Classification on MNLI (ECE, BS)
1.85ECETemperature Scaling
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Temperature ScalingLLM=Phi-4-mini (3.8B)2026.06 | 1.85 | 16.82 | |
| CaliDist (As.)LLM=Phi-4-mini (3.8B)2026.06 | 3.02 | 16.99 | |
| CaliDist (Pr.)LLM=Phi-4-mini (3.8B)2026.06 | 3.4 | 17.03 | |
| CaliDist (Sa.)LLM=Phi-4-mini (3.8B)2026.06 | 3.68 | 17.31 | |
| CaliDist (Pr.)LLM=Gemma-3 (4B)2026.06 | 3.71 | 23.29 | |
| CaliDist (Sa.)LLM=Gemma-3 (4B)2026.06 | 5.77 | 24.1 | |
| CaliDist (As.)LLM=Gemma-3 (4B)2026.06 | 6.45 | 24.51 | |
| VanillaLLM=Phi-4-mini (3.8B)2026.06 | 8.05 | 17.66 | |
| ConsistencyLLM=Phi-4-mini (3.8B)2026.06 | 9.7 | 18.53 | |
| FSDLLM=Phi-4-mini (3.8B)2026.06 | 15.16 | 21.69 | |
| EntropyLLM=Phi-4-mini (3.8B)2026.06 | 28.09 | 28.65 | |
| Temperature ScalingLLM=Gemma-3 (4B)2026.06 | 38.47 | 38.61 | |
| EntropyLLM=Gemma-3 (4B)2026.06 | 39.75 | 39.9 | |
| FSDLLM=Gemma-3 (4B)2026.06 | 39.8 | 39.98 | |
| ConsistencyLLM=Gemma-3 (4B)2026.06 | 40.21 | 40.25 | |
| VanillaLLM=Gemma-3 (4B)2026.06 | 40.22 | 40.16 |