InternAgent: When Agent Becomes the Scientist -- Building Closed-Loop System from Hypothesis to Verification

TL;DR

InternAgent introduces a unified closed-loop multi-agent framework for Autonomous Scientific Research (ASR) that spans 12 diverse tasks, enabling autonomous hypothesis generation, methodology construction, and iterative experimental validation. It combines self-evolving idea generation with human-interactive feedback and a robust idea-to-methodology pipeline, supported by evolutionary planning and an exception-guided debugging framework. Empirical results show consistent improvements over baselines and prior auto-research systems across multiple domains, along with analyses on modes of literature search, idea evolution, and planning, as well as qualitative case studies and human evaluations. The work lays a foundation for scalable, interactive, and cost-efficient AI-driven scientific discovery and outlines directions for knowledge retrieval, representation, and agent enhancement.

Abstract

Artificial Intelligence (AI) is accelerating the transformation of scientific research paradigms, not only enhancing research efficiency but also driving innovation. We introduce InternAgent, a unified closed-loop multi-agent framework to conduct Autonomous Scientific Research (ASR) across various scientific research fields, enabling researchers to tackle complicated problems in these fields with unprecedented speed and precision. InternAgent highlights three key advantages: 1) Scalability: InternAgent has demonstrated its versatility across 12 scientific research tasks, capable of generating innovative ideas to enhance the performance of baseline code. 2) Interactivity: InternAgent provides an interface for human expert feedback and multi-agent interaction in automated end-to-end processes, allowing for the seamless integration of domain expert knowledge. 3) Efficiency: InternAgent has achieved promising performance gains in several scientific fields with significantly less time cost compared to human efforts. For instance, in reaction yield prediction, it increased from 27.6% to 35.4% in just 12 hours; in enhancer activity prediction, accuracy rose from 0.65 to 0.79 with only 4 hours of processing; and in 2D semantic segmentation, precision advanced from 78.8% to 81.0% in a mere 30 hours.

Figures (13)

• Figure 1: InternAgent can support 12 types of scientific research tasks ranging from the AI field to the science field, including reaction yield prediction, molecular dynamics, power flow estimation, time series forecasting, transcription prediction, enhancer activity prediction, sentiment classification, 2D image classification, 3D point classification, 2D semantic segmentation, 3D autonomous driving, large vision-language model fine-tuning.
• Figure 2: InternAgent covers three main capabilities: 1) Self-evolving Idea Generation with Human-interactive Feedback, 2) Idea-to-Methodology Construction, and 3) Evolutionary Experimental Planning and Execution.
• Figure 3: InternAgent Self-evolutionary path of ideas for reaction yield prediction task.
• Figure 4: Analysis of two modes on survey agent.
• Figure 5: Visual Examples of AutoRYP Task.