High-quality creative writing data for language models remains dominated by story-centric data, limiting models’ ability to follow the structural and functional conventions of diverse creative formats. We propose an attribute-guided genre expansion framework for scaling creative writing data beyond story generation. By separating thematic breadth from genre-form control, our framework leverages human-authored story prompts as diverse creative seeds, while utilizing manually curated genre attributes to enforce distinct structural, stylistic, and formatting conventions. We combine these to prompt strong models for genre-faithful query–response pairs, which are then quality-filtered. Applying this framework, we construct the Multi-Genre Collection, a 50K-example corpus spanning 13 creative genres, including story, lyrics, game design, and other creative formats. Experiments demonstrate that models finetuned on our data consistently surpass not only base models and writing-specialized baselines, but also models trained on existing writing corpora. Genre-count ablations further indicate that genre expansion is a key driver of robust creative writing capability.</p>\n","updatedAt":"2026-08-20T04:24:08.597Z","author":{"_id":"647c4a2692182942d7c2e698","avatarUrl":"/avatars/bcddf5fe49aa092a2645f70812108348.svg","fullname":"HWANCHANG","name":"HwanChang0106","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false}},"numEdits":0,"identifiedLanguage":{"language":"en","probability":0.8709279298782349},"editors":["HwanChang0106"],"editorAvatarUrls":["/avatars/bcddf5fe49aa092a2645f70812108348.svg"],"reactions":[],"isReport":false}}],"primaryEmailConfirmed":false,"paper":{"id":"2608.13947","authors":[{"_id":"6a83272d675db694db8cd2dc","user":{"_id":"647c4a2692182942d7c2e698","avatarUrl":"/avatars/bcddf5fe49aa092a2645f70812108348.svg","isPro":false,"fullname":"HWANCHANG","user":"HwanChang0106","type":"user","name":"HwanChang0106"},"name":"Hwan Chang","status":"claimed_verified","statusLastChangedAt":"2026-08-17T16:45:04.554Z","hidden":false},{"_id":"6a83272d675db694db8cd2dd","name":"Yongil Kim","hidden":false},{"_id":"6a83272d675db694db8cd2de","name":"Heuiyeen Yeen","hidden":false},{"_id":"6a83272d675db694db8cd2df","name":"Yireun Kim","hidden":false},{"_id":"6a83272d675db694db8cd2e0","name":"Jinsik Lee","hidden":false},{"_id":"6a83272d675db694db8cd2e1","name":"Hwanhee Lee","hidden":false}],"publishedAt":"2026-08-14T00:00:00.000Z","submittedOnDailyAt":"2026-08-20T00:00:00.000Z","title":"Scaling Creative Writing Beyond Story-Centric Data with Attribute-Guided Genre Expansion","submittedOnDailyBy":{"_id":"647c4a2692182942d7c2e698","avatarUrl":"/avatars/bcddf5fe49aa092a2645f70812108348.svg","isPro":false,"fullname":"HWANCHANG","user":"HwanChang0106","type":"user","name":"HwanChang0106"},"summary":"High-quality creative writing data for large language models (LLMs) remains dominated by story-centric data, limiting models' ability to follow the structural and functional conventions of diverse creative formats. We propose an attribute-guided genre expansion framework for scaling creative writing data beyond story generation. By separating thematic breadth from genre-form control, our framework leverages human-authored story prompts as diverse creative seeds, while utilizing manually curated genre attributes to enforce distinct structural, stylistic, and formatting conventions. We combine these to prompt strong LLMs for genre-faithful query-response pairs, which are then quality-filtered. Applying this framework, we construct the Multi-Genre Collection, a 50K-example corpus spanning 13 creative genres, including story, rap, lyrics, scripts, game design, character design, and other creative formats. Experiments across out-of-distribution writing benchmarks and held-out genre diagnostics demonstrate that models fine-tuned on our data consistently surpass not only base models and writing-specialized baselines, but also models trained on existing writing corpora. Genre-count ablations further indicate that controlled genre expansion, rather than story-centric scaling alone, is a key driver of robust creative writing capability.","upvotes":12,"discussionId":"6a83272e675db694db8cd2e2","ai_summary":"A framework that separates thematic seeds from genre-form controls generates diverse, high-quality creative writing data across 13 genres and improves LLM creative writing performance.","ai_keywords":["attribute-guided genre expansion","genre-form control","genre attributes","query-response pairs","Multi-Genre Collection","out-of-distribution writing benchmarks","genre-count ablations"],"ai_summary_model":"thinkingmachines/Inkling-Small"},"canReadDatabase":false,"canManagePapers":false,"canSubmit":false,"hasHfLevelAccess":false,"upvoted":false,"upvoters":[{"_id":"6a6a831d726441725a2ed822","avatarUrl":"/avatars/45ba67b0e43f5f2e65aff93418dcdba2.svg","isPro":false,"fullname":"Charles Thompson","user":"charlesthompson","type":"user"},{"_id":"6a6aa6de726441725a30a027","avatarUrl":"/avatars/2e777075d9d0d6dcdaaf0e97860f551e.svg","isPro":false,"fullname":"Sarah Clark","user":"granitecore","type":"user"},{"_id":"6a6aa41d4c287dbb8e805dd3","avatarUrl":"/avatars/f109124053777f7667ff3235cd8aaf79.svg","isPro":false,"fullname":"Sarah Thompson","user":"rapidWing","type":"user"},{"_id":"6a6c83c8bd71e954a590ca0d","avatarUrl":"/avatars/fa4e69ebce443312d7b1962d551b29d7.svg","isPro":false,"fullname":"Barbara Brown","user":"GraniteLens","type":"user"},{"_id":"6a6c8c2bf8982269d7d8b20e","avatarUrl":"/avatars/be8ab2e5f32f028fc7af8c3245465ed5.svg","isPro":false,"fullname":"Joseph White","user":"Rapid-Joseph","type":"user"},{"_id":"6a6c9c217d5ca057d90c7cf1","avatarUrl":"/avatars/2a27fb930a6493f2b9aa26516ae4a99e.svg","isPro":false,"fullname":"John Taylor","user":"AzureWisp","type":"user"},{"_id":"6a6de4482324580e33e1f8c1","avatarUrl":"/avatars/e21e3f3a3b3758766aa0c3bf514ca9c3.svg","isPro":false,"fullname":"Elizabeth Johnson","user":"Elizabeth-Johnson","type":"user"},{"_id":"6a6deb3bc51edbf08f1177d1","avatarUrl":"/avatars/21984308f858768e1770697c4e5b2456.svg","isPro":false,"fullname":"Mark Perez","user":"Sable-Theo","type":"user"},{"_id":"6a701f86bc4016943249be07","avatarUrl":"/avatars/928c7a44ae1bcec57c75503d8fe7adb7.svg","isPro":false,"fullname":"Jennifer Clark","user":"RapidBeacon","type":"user"},{"_id":"6a7e7f71c69e6d401c4348de","avatarUrl":"/avatars/c6721bf7013cfbcfa681ed0bd7c72ae5.svg","isPro":false,"fullname":"Jingyi Sun","user":"GraniteJingyi","type":"user"},{"_id":"6a82621b6f3e370df0aeb3ce","avatarUrl":"/avatars/a9547a8c8818a0cfaaa06eba18bdad72.svg","isPro":false,"fullname":"Xinyi Li","user":"ZenithRidge67","type":"user"},{"_id":"647c4a2692182942d7c2e698","avatarUrl":"/avatars/bcddf5fe49aa092a2645f70812108348.svg","isPro":false,"fullname":"HWANCHANG","user":"HwanChang0106","type":"user"}],"acceptLanguages":["en"],"dailyPaperRank":0,"markdownContentUrl":"https://huggingface.co/buckets/huggingchat/papers-content/resolve/2608/2608.13947.md","query":{}}">
Scaling Creative Writing Beyond Story-Centric Data with Attribute-Guided Genre Expansion
Abstract
A framework that separates thematic seeds from genre-form controls generates diverse, high-quality creative writing data across 13 genres and improves LLM creative writing performance.
High-quality creative writing data for large language models (LLMs) remains dominated by story-centric data, limiting models' ability to follow the structural and functional conventions of diverse creative formats. We propose an attribute-guided genre expansion framework for scaling creative writing data beyond story generation. By separating thematic breadth from genre-form control, our framework leverages human-authored story prompts as diverse creative seeds, while utilizing manually curated genre attributes to enforce distinct structural, stylistic, and formatting conventions. We combine these to prompt strong LLMs for genre-faithful query-response pairs, which are then quality-filtered. Applying this framework, we construct the Multi-Genre Collection, a 50K-example corpus spanning 13 creative genres, including story, rap, lyrics, scripts, game design, character design, and other creative formats. Experiments across out-of-distribution writing benchmarks and held-out genre diagnostics demonstrate that models fine-tuned on our data consistently surpass not only base models and writing-specialized baselines, but also models trained on existing writing corpora. Genre-count ablations further indicate that controlled genre expansion, rather than story-centric scaling alone, is a key driver of robust creative writing capability.
Community
High-quality creative writing data for language models remains dominated by story-centric data, limiting models’ ability to follow the structural and functional conventions of diverse creative formats. We propose an attribute-guided genre expansion framework for scaling creative writing data beyond story generation. By separating thematic breadth from genre-form control, our framework leverages human-authored story prompts as diverse creative seeds, while utilizing manually curated genre attributes to enforce distinct structural, stylistic, and formatting conventions. We combine these to prompt strong models for genre-faithful query–response pairs, which are then quality-filtered. Applying this framework, we construct the Multi-Genre Collection, a 50K-example corpus spanning 13 creative genres, including story, lyrics, game design, and other creative formats. Experiments demonstrate that models finetuned on our data consistently surpass not only base models and writing-specialized baselines, but also models trained on existing writing corpora. Genre-count ablations further indicate that genre expansion is a key driver of robust creative writing capability.
Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.
Tap or paste here to upload images
Cite arxiv.org/abs/2608.13947 in a model README.md to link it from this page.
Cite arxiv.org/abs/2608.13947 in a dataset README.md to link it from this page.
Cite arxiv.org/abs/2608.13947 in a Space README.md to link it from this page.
Discussion (0)
Sign in to join the discussion. Free account, 30 seconds — email code or GitHub.
Sign in →No comments yet. Sign in and be the first to say something.