List your company

Hallucination prediction on Mintaka refined by question type and domain

75.51AUROCConf (SCAO)
65.983668.456870.9373.4032Sep 18, 2025

Evaluation results

MethodLinks
Conf (SCAO)Model Size=8B, SCAO=true2025.0975.51—
Conf (SCAO)Model Size=8B2025.0975.51—
Conf + Probe (SCAO)Model Size=8B, SCAO=true2025.0974.8913.27
Conf + Probe (SCAO)Model Size=8B2025.0974.8913.27
Conf + Probe (SCAO)Model Size=70B, SCAO=true2025.0972.8914.12
Conf + Probe (SCAO)Model Size=70B2025.0972.8914.12
Probe_dnnModel Size=8B2025.0972.7911.17
Probe_dnnModel Size=8B2025.0972.7911.17
Conf (SCAO)Model Size=70B, SCAO=true2025.0972.46—
Conf (SCAO)Model Size=70B2025.0972.46—
Probe_dnnModel Size=70B2025.0971.0312.26
Probe_dnnModel Size=70B2025.0971.0312.26
Conf + ProbeModel Size=70B2025.0970.812.03
Conf + ProbeModel Size=70B2025.0970.812.03
Conf + ProbeModel Size=8B2025.0970.779.15
Conf + ProbeModel Size=8B2025.0970.779.15
ConfModel Size=8B2025.0969.23—
ConfModel Size=8B2025.0969.23—
ConfModel Size=70B2025.0966.35—
ConfModel Size=70B2025.0966.35—