REVERE: Reflective Evolving Research Engineer for Scientific Workflows

TL;DR

REVERE is a framework that enables scientific workflow agents to continually learn from experience, recognize recurring issues, and improve their performance by making targeted, knowledge-preserving edits across prompts and resources.

Contribution

It introduces a reflective, continual learning approach for research coding workflows that enhances performance by recognizing patterns and distilling reusable heuristics.

Findings

REVERE improves performance on research coding benchmarks by 3.5-4.9%.

It demonstrates the effectiveness of continual, reflective learning in scientific workflows.

Agents with REVERE evolve capabilities over time through global memory consolidation.

Abstract

Existing prompt-optimization techniques rely on local signals to update behavior, often neglecting broader and recurring patterns across tasks, leading to poor generalization; they further rely on full-prompt rewrites or unstructured merges, resulting in knowledge loss. These limitations are magnified in research-coding workflows, which involve heterogeneous repositories, underspecified environments, and weak feedback, where reproducing results from public codebases is an established evaluation regime. We introduce Reflective Evolving Research Engineer (REVERE), a framework that continuously learns from Global Training Context, recognizes recurring failure modes in cross-repository execution trajectories, distills them into reusable heuristics, and performs targeted edits across three configurable fields: the system prompt, a task-prompt template, and a cumulative cheatsheet. REVERE, via this reflective optimization framework, improves performance over prior state-of-the-art expert-crafted instructions on research coding tasks by 4.50% on SUPER, 3.51% on ResearchCodeBench, and 4.89% on ScienceAgentBench across their respective metrics. These results demonstrate that agents equipped with mechanisms for continual learning and global memory consolidation can meaningfully evolve their capabilities over time.