id int64 1 10 | poet stringclasses 7
values | title stringclasses 10
values | form stringclasses 6
values | verse_count int64 2 8 | theme listlengths 2 2 | meter stringclasses 9
values | verified bool 1
class | notes stringclasses 6
values | selected_section stringclasses 1
value | text listlengths 2 8 |
|---|---|---|---|---|---|---|---|---|---|---|
1 | حافظ | الا یا ایها الساقی | غزل | 7 | [
"عاشقانه",
"عرفانی"
] | مفاعیلن مفاعیلن مفاعیلن مفاعیلن (هزج مثمن سالم) | true | بیت آخر به عربی است (بخش اصیل شعر). قبلاً در نسخهی اولیه دو بیت پایانی جا افتاده بود؛ اکنون کامل شده. | [
"الا یا ایها الساقی ادر کأساً و ناولها / که عشق آسان نمود اول ولی افتاد مشکلها",
"به بوی نافهای کاخر صبا زان طره بگشاید / ز تاب جعد مشکینش چه خون افتاد در دلها",
"مرا در منزل جانان چه امن عیش چون هر دم / جرس فریاد میدارد که بربندید محملها",
"به می سجاده رنگین کن گرت پیر مغان گوید / که سالک بیخبر نبود ز ... | |
2 | خیام | رباعی شمارهٔ ۶۴ | رباعی | 2 | [
"فلسفی",
"هستیشناسی"
] | مفعول مفاعیل مفاعیل فعل (وزن رباعی) | true | null | [
"ای بس که نباشیم و جهان خواهد بود / نی نام ز ما و نی نشان خواهد بود",
"زین پیش نبودیم و نَبُد هیچ خِلَل / زین پس چو نباشیم همان خواهد بود"
] | |
3 | خیام | رباعی شمارهٔ ۶۶ | رباعی | 2 | [
"فلسفی",
"دمغنیمتی"
] | مفعول مفاعیل مفاعیل فعل (وزن رباعی) | true | null | [
"این قافلهٔ عمر عجب میگذرد / دریاب دمی که با طَرَب میگذرد",
"ساقی، غم فردای حریفان چه خوری / پیش آر پیاله را، که شب میگذرد"
] | |
4 | باباطاهر | دوبیتی شمارهٔ ۱۰۹ | دوبیتی | 2 | [
"عاشقانه",
"انتظار"
] | مفاعیلن مفاعیلن فعولن (هزج مسدس محذوف) | true | به گویش محلی (فهلویات همدانی/لری) است، نه فارسی معیار — این ویژگی اصیل و مهم این قالب است. | [
"همه شو تا سحر اختر شمارِم / که ماه رویت آیو در کنارِم",
"شوان گوشِم به در چَشمِم به راهِت / گذاری تا به کی در انتظارِم؟"
] | |
5 | سعدی | قطعه شمارهٔ ۳ | قطعه | 2 | [
"طنز",
"اجتماعی"
] | مضارع مثمن اخرب مکفوف محذوف (مفعول فاعلات مفاعیل فاعلن) | true | null | [
"گفتم چه کردهام که نگاهم نمیکنی؟ / وآن دوستی که داشتی اول چرا کم است؟",
"گفتا به جرم آنکه به هفتاد سالگی / سودای سور میپزی و جای ماتم است؟"
] | |
6 | حافظ | قطعه شمارهٔ ۱ | قطعه | 2 | [
"اخلاقی",
"دینی"
] | نامشخص (بیت دوم عربی و برگرفته از قرآن است) | true | بیت دوم عیناً آیهای از قرآن (سوره طلاق) است که حافظ در شعرش گنجانده. | [
"تو نیک و بد خود هم از خود بپرس / چرا بایدت دیگری محتسب",
"وَمَن یَتَّقِ اللَّهَ یَجْعَل لَّهُ مَخْرَجًا / وَیَرْزُقْهُ مِنْ حَیْثُ لَا یَحْتَسِبُ"
] | |
7 | سعدی | به جهان خرم از آنم (غزل شمارهٔ ۱۳، بخش مواعظ) | غزل | 8 | [
"عرفانی",
"عاشقانه"
] | فعلاتن فعلاتن فعلاتن فعلن (رمل مثمن مخبون محذوف) | true | توجه: این شعر ۸ بیت دارد، یک بیت بیشتر از سقف بازهٔ سطح اول (۱-۷). بهدلیل کیفیت استثنایی و شهرت شعر نگه داشته شد؛ در صورت نیاز به رعایت دقیق بازه، حذف یا به سطح دوم منتقل شود. | [
"به جهان خرم از آنم که جهان خرم از اوست / عاشقم بر همه عالم که همه عالم از اوست",
"به غنیمت شمر ای دوست دم عیسی صبح / تا دل مرده مگر زنده کنی کاین دم از اوست",
"نه فلک راست مسلم نه ملک را حاصل / آنچه در سر سویدای بنیآدم از اوست",
"به حلاوت بخورم زهر که شاهد ساقیست / به ارادت ببرم درد که درمان هم از اوست",
... | |
8 | فردوسی | آغاز کتاب شاهنامه (۷ بیت نخست) | مثنوی | 7 | [
"حکمی",
"دینی"
] | فعولن فعولن فعولن فعل (متقارب مثمن محذوف، وزن شاهنامه) | true | این ۷ بیت، آغازِ بخش «آغاز کتاب» شاهنامه است که در کل ۱۵ بیت دارد؛ اینجا بهعنوان یک برش مستقل و معنادار از یک اثر بلندتر آمده، نه یک شعر کاملاً مستقل. | [
"به نام خداوند جان و خرد / کز این برتر اندیشه بر نگذرد",
"خداوند نام و خداوند جای / خداوند روزی ده رهنمای",
"خداوند کیوان و گردان سپهر / فروزندهٔ ماه و ناهید و مهر",
"ز نام و نشان و گمان برتر است / نگارندهٔ بر شده پیکر است",
"به بینندگان آفریننده را / نبینی مرنجان دو بیننده را",
"نیابد بدو نیز اندیشه راه ... | |
9 | مولانا | رباعی شمارهٔ ۱ (دیوان شمس) | رباعی | 2 | [
"عرفانی",
"وحدت وجود"
] | وزن رباعی | true | null | [
"آن دل که شد او قابلِ انوارِ خدا / پُر باشد جانِ او ز اسرارِ خدا",
"زنهار تنِ مرا چو تنها مَشمُر / کو جمله نمک شد بنمکزارِ خدا"
] | |
10 | پروین اعتصامی | گل و خار (شمارهٔ ۱۳۱) | قطعه/تمثیل | 7 | [
"اخلاقی",
"اجتماعی"
] | نامشخص در منبع | true | این مورد جایگزینِ آیتم جعلیِ «گوهر انسان» در نسخهی اولیه شد. | [
"در باغ، وقت صبح چنین گفت گل به خار / کز خویش، هیچ نایدت ای زشت روی عار",
"گلزار، خانهٔ گل و ریحان و سوسن است / آن به که خار، جای گزیند به شورهزار",
"پژمرده خاطر است و سرافکنده و نژند / در باغ، هر که را نبود رنگ و بو و بار",
"با من ترا چه دعوی مهر است و همسری / ناچیزی توام، همه جا کرد شرمسار",
"در صحبت تو،... |
Khayyam Challenge
Khayyam Challenge is a benchmark for evaluating whether AI models can move between classical Persian poetry and natural-language Persian prose without losing meaning, imagery, and literary structure. It supports two directions:
- Poem → Prose: given a classical Persian poem, produce a faithful prose description.
- Prose → Poem: given a prose description (with the original poem hidden), reconstruct a new poem that preserves the source poem's meaning, imagery, and — ideally — its poetic structure.
The dataset pairs 20 classical Persian poems (Hafez, Ferdowsi, and others) with two independently AI-generated prose "descriptions" of each poem, produced by two different LLMs (internally referred to as Luna and Terra). The prose versions strip away verse form and direct wording while preserving core meaning, so a model attempting the Prose → Poem direction cannot simply copy the original.
This dataset accompanies the Persian Poetry Challenge framework, which also provides an automated pipeline for running generation + AI-judge scoring against this data.
Dataset structure
The dataset has three configs, each split into low / medium / long by poem length:
| Config | Description | Rows per split |
|---|---|---|
poems |
The 20 source poems with metadata | low: 10, medium: 7, long: 3 |
prose_luna |
Poems + prose description generated by the "Luna" model | low: 10, medium: 7, long: 3 |
prose_terra |
Poems + prose description generated by the "Terra" model | low: 10, medium: 7, long: 3 |
from datasets import load_dataset
poems = load_dataset("artindnr/khayyam-challenge", "poems")
prose_luna = load_dataset("artindnr/khayyam-challenge", "prose_luna")
prose_terra = load_dataset("artindnr/khayyam-challenge", "prose_terra")
poem = poems["low"][0]
prose = prose_luna["low"][0]
assert poem["id"] == prose["id"] # same poem, join key
print(prose["conversion"]["prose"])
Records in poems, prose_luna, and prose_terra are joined on id within each split. id is only unique within a split (low/medium/long), not across splits — always join within the same split.
Length categories
| Split | # poems | Verses per poem | Notes |
|---|---|---|---|
low |
10 | 1–7 beyts | Short ghazals/quatrains; good for quick iteration |
medium |
7 | 7–20 beyts | Longer ghazals; more complex imagery and structure |
long |
3 | 20–50 beyts | Excerpts from epic mathnavi (e.g. Ferdowsi's Shahnameh); narrative + philosophical depth |
Schema
Every split shares the same columns (required so datasets/the Hub viewer can build each config), but not every field is meaningful in every split — this reflects the source material, not a data error:
- All splits:
id,poet,title,form,verse_count,theme,text(list of verses, hemistichs joined with/). meter,notes: populated only forlow;""(empty string) elsewhere.verified: populated only forlow(whether the transcription was manually checked against a reference edition);falseelsewhere — this is a "not applicable" default, not a claim thatmedium/longpoems are unverified.selected_section: populated only forlong(which excerpt of a longer narrative poem this record covers, since full epics run to thousands of beyts);""elsewhere.
If you're filtering or displaying these fields, treat "" / false in the non-native split as "not applicable" rather than as real values.
prose_luna and prose_terra records contain everything in poems plus a conversion object:
"conversion": {
"model": "gpt-5.6-luna", // or gpt-5.6-terra
"version": "1",
"prose": "..." // the generated Persian prose description
}
Intended use: the two benchmark tasks
1. Poem → Prose. Give a model poems[split][i]["text"] and ask it to produce a fluent Persian prose description that preserves meaning, imagery, emotional register, and narrative/philosophical content without directly quoting the verse. Score the output against prose_luna/prose_terra as references, or with a judge model using a rubric (see below).
2. Prose → Poem. Give a model a prose description from prose_luna or prose_terra (with the source poem withheld) and ask it to generate a new Persian poem conveying the same meaning. Score the generated poem against the original in poems using a judge model.
Suggested scoring rubric
The reference implementation (evaluate_poetry.py) scores Prose → Poem generations on a 0–10 scale across five weighted criteria:
| Criterion | Weight | What it measures |
|---|---|---|
| Semantic similarity | 30% | Meaning preservation vs. the source prose/poem |
| Rhyme quality | 20% | Correct rhyme patterns, sound harmony, consistency |
| Meter accuracy | 20% | Persian prosody, rhythm, correct poetic meter |
| Literary devices | 15% | Use of metaphor, simile, symbolism, imagery |
| Semantic depth | 15% | Philosophical/emotional depth, originality |
final_score = 0.30·semantic + 0.20·rhyme + 0.20·meter + 0.15·literary + 0.15·depth
This dataset does not ship pre-computed judge scores — scoring is meant to run against whatever generation model and judge model you're evaluating. See the companion repo for a runnable harness (config.json + evaluate_poetry.py) that takes any OpenAI-compatible generation/judge endpoint.
Data collection
- Poems: classical Persian poems (public domain), manually curated and, for the
lowsplit, cross-checked against reference editions (seeverified/notes). - Prose: generated by prompting two different LLMs (internally labeled Luna and Terra) to render each poem as Persian prose, preserving meaning while removing verse form. Each poem has one prose version per model.
Known limitations
- Small dataset (20 poems total) — intended as a focused diagnostic benchmark, not a large-scale training corpus.
- Prose descriptions are themselves AI-generated and not independently human-verified; they may carry over interpretive choices or errors from the generating model.
verified/notesin thelowsplit describe verification of the poem transcription, not the prose description.- Meter/rhyme scoring in the suggested rubric depends on the judge model's own competence at Persian prosody, which varies significantly across models — treat absolute scores cautiously and prefer relative comparisons between systems evaluated with the same judge.
Licensing
- The source poems are classical Persian works in the public domain.
- The compilation, curation, metadata, and AI-generated prose in this dataset are released under CC BY 4.0. Attribution: cite this dataset (see below) when using it.
Citation
If you use this dataset, please cite:
@misc{khayyam-challenge,
title = {Khayyam Challenge: A Persian Poem-Prose Benchmark},
author = {Artin},
year = {2026},
howpublished = {\url{https://huggingface.co/datasets/artindnr/khayyam-challenge}},
note = {Companion framework: \url{https://github.com/Artin200912/Persian-Poetry-Challenge}}
}
Links
- Evaluation framework / harness: https://github.com/Artin200912/Persian-Poetry-Challenge
- Downloads last month
- 18