conftrace_

dangerous capability

1 paper

Explore in graph

Co-occurring keywords · papers mentioning both

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

Papers

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