b1a5a88cb9
- Reject advisor result when Pass 3 scores < 6 or is_acceptable=false,
falls back to simple column classification instead of using bad output
- Improved Pass 2 prompt: explicit rules for label_mapping completeness,
{column_name} vs {column_name_name} for mapped labels, column_roles
must match which template uses them
- Build suggested_mapping from ALL template-referenced columns (not just
first match per role) — fixes hypothesis being dropped from SNLI mapping
- Guard against LLM returning literal string "null" for revised_system_prompt
- Always show AI Assist button when available, even when mapping looks complete
|
||
|---|---|---|
| .. | ||
| datasets | ||
| hardware | ||
| inference | ||
| models | ||
| paths | ||
| .gitkeep | ||
| cache_cleanup.py | ||
| transformers_version.py | ||
| utils.py | ||