Metraipolis › archive › dc_bc336921f375
ItColBERT: An Italian-Specialised Late-Interaction Retriever — arXiv:2609.26856
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.26856
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-08T12:27:45.899Z
- SHA-256 of the archived text
b258b0a139c0160b651c43a499add5844ee8b4d25258299107b4e7738dd06cb6- Stable URL of this page
https://metraipolis.com/documento/dc_bc336921f375
How to cite
ItColBERT: An Italian-Specialised Late-Interaction Retriever — arXiv:2609.26856 — arXiv (en), https://arxiv.org/abs/2609.26856 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_bc336921f375 (Metraipolis archive dc_bc336921f375, sha256 b258b0a139c0160b651c43a499add5844ee8b4d25258299107b4e7738dd06cb6)
BibTeX
@misc{dcbc336921f375,
title = {ItColBERT: An Italian-Specialised Late-Interaction Retriever — arXiv:2609.26856},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.26856},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_bc336921f375; sha256 b258b0a139c0160b651c43a499add5844ee8b4d25258299107b4e7738dd06cb6},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.26856. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Neural information retrieval for Italian is served almost entirely by multilingual models. Several multi-vector (late-interaction) retrievers include Italian among dozens of languages, and several strong Italian dense embedders exist, but as of August 2026 no late-interaction retriever specialised on Italian had been released. We present ItColBERT, a 135M-parameter Italian ColBERT trained with PyLate following the ColBERT-Zero recipe: initialise from a checkpoint that already retrieves, then apply supervised contrastive training followed by single-teacher distillation, for a total of roughly 14.5 GPU-hours on one RTX 3090. Across four Italian retrieval benchmarks it outperforms every general-purpose late-interaction baseline we tested except one (mLateOn), at 2-4.4x fewer parameters than every baseline but one of comparable size. Our principal empirical finding is methodological and partly negative. On the only cleanly out-of-domain benchmark (MLDR-it), an inference-time chunking recipe applied to an unchanged checkpoint yields +0.0602 nDCG@10 (p = 0.0225), a larger effect than anything two further rounds of training produced. Self-mined hard negatives and native 1024-token training were both evaluated against pre-registered decision gates and both failed. We report every comparison with paired bootstrap tests against an empirically measured noise floor of 0.0030 nDCG@10, and we release the weights, the training and evaluation code, and the complete experimental record including the rejected rounds. Tema degli esploratori: statistica e scienza dei dati (retro-catalogo arXiv: annunci del 2026-09-24 in cs.IR). Categorie arXiv: Information Retrieval (cs.IR). Pagina del lavoro: https://arxiv.org/abs/2609.26856 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-08T12:11:32.107Z dalla pagina https://arxiv.org/catchup/cs.IR/2026-09-24?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).