Paper Overview
Field: NLP Authors: Shuhan Xue, Jianyuan Zhong, Ziyuan Nan, Wenbin Li, Zhaochen Yu, Jinchao Ding, Qiang Gao, Pengyu Zhan, Yuntong Zhang, Tian Cheng, Zhenfei Yin, Yingcheng Wu, Ling Yang Release date: 2026-09-15 arXiv: 2609.17523
Abstract
We introduce and release ScienceBuddy, an interactive scientific research workspace that brings continually improving scientific agents into researchers' everyday workflows. ScienceBuddy supports researchers in carrying out scientific tasks while transforming their requests, feedback, and execution evidence into tasks and evaluation rubrics for continual learning.
At its core is recursive-in-recursive self-improvement, a paradigm that couples harness evolution with model reinforcement learning:
- The inner recursion improves the harness (toolchain) with the model fixed.
- The outer recursion trains the model under the improved harness.
- Paper: https://arxiv.org/abs/2609.17523
- Website: http://science-buddy.io
Harness evolution shapes training experience, and model learning creates new opportunities for harness adaptation.
The authors present case studies of researcher interaction, toolchain optimization, and model learning, with benchmarks covering four scientific task families. By releasing ScienceBuddy as a research product, the team makes this paradigm available to the scientific community and takes a step toward discovery intelligence: scientific AI that continually improves through collaboration with researchers and co-evolves with the research it supports.