List your company
ResearchFormerly WizWand

Attribution faithfulness on LongRA

10.8Soft-NC ScoreHETA
-1.97121.34444.667.9756Apr 14, 2026
Updated 6mo ago

Evaluation results

MethodLinks
HETAModel=Phi-3-Medium-14B2026.0410.82.35
HETAModel=GPT-J 6B2026.0410.32.31
HETABackbone=Qwen2.5 3B2026.0410.12.5
HETAModel=LLaMA-3.1 70B2026.049.92.6
Peering (PML)Model=Phi-3-Medium-14B2026.042.150.49
Peering (PML)Backbone=Qwen2.5 3B2026.042.050.61
Peering (PML)Model=GPT-J 6B2026.042.050.5
Integrated GradientsModel=Phi-3-Medium-14B2026.041.950.44
ContextCiteBackbone=Qwen2.5 3B2026.041.90.56
Integrated GradientsModel=GPT-J 6B2026.041.870.45
ReAgentModel=GPT-J 6B2026.041.680.37
ReAgentBackbone=Qwen2.5 3B2026.041.660.38
ReAgentModel=Phi-3-Medium-14B2026.041.660.38
ReAgentModel=LLaMA-3.1 70B2026.041.550.36
SEA-CoTModel=GPT-J 6B2026.041.540.32
SEA-CoTBackbone=Qwen2.5 3B2026.041.50.31
ContextCiteModel=Phi-3-Medium-14B2026.041.50.04
SEA-CoTModel=Phi-3-Medium-14B2026.041.50.31
ContextCiteModel=GPT-J 6B2026.041.420.03
Progressive InferenceModel=GPT-J 6B2026.041.350.28
SEA-CoTModel=LLaMA-3.1 70B2026.041.350.3
Progressive InferenceBackbone=Qwen2.5 3B2026.041.30.25
Progressive InferenceModel=Phi-3-Medium-14B2026.041.30.25
Integrated GradientsBackbone=Qwen2.5 3B2026.041.20.14
Progressive InferenceModel=LLaMA-3.1 70B2026.041.20.24
ContextCiteModel=LLaMA-3.1 70B2026.041.170.58
TDD-backwardModel=GPT-J 6B2026.041.1-0.12
TDD-backwardBackbone=Qwen2.5 3B2026.041.05-0.08
TDD-backwardModel=Phi-3-Medium-14B2026.041.05-0.1
fAMLModel=LLaMA-3.1 70B2026.040.46-0.21
Attention RolloutModel=GPT-J 6B2026.040.41-0.01
Attention RolloutModel=Phi-3-Medium-14B2026.040.39-0.02
Attention RolloutBackbone=Qwen2.5 3B2026.040.38-0.03
fAMLBackbone=Qwen2.5 3B2026.040.23-0.09
fAMLModel=Phi-3-Medium-14B2026.040.23-0.09
fAMLModel=GPT-J 6B2026.040.21-0.1
Peering (PML)Model=LLaMA-3.1 70B2026.040.150.15
Integrated GradientsModel=LLaMA-3.1 70B2026.040.130.13
TDD-backwardModel=LLaMA-3.1 70B2026.04-0.02-0.11
Attention RolloutModel=LLaMA-3.1 70B2026.04-1.480.01