More capable models can better recognize the specific circumstances under which they are trained. Because of this, they are more likely to learn to act as expected in precisely those circumstances while behaving competently but unexpectedly in others. This can surface in the form of problems that Perez et al. (2022) call sycophancy, where a model answers subjective questions in a way that flatters their user’s stated beliefs, and sandbagging, where models are more likely to endorse common misconceptions when their user appears to be less educated.
Recent articles
- CaMeL offers a promising new direction for mitigating prompt injection attacks - 11th April 2025
- Model Context Protocol has prompt injection security problems - 9th April 2025
- Long context support in LLM 0.24 using fragments and template plugins - 7th April 2025