Tag
论文提出 Magnolia，一个面向计算化学的持久记忆 AI 智能体框架，采用三层记忆系统（会话、项目、跨项目规则），整合对接、分子动力学与量子化学计算流程；在四个案例中验证了记忆机制显著提升任务检查通过率。
This paper introduces SO(3) Equivariant Neural Kalman Networks (SENK) for accurate and transferable vibrational spectral prediction, outperforming existing methods and integrating tensor prediction with physics-informed calibration.
EnSol is an environment-aware graph neural network that predicts molecular solubility by representing solutes and solvents as graphs and using cross-attention to model interactions, with probabilistic outputs to capture temperature effects and experimental uncertainty. It achieves state-of-the-art performance on benchmark datasets, validated experimentally.
GRACE is a novel 3D predictor for collision cross sections in ion mobility mass spectrometry, using geometric residual adduct conditioning via early fusion to improve accuracy and generalization.
ARCHE is an autonomous agentic system that integrates reasoning and computational models to automate the discovery and validation of chemical reaction mechanisms, validated through various chemical scenarios.
The paper introduces MolEmb, a lightweight framework that adapts multimodal large language models for general molecular embedding, enabling context-aware representations and cross-modal retrieval, along with a diagnostic benchmark MolCAR.
The author rants about the clunkiness of using SSH and VPN to access university HPC clusters for computational chemistry, arguing that the friction discourages independent researchers from doing work and advocates for local computations.
MDForge is an LLM agent that automates the design of molecular dynamics pipelines for host-guest binding free-energy calculations, achieving human-expert competitive results and discovering a novel high-affinity binder.
This ICML 2026 paper introduces Derivative Informed XC-Loss (DI-Loss), a training approach for machine-learned exchange-correlation functionals that incorporates first and second derivative supervision on the Grassmannian of density matrices. Across four architectures, DI-Loss reduces total-energy MAE by 66% compared to energy and density supervision alone, and improves excited-state predictions in TDDFT calculations.
SITA (Scalable Inference-Time Annealing) introduces a method for efficiently sampling molecular Boltzmann distributions by retraining flow-based models along a temperature ladder using energy-based surrogate likelihoods, avoiding costly divergence computations. The approach achieves state-of-the-art performance on Alanine Dipeptide and Tripeptide benchmarks.
The article introduces ARMOR, an agentic framework for predicting chemical reaction feasibility by adaptively prioritizing and resolving conflicts among multiple AI tools. It demonstrates superior performance over single-tool and aggregation methods on public datasets.
This paper presents an agentic system using Large Language Models to automate the discovery of exchange-correlation functionals in Density Functional Theory, achieving improvements over human-designed baselines while highlighting challenges with benchmark overfitting.
Researchers from Universitat Rovira i Virgili published a paper in Nature Machine Intelligence introducing CoCoGraph, an AI tool that generates chemically valid novel molecules using a constrained discrete diffusion process.
Microsoft Research releases Skala, a deep-learning exchange-correlation functional for DFT that achieves 2.8 kcal/mol accuracy on GMTKN55 at semi-local cost, outperforming traditional functionals across broad chemistry benchmarks.