Benchmarking Generation and Evaluation Capabilities of Large Language Models for Instruction Controllable Summarization | Synapse