List your company

First-error step detection on MoreHopQA

0.7885AUROCGIBS
0.2813960.4130480.54470.676352May 19, 2026

Evaluation results

MethodLinks
GIBSModel=Deepseek2026.050.7885
NLI-MaxModel=Deepseek2026.050.711
GIBSModel=Phi42026.050.6841
EntropyModel=Phi42026.050.6837
NLI-MaxModel=Phi42026.050.6632
NLI-MeanModel=Deepseek2026.050.6544
GIBSModel=Llama3.1-8b2026.050.6164
EntropyModel=Deepseek2026.050.6004
NLI-MeanModel=Phi42026.050.5824
EntropyModel=Llama3.1-8b2026.050.557
NLI-MaxModel=Llama3.1-8b2026.050.5375
Cos-MeanModel=Deepseek2026.050.5313
NLI-MeanModel=Llama3.1-8b2026.050.5285
P(true)Model=Deepseek2026.050.5152
P(true)Model=Llama3.1-8b2026.050.5126
Cos-MeanModel=Phi42026.050.5099
P(true)Model=Phi42026.050.509
SL(norm)Model=Llama3.1-8b2026.050.4941
Cos-MeanModel=Llama3.1-8b2026.050.4863
Cos-MaxModel=Phi42026.050.4772
SL(norm)Model=Phi42026.050.4724
Cos-MaxModel=Deepseek2026.050.4529
Cos-MaxModel=Llama3.1-8b2026.050.4275
LECOModel=Phi42026.050.3884
LECOModel=Llama3.1-8b2026.050.3839
LECOModel=Deepseek2026.050.3307
SL(norm)Model=Deepseek2026.050.3009