AgentRxiv: Towards Collaborative Autonomous Research

2025-03-23

AI Summary

Current autonomous scientific research workflows lack cross-agent collaboration and knowledge iteration capabilities, resulting in isolated, non-reusable outputs. This paper introduces the first LLM-agent-based research framework enabling sustained collaborative science: it establishes a shared preprint server where multi-agent laboratories can upload structured research reports, perform semantic retrieval over existing work, and iteratively refine prompts and learn reasoning strategies to enable cross-agent knowledge accumulation and co-evolution. The framework breaks the traditional single-agent, closed-loop research paradigm, achieving for the first time traceable, reusable, and incrementally improvable collective scientific inquiry among agents. Experiments demonstrate that multi-laboratory collaboration improves performance by 13.7% on MATH-500; single-laboratory reuse yields a 11.4% gain; and cross-domain generalization improves by an average of 3.3%.

Technology Category

• Multiagent Systems: Teamwork
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Agent Architectures

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work

Abstract

Progress in scientific discovery is rarely the result of a single"Eureka"moment, but is rather the product of hundreds of scientists incrementally working together toward a common goal. While existing agent workflows are capable of producing research autonomously, they do so in isolation, without the ability to continuously improve upon prior research results. To address these challenges, we introduce AgentRxiv-a framework that lets LLM agent laboratories upload and retrieve reports from a shared preprint server in order to collaborate, share insights, and iteratively build on each other's research. We task agent laboratories to develop new reasoning and prompting techniques and find that agents with access to their prior research achieve higher performance improvements compared to agents operating in isolation (11.4% relative improvement over baseline on MATH-500). We find that the best performing strategy generalizes to benchmarks in other domains (improving on average by 3.3%). Multiple agent laboratories sharing research through AgentRxiv are able to work together towards a common goal, progressing more rapidly than isolated laboratories, achieving higher overall accuracy (13.7% relative improvement over baseline on MATH-500). These findings suggest that autonomous agents may play a role in designing future AI systems alongside humans. We hope that AgentRxiv allows agents to collaborate toward research goals and enables researchers to accelerate discovery.

Problem

• Enabling autonomous agents to collaboratively improve research outcomes
• Facilitating shared knowledge among AI labs via preprint server
• Enhancing agent performance through iterative research building

Innovation

• LLM agents share research via preprint server
• Agents iteratively improve prior research results
• Collaborative agents achieve higher performance gains