StoryCoder: Narrative Reformulation for Structured Reasoning in LLM Code Generation

2026-04-16 · Source: Computation and Language · 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 descriptions into structured natural language narratives. This approach, inspired by human information organization, provides richer contextual structure than simple rephrasing. Each StoryCoder narrative comprises a task overview, constraints, and example test cases, tailored by algorithm and genre. Across 11 different LLMs, StoryCoder consistently improved zero-shot pass@10 scores by an average of 18.7% on benchmarks including HumanEval, LiveCodeBench, and CodeForces. Analysis indicates that narrative reformulation 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 and Research Scientists developing LLM-based code generation systems, integrating narrative reformulation frameworks like StoryCoder can significantly boost performance. By structuring problem descriptions into coherent narratives with explicit constraints and examples, you can guide models toward better algorithmic choices and reduce common implementation errors. Consider adopting this approach to improve code quality and reduce debugging cycles in your LLM-powered development workflows.

Key insights

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

Principles

• Structured problem representation is crucial.
• Narrative coherence and genre alignment matter.

Method

StoryCoder transforms code generation questions into coherent narratives with task overviews, constraints, and test cases, guided by algorithm and genre.

In practice

• Use structured narratives for LLM prompts.
• Tailor narratives to specific algorithms.
• Include constraints and test cases explicitly.

Topics

• StoryCoder
• Narrative Reformulation
• LLM Code Generation
• Structured Reasoning
• Algorithmic Strategies

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 context.
• 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.
• 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.
• Unlocking LLM Code Correction with Iterative Feedback Loops — Iterative feedback significantly enhances LLM code correction, with reasoning models excelling at leveraging execution feedback.
• 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.
• DeNovoSWE: Scaling Long-Horizon Environments for Generating Entire Repositories from Scratch — Automated, structured data generation with difficulty-aware filtering scales long-horizon software engineering training for LLM agents.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.