MOOSE-Copilot: A Web-Based Interactive Assistant for Unified Exploratory and Fine-Grained Scientific Hypothesis Discovery
This work addresses the need for human-guided, end-to-end scientific hypothesis discovery, enabling interdisciplinary researchers to leverage LLMs without steep learning curves.
MOOSE-Copilot introduces a unified framework for scientific hypothesis discovery that integrates divergent exploratory ideation and convergent fine-grained refinement through a formalized human-AI interaction protocol. Quantitative evaluations show that structured expert signals significantly outperform purely autonomous baselines, establishing a performance ceiling under oracle guidance.
Large language models (LLMs) show remarkable potential in scientific hypothesis discovery. However, existing approaches face two critical limitations: they treat divergent exploratory ideation and convergent fine-grained refinement as isolated tasks, and they operate autonomously with little to no human guidance. We present MOOSE-Copilot, the first unified framework to bridge this abstraction gap through a formalized human-AI interaction (HAII) protocol. Our system empowers scientists to steer the generative process via three explicit signals: initial blueprints, inter-stage routing, and regenerative feedback. Quantitative evaluations demonstrate that injecting these structured expert signals significantly outperforms purely autonomous baselines, establishing a performance ceiling under oracle guidance. Furthermore, to democratize this paradigm, we develop an intuitive web-based interface featuring interactive tree visualization. This explicitly eliminates the steep learning curve of complex command-line agentic tools, empowering interdisciplinary researchers to directly leverage, visually orchestrate, and accelerate end-to-end scientific breakthroughs.