List your company

Stepwise error detection on MoreHopQA (test)

0.808AUROCGIBS
0.3920.50.6080.716May 19, 2026

Evaluation results

MethodLinks
GIBSLLM backbone=Deepseek2026.050.8080.83570.5
NIBSLLM backbone=Deepseek2026.050.6660.77664.9
GIBSLLM backbone=Phi42026.050.6610.68670.5
GIBSLLM backbone=Llama3.1-8b2026.050.6470.66956
PARCLLM backbone=Deepseek2026.050.6370.68665.9
PARCLLM backbone=Phi42026.050.6060.67665
PARCLLM backbone=Llama3.1-8b2026.050.5890.60851.1
NIBSLLM backbone=Phi42026.050.580.66369.6
ROSCOE-saLLM backbone=Llama3.1-8b2026.050.5240.56554.1
NIBSLLM backbone=Llama3.1-8b2026.050.5120.55353.1
ROSCOE-saLLM backbone=Deepseek2026.050.4920.661.2
ROSCOE-saLLM backbone=Phi42026.050.4880.59461.3
ROSCOE-ssLLM backbone=Llama3.1-8b2026.050.4770.51750.3
ROSCOE-ssLLM backbone=Phi42026.050.4210.60354.2
ROSCOE-ssLLM backbone=Deepseek2026.050.4080.58452.8