BayanSimplify-v0.2-Fast

Model 2 of the Bayan project, compact: AraBART on the same 22,733-pair mix, the app's fast default.

What it does

The same data and tags as BayanSimplify-v0.2 on AraBART (139M parameters, a third of the size), so it fits a 250 MB download as an int8 bundle (222 MB) and shows the first word on a mid-range phone in 82–172 ms.

Input: بسط: (without the shadda, which AraBART's tokenizer cannot read), then an optional tag such as [S2] , then the text.

BayanBench v2.0, test, core items:

Setting	Meaning kept	Longest clause, words cut
[S2], float	40.1%	6.7
no tag, float	51.4%	4.9
no tag, int8 in the app	50.5%	4.8

Fast, but it keeps the meaning less often than the AraT5v2 models; the app offers BayanSimplify-v0.3 as its large model.

Use

from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

repo = "Congi-libya/BayanSimplify-v0.2-Fast"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForSeq2SeqLM.from_pretrained(repo)

text = "وقد أدى التوسع العمراني السريع الذي شهدته المدينة خلال العقدين الماضيين إلى ازدحام مروري خانق في ساعات الذروة."
x = tok("بسط: " + text, return_tensors="pt", max_length=256, truncation=True)
y = model.generate(**x, num_beams=1, no_repeat_ngram_size=3, max_length=256)
print(tok.decode(y[0], skip_special_tokens=True))

The BayanSimplify family

Model	What it is	Base model	Training data	In the Bayan app
BayanSimplify-v0.1	Model 1	AraT5v2-base-1024	SAMER, level 5 → 3	not shipped
BayanSimplify-v0.2	Model 2	AraT5v2-base-1024	22,733-pair mix (SAMER, Baseet, DAASI) with strength tags	"Earlier large model"
BayanSimplify-v0.2-Fast	Model 2, compact	AraBART	the same mix	"Fast model", the default download
BayanSimplify-v0.3	Model 3, our best	AraT5v2-base-1024	Bayan corpus v1, 14,975 rows	"Large model", four beams
BayanSimplify-ONNX	The int8 bundles the app downloads			v0.2-Fast, v0.2, v0.3

Licence and data

CC BY-NC 4.0 (non-commercial), in line with the training data's terms. SAMER is used under the CAMeL Lab's permission to fine-tune and share weights for non-commercial use; its text is not redistributed here.

About Bayan

Bayan (بيان) simplifies Arabic text for readers with dyslexia, entirely on an Android phone: select text in any app, choose «تبسيط» (Simplify), and a simpler version appears over the page. Built by Team Cogni for the Samsung Innovation Campus AI capstone, 2026. Every number on this card comes from the team's final report and BayanBench v2.0, the benchmark built for the task (meaning scored by Gemma 4 31B, checked against human raters, AUC 0.85).