From Transcription to Reformulation: Defining and Benchmarking Grounded Generative Speech Recognition | ArxivCSExplorer