Tag
This paper investigates whether LLMs provide grounded pronunciation feedback to L2 English learners, finding that LLMs often rely on stereotypes and prior knowledge rather than acoustic evidence, leading to inaccurate but coherent feedback.