An agentic framework ran closed-loop measurements on a real quantum-materials instrument and revised its own plans based on the data it collected
AspireAI community · arXiv (Rha, Chotrattanapituk, Cheng, Chou, Ye, Dale, Ju, Liu, Li)
Characterizing quantum materials needs experimental expertise, iterative decision-making and accumulated lab knowledge that is hard to document and transfer. QUANTA is an agentic AI framework for autonomous quantum-materials characterization, built on a widely adopted Physical Property Measurement System. It combines retrieval augmentation over scientific literature, experimental protocols and prior measurements with specialized agents for experimental planning, troubleshooting and data interpretation, while a separate deterministic control layer validates instrument actions. Guided by incoming data, QUANTA evaluates measurement quality, revises experimental plans and selects subsequent measurements. The demonstrations are transport measurements of the topological Weyl semimetal TaP, and investigations of nonlinear responses in the magnetic material Co2MnGa. The result establishes QUANTA as a generalizable framework for evidence-driven, safeguarded and closed-loop characterization. The limitation is that the deterministic control layer doing the validating is doing real work - the agent proposes and revises, but the instrument only moves because something else checked it - and the paper reports two material systems rather than a broad lab record.
Read the story and explore more from the AspireAI community.
Two sparse-autoencoder analyses have resisted scaling: matching semantically similar features across layers, and compressing large feature circuits into interpretable supernodes. This paper argues both are the same underlying problem, estimating semantic distances between SAE features that sit on different activation manifolds. Its representation treats each feature not as a single decoder vector but as an activation-weighted distribution over the hidden states expressing it, projects those distributions into a shared reference space, and compares them with Wasserstein distance. The representation is proved invariant to activation rescaling and stable under perturbation, and recovers true matches under finite-sample margin conditions. Empirically it outperformed decoder-vector and LLM-based baselines and captured subtle functional distinctions between related features, and it compressed large feature circuits into interpretable supernodes automatically. The limitation is that this is interpretability tooling: the paper demonstrates better feature matching and circuit compression, not that downstream interpretability claims become more accurate.