List your company

Faithfulness Evaluation on Fakepedia Conflicting long-context

849.1AOPCcompATTN
604.7668.15731.6795.05Oct 3, 2025

Evaluation results

MethodLinks
ATTNModel=Qwen2.5-7B2025.10849.11,028.22
FAModel=Qwen2.5-7B2025.10843.71,047.21
IGModel=Qwen2.5-7B2025.10834.61,052.99
MechLightModel=Qwen2.5-7B2025.108191,019.6
FAModel=Pythia-6.9B2025.10815.61,100.21
IGModel=Pythia-6.9B2025.10811.11,111.49
MechLightModel=Pythia-6.9B2025.10691.11,108.89
ATTNModel=Pythia-6.9B2025.10614.11,115.6