Multi-Task Inference: Can Large Language Models Follow Multiple Instructions at Once?