Messages-format version of the cleaned WildChat Qwen dataset.
Each row has a messages list:
69688 rows have the original two-message conversation: user, assistant.
30000 rows additionally include a simulated follow-up user turn and a generated assistant response: user, assistant, user, assistant.
Total rows: 99688.
The system prompt used for assistant generation is not included in the dataset.
See metadata/generation_report.json for generation metadata.
