Theorem Question Answering on ThmQA
26.12Pass@1Teacher
Evaluation Results
| Method | Links | |
|---|---|---|
| TeacherModel=Gemma3-4B2026.05 | 26.12 | |
| Bridge-Garden (Skew FKL)Distillation Family=Distance & Skewed Family2026.05 | 5.24 | |
| Skew FKLDistillation Family=Distance & Skewed Family2026.05 | 5.01 | |
| Bridge-Garden (Reverse KL)Distillation Family=KL Divergence Family2026.05 | 4.57 | |
| Bridge-Garden (α-β divergence)Distillation Family=Unified Divergence Family2026.05 | 4.54 | |
| α-β divergenceDistillation Family=Unified Divergence Family2026.05 | 4.34 | |
| Forward KLDistillation Family=KL Divergence Family2026.05 | 4.25 | |
| Bridge-Garden (Forward KL)Distillation Family=KL Divergence Family2026.05 | 4.22 | |
| Bridge-Garden (Total Variation)Distillation Family=Distance & Skewed Family2026.05 | 4 | |
| Bridge-Garden (Skew RKL)Distillation Family=Distance & Skewed Family2026.05 | 3.41 | |
| Skew RKLDistillation Family=Distance & Skewed Family2026.05 | 3.38 | |
| Hard KDTeacher Model=Gemma3-4B, Student Model=Gemma3-1B2026.05 | 3.15 | |
| Total VariationDistillation Family=Distance & Skewed Family2026.05 | 1.29 | |
| Reverse KLDistillation Family=KL Divergence Family2026.05 | 1.26 | |
| Student (No Distill)Model=Gemma3-1B2026.05 | 0.95 |