【Conference】BENCHMARKING JAPANESE SPEECH RECOGNITION ON ASR-LLM SETUPS WITH MULTI-PASS AUGMENTED GENERATIVE ERROR CORRECTION