Fine-tuning Robustness against Harmful Data Attacks on GSM8K
0.95Harmful Score (Clean)
Lisa
Updated 1mo ago
Evaluation Results
| Method | Links | ||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
2026.02 | 0.95 | 3.05 | 3.82 | 4.62 | 5.86 | 8.16 | 4.41 | 9.97 | 9.57 | 9.6 | 8.9 | 9.23 | 9.2 | 9.41 | |
2026.02 | 0.95 | 1.14 | 0.95 | 1.14 | 1.24 | 1.29 | 1.12 | 15.83 | 14.57 | 15.4 | 15.4 | 15.07 | 14.37 | 15.12 | |
2026.02 | 1.29 | 7.15 | 13.07 | 19.55 | 23.6 | 29.33 | 15.67 | 12.27 | 12.23 | 11.77 | 11.77 | 11.77 | 11.77 | 11.93 | |
2026.02 | 1.38 | 1.76 | 2.91 | 5.1 | 9.06 | 13.49 | 5.62 | 16.53 | 15.97 | 16.5 | 16.3 | 16.43 | 15.7 | 16.24 | |
2026.02 | 2.05 | 10.68 | 15.59 | 19.06 | 23.94 | 27.85 | 16.53 | 12.23 | 11.83 | 11.43 | 10.9 | 10.9 | 10.53 | 11.3 |