Classification Calibration on HellaSwag
2.38ECECaliDist (Pr.)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CaliDist (Pr.)LLM=Gemma-3 (4B)2026.06 | 2.38 | 23.97 | |
| CaliDist (As.)LLM=Gemma-3 (4B)2026.06 | 2.89 | 24.69 | |
| CaliDist (Sa.)LLM=Gemma-3 (4B)2026.06 | 3.08 | 24.08 | |
| Temperature ScalingLLM=Phi-4-mini (3.8B)2026.06 | 3.5 | 18.65 | |
| CaliDist (As.)LLM=Phi-4-mini (3.8B)2026.06 | 5.75 | 18.9 | |
| CaliDist (Sa.)LLM=Phi-4-mini (3.8B)2026.06 | 5.78 | 18.91 | |
| CaliDist (Pr.)LLM=Phi-4-mini (3.8B)2026.06 | 6.09 | 18.93 | |
| VanillaLLM=Phi-4-mini (3.8B)2026.06 | 8.8 | 19.31 | |
| ConsistencyLLM=Phi-4-mini (3.8B)2026.06 | 11.69 | 20.96 | |
| FSDLLM=Phi-4-mini (3.8B)2026.06 | 13.48 | 22.43 | |
| EntropyLLM=Phi-4-mini (3.8B)2026.06 | 25.23 | 27.64 | |
| Temperature ScalingLLM=Gemma-3 (4B)2026.06 | 41.6 | 41.73 | |
| EntropyLLM=Gemma-3 (4B)2026.06 | 43.32 | 43.79 | |
| FSDLLM=Gemma-3 (4B)2026.06 | 43.61 | 44.05 | |
| ConsistencyLLM=Gemma-3 (4B)2026.06 | 44.64 | 44.55 | |
| VanillaLLM=Gemma-3 (4B)2026.06 | 44.68 | 44.55 |