List your company

Abstention on AbstentionBench KUQ Cont

100Abstention F1Gemini 3
91.99294.07196.1598.229May 25, 2026

Evaluation results

MethodLinks
Gemini 3Model Type=Proprietary API Model2026.0510010010050
GPT-5.2Model Type=Proprietary API Model2026.0592.992.992.975
TIARBackbone=Llama-3.1-8B-Instruct2026.0592.992.992.975
Claude Sonnet 4.5Model Type=Proprietary API Model2026.0592.385.7100100