Trustworthy Agentic Systems

List of Papers

1. Defining Trust in AI
   1. A Survey on Trustworthy LLM Agents: Threats and Countermeasures
   2. Agentic AI Systems: Opportunities, Challenges, and Trustworthiness
2. Agency vs. Tool Use
   1. ToolReflection: Improving Large Language Models for Real-World API Calls with Self-Generated Data
   2. AgentDojo: A Dynamic Environment to Evaluate Prompt Injection Attacks and Defenses for LLM Agents
   3. INJECAGENT: Benchmarking Indirect Prompt Injections in Tool-Integrated Large Language Model Agents
3. Uncertainty and Robustness
   1. Corrupted by Reasoning: Reasoning Language Models Become Free-Riders in Public Goods Games
   2. Are You Sure You’re Positive? Consolidating Chain-of-Thought Agents with Uncertainty Quantification for Aspect-Category Sentiment Analysis
4. Interpretability and Transparency
   1. Because we have LLMs, we Can and Should Pursue Agentic Interpretability
   2. LLM-Driven Social Influence for Cooperative Behavior in Multi-Agent Systems
5. Multi-Agent Safety / Cross Agent Trust Networks
   1. TrustAgent: Towards Safe and Trustworthy LLM-based Agents
   2. Evaluating And Mitigating The Safety awareness-execution Gaps Of LM Agents
   3. Malibu Benchmark: Multi-agent LLM Implicit Bias Uncovered
   4. Safe in Isolation, Dangerous Together: Agent-Driven Multi-Turn Decomposition Jailbreaks on LLMs
   5. Agents Under Siege: Breaking Pragmatic Multi-Agent LLM Systems with Optimized Prompt Attacks
6. Security by Design for Agentic Systems
   1. Defeating Prompt Injections by Design
   2. AirGapAgent: Protecting Privacy-Conscious Conversational Agents
   3. Privacy Awareness for Information-Sharing Assistants: A Case-study on Form-filling with Contextual Integrity
7. Human-Agent Collaboration and Oversight
   1. Facilitating Trustworthy Human-Agent Collaboration in LLM-based Multi-Agent System oriented Software Engineering
   2. MATAGENT: A Human-in-the-loop Multi-agent LLM Framework For Accelerating The Material Science Discovery Cycle
   3. Exploring The Role Of Agentic AI In Enhancing Human-AI Collaboration
8. Trust in Browser and Web Agents
   1. Aligned LLMs Are Not Aligned Browser Agents
   2. Context manipulation attacks : Web agents are susceptible to corrupted memory
9. Long-term agent memory
   1. Unveiling Privacy Risks in LLM Agent Memory
   2. AGENTPOISON: Red-teaming LLM Agents via Poisoning Memory or Knowledge Bases
   3. Evaluating Very Long-Term Conversational Memory of LLM Agents
10. Deployment and Monitoring of Agents
    1. Security Challenges in AI Agent Deployment: Insights from a Large Scale Public Competition
    2. A Survey on AgentOps: Categorization, Challenges, and Future Directions
11. Societal Implications of Agentic AI
    1. Cooperate or Collapse: Emergence of Sustainable Cooperation in a Society of LLM Agents
    2. Can Large Language Model Agents Simulate Human Trust Behavior?