Reproduction Beyond Benchmarks: ConstBERT and ColBERT-v2 Across Backends and Query Distributions
For researchers using multi-vector retrieval models, this work reveals fundamental architectural constraints that cannot be fixed by adaptation, highlighting reproducibility issues beyond numerical accuracy.
ConstBERT reproduces within 0.05% MRR@10 on MS-MARCO, but both ConstBERT and ColBERT-v2 suffer 86-97% drops on long narrative queries due to architectural limitations of the MaxSim operator, which plateaus at 20 words. Undocumented backend parameters cause an 8-point gap, and more fine-tuning data degrades performance by up to 29%.
Reproducibility must validate architectural robustness, not just numerical accuracy. We evaluate ColBERT-v2 and ConstBERT across five dimensions, finding that while ConstBERT reproduces within 0.05% MRR@10 on MS-MARCO, both models show a drop of 86-97% on long, narrative queries (TREC ToT 2025). Ablations prove this failure is architectural: performance plateaus at 20 words because the MaxSim operator's uniform token weighting cannot distinguish signal from filler noise. Furthermore, undocumented backend parameters create an 8-point gap due to ConstBERT's sparse centroid coverage, and fine-tuning with 3x more data actually degrades performance by up to 29%. We conclude that architectural constraints in multi-vector retrieval cannot be overcome by adaptation alone. Code: https://github.com/utshabkg/multi-vector-reproducibility.