Result Inconsistency

[theartpiece]

Relevant library versions-

1. transformers==4.16.2
2. PyTorch==1.10.1
3. Python==3.7.0

Task- Paper-Label(L2) classification task

Dataset: CS subset

Observation
Results mentioned in the paper (avg across 5 runs)- (NDCG) 0.403±.041 (MRR) 0.439±.078
Results I got (average across 2 runs)- (NDCG) 0.44 (MRR) 0.52

Doubt
Your upper bound (for eg 0.403+0.041 is close to my average 0.44, etc) is my average.

Wanted to know why this difference occurs. If you know something about this, please let me know.