ICAE-Bench: Evaluating Coding Agents as Interactive Project Builders

2026-07-23

AI Summary

Existing benchmarks for evaluating code-generating agents are largely confined to static, fully specified tasks, making them inadequate for assessing agents’ capabilities in interactive project construction under ambiguous requirements. This work proposes a novel evaluation benchmark that generates realistic, underspecified product requirements from actual open-source repositories and employs an automated user agent to simulate authentic developer dialogues. It introduces a multidimensional black-box evaluation framework that integrates functional correctness testing, design quality analysis—encompassing semantic and API similarity as well as structural fidelity—and interaction quality diagnostics. For the first time, this framework enables comprehensive, fair, and reproducible assessment of agents’ integrated abilities in planning, requirement clarification, debugging, and repository-scale software construction.

Abstract

The recent emergence of vibe-coding workflows is changing what coding agents are expected to do. Instead of merely completing code under fully specified instructions, agents are increasingly expected to transform incomplete product intent into working software by combining various abilities including planning, requirement clarification, tool use, debugging, and repository-level construction. Yet existing benchmarks have not fully caught up with this shift, evaluating agents on static, fully specified tasks. In this paper, we introduce ICAE-Bench, a benchmark for evaluating coding agents under interactive project-building settings. The basic idea is to start from a fuzzy product requirement, simulating the dynamic paradigm with an automated User Agent. To make this setting both realistic and evaluable, ICAE-Bench introduces three key designs. First, to avoid the ambiguity of unconstrained fuzzy requirements, each task derives ambiguity from a precise real open-source repository with executable behavior. Second, to ensure high-quality and reproducible user simulation, ICAE-Bench grounds interaction through User Agent Data, allowing the User Agent to reveal hidden constraints without inventing new requirements or leaking implementation artifacts. Third, to evaluate open-ended repositories fairly, ICAE-Bench uses standardized black-box tests together with multi-dimensional diagnostics, including functional correctness, semantic and API similarity, structural fidelity, design quality, and interaction quality.

Problem

• coding agents
• interactive project building
• fuzzy requirements
• benchmark evaluation
• software construction

Innovation

• interactive coding agents
• project-building benchmark
• fuzzy requirements
• user agent simulation
• multi-dimensional evaluation