Experimental artifacts for "From Experiments to Expertise: Scientific Knowledge Consolidation for AI-Driven Computational Physics" (QMatSuite)

This record contains the experimental artifacts accompanying the paper "From Experiments to Expertise: Scientific Knowledge Consolidation for AI-Driven Computational Physics" (AI4Physics Workshop, ICML 2026), which introduces QMatSuite, an open-source platform that gives LLM agents persistent, provenance-tracked scientific memory for computational materials science. The archive includes agent execution traces, knowledge-database snapshots, reviewed-insight tables, and the supporting Quantum ESPRESSO / Wannier90 / ORCA calculation outputs for the experiments reported in the paper: the 135-material solid-state scale validation, the 98-molecule ORCA cross-engine validation, the bcc-Fe anomalous Hall conductivity learning curve, the Fe→Ni cross-material transfer, the knowledge-consolidation chain, and the pattern-guided design experiment. All personal identifiers have been redacted. See MANIFEST.json for the file index and per-file provenance. The QMatSuite platform is available at https://github.com/QMatSuite/QMatSuite.