Records · kv:2610.00038v1

Reproduction of FlagOpen/FlagEmbedding: BGE-base-en-v1.5 on BEIR SciFact, NFCorpus and FiQA

K-Veritas Team

Submitted by K-Veritas Team

Published 2026-10-11

Independent reproduction · Artifact evaluation · Natural language processing

PaperReproduction of: original code

Abstract

Evaluated the BAAI/bge-base-en-v1.5 embedding model with the repository's BEIR evaluation module (FlagEmbedding.evaluation.beir) at the pinned commit: dense retrieval without a reranker on the test splits of SciFact, NFCorpus and FiQA, top 1000 retrieved, reporting nDCG@10 and Recall@100. Corpus embeddings were computed during the run. The research/ and examples/finetune/ folders are withheld as hash-only entries since this run does not use them. PyTorch 2.11 (CUDA 12.8). One NVIDIA RTX 5060 Ti.

Sealed report

Cite this record

@misc{kveritas261000038v1,
  title={Reproduction of FlagOpen/FlagEmbedding: BGE-base-en-v1.5 on BEIR SciFact, NFCorpus and FiQA},
  author={K-Veritas Team},
  year={2026},
  howpublished={K-Veritas Records},
  note={kv:2610.00038v1},
  url={https://kveritas.org/records/2610.00038v1},
}

References

1. [1] Mamadou K. Keita and Christopher Homan. Computer Science Conferences Should Require Nonrepudiable Experimental Results. NeurIPS 2026 Position Paper Track. arXiv:2605.08586.

Sealed with K-Veritas [1]. Listed, not endorsed. A record shows these results came from the sealed code, unchanged. Whether the work is sound is for the reader to judge.