Can you have top - notch scientific researchers as your assistants? DP Technology, in collaboration with Shanghai Jiao Tong University, has released the general scientific research intelligent agent, SciMaster.

On July 26, Shanghai Jiao Tong University, DP Technology, and the Shanghai Institute for Algorithmic Innovation jointly released the truly general - purpose scientific foundation large - model Innovator, and launched the "General Scientific Research Agent" SciMaster based on Innovator.

SciMaster is not just another cold tool, but an accessible "Master - expert - level" scientific research assistant, turning the grand vision of AI for Science into a new scientific research reality that everyone can participate in.

The "Super Agent" Born to Address Scientific Research Pain Points —— SciMaster

The development of science relies on the tireless exploration of researchers across generations and the inheritance of a vast amount of ancient books and literature.

However, statistics show that since the 19th century, the number of human literatures has increased by 4000 times, while the speed at which people read and study literatures has only doubled. Facing infinite scientific research needs and vast exploration spaces, researchers have only limited exploration means, which seriously restricts the speed of scientific discovery and limits the occurrence of more innovations.

In addition, current scientific research increasingly requires the simultaneous participation of multi - disciplinary researchers, and the types of tools needed to solve problems are growing and becoming more complex, forcing researchers into efficiency bottlenecks caused by issues such as knowledge barriers.

To address these scientific research challenges, SciMaster emerged, aiming to reshape the scientific research process. It is no longer satisfied with providing isolated fragments of answers, but builds a transparent, controllable, and collaborative intelligent research ecosystem based on "reading, calculating, and doing".

Core Function: Expert - Level In - Depth Research Ability

In generating "in - depth research reports", SciMaster demonstrates "expert - level" in - depth research ability.

When a researcher poses a scientific question, such as "What is the typical process of molecular dynamics in drug screening?" or "What are the algorithms for solving differential equations?", SciMaster will evaluate and analyze the question, break it down into multiple subtasks, and flexibly use three retrieval methods: WebSearch, WebParse, and PaperSearch to search the entire Internet and all literatures, collecting relevant information, data, papers, patents, and other materials in the field. Thus, SciMaster will ultimately deliver a detailed and reliable research report to the user.

Just one day after its launch, SciMaster had generated thousands of such customized high - quality scientific research reports, which support one - click export and free sharing.

Flexible Tool Invocation

Notably, SciMaster supports the thought - chain editing function, so researchers can actively intervene in SciMaster's execution logic. When SciMaster is conducting in - depth thinking and planning tasks, researchers can act like a "commander - in - chief", intervening at any time to click pause, modifying the task logic and content within the "in - depth thinking" box, and jointly thinking and iteratively optimizing with SciMaster to more accurately and reasonably meet research needs.

SciMaster also integrates many scientific - specific tools, and there are two ways to invoke them: "active invocation", where researchers can actively call the required calculation tools in the dialog box according to task requirements to assist in task completion; "automatic invocation", where after analyzing the questions raised by researchers, SciMaster screens and invokes relevant tools according to the situation to accurately complete tasks.

Reshaping the Scientific Research Paradigm and Closing the Loop between Wet and Dry Experiments

Meanwhile, SciMaster is reshaping the scientific research paradigm in universities. DP Technology has reached cooperation agreements with more than 40 universities, including Peking University, Shanghai Jiao Tong University, and Xiamen University. Among them, the "Super Scientific Research Platform" built by Shanghai Jiao Tong University based on its X Laboratory is a prime example.

Supported by the Bohr Scientific Research Space Station, the platform deeply integrates the university library resources, the computing center, and the analysis and testing center, creating a dedicated agent toolchain for all teachers and students in the university. As the agent steward of the super scientific research platform, SciMaster can access the private data of each research group, integrate the self - built agents of teachers and students, and accelerate collaborative innovation.

SciMaster can directly call the MCP service provided by Uni - Lab, which seamlessly integrates various laboratory instruments and software systems, efficiently builds and manages laboratory workflows, and constructs a complete "wet - dry closed - loop" experimental ecosystem covering software, hardware, researchers, and scientific research data.

On this basis, the research team led by Professor Zhang Wanbin used the T5 - AsymCat agent to accelerate the process research and development of several blockbuster chiral drugs; Professor Dong Jiajia combined click chemistry with the molecular building block library, making AI an accelerator for demand - driven molecular design.

In addition, the platform also realizes cloud - based linkage with the wet - dry closed - loop device of the Yibin New Energy and New Materials Digital and Intelligent Innovation Center. After inputting the performance requirements of the electrolyte formula, it can conduct closed - loop optimization and iteration in combination with the dedicated knowledge base, the Uni - ELF formula large - model, a full set of automated stations, and the intelligent experimental system, increasing the R & D and delivery efficiency of the formula by at least 10 times.

The Open - Source Agent Platform

Open - source and openness are the future of scientific development, and SciMaster always embraces open - source and openness. Based on SciMaster, anyone can upgrade existing scientific computing tools to agents with just one click.

SciMaster has been connected to the DeepModeling open - source community, with an open contribution evaluation and benefit feedback mechanism to quantify the value of human - machine interaction. It has spawned thousands of Agent - ready tools and hundreds of Agent applications, and given birth to sub - agents of SciMaster such as ChemMaster, making R & D and design at the atomic and molecular scales accessible; the literature and patent reading, omics data mining and analysis modules of the virtual disease biologist system "Yuansheng" of the Lingang Laboratory are also open for global access.

SciMaster enables scientists to break through the limitations of original disciplinary boundaries and tools, improving scientific research efficiency; it also promotes the transformation of scientific research collaboration models.

SciMaster: The Path to the "Revolutionary Tool"

After humans experienced the reshaping of the physical world by steam, oil, and electricity, and the reconstruction of the information world by the Internet and supercomputers, a deeper - level transformation that connects the information world and the physical world is quietly approaching - artificial intelligence (AI) and scientific research are intertwining and co - evolving with unprecedented depth and breadth.

What supports SciMaster's excellent performance is the powerful scientific foundation large - model Innovator behind it and the Bohr Scientific Research Space Station that connects "reading, calculating, and doing".

The three work closely together to help scientists enhance their abilities in understanding scientific data, reasoning about scientific problems, and invoking scientific tools. Particularly crucially, Each human - machine interaction between researchers and SciMaster - from the perception and understanding of scientific problems, to complex logical reasoning, and then to precise tool invocation - continuously generates valuable scientific research data "nutrients". These data are continuously fed back to Innovator, further enhancing SciMaster's "reading, calculating, and doing" abilities.

This is an important way towards Artificial General Intelligence for Science (AGI for Science).

"The transformation of production tools (the revolution of tools) will inevitably lead to fundamental changes in social relations (the revolutionary tool)," Marx's judgment nearly two hundred years ago is being realized again in this profound transformation triggered by AI. Every technological revolution is a spiral leap in the boundaries of human cognition.

Welcome to join hands with SciMaster to embark on this new era towards Scientific General Intelligence (SGI). Your questions are the key "engine" to activate this leap.