List your company

Question Answering on SQuAD (F1 and Time)

0.839F1 ScoreUnwatermarked
0.378280.497890.61750.73711May 28, 2026

Evaluation results

MethodLinks
UnwatermarkedBase Model=Ministral3-8B2026.050.8391
De-markBase Model=Ministral3-8B2026.050.83236.42
WASHBase Model=Ministral3-8B, n=52026.050.8211.3
WASHBase Model=Ministral3-8B, n=42026.050.8071.3
WASHBase Model=Ministral3-8B, n=32026.050.8051.19
UnwatermarkedBase Model=Llama3.1-8B2026.050.7951
Watermarked BaselineBase Model=Ministral3-8B2026.050.7951
ToBlendBase Model=Ministral3-8B2026.050.77515.92
WASHBase Model=Llama3.1-8B, n=42026.050.7651.53
WASHBase Model=Llama3.1-8B, n=32026.050.7641.49
WASHBase Model=Llama3.1-8B, n=52026.050.7641.54
Watermarked BaselineBase Model=Llama3.1-8B2026.050.7491
De-markBase Model=Llama3.1-8B2026.050.64232.5
ToBlendBase Model=Llama3.1-8B2026.050.63317.55
WASHBase Model=Qwen3-8B, n=42026.050.6291.82
WASHBase Model=Qwen3-8B, n=52026.050.6121.96
WASHBase Model=Qwen3-8B, n=32026.050.611.69
ToBlendBase Model=Qwen3-8B2026.050.47419.18
De-markBase Model=Qwen3-8B2026.050.43430.98
UnwatermarkedBase Model=Qwen3-8B2026.050.4211
Watermarked BaselineBase Model=Qwen3-8B2026.050.3961