Question Answering on HotpotQA (test) (Accuracy, F1, Reference, and ASR Metrics)
79.85AccuracyML Routing
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| ML RoutingAdversarial attack=TextAttack2026.06 | 79.85 | 81.2 | 89.96 | 10 | |
| Geometric-Based ModelAdversarial attack=TextAttack2026.06 | 76.3 | 74.53 | — | 25.59 | |
| LWV FusionAdversarial attack=TextAttack2026.06 | 75.21 | 76.48 | 76.42 | 23.48 | |
| Entropy-Based ModelAdversarial attack=TextAttack2026.06 | 59.9 | 58.92 | 47.38 | 41.01 | |
| Uncertainty-Based ModelAdversarial attack=TextAttack2026.06 | 56.9 | 41.47 | 22.48 | 47.52 | |
| Base Model (no defence)Adversarial attack=TextAttack2026.06 | 50.99 | 49.26 | — | 52.62 | |
| FreeLBAdversarial attack=TextAttack2026.06 | 43.76 | 43.76 | — | 60.95 |