Chirak

Runs

Each run turns a task description and a few examples into a LoRA-tuned small model.

RunStatusTarget BaselineBestIterations
Store review triage (BG) v2 r2-jev
3 weeks ago
completed macro-f1 95% 92.6% 97.7% 1 Report
Store review triage (BG) v2 r2-human
3 weeks ago
completed macro-f1 95% 91.6% 97.2% 1 Report
Store review triage (BG) v2 + own tickets
3 weeks ago
completed macro-f1 95% 90.3% 97.7% 1 Report
Store review triage (BG) v2 + Jev-cleaned dev
3 weeks ago
completed macro-f1 95% 88.3% 97.2% 1 Report
Store review triage (BG) v2
3 weeks ago
completed macro-f1 95% 94.0% 95.8% 1 Report
Store review triage (BG) + own tickets
3 weeks ago
completed macro-f1 95% 90.3% 97.7% 1 Report
Store review triage (BG)
3 weeks ago
completed macro-f1 95% 90.3% 98.1% 1 Report
Merchant ticket triage (BG/EN), guidelines v2 + real tickets (ABCD train)
3 weeks ago
completed exact-match 92% 65.1% 92.6% 2 Report
Merchant ticket triage (BG/EN), guidelines v2, target exact match
3 weeks ago
completed exact-match 92% 65.6% 94.9% 1 Report
Merchant ticket triage (BG/EN), guidelines v2
3 weeks ago
not worth it macro-f1 92% 93.5% — 0 Report
Merchant ticket triage (BG/EN)
3 weeks ago
completed macro-f1 92% 84.1% 98.2% 3 Report