arXiv subjects

Jinzhe Ma

Publications and source records attributed to Jinzhe Ma.

3 recordsLinked to original sources

arXiv2026

VASP Agent: An Agentic Framework for Autonomous First-principles Calculations

Large Language Models (LLMs) are increasingly embedded in agentic frameworks for scientific discovery. First-principles materials computation imposes a demanding standard for autonomy: successful execution depends on internally consistent inputs, supervision of long-running calculations, and verified outputs. Here we present VASP Agent, a coding-agent-centered system that combines reusable domain skills, deterministic tools, workspace-state inspection, runtime evidence, and scientific guardrails to execute multi-step VASP calculations. The system is evaluated across multiple tasks including structural relaxation, bandgap calculation, equilibrium lattice constant determination, and CO/Pt(111) adsorption. VASP Agent completes all evaluated cases, and its computed numerical results are compared with those obtained using pymatgen and other agentic tools. When large deviations occur, the calculation parameters produced by VASP Agent are more appropriate than those produced by LLM-based workflows. Failure analysis shows that errors that terminate fixed pipelines can be diagnosed and recovered under agentic control.

cs.AI↗

arXiv2025

Probing Scientific General Intelligence of LLMs with Scientist-Aligned Workflows

Despite advances in scientific AI, a coherent framework for Scientific General Intelligence (SGI)-the ability to autonomously conceive, investigate, and reason across scientific domains-remains lacking. We present an operational SGI definition grounded in the Practical Inquiry Model (PIM: Deliberation, Conception, Action, Perception) and operationalize it via four scientist-aligned tasks: deep research, idea generation, dry/wet experiments, and experimental reasoning. SGI-Bench comprises over 1,000 expert-curated, cross-disciplinary samples inspired by Science's 125 Big Questions, enabling systematic evaluation of state-of-the-art LLMs. Results reveal gaps: low exact match (10--20%) in deep research despite step-level alignment; ideas lacking feasibility and detail; high code executability but low execution result accuracy in dry experiments; low sequence fidelity in wet protocols; and persistent multimodal comparative-reasoning challenges. We further introduce Test-Time Reinforcement Learning (TTRL), which optimizes retrieval-augmented novelty rewards at inference, enhancing hypothesis novelty without reference answer. Together, our PIM-grounded definition, workflow-centric benchmark, and empirical insights establish a foundation for AI systems that genuinely participate in scientific discovery.

cs.AI↗

arXiv2025

Evidential Deep Learning for Interatomic Potentials

Machine learning interatomic potentials (MLIPs) have been widely used to facilitate large-scale molecular simulations with accuracy comparable to ab initio methods. In practice, MLIP-based molecular simulations often encounter the issue of collapse due to reduced prediction accuracy for out-of-distribution (OOD) data. Addressing this issue requires enriching the training dataset through active learning, where uncertainty serves as a critical indicator for identifying and collecting OOD data. However, existing uncertainty quantification (UQ) methods tend to involve either expensive computations or compromise prediction accuracy. In this work, we introduce evidential deep learning for interatomic potentials (eIP) with a physics-inspired design. Our experiments indicate that eIP provides reliable UQ results without significant computational overhead or decreased prediction accuracy, consistently outperforming other UQ methods across a variety of datasets. Furthermore, we demonstrate the applications of eIP in exploring diverse atomic configurations, using examples including water and universal potentials. These results highlight the potential of eIP as a robust and efficient alternative for UQ in molecular simulations.

physics.comp-ph↗