List your company

Fact-checking on Causal and Downstream Robustness Ablation Suite Averaged over 4 models

3.7Fact EMΔHETA
0.3721.2362.12.964Apr 14, 2026

Evaluation results

MethodLinks
HETAMethod Variant=Full2026.043.7
HETAMethod Variant=LR+WIN2026.043.3
HETAMethod Variant=w/o Hessian2026.042.5
HETAMethod Variant=w/o KL2026.042
ReAGentMethod Variant=Standard2026.042
SEA-CoTMethod Variant=Standard2026.041.7
HETAMethod Variant=w/o Transition2026.041.6
Progressive InferenceMethod Variant=Standard2026.041.5
fAMLMethod Variant=Standard2026.041.3
ContextCiteMethod Variant=Standard2026.041.2
TDD-backwardMethod Variant=Standard2026.041.1
Peering (PML)Method Variant=Standard2026.041
Integrated GradientsMethod Variant=Standard2026.040.9
Attention RolloutMethod Variant=Standard2026.040.5