List your company

Abstention on AbstentionBench BBQ

97Abstention F1Gemini 3
85.5688.5391.594.47May 25, 2026

Evaluation results

MethodLinks
Gemini 3Model Type=Proprietary API Model2026.059796.19867.3
Claude Sonnet 4.5Model Type=Proprietary API Model2026.0587.910078.559.2
TIARBackbone=Llama-3.1-8B-Instruct2026.0587.596.180.373.5
GPT-5.2Model Type=Proprietary API Model2026.058684.387.873.5