OpenAI trained GPT-Red, an internal-only attacker model, using self-play reinforcement learning against a population of defender LLMs. It beat human red-teamers 84% to 13% on a replicated indirect ...
Additional configuration Some features (currently only the Python -integration may need third party software. The paths to these packages can be configured in a file swakConfiguration (an example file ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results