidence extraction, selective semantic verification, and consistency-preserving adjudication. We evaluate SkillGuard-Robust on SkillGuardBench and two public-ecosystem extensions through five large evaluation views ranging from 254 to 404 packages. On the 404-package held-out aggregate, SkillGuard-Robust reaches 97.30% overall exact match, 98.33% malicious-risk recall, and 98.89% attack exact consistency. On the 254-package external-ecosystem view, it reaches 99.66%, 100.00%, and 100.00%, respectively. These results support a bounded conclusion: factorized package auditing materially improves frozen and public-ecosystem robustness, while harsher external-source transfer remains an open challenge.

RouteGuard: Internal-Signal Detection of Skill Poisoning in LLM Agents

Agent skills introduce a new and more severe form of indirect injection for LLM agents: unlike traditional indirect prompt injection, attackers can hide malicious instructions inside a dense, action-oriented skill that already functions as a legitimate instruction source. We study pre-execution skill-poison detection and show that successful skill poisoning induces a structured internal effect, attention hijacking, in which response-time attention shifts from trusted context to malicious skill spans and drives harmful behavior. Motivated by this mechanism, we propose RouteGuard, a frozen-backbone detector that combines response-conditioned attention and hidden-state alignment through reliability-gated late fusion. Across both real and synthetic open-source skill benchmarks, RouteGuard is consistently the strongest or most robust detector; on the critical Skill-Inject channel slice, it reaches 0.8834 F1 and recovers 90.51% of description attacks missed by lexical screening, showing that defending against skill poisoning requires internal-signal detection rather than text-only filtering

FABLE: Fine-grained Fact Anchoring for Unstructured Model Editing

Unstructured model editing aims to update models with real-world text, yet existing methods often memorize text holistically without reliable fine-grained fact access. To address this, we propose FABLE, a hierarchical framework that decouples fine-grained fact injection from holistic text generation. FABLE follows a two-stage, fact-first strategy: discrete facts are anchored in shallow layers, followed by minimal updates to deeper layers to produce coherent text. This decoupling resolves the mismatch between holistic recall and fine-grained fact access, reflecting the unidirectional Transformer flow in which surface-form generation amplifies rather than corrects underlying fact representations. We also introduce UnFine, a diagnostic benchmark with fine-grained question-answer pairs and fact-level metrics for systematic evaluation. Experiments show that FABLE substantially improves fine-grained question answering while maintaining state-of-the-art holistic editing performance. Our code is publicly available at https://github.com/caskcsg/FABLE.

RecBundle: A Next-Generation Geometric Paradigm for Explainable Recommender Systems

Recommender systems are inherently dynamic feedback loops where prolonged local interactions accumulate into macroscopic structural degradation such as information cocoons. Existing representation learning paradigms are universally constrained by the assumption of a single flat space, forcing topologically grounded user associations and semantically driven historical interactions to be fitted within the same vector space. This excessive coupling of heterogeneous information renders it impossible for researchers to mechanistically distinguish and identify the sources of systemic bias. To overcome this theoretical bottleneck, we introduce Fiber Bundle from modern differential geometry and propose a novel geometric analysis paradigm for recommender systems. This theory naturally decouples the system space into two hierarchical layers: the base manifold formed by user interaction networks, and the fibers attached to individual user nodes that carry their dynamic preferences. Building upon this, we construct RecBundle, a framework oriented toward next-generation recommender systems that formalizes user collaboration as geometric connection and parallel transport on the base manifold, while mapping content evolution to holonomy transformations on fibers. From this foundation, we identify future application directions encompassing quantitative mechanisms for information cocoons and evolutionary bias, geometric meta-theory for adaptive recommendation, and novel inference architectures integrating large language models (LLMs). Empirical analysis on real-world MovieLens and Amazon Beauty datasets validates the effectiveness of this geometric framework.

A Cognitive Distribution and Behavior-Consistent Framework for Black-Box Attacks on Recommender Systems

With the growing deployment of sequential recommender systems in e-commerce and other fields, their black-box interfaces raise security concerns: models are vulnerable to extraction and subsequent adversarial manipulation. Existing black-box extraction attacks primarily rely on hard labels or pairwise learning, often ignoring the importance of ranking positions, which results in incomplete knowledge transfer. Moreover, adversarial sequences generated via pure gradient methods lack semantic consistency with real user behavior, making them easily detectable. To overcome these limitations, this paper proposes a dual-enhanced attack framework. First, drawing on primacy effects and position bias, we introduce a cognitive distribution-driven extraction mechanism that maps discrete rankings into continuous value distributions with position-aware decay, thereby advancing from order alignment to cognitive distribution alignment. Second, we design a behavior-aware noisy item generation strategy that jointly optimizes collaborative signals and gradient signals. This ensures both semantic coherence and statistical stealth while effectively promoting target item rankings. Extensive experiments on multiple datasets demonstrate that our approach significantly outperforms existing methods in both attack success rate and evasion rate, validating the value of integrating cognitive modeling and behavioral consistency for secure recommender systems.

Re-Align: Structured Reasoning-guided Alignment for In-Context Image Generation and Editing

In-context image generation and editing (ICGE) enables users to specify visual concepts through interleaved image-text prompts, demanding precise understanding and faithful execution of user intent. Although recent unified multimodal models exhibit promising understanding capabilities, these strengths often fail to transfer effectively to image generation. We introduce Re-Align, a unified framework that bridges the gap between understanding and generation through structured reasoning-guided alignment. At its core lies the In-Context Chain-of-Thought (IC-CoT), a structured reasoning paradigm that decouples semantic guidance and reference association, providing clear textual target and mitigating confusion among reference images. Furthermore, Re-Align introduces an effective RL training scheme that leverages a surrogate reward to measure the alignment between structured reasoning text and the generated image, thereby improving the model's overall performance on ICGE tasks. Extensive experiments verify that Re-Align outperforms competitive methods of comparable model scale and resources on both in-context image generation and editing tasks.

Exploiting Synergistic Cognitive Biases to Bypass Safety in LLMs

Large Language Models (LLMs) demonstrate impressive capabilities across a wide range of tasks, yet their safety mechanisms remain susceptible to adversarial attacks that exploit cognitive biases -- systematic deviations from rational judgment. Unlike prior jailbreaking approaches focused on prompt engineering or algorithmic manipulation, this work highlights the overlooked power of multi-bias interactions in undermining LLM safeguards. We propose CognitiveAttack, a novel red-teaming framework that systematically leverages both individual and combined cognitive biases. By integrating supervised fine-tuning and reinforcement learning, CognitiveAttack generates prompts that embed optimized bias combinations, effectively bypassing safety protocols while maintaining high attack success rates. Experimental results reveal significant vulnerabilities across 30 diverse LLMs, particularly in open-source models. CognitiveAttack achieves a substantially higher attack success rate compared to the SOTA black-box method PAP (60.1% vs. 31.6%), exposing critical limitations in current defense mechanisms. These findings highlight multi-bias interactions as a powerful yet underexplored attack vector. This work introduces a novel interdisciplinary perspective by bridging cognitive science and LLM safety, paving the way for more robust and human-aligned AI systems.

Paper Summary Attack: Jailbreaking LLMs through LLM Safety Papers

The safety of large language models (LLMs) has garnered significant research attention. In this paper, we argue that previous empirical studies demonstrate LLMs exhibit a propensity to trust information from authoritative sources, such as academic papers, implying new possible vulnerabilities. To verify this possibility, a preliminary analysis is designed to illustrate our two findings. Based on this insight, a novel jailbreaking method, Paper Summary Attack (\llmname{PSA}), is proposed. It systematically synthesizes content from either attack-focused or defense-focused LLM safety paper to construct an adversarial prompt template, while strategically infilling harmful query as adversarial payloads within predefined subsections. Extensive experiments show significant vulnerabilities not only in base LLMs, but also in state-of-the-art reasoning model like Deepseek-R1. PSA achieves a 97\% attack success rate (ASR) on well-aligned models like Claude3.5-Sonnet and an even higher 98\% ASR on Deepseek-R1. More intriguingly, our work has further revealed diametrically opposed vulnerability bias across different base models, and even between different versions of the same model, when exposed to either attack-focused or defense-focused papers. This phenomenon potentially indicates future research clues for both adversarial methodologies and safety alignment.Code is available at https://github.com/233liang/Paper-Summary-Attack

Align Beyond Prompts: Evaluating World Knowledge Alignment in Text-to-Image Generation

Recent text-to-image (T2I) generation models have advanced significantly, enabling the creation of high-fidelity images from textual prompts. However, existing evaluation benchmarks primarily focus on the explicit alignment between generated images and prompts, neglecting the alignment with real-world knowledge beyond prompts. To address this gap, we introduce Align Beyond Prompts (ABP), a comprehensive benchmark designed to measure the alignment of generated images with real-world knowledge that extends beyond the explicit user prompts. ABP comprises over 2,000 meticulously crafted prompts, covering real-world knowledge across six distinct scenarios. We further introduce ABPScore, a metric that utilizes existing Multimodal Large Language Models (MLLMs) to assess the alignment between generated images and world knowledge beyond prompts, which demonstrates strong correlations with human judgments. Through a comprehensive evaluation of 8 popular T2I models using ABP, we find that even state-of-the-art models, such as GPT-4o, face limitations in integrating simple real-world knowledge into generated images. To mitigate this issue, we introduce a training-free strategy within ABP, named Inference-Time Knowledge Injection (ITKI). By applying this strategy to optimize 200 challenging samples, we achieved an improvement of approximately 43% in ABPScore. The dataset and code are available in https://github.com/smile365317/ABP.

LyapLock: Bounded Knowledge Preservation in Sequential Large Language Model Editing

Large Language Models often contain factually incorrect or outdated knowledge, giving rise to model editing methods for precise knowledge updates. However, current mainstream locate-then-edit approaches exhibit a progressive performance decline during sequential editing, due to inadequate mechanisms for long-term knowledge preservation. To tackle this, we model the sequential editing as a constrained stochastic programming. Given the challenges posed by the cumulative preservation error constraint and the gradually revealed editing tasks, \textbf{LyapLock} is proposed. It integrates queuing theory and Lyapunov optimization to decompose the long-term constrained programming into tractable stepwise subproblems for efficient solving. This is the first model editing framework with rigorous theoretical guarantees, achieving asymptotic optimal editing performance while meeting the constraints of long-term knowledge preservation. Experimental results show that our framework scales sequential editing capacity to over 10,000 edits while stabilizing general capabilities and boosting average editing efficacy by 11.89\% over SOTA baselines. Furthermore, it can be leveraged to enhance the performance of baseline methods. Our code is released on https://github.com/caskcsg/LyapLock.