Contract-Coding: Towards Repo-Level Generation via Structured Symbolic Paradigm

arXiv cs.AI / 4/16/2026

💬 OpinionSignals & Early TrendsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper argues that intent-driven “Vibe Coding” breaks down on repo-level tasks due to a context-fidelity trade-off that causes architectural failure as reasoning chains become overwhelmed by vague intents.
It introduces “Contract-Coding,” a structured symbolic approach that turns ambiguous user intent into a formal Language Contract acting as a single source of truth between intent and executable code.
The method uses Autonomous Symbolic Grounding to isolate inter-module implementation details, reduce topological execution depth, and enable architectural parallelism during code generation.
On the Greenfield-5 benchmark, it reports 47% functional success while keeping near-perfect structural integrity, contrasting with different hallucination patterns seen in state-of-the-art agents.
The authors position Contract-Coding as a step toward repository-scale autonomous engineering by moving from brittle specification-following toward more robust, intent-driven architecture synthesis, and they provide code publicly.

Abstract

The shift toward intent-driven software engineering (often termed "Vibe Coding") exposes a critical Context-Fidelity Trade-off: vague user intents overwhelm linear reasoning chains, leading to architectural collapse in complex repo-level generation. We propose Contract-Coding, a structured symbolic paradigm that bridges unstructured intent and executable code via Autonomous Symbolic Grounding. By projecting ambiguous intents into a formal Language Contract, our framework serves as a Single Source of Truth (SSOT) that enforces topological independence, effectively isolating inter-module implementation details, decreasing topological execution depth and unlocking Architectural Parallelism. Empirically, while state-of-the-art agents suffer from different hallucinations on the Greenfield-5 benchmark, Contract-Coding achieves 47\% functional success while maintaining near-perfect structural integrity. Our work marks a critical step towards repository-scale autonomous engineering: transitioning from strict "specification-following" to robust, intent-driven architecture synthesis. Our code is available at https://github.com/imliinyi/Contract-Coding.