List your company

Question Answering on QA domain average

85.2Best AccuracyTask-level fine-tuning
79.89681.27382.6584.027Jun 6, 2025

Evaluation results

MethodLinks
Task-level fine-tuning2025.0685.2—
Many-shot (ours)Masking Strategy=mask all targets2025.0684.584.5
Many-shot (ours)Masking Strategy=mask last target2025.068484
Few-shotMasking Strategy=mask all targets2025.0682.282.2
Zero shot2025.0681.881.3
Few-shotMasking Strategy=mask last target2025.0681.881.8
Many-shotStrategy=autoregressive2025.0680.880.8
No fine-tuning2025.0680.179.8