PARC: An Autonomous Self-Reflective Coding Agent for Robust Execution of Long-Horizon Tasks

TL;DR

PARC is a hierarchical multi-agent coding system that autonomously executes, monitors, and corrects long-horizon scientific and data analysis tasks, reducing human intervention and improving robustness.

Contribution

It introduces a novel self-reflective architecture enabling autonomous management and correction of complex computational tasks across scientific domains.

Findings

Successfully reproduces key scientific results in materials science.

Manages and coordinates extensive parallel simulations over 43 hours each.

Produces competitive data analysis solutions from minimal instructions.

Abstract

We introduce PARC, a coding agent for the autonomous and robust execution of long-horizon computational tasks. PARC is built on a hierarchical multi-agent architecture incorporating task planning, execution, and a mechanism that evaluates its own actions and their outcomes from an independent context and provides feedback, namely self-assessment and self-feedback. This design enables PARC to detect and correct high-level strategic errors and sustain progress without human intervention. We evaluate PARC across computational science and data science tasks. In materials science, it autonomously reproduces key results from studies on lithium-ion conduction and alloy segregation. In particular, it coordinates dozens of parallel simulation tasks, each requiring roughly 43 hours of computation, managing orchestration, monitoring, and error correction end-to-end. In Kaggle-based experiments, starting from minimal natural-language instructions, PARC conducts data analysis and implements search strategies, producing solutions competitive with human-engineered baselines. These results highlight the potential of integrating a hierarchical multi-agent system with self-assessment and self-feedback to enable AI systems capable of independent, large-scale scientific and analytical work.