Large Language Model Agents - Connect Session

ResearchTrend.AI LLM Agent Connect Session: Personality & Safety Compliance!

We are excited to announce our upcoming Large Language Model Agent (LLM Agent) Connect Session on ResearchTrend.AI!

Meet the Speakers 🧑‍🔬
Jivnesh Sandhan: Postdoc at the Graduate School of Informatics, Kyoto University
Haibo Jin: PhD student at the University of Illinois Urbana-Champaign (UIUC)

This virtual session 💻 features two essential presentations from leading researchers, diving into the critical challenges of personality consistency and ethical compliance in autonomous LLM agents.

Agenda (UTC) - Wednesday, December 10th

17:00 - 17:30: Jivnesh Sandhan (Kyoto University)
📄 Paper: CAPE: Context-Aware Personality Evaluation Framework for Large Language Models
💡 Abstract: Traditional personality tests for LLMs (like the "Disney World test") ignore conversational history, which fundamentally shapes real-world agent behavior. Jivnesh introduces the Context-Aware Personality Evaluation (CAPE) framework, the first to incorporate prior interactions. Experiments show that context enhances consistency but can also induce extreme personality shifts (especially in GPT models), highlighting the dynamic and sensitive nature of personality in conversational agents.

17:30 - 18:00: Haibo Jin (UIUC)
📄 Paper: GUARD: Guideline Upholding Test through Adaptive Role-play and Jailbreak Diagnostics for LLMs
💡 Abstract: Translating high-level government ethics guidelines into actionable tests for LLM compliance is a major gap. Haibo will present GUARD, a testing method that operationalizes guidelines into specific questions. Crucially, GUARD integrates Jailbreak Diagnostics (GUARD-JD), using adaptive role-play to provoke unethical scenarios and diagnose potential bypasses of safety mechanisms. This method culminates in a comprehensive compliance report, validated across seven major LLMs.

🌟 This is a fantastic opportunity to engage directly with research that is critical for building trustworthy, consistent, and ethically compliant LLM-based applications.

🗓️ Time: 5:00 PM - 6:00 PM UTC 📍 Location: Virtual
👉 Register for this event here: https://lnkd.in/eF6gzFus

Don't miss our future sessions! 📆 Find out more about upcoming events: https://lnkd.in/g7-iczUp