Google T5Gemma Redefining LLM Efficiency With Encoder-Decoder Architecture

T5Gemma uses an asymmetric encoder-decoder architecture based on Gemma 2 to optimize latency and context processing.