Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

POLARIS: Guiding Small Models to Write Long Stories

About

Small open-weight models struggle at long-form creative writing: their generated stories either fall far short of the requested length, or their quality significantly degrades as length increases, especially when compared to frontier models. We present POLARIS (Policy Optimization with LLM-as-a-judge rewards and Anchored-Reference Injection for Storywriting), a lower-compute GRPO recipe with two key ingredients: a frontier LLM judge with a structured Story Quality rubric as the online reward, and human-reference injection (HRI), where a teacher-forced human-written story serves as a high-reward anchor within each GRPO group. By applying our training recipe to Qwen3.5-9B, using a dataset of approximately 1.4K prompt-story pairs derived from 100 short-story anthologies and 4 A100 GPUs, we obtain POLARIS-9B. Across five benchmarks spanning in-distribution and out-of-distribution prompts and rubrics, POLARIS-9B is competitive with much larger open-weight models while following length instructions more closely. A blinded human evaluation confirms that POLARIS-9B is preferred to the base Qwen3.5-9B and on par with Qwen3.5-27B. Despite training only on stories up to 4k words, POLARIS-9B preserves quality on prompts requesting stories up to 3 times the training length, a regime where most open-weight models degrade substantially in quality, length adherence, or both. More broadly, our results suggest that length generalization is a meaningful stress test for creative-writing models and a useful lens for distinguishing otherwise close models.

Rishanth Rajendhran, Jenna Russell, Mohit Iyyer, John Frederick Wieting• 2026

Related benchmarks

TaskDatasetResultRank
Long-form writingLongBench-Write (test)
Overall Quality Score (S_bar)81.2
25
Creative WritingEQ-Bench Longform (Creative Writing) (32 prompts)
Elo Rating1.42e+3
17
Instruction FollowingEQ-Bench Longform (ID prompts) (45 prompts)
Elo Score1.66e+3
16
Long-form Writing EvaluationWritingBench In-domain (test)
Elo Rating1.42e+3
15
Story Generation Quality180 prompts Aggregate (held-out)
Story Quality Score52.1
15
Creative Writing EvaluationEQ-Bench Creative (OOD)
Elo Rating1.66e+3
14
Creative Writing EvaluationEQ-Bench Creative
S Score70.3
14
Long-form writingWritingBench D4 English literature/arts
Score (D4 Lit/Arts)7.9
14
Human Preference EvaluationEQ-Bench Creative and LongBench Write (60 pairs)
Winrate67.5
2
Long-form Emotional Intelligence EvaluationEQ-Bench Longform Aggregate
EQ-Bench Score59.8
1
Showing 10 of 20 rows

Other info

Follow for update