|
|
|
|
|
|
by roh26it
545 days ago
|
|
|
The most interesting findings (at least from my perspective)
- OpenAI measured "honesty" that could become an interesting model eval others should adopt.
- Using instruction heirarchy to _avoid_ prompt injections
- When pushed to pursue misaligned goals, it tried to disable oversight in 5% of tests! |
|
|