StoryCoder: Narrative Reformulation for Structured Reasoning in LLM Code Generation

2026-04-16 · Source: Takara TLDR - Daily AI Papers · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Expert, quick

Summary

StoryCoder is a novel narrative reformulation framework designed to enhance large language model (LLM) code generation by transforming problem statements into coherent natural language narratives. Developed by Geonhui Jang, Dongyoon Han, and YoungJoon Yoo, this framework structures code generation questions into a task overview, constraints, and example test cases, guided by a selected algorithm and genre. This approach provides richer contextual structure compared to simple rephrasing, drawing inspiration from human information organization. Experiments conducted across 11 different LLMs on benchmarks such as HumanEval, LiveCodeBench, and CodeForces demonstrated an average improvement of 18.7% in zero-shot pass@10. The analysis indicates that StoryCoder guides models toward correct algorithmic strategies, reduces implementation errors, and promotes modular code structures, with benefits tied to narrative coherence and genre alignment.

Key takeaway

For AI Engineers developing LLM-powered code generation tools, integrating narrative reformulation frameworks like StoryCoder can substantially improve output quality. Your efforts to structure problem statements into coherent narratives, including task overviews, constraints, and test cases, will lead to more accurate and modular code. Consider aligning narrative genre and coherence to maximize performance gains, regardless of the underlying model's scale or architecture.

Key insights

Narrative reformulation of code problems significantly improves LLM code generation by structuring context.

Principles

• Structured problem representation is crucial for code generation.
• Narrative coherence and genre alignment enhance model performance.

Method

StoryCoder reformulates code generation questions into natural language narratives comprising a task overview, constraints, and example test cases, guided by algorithm and genre selection.

In practice

• Apply narrative reformulation to improve LLM code generation.
• Structure problem statements with task, constraints, and examples.

Topics

• LLM Code Generation
• Narrative Reformulation
• StoryCoder Framework
• Structured Reasoning
• HumanEval

Code references

• gu-ni/StoryCoder

Best for: AI Engineer, Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• StoryCoder: Narrative Reformulation for Structured Reasoning in LLM Code Generation — Narrative reformulation of code problems significantly improves LLM code generation by structuring reasoning.
• Improving Collaborative Storytelling with a Multi-Agent Framework Based on Large Language Models — An iterative LLM Writer-Editor framework significantly improves collaborative storytelling quality, even with few refinement steps.
• How Generation Architecture Shapes Code Complexity in Multi-Agent LLM Systems: A Paired Study on HumanEval — Multi-agent LLM code generation architectures significantly impact code complexity without necessarily improving functional correctness.
• Mastering AI Storytelling: Crafting Prompts for Captivating Narratives — Mastering AI storytelling prompts is a core competency for generating compelling narratives with generative AI.
• LeGo-Code: Can Modular Curriculum Learning Advance Complex Code Generation? Insights from Text-to-SQL — Modular curriculum learning with tier-specific adapters improves LLM performance on complex Text-to-SQL tasks.
• Beyond Code Pairs: Dialogue-Based Data Generation for LLM Code Translation — A dual-LLM Questioner–Solver generates dialogue-based data, improving LLM code translation in low-resource domains.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Takara TLDR - Daily AI Papers.