Novel Operator Test

The Novel Operator Test, developed by Lawrence Livermore National Lab and CESER, is a benchmark that reveals AI models can execute correct reasoning steps but still produce incorrect final answers.

Two-hop subgraph: this entity, every entity it directly relates to, and every entity those neighbors relate to. Drag a node, scroll to zoom, click to inspect — or click any neighbor and re-center the atlas there.