Commonsense Reasoning on CSQA (ASRr and ACCc)
92.6ASRrCFT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CFTDefense scheme=CFT, Backbone=GPT-OSS-20B2026.04 | 92.6 | 89.6 | |
| No defenseDefense scheme=No defense, Backbone=GPT-OSS-20B2026.04 | 90.4 | 88.1 | |
| ONIONDefense scheme=ONION, Backbone=GPT-OSS-20B2026.04 | 64.6 | 86.5 | |
| CoSDefense scheme=CoS, Backbone=GPT-OSS-20B2026.04 | 64.2 | 84.3 | |
| ShuffleDefense scheme=Shuffle, Backbone=GPT-OSS-20B2026.04 | 38.2 | 75.3 | |
| Shuffle++Defense scheme=Shuffle++, Backbone=GPT-OSS-20B2026.04 | 3.8 | 46.3 | |
| Critical-CoTDefense scheme=Ours, Backbone=GPT-OSS-20B2026.04 | 0.72 | 86.8 |