Fabien Roger

2 papers · 2024–2024 · 2 conferences · across top CS/AI conferences

Achievements

🌉 Interdisciplinary Bridge 🧭 Keyword Pioneer 🌍 Conference Polyglot (2) 🐝 Cross-Pollinator (12) 🐣 Hot Topic Early Bird

Conferences

ICML (1) NIPS (1)

Top co-authors

Ryan Greenblatt (2) David Krueger (1) Kshitij Sachan (1) Buck Shlegeris (1) Dmitrii Krasheninnikov (1)

Keywords

reinforcement learning (1) model evaluation (1) model safety (1) capability elicitation (1) password-locked model (1) safety evaluation (1) hidden capability (1) llm alignment (1) large language model (1) dangerous capability (1) fine-tuning evaluation (1)

Papers

Stress-Testing Capability Elicitation With Password-Locked Models NIPS 2024

AI Control: Improving Safety Despite Intentional Subversion ICML 2024