Metraipolis › archive › dc_eeb5b45bc49f
AQL-Lite: Enabling Efficient Knowledge Acquisition for Arabic LLMs via Data-Centric Continual Pretraining — arXiv:2610.30344
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.30344
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-23T15:20:24.784Z
- SHA-256 of the archived text
41a3c744ff7f56c8fc812b20e8a8a087579e64f5261faeffdf32267aa9c67b0a- Stable URL of this page
https://metraipolis.com/documento/dc_eeb5b45bc49f
How to cite
AQL-Lite: Enabling Efficient Knowledge Acquisition for Arabic LLMs via Data-Centric Continual Pretraining — arXiv:2610.30344 — arXiv (en), https://arxiv.org/abs/2610.30344 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_eeb5b45bc49f (Metraipolis archive dc_eeb5b45bc49f, sha256 41a3c744ff7f56c8fc812b20e8a8a087579e64f5261faeffdf32267aa9c67b0a)
BibTeX
@misc{dceeb5b45bc49f,
title = {AQL-Lite: Enabling Efficient Knowledge Acquisition for Arabic LLMs via Data-Centric Continual Pretraining — arXiv:2610.30344},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.30344},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_eeb5b45bc49f; sha256 41a3c744ff7f56c8fc812b20e8a8a087579e64f5261faeffdf32267aa9c67b0a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.30344. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Training Arabic large language models faces challenges from noisy web data, dialectal variation, and the high cost of continual pretraining (CPT). We introduce AQL-Lite (AQL: Arabic Quality Language), a time- and resource-efficient framework for adapting a multilingual LLM to Modern Standard Arabic (MSA). AQL-Lite couples a quality-first curation pipeline (structural cleaning, heuristic and model-based filtering, MSA selection, and deduplication) with a partial-parameter CPT recipe that trains RS-LoRA adapters across the network together with the last eight transformer blocks and the embedding matrix, under layer-wise learning rates. At a matched 1B-token budget, CPT on AQL-curated data raises a four-benchmark Arabic mean by 3.51 points over CPT on raw web text, which yields no gain over the base model. In a controlled 5B-token comparison, AQL-Lite recovers 94% of the gain of full-parameter CPT (mean 73.19 vs. 73.58) in 52% of the wall-clock time and outperforms adapter-only baselines on every benchmark, and a component ablation shows that adapters, unfrozen blocks, and embedding training each contribute. Scaled to 40B MSA tokens and instruction-tuned on one million examples, the resulting 12B model scores 75.3 on the seven-benchmark Open Arabic LLM Leaderboard v2 (OALL v2) suite. These results suggest that partial-parameter CPT on a carefully curated corpus is a practical route to efficient MSA adaptation. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL) ; Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2610.30344 — 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 2026-10-23T15:09:38.309Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).