J-Lens-informed reasoning observability skill for OpenClaw, Claude, Codex, and AI harnesses
The library's lenses default to final layer. If you read the replication only, then go straight to the library, you might think the lens is penultimate fitted. That is all.
Pre-registered Jacobian-lens probe: does a verbatim-stated trait ("Maria is generous") persist in a model's J-space differently than an inferred one, as neutral text intervenes? Gema-3-4B-IT.
Pre-registered Jacobian-lens probe: does a verbatim-stated trait ("Maria is generous") persist in a model's J-space differently than an inferred one, as neutral text intervenes? Gema-3-4B-IT. No directional prediction.
Independent interpretability research (DRAFT — not yet run): is an inferred character trait stored, or reconstructed from the scene on demand? Across model scale + a scene-masked direction probe
Add a description, image, and links to the j-lens topic page so that developers can more easily learn about it.
To associate your repository with the j-lens topic, visit your repo's landing page and select "manage topics."