license: apache-2.0
language:

• ar
  size_categories:
• 100K<n<1M
  task_categories:
• text-generation
  tags:
• arabic
• math
• reasoning
• chain-of-thought
• gsm8k
• synthetic
  configs:
• config_name: default
  data_files:
  • split: train
    path: arabic_math_reasoning_synth.parquet

Arabic Math Reasoning (synthetic) — مسائل رياضيات عربية مع خطوات الحل

120,462 Arabic grade-school math word problems, each with a step-by-step derivation and a
concluding sentence. Generated with gemma-3-12b-it and Qwen3.8-27B-Uncensored-NVFP4 and
arithmetically verified — every equation the reasoning states was re-evaluated, and rows whose
own arithmetic does not check out were dropped.

بالعربية: 120,462 مسألة حسابية عربية مع خطوات حل مفصّلة وجملة إجابة نهائية. جميع المعادلات
داخل خطوات الحل تم التحقق منها حسابيًا، والصفوف التي تحتوي على خطأ حسابي مستبعدة.

Schema

It matches the {instruction, reasoning, answer} schema used by the Nawah-Reasoning training
scripts, so it drops straight into ChatML + <think> SFT.

How it was built

Prompts are drawn from a variation grid — 30 domains × 23 operation types ×
step counts × number styles × 12 country/currency pairs × 8 structural twists —
seeded deterministically from the task index. This is a deliberate reaction to the corpus this was
built to supplement: a machine-translated Arabic GSM8K whose 142,969 rows expand from only
2,814 underlying question templates. Here, 120,462 distinct templates survive
number-masked deduplication.

The operation axis is drawn from two disjoint pools. The first 100,323 rows use 15
general operations (totals, percentages, unit rates, remainders, …). A later pass added a
relational pool of 8 operations — ضعف / أضعاف, نصف, ثلث و ربع, absolute increase and
decrease, three-way chains, and the inverse direction — contributing 20,139 rows. These
express a quantity relative to another entity ("Ziad bought double that number"), which
forces a derivation step that a percentage problem does not. In the original corpus ضعف /
أضعاف appeared in 1.34% of rows and أكثر بـ / أقل بـ in 0.26% — against 17.6% for percentages —
and that gap is the documented weak spot of models trained on it, so the pool was generated
deliberately rather than left to the grid.

Verification

Generated math is fluent and often wrong, so acceptance is mechanical rather than stylistic:

1. every a op b = c in the reasoning is re-evaluated with real arithmetic — one bad equation
   rejects the row;
2. the number in the answer sentence must equal the last computed result (a chain that is
   internally correct but ends on a different number is the failure that survives fluency checks);
3. no Latin characters, length bounds, and the answer must open with a conclusion marker.

64.3% of parsed items passed. Rejections:

out_synth/rejects.jsonl in the source repo keeps every rejected item with its reason, so the
filter itself can be audited.

Limitations

This is synthetic data: the problems are machine-written and reflect that model's habits of
phrasing, its distribution of scenarios, and its idea of what a "typical" Arabic math problem
looks like. Verification proves each row's arithmetic is self-consistent — it does not prove
the problem is well-posed, that the setup is the only reasonable reading, or that the wording is
natural to any particular dialect region (the axis_region field sets currency and context, not
dialect; everything is MSA). Rows whose reasoning states no explicit equation are dropped, which
biases the corpus toward problems that decompose into clean arithmetic steps.

Trained model

This dataset is part of the training mix for
oddadmix/Nawah-Math-Reasoning — a 51.8M-parameter Arabic model that reasons step by step
inside <think>…</think> before answering. The full training code, including the pipeline that
built this dataset, ships in code/ in that repo, and
there is a live demo Space.