Task
Document the new benchmark matrix flow clearly.
Requirements
- Where to run Q1~Q8 matrix
- Single-node baseline commands
- Expected output layout and fields
- How to aggregate + upload to HF
- Troubleshooting section (missing deps, partial failures)
Acceptance Criteria
- README/docs have copy-paste command path
- New users can reproduce matrix run and upload end-to-end
Task
Document the new benchmark matrix flow clearly.
Requirements
Acceptance Criteria