Advancements in AI-Driven Research and Development

The field of AI-driven research and development is rapidly evolving, with a focus on improving the efficiency, reproducibility, and transparency of scientific experiments and software development. Recent studies have highlighted the importance of standardized evaluation protocols, robust benchmarking, and systematic approaches to integrating domain knowledge into requirements engineering. Notably, innovative frameworks and tools, such as data-centric infrastructures and semantic-aware digital twins, are being developed to support these efforts. Meanwhile, researchers are exploring the application of agile management methodologies to machine learning-enabled systems, aiming to address the unique challenges posed by these dynamic and rapidly changing environments. Overall, the field is moving towards greater emphasis on rigor, reproducibility, and collaboration, with AI-driven tools and approaches playing an increasingly crucial role in accelerating scientific progress. Noteworthy papers include OAgents, which introduced a new foundation agent framework that achieves state-of-the-art performance, and Doc2Agent, a scalable pipeline for building tool-using agents from API documentation. Additionally, the AutoExperiment benchmark was proposed to evaluate AI agents' ability to implement and run machine learning experiments based on natural language descriptions, and the Define-ML framework was presented as a systematic approach to ideating machine learning-enabled systems.

Sources

OAgents: An Empirical Study of Building Effective Agents

OSWorld-Human: Benchmarking the Efficiency of Computer-Use Agents

From Data to Decision: Data-Centric Infrastructure for Reproducible ML in Collaborative eScience

Simulating the Waterfall Model: A Systematic Review

From Reproduction to Replication: Evaluating Research Agents with Progressive Code Masking

Doc2Agent: Scalable Generation of Tool-Using Agents from API Documentation

AI Copilots for Reproducibility in Science: A Case Study

AI and Agile Software Development: From Frustration to Success -- XP2025 Workshop Summary

Ten simple rules for PIs to integrate Research Software Engineering into their research group

The Composition of Digital Twins for Systems-of-Systems: a Systematic Literature Review

Define-ML: An Approach to Ideate Machine Learning-Enabled Systems

Domain Knowledge in Requirements Engineering: A Systematic Mapping Study

Agile Management for Machine Learning: A Systematic Mapping Study

Semantic-aware Digital Twin for AI-based CSI Acquisition

Built with on top of