arXiv cs.CL· Ali Khairallah, Arkaitz Zubiaga·· 11 小时前AI 评分18
AK_QMUL 在 ArGuard 共享任务：单一微调 JAIS 用于阿拉伯语有害提示检测
AK_QMUL at ArGuard Shared Tasks: A Single Fine-Tuned JAIS for Arabic Harmful-Prompt Detection
AI 导读
AK_QMUL 在 ArGuard-2026 阿拉伯语有害提示检测两项共享任务中均排名第一，两个提交共用同一骨干 JAIS-6.7B，采用 4-bit QLoRA 微调并在末位 token 表示上接线性分类头。
正文 · 原文
Abstract:We present AK_QMUL's submissions to the ArGuard-2026 shared tasks on Arabic harmful-prompt detection: a binary safe/unsafe task (B1) and a seven-way harm-domain task (B2). Both submissions use one backbone -- JAIS-6.7B fine-tuned with 4-bit QLoRA and a linear classification head over its last-token representation. For B1 we train a four-way head that factors the label by dialect register {safe, unsafe} x {MSA, other}, we augment the training fold with MSA prompts from AraSafe, and decode with a plain argmax to score safe/unsafe. For B2 we fine-tune the same backbone as a seven-way classifier and average the softmax outputs of five random seeds. Both submissions ranked first on their track, with macro-F1 0.984 on B1 and 0.790 on B2. We report results by dialect and by class: on B1 the model is accurate across all six dialects (93.6--99.8%) and its predicted unsafe rate tracks the true rate closely; on B2 six of the seven classes reach F1 between 0.73 and 0.98, with the residual error concentrated in the Bully class, which overlaps semantically with harassment and harm-to-others.
| Comments: | Published at ArabicNLP 2026 (Shared Tasks), co-located with EMNLP 2026. ACL Anthology: this https URL |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.36360 [cs.CL] |
| (or arXiv:2610.36360v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.36360 arXiv-issued DOI via DataCite (pending registration) |
|
| Journal reference: | Proceedings of The Fourth Arabic Natural Language Processing Conference: Shared Tasks, pages 347-351 (2026) |
Submission history
From: Ali Khairallah [view email]
[v1]
Mon, 26 Oct 2026 22:25:18 UTC (25 KB)
来源：arXiv cs.CL · arxiv.org