Answer Correctness Prediction on TriviaQA, HotpotQA, and MedMCQA Average (val)
0.68Head EntropyHEAD ENTROPY
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HEAD ENTROPYModel=Llama 3.2 3B2026.02 | 0.68 | — | 0.03 | |
| HEAD ENTROPYModel=Llama 3.1 8B2026.02 | 0.68 | — | 0.05 | |
| HEAD ENTROPYModel=Average2026.02 | 0.76 | — | 0.06 | |
| HEAD ENTROPYModel=Qwen3 1.7B2026.02 | 0.8 | — | 0.12 | |
| HEAD ENTROPYModel=Qwen3 32B2026.02 | 0.8 | — | 0.01 | |
| HEAD ENTROPYModel=Qwen3 8B2026.02 | 0.81 | — | 0.09 | |
| LookbackLensModel=Llama 3.2 3B2026.02 | — | 0.65 | — | |
| LookbackLensModel=Llama 3.1 8B2026.02 | — | 0.63 | — | |
| LookbackLensModel=Qwen3 1.7B2026.02 | — | 0.69 | — | |
| LookbackLensModel=Qwen3 8B2026.02 | — | 0.73 | — | |
| LookbackLensModel=Qwen3 32B2026.02 | — | 0.79 | — | |
| LookbackLensModel=Average2026.02 | — | 0.7 | — |