List your company

Abstention on AbstentionBench FreshQA

75.2Abstention F1Claude Sonnet 4.5
31.10442.5525465.448May 25, 2026

Evaluation results

MethodLinks
Claude Sonnet 4.5Model Type=Proprietary API Model2026.0575.273.277.463.6
TIARBackbone=Llama-3.1-8B-Instruct2026.0568.471.465.652.3
GPT-5.2Model Type=Proprietary API Model2026.0548.835.776.970.5
Gemini 3Model Type=Proprietary API Model2026.0532.819.610081.8