List your company

Reading Comprehension on SQuAD v2.0

50.64Disambiguation ScoreRandom
15.716824.783433.8542.9166Mar 20, 2026

Evaluation results

MethodLinks
RandomModel=Llama-3.1-8B-Instruct2026.0350.6452.391.75
IFRModel=Llama-3.1-8B-Instruct2026.0350.1662.5112.35
GradientModel=Llama-3.1-8B-Instruct2026.0349.9546.68-3.27
RolloutModel=Llama-3.1-8B-Instruct2026.0340.5869.4328.85
Input X GradientModel=Llama-3.1-8B-Instruct2026.0334.0763.2229.15
Last layerModel=Llama-3.1-8B-Instruct2026.0321.9682.0460.08
DePassModel=Llama-3.1-8B-Instruct2026.0319.2978.4559.16
Integrated GradientsModel=Llama-3.1-8B-Instruct2026.0317.9279.5361.61
Dual Path Attribution (DPA)Model=Llama-3.1-8B-Instruct2026.0317.7392.5174.78
MeanModel=Llama-3.1-8B-Instruct2026.0317.0689.4672.4