Credit-Budgeted ICPC-Style Coding: When Agents Must Pay for Every Decision

2026-04-11

AI Summary

This work addresses the critical gap in existing evaluations of programming agents, which often neglect real-world computational and temporal resource constraints, thereby failing to account for budget exhaustion risks in large-scale deployment. We propose USACOArena—a novel ACM-ICPC–style programming arena grounded in a strict “credit” economy—that introduces explicit resource budgeting into agent evaluation for the first time. In this framework, every token generation, local test execution, and unit of elapsed time consumes a fixed amount of budget, compelling agents to strategically balance accuracy against resource expenditure. Leveraging an interactive arena architecture, fine-grained resource tracking, and behavioral analysis, we demonstrate that state-of-the-art agents exhibit suboptimal performance and path-dependent behaviors under resource constraints, thereby validating USACOArena as a pivotal platform for training efficient, resource-aware programming agents.

Technology Category

• Multiagent Systems: Mechanism Design
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Current evaluations of autonomous coding agents assume an unrealistic, infinite-resource environment. However, real-world software engineering is a resource-bound competition. As we scale toward large agent swarms, ignoring compute and time costs risks catastrophic budget exhaustion. To shift the focus from isolated accuracy to cost-aware problem-solving, we introduce USACOArena, an interactive ACM-ICPC-style arena driven by a strict "credit" economy. Every generated token, local test, and elapsed second depletes a fixed budget, forcing agents to make strategic trade-offs. Our comprehensive profiling reveals that frontier single agents and swarms currently fail to optimally balance accuracy with these constraints, exhibiting divergent, path-dependent behaviors. Ultimately, USACOArena provides an essential dynamic training ground for developing highly efficient, resource-aware agent architectures.

Problem

• resource-constrained coding
• credit budget
• autonomous coding agents
• cost-aware problem-solving
• computational efficiency

Innovation

• credit-budgeted evaluation
• resource-aware agents
• ICPC-style coding
• agent swarms
• cost-aware problem-solving