HyperAI

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

NovelSeek Team, Bo Zhang, Shiyang Feng, Xiangchao Yan, Jiakang Yuan, Zhiyin Yu, Xiaohan He, Songtao Huang, Shaowei Hou, Zheng Nie, Zhilong Wang, Jinyao Liu, Runmin Ma, Tianshuo Peng, Peng Ye, Dongzhan Zhou, Shufei Zhang, Xiaosong Wang, Yilan Zhang, Meng Li, Zhongying Tu, Xiangyu Yue, Wangli Ouyang, Bowen Zhou, Lei Bai
Release Date: 5/25/2025
NovelSeek: When Agent Becomes the Scientist -- Building Closed-Loop
  System from Hypothesis to Verification
Abstract

Artificial Intelligence (AI) is accelerating the transformation of scientificresearch paradigms, not only enhancing research efficiency but also drivinginnovation. We introduce NovelSeek, a unified closed-loop multi-agent frameworkto conduct Autonomous Scientific Research (ASR) across various scientificresearch fields, enabling researchers to tackle complicated problems in thesefields with unprecedented speed and precision. NovelSeek highlights three keyadvantages: 1) Scalability: NovelSeek has demonstrated its versatility across12 scientific research tasks, capable of generating innovative ideas to enhancethe performance of baseline code. 2) Interactivity: NovelSeek provides aninterface for human expert feedback and multi-agent interaction in automatedend-to-end processes, allowing for the seamless integration of domain expertknowledge. 3) Efficiency: NovelSeek has achieved promising performance gains inseveral scientific fields with significantly less time cost compared to humanefforts. For instance, in reaction yield prediction, it increased from 27.6% to35.4% in just 12 hours; in enhancer activity prediction, accuracy rose from0.52 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.