Alibaba's HDPO framework trains AI agents to skip unnecessary tool calls, cutting redundant invocations from 98% to 2% while ...

If OpenAI can accidentally train its flagship model to obsess over goblins, what other more subtle and potentially harmful biases are being reinforced through the same feedback loops?