Answer Correctness Prediction on HotpotQA (val)
0.6Head EntropyHEAD ENTROPY
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| HEAD ENTROPYprediction_input=Question tokens only, prediction_stage=Pre-generation, aggregation=Average across Llama-3.2-3B-Instruct, Llama-3.1-8B-Instruct, Qwen3-1.7B, Qwen3-8B, Qwen3-32B2026.02 | 0.6 | — | 0.05 | |
| Hidden State LRprediction_input=Question tokens only, prediction_stage=Pre-generation, aggregation=Average across Llama-3.2-3B-Instruct, Llama-3.1-8B-Instruct, Qwen3-1.7B, Qwen3-8B, Qwen3-32B2026.02 | — | 0.56 | — |