List your company

Tool Use on Causal and Downstream Robustness Ablation Suite Averaged over 4 models

4.1Tool Hit@1ΔHETA
0.461.4052.353.295Apr 14, 2026

Evaluation results

MethodLinks
HETAMethod Variant=Full2026.044.1
HETAMethod Variant=LR+WIN2026.043.8
HETAMethod Variant=w/o Hessian2026.042.7
HETAMethod Variant=w/o KL2026.042.3
ReAGentMethod Variant=Standard2026.042.1
HETAMethod Variant=w/o Transition2026.041.9
SEA-CoTMethod Variant=Standard2026.041.8
Progressive InferenceMethod Variant=Standard2026.041.6
fAMLMethod Variant=Standard2026.041.5
ContextCiteMethod Variant=Standard2026.041.4
TDD-backwardMethod Variant=Standard2026.041.3
Peering (PML)Method Variant=Standard2026.041.2
Integrated GradientsMethod Variant=Standard2026.041.1
Attention RolloutMethod Variant=Standard2026.040.6