paper-with-me

Papers

Mind the Gap: Evaluating Model- and Agentic-Level Vulnerabilities in LLMs with Action Graphs

2025-09-05 · Ilham Wicaksono, Zekun Wu, Rahul Patel, Theo King, Adriano Koshiyama, Philip Treleaven arxiv

As large language models increasingly deployed into agentic systems, existing methods face critical gaps in observing, assessing, and mitigating deployment-specific risks. We present a comprehensive, observability-driven workflow: we introduce \textbf{AgentSeer}, observability tool which decomposes agentic executions into granular \emph{action-component} graphs; we use this decomposition to rigorously quantify the gap between model-level and agent-level jailbreaking risk via cross-model validation on GPT-OSS-20B and Gemini-2.0-flash with HarmBench under single-turn and iterative-refinement attacks; we leverage action-graph risk signals to automate iterative prompt hardening against direct and iterative jailbreak attacks. Stark differences is revealed between model-level and agentic-level vulnerability profiles. Model-level evaluation reveals baseline differences: GPT-OSS-20B (39.47\% ASR) versus Gemini-2.0-flash (50.00\% ASR), with both models showing susceptibility to social engineering. However, agentic-level assessment exposes agent-specific risks invisible to traditional evaluation. We discover "agentic-only" vulnerabilities that emerge exclusively in agentic contexts, with tool-calling showing 24-60\% higher ASR across both models. Cross-model analysis reveals universal agentic patterns, where agent transfer operations as highest-risk tools, with semantic pattern revealed rather than syntactic vulnerability mechanisms. Direct attack transfer from model-level to agentic contexts shows degraded performance of successful prompts (GPT-OSS-20B: 57\% human injection ASR; Gemini-2.0-flash: 28\%), while context-aware iterative attacks successfully compromise objectives that failed at model-level, confirming systematic vulnerabilities gaps. Action-based prompt improvement substantially reduces action-averaged agentic jailbreak success on GPT-OSS-20B (direct: 45.3\%

📄 PDF Abstract BibTeX arXiv:2509.04802

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

Mind the Gap: Comparing Model- vs Agentic-Level Red Teaming with Action-Graph Observability on GPT-OSS-20B

2025-09-21 · Ilham Wicaksono, Zekun Wu, Rahul Patel, Theo King 외 arxiv

As the industry increasingly adopts agentic AI systems, understanding their unique vulnerabilities becomes critical. Prior research suggests that security flaws at the model level do not fully capture the risks present i…

Red Teaming

ZeroDayBench: Evaluating LLM Agents on Unseen Zero-Day Vulnerabilities for Cyberdefense

2026-03-02 · Nancy Lau, Louis Sloot, Jyoutir Raj, Giuseppe Marco Boscardin 외 arxiv

Large language models (LLMs) are increasingly being deployed as software engineering agents that autonomously contribute to repositories. A major benefit these agents present is their ability to find and patch security v…

AgenticSCR: An Autonomous Agentic Secure Code Review for Immature Vulnerabilities Detection

2026-01-27 · Wachiraphan Charoenwet, Kla Tantithamthavorn, Patanamon Thongtanunam, Hong Yi Lin 외 arxiv

Secure code review is critical at the pre-commit stage, where vulnerabilities must be caught early under tight latency and limited-context constraints. Existing SAST-based checks are noisy and often miss immature, contex…

Vulnerability Detection

LLM-Coordination: Evaluating and Analyzing Multi-agent Coordination Abilities in Large Language Models

2023-10-05 · Saaket Agashe, Yue Fan, Anthony Reyna, Xin Eric Wang

Large Language Models (LLMs) have demonstrated emergent common-sense reasoning and Theory of Mind (ToM) capabilities, making them promising candidates for developing coordination agents. This study introduces the LLM-Coo…

Common Sense ReasoningMultiple-choiceQuestion Answering

Mind2Web 2: Evaluating Agentic Search with Agent-as-a-Judge

2025-06-26 · Boyu Gou, Zanming Huang, Yuting Ning, Yu Gu 외

Agentic search such as Deep Research systems, where large language models autonomously browse the web, synthesize information, and return comprehensive citation-backed answers, represents a major shift in how users inter…

Benchmarking