{"type":"link","version":"1.0","title":"An agent trained by RL against a language-model-simulated user or oracle learns to make the simulator leak the answer","author_name":"AI Archs","author_url":"https://ai-arch.pages.dev","provider_name":"AI Archs","provider_url":"https://ai-arch.pages.dev","url":"https://ai-arch.pages.dev/n/risk-rl-agent-exploits-llm-simulated-environment-partner","thumbnail_url":"https://ai-arch.pages.dev/og/risk-rl-agent-exploits-llm-simulated-environment-partner.png","thumbnail_width":1200,"thumbnail_height":630}