Materials Science Agent MatOS Officially Open-Sourced! Significantly Boosts Success Rate of Materials R&D Tasks

Screening materials, finding formulations, developing new models… tackling multiple research tasks
Yunzhong, from Feisi, QuantumBit | WeChat public account QbitAI
QuantumBit | WeChat public account QbitAI
Large model agents can already retrieve literature, invoke simulation software, and even collaborate with automated laboratories. But getting them to independently complete a full materials research project still faces challenges:
A single study may require progressively screening tens of millions of candidate structures, or repeatedly searching through hundreds of alloy formulations over several days, continuously adjusting plans based on intermediate results; each time a new database or software is integrated, the pipeline often has to be rebuilt; and the experience accumulated in one study is difficult to reuse in others.
To this end, the Beijing Zhongguancun Academy and the Zhongguancun Institute of Artificial Intelligence (Zhongguancun Two Institutes) released their annual research progress in September this year — MatOS, a research agent system for materials R&D.
The project was built by the AI Materials Research Institute team at the Zhongguancun Two Institutes. It organizes data, knowledge, tools, skills, roles, and workflows into unified scientific assets, supports researchers and MatOS itself in building new capabilities on top of existing assets, executes research tasks through stateful workflows, and applies inspected and reviewed asset revisions to subsequent research, enabling research capabilities to continuously evolve through use.
At present, MatOS's asset library contains 3,135 datasets, 386 skills, 104 workflows, and 51 expert roles, and integrates 48 materials computation software packages; the technical report has been published on arXiv, and the code has been open-sourced on GitHub.

Across 3 agent benchmarks and 8 research cases, MatOS completed a variety of tasks ranging from end-to-end autonomous research and tool invocation to large-scale screening, formulation design, and new method development, for example:
On the end-to-end autonomous research benchmark ResearchClawBench, compared with Codex CLI using the same base model, the average score is 33.6% higher, and on materials tasks 80.8% higher;
On the materials software usage benchmark MatTools, when using the same base model as Codex CLI and Claude Code, the complete-question success rate reached 97.96%, the highest of the three;
Building a hierarchical screening pipeline, it screened 97 solid-state electrolyte candidates from more than 30 million candidate structures;
It supported the development of the alloy melting point prediction model DuoCompNet, reducing the mean absolute error on a fixed test set to 3.18°C, 13.6% lower than the second-best model.

From answering questions to conducting research, and on to continuously producing high-quality scientific assets
In recent years, work such as ChemCrow and LLaMP has enabled large models to invoke specialized tools, and systems such as MARS connect agents with automated experiments; research agents are moving from answering questions toward conducting research.
Materials research imposes three requirements on agent systems:
Databases, simulation software, and machine learning potentials (methods that use machine learning models to approximate interatomic forces and accelerate simulations) keep emerging, requiring the system to integrate new capabilities without rebuilding;
Research spans literature analysis, computation, and experiments, requiring the system to preserve state and check results in long-duration tasks;
Experience such as simulation duration and sampling strategies mostly comes from practice, and the system must be able to store this experience and effectively reuse it in subsequent tasks.
MatOS responds to these requirements with three major designs: a hierarchical scientific asset graph that organizes and extends R&D capabilities, stateful adaptive workflows that invoke assets to execute tasks and preserve evidence, and evidence-driven asset evolution that turns reusable experience into inspected and reviewed asset revisions.

Overall architecture of MatOS: hierarchical scientific asset graph, stateful adaptive workflows, and evidence-driven asset evolution.
Three major designs: construction, execution, and evolution
Hierarchical scientific asset graph: combining existing capabilities to build new ones
MatOS organizes data, knowledge, tools, skills, roles, and workflows into six categories of scientific assets.
Each asset has a unique identifier, declared dependencies, and provenance records, and can either store content directly or reference external databases or computation software.
Assets are connected through typed relations such as dependency, invocation, evidence, and validation; when an agent receives a task, it combines the required capabilities according to these relations.
Among them, a skill is equivalent to a set of operating procedures for a class of research problems, specifying when to use it, how to select methods, what prerequisites are needed, how to execute, and how to check results; a role is an expert agent defined by a responsibility description and invocable assets; a workflow describes how multiple research stages connect. The same skill can be used by multiple roles, and the same role can participate in different workflows.
Both researchers and MatOS itself can build new capabilities on top of existing assets.
Researchers can register databases, software, or methods as assets and combine them into skills, roles, or workflows according to interfaces and dependencies; MatOS can also create or revise assets based on task evidence, which, after inspection and review, are reused in subsequent tasks. The use and revision of assets are all recorded, and every improvement is traceable.

How MatOS organizes scientific assets: (a) six categories of assets belong to four semantic layers, abstracting layer by layer from data, knowledge, and tools to skills, roles, and workflows; (b) taking the battery role as an example, the role selects and combines data, knowledge, tools, skills, and workflows from the asset graph.
Stateful adaptive workflows: stable execution and on-demand adjustment
Workflow assets are reusable process descriptions, while a specific study runs through an execution graph. Each node in the execution graph specifies its goal, inputs and outputs, completion conditions, and the assets it references. The system schedules and checks according to the graph, delegating to the agent only the steps that require judgment, such as method selection, plan adjustment, and anomaly interpretation; steps requiring researcher confirmation can be set as human nodes.
Because large models may produce different outputs for the same input, decisions in long tasks can easily drift from the goal as context changes. MatOS therefore reduces execution's dependence on model sessions through explicit steps and check conditions, lowering the risk of deviating from research goals in multi-step execution.
Execution graph snapshots, node states, results, and event records are all stored in the system, independent of the model's session.
Computations that take several days are handed off to background execution; when they complete, need input, or become unresponsive, the system proactively notifies the agent; tasks waiting for external results or human input retain their progress and continue execution afterward. Every output is associated with the node and execution attempt that produced it.
Intermediate results may require adjusting subsequent steps, for example adding a verification step first and then having subsequent analysis rely on the verification result. At this point, the agent or researcher can propose adjustments to the execution graph, adding or modifying nodes and changing dependencies. Adjustments must pass three checks — structural, policy, and compilation — after which a new version of the execution graph is formed; if they fail, the original execution graph continues to run. The proposal, check results, and adoption decision of each adjustment are all recorded, and the evolution of the execution graph is fully traceable.
Adjustments to the execution graph primarily serve the current task. Process improvements with reuse value enter asset evolution as evidence and, after inspection and review, are used to revise the corresponding workflow assets for invocation in subsequent similar tasks.

Stateful adaptive workflows: node types, execution graphs and their dynamic adjustment, and persistent state.
Evidence-driven asset evolution: continuous improvement through use
User corrections, failed tool invocations, recurring operations, and workflow results may all indicate that an asset needs improvement.
MatOS first assesses whether the evidence is sufficient, whether the problem holds, and whether the experience is reusable, then locates the asset responsible for the corresponding function: a wrong method choice may require revising knowledge or skills, wrong tool behavior requires fixing the tool, and missing research steps require revising the workflow; when necessary, it also creates new assets or revises multiple related assets together.
For example, if feedback indicates that the equilibration time of a molecular dynamics simulation is too short, the system can change the equilibration duration in the corresponding skill from 50ps to at least 200ps, and add a check to confirm that atomic diffusion has reached a stable state.
Candidate revisions must pass a structural check and a model-based semantic review: the structural check mainly determines whether new structural, interface, or composition errors are introduced; the semantic review determines whether the revision addresses the original evidence, preserves the method's applicability conditions, and stays consistent with related assets.
Only revisions that pass inspection and review are adopted, and the results of subsequent tasks become evidence for the next round of evaluation. This experience is stored in assets outside the model, and the parameters of the base language model remain unchanged; assets are also not bound to a specific model — in the evaluation, the same MatOS was paired with GPT-5.4 and the DeepSeek series models respectively.

Evidence-driven asset evolution: evidence assessment, target localization, local revision, structural check and semantic review, version adoption.
Benchmark evaluation results
ResearchClawBench requires the system to conduct research based on literature and raw data and submit a report. The team selected 12 tasks in the fields of chemistry, energy, and materials, and compared with 7 research agents using GPT-5.4, including MIRA, ResearchClaw, EvoScientist, and Codex CLI.
When using GPT-5.4, MatOS achieved an average score of 19.46, 11.15% higher than the best baseline MIRA and 33.6% higher than Codex CLI. Performance on materials tasks was even more prominent: 18.7% higher than MIRA and 80.8% higher than Codex CLI. When switching to the lower-cost DeepSeek-V4-Flash, MatOS achieved an average score of 19.20, also higher than all baselines.

Per-domain average scores and overall average scores for the chemistry, energy, and materials tasks of ResearchClawBench.
Title: Materials science agent MatOS officially open-sourced! Significantly improves success rate of materials R&D tasks
Body: In terms of cost, the DeepSeek-V4-Flash configuration has an average cost per task of about $0.85, less than one-third of Codex CLI ($2.70), with a higher score; the GPT-5.4 configuration achieves the highest score at a cost of about $3.78. The two MatOS configurations together expand the cost-score Pareto frontier in this evaluation (that is, among multiple objectives, finding a set of optimal choices that each involve trade-offs but none of which can be comprehensively outperformed by other solutions).

Average cost per task and average score of each system (cost on a logarithmic scale).
On the materials software usage benchmark MatTools and the multi-step tool calling benchmark SciAgentGym, all three systems—MatOS, Codex CLI, and Claude Code—use DeepSeek-V4-Pro as the base model.
MatOS's full-problem success rate on MatTools is 97.96%, higher than Codex CLI's 95.92% and Claude Code's 89.80%; its subtask success rate is 97.10%, compared with 95.65% and 91.30% for the other two, respectively. On SciAgentGym, MatOS's case success rate is 59.5%, higher than the other two systems' 57.1%.

With DeepSeek-V4-Pro as the common base model, the success rates of the three systems on MatTools and SciAgentGym.
Research cases: from large-scale screening to new method development
The 8 research cases cover materials directions such as machine learning potentials, solid and liquid electrolytes, and alloys, and extend to fluid mechanics. Task types include large-scale screening, molecular simulation, formulation design, and new method development.
In large-scale screening, MatOS follows the hierarchical screening framework proposed by Du et al.: first, a model based only on chemical composition is used for rapid coarse screening; then a graph neural network that predicts properties from crystal structure is used for fine screening; finally, molecular dynamics simulations are used to model ion motion, progressively examining conductivity, electrochemical stability, and mechanical properties, narrowing 30,364,908 initial structures down to 97 solid electrolyte candidates, with calculated conductivity up to 59.0 mS/cm.

Hierarchical screening of solid electrolytes: 30,364,908 initial structures are ultimately narrowed to 97 candidates.
In formulation design, MatOS targets semiconductor packaging and searches for alloys that begin to melt at around 300°C and have a narrow melting temperature range: the large model proposes candidate formulations in four parallel groups, and a surrogate model (a model for fast approximate computation) predicts the proportion of solid and liquid phases of each formulation at different temperatures, with scoring results guiding the next round of search.
Among the 286 formulations evaluated, the highest score increased from 0.534 for the starting formulation to 0.927; the highest-scoring Pb–Ag–Pt–Sn alloy is expected to begin melting at 297.75°C, and after heating by 4.55°C the liquid phase proportion reaches 80%.

Alloy formulation search targeting melting at 300°C: (a) scores of formulations in four parallel searches, with the black line showing the current highest score; (b) predicted liquid phase proportion near 300°C for the starting formulation and the highest-scoring formulation.
In new method development, the research team used MatOS to build an active learning workflow: the system selects alloy formulations that need thermodynamic calculations, and after the calculation results are returned, the model is retrained, with data and model expanding in sync.
DuoCompNet, developed in this process, is used to predict the midpoint of the temperature range from the start of melting to complete melting for tin-containing alloys. As training data increased, its mean absolute error on a fixed test set dropped from 28.40°C to 3.18°C, the lowest among the six models compared, 13.6% lower than the second-best Roost; the test set consists of 1,401 tin-containing formulations composed of two or three elements with melting range midpoints between 200°C and 400°C, with reference values from thermodynamic calculations.
With the training data selection method RarityGuard-MP developed using MatOS, 100,000 structures were selected from about 1.26 million structures for training machine learning potentials. At the same data volume, the combined energy, force, and stress error of the trained model is 3.10% lower than that of the DIRECT method.
Summary and outlook
MatOS connects capability building, task execution, and asset evolution:
Researchers can plug in their own data, software, and methods; the system can propose asset revisions from task evidence; new assets and revisions adopted after inspection and review become capabilities that can be directly invoked in subsequent research, so new research does not have to start from scratch.
Next, MatOS will connect instruments and experimental workflows, so that the evidence of a study from computational screening to experimental validation can be linked within the same system.
Technical report: MatOS: An Agent Operating System for Materials Research Report link: https://arxiv.org/abs/2610.34203 Code: https://github.com/DingyangLyu/MatOS AI Materials Research Institute contact email: [email protected]
*This article is published by QbitAI with authorization; the views are solely those of the original author.
- Aimotion, joined by Li Mu, what kind of AI company is it? 2026-11-17
- Loongson and Hygon veterans assemble, the AGI chip company incubated by Tongyan Institute emerges! Raised hundreds of millions in consecutive rounds within half a year 2026-11-16
- 2.48x faster without compromising image quality, AI video "cost reduction and efficiency improvement" has a new solution 2026-11-16
- High-difficulty blind test surpasses GPT-6-Astra and Opus-5! Wujie Frontier open-sources Fx-Bio-1031, bringing RSI into life sciences 2026-11-13
Related reading

This wood is sharper than steel and ceramics, easily cuts medium-rare steak, nails through three layers of wooden boards, and never rusts | Cell sub-journal
After being hardened 23 times, this piece of wood became extraordinary.