Closed guotong1988 closed 5 years ago
Hi, No, unfortunately I was not able to reproduce DRMM results for Robust 04 - in my current research on neural IR models using MS MARCO I also observed DRMM is not outperforming BM25 (In contrast other models do outperform BM25 by a large margin on the MS MARCO passage re-ranking task). If you want to know more about the better models visit (with re-implementations in PyTorch): https://github.com/sebastian-hofstaetter/sigir19-neural-ir
Thank you!
Result in paper. Thank you! @sebastian-hofstaetter