List your company

Instruction Following on IFBench (Loose and Strict Success Rates)

72.2Loose Prompt Success RateVP2O
63.3665.65567.9570.245Jun 6, 2026

Evaluation results

MethodLinks
VP2OContext Length=16K2026.0672.274.966.269
VP2OContext Length=8K2026.0668.37161.764.6
GRPO-baselineContext Length=16K2026.0666.569.36163.7
GRPO-baselineContext Length=8K2026.0663.766.358.160.8