# Benchmark Summary ## Gov Chatbot Red-Team v2 | Suite | Full F1 | Full Exact F1 | Full Failures | Q8 F1 | Q8 Exact F1 | Q8 Failures | |---|---:|---:|---:|---:|---:|---:| | Gov chatbot red-team v2 | 1.0000 | 1.0000 | 0.0000 | 1.0000 | 1.0000 | 0.0000 | | Gov chatbot gap holdout v2 | 1.0000 | 1.0000 | 0.0000 | 1.0000 | 1.0000 | 0.0000 | ## ONNX q8 | Suite | F1 | Examples/s | |---|---:|---:| | Irish core | 1.0000 | 221.5743 | | UAT exact | 1.0000 | 188.6537 | | Fresh holdout | 0.9841 | 145.9336 | | Hardening exact | 1.0000 | 112.4948 | | Edge | 1.0000 | 235.2888 | | Multilingual PPSN overall | 0.9333 | 224.6955 | | Multilingual PPSN label-only | 1.0000 | — | ## Full checkpoint | Suite | F1 | Examples/s | |---|---:|---:| | Irish core | 1.0000 | 79.8593 | | UAT exact | 1.0000 | 29.1305 | | Fresh holdout | 0.9841 | 31.4254 | | Hardening exact | 1.0000 | 26.6732 | | Edge | 1.0000 | 52.4220 | | Multilingual PPSN overall | 0.9333 | 89.3536 | | Multilingual PPSN label-only | 1.0000 | — |