OOD Detection on Spam (test)
84.8Length-Matched AUROCMaha-Traj
Evaluation Results
| Method | Links | |
|---|---|---|
| Maha-Trajcategory=Trajectory, backbone=Qwen3-0.6B, evaluation_mode=length-matched2026.04 | 84.8 | |
| Best Singlecategory=Trajectory, backbone=Qwen3-0.6B, evaluation_mode=length-matched2026.04 | 77.4 | |
| RAUQcategory=Attn-Based, backbone=Qwen3-0.6B, evaluation_mode=length-matched2026.04 | 52 | |
| Attn Ent.category=Attn-Based, backbone=Qwen3-0.6B, evaluation_mode=length-matched2026.04 | 44.8 | |
| CEDcategory=Attn-Based, backbone=Qwen3-0.6B, evaluation_mode=length-matched2026.04 | 43.2 |