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
- Embracing the parallel coding agent lifestyle - 5th October 2025
- Designing agentic loops - 30th September 2025
- Claude Sonnet 4.5 is probably the "best coding model in the world" (at least for now) - 29th September 2025