QMD Query Expansion 1.7B (SFT)
Updated Qwen3-1.7B model for query expansion using the production SFT pipeline.
Prompt Format
The model is trained on messages formatted with the Qwen3 chat template using:
/no_think Expand this search query: <query>
Notes
This checkpoint is the SFT-only trained version (GRPO is not part of the default pipeline).
