In the Eye of MLLM: Benchmarking Egocentric Video Intent Understanding with Gaze-Guided Prompting | Read Paper on Bytez