POLARIS trains small models for better long-form story writing

Researchers have developed POLARIS, a new training method designed to improve the long-form creative writing capabilities of smaller open-weight language models. This method utilizes a frontier LLM as a judge with a structured quality rubric and incorporates human-written story references as high-reward anchors during training. Applied to Qwen3.5-9B, the resulting POLARIS-9B model demonstrates competitive performance against larger models and shows improved adherence to length instructions, even for stories exceeding its training length. AI

IMPACT Enhances the creative writing capabilities of smaller, more accessible language models, potentially democratizing advanced AI content generation.

RANK_REASON The cluster contains an academic paper detailing a new method for training language models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• POLARIS
• Qwen3.5-27B
• Qwen3.5-9B
• Rishanth Rajendhran

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →