WASIL: In-the-Wild Arabic Spoken Interactions with LLMs

In three linesWASIL is a dataset of 8,529 turns of Arabic voice interactions with LLMs, including audio, ASR hypotheses, responses, and user feedback (14.2% dislikes). Covers Modern Standard Arabic and 4 major dialects. Enables isolating ASR errors from intrinsic unanswerability via annotation and multi-judge LLM evaluation.

Summary generated by Claude — human-verified