Repository navigation
FAQ
No. The explorer shows ten historical configurations and unranked development observations. No qualified held-out entry or PBI score is published.
No. The toolkit validates supplied evidence and calculates scoped scores. Workbench is the document-review application whose workflow helps motivate the research. This release does not qualify a new Workbench installer or end-to-end app campaign.
No. Schema compliance, source citation existence and semantic correctness are distinct. An output can satisfy the first two while changing an obligation.
Not for the synthetic evaluator or archived offline replay. Python dependencies must be installed. Fresh inference experiments have separate runtime, access and cost requirements.
The evaluator cannot establish reviewer independence, external custody or trusted registration from submitter-supplied records. Qualification needs an independent, scoped decision that is not yet operational.
No. They were used during PALO development. Their size does not establish independent generalisation or successful reviewed exports.
It is a proposed future composite. Formula, weights and eligibility gates require preregistration and validation. The eight dimensions remain a research framework, not an implemented aggregate score.
Do not post them publicly. Begin with a public URL or synthetic example and discuss rights, scope and controlled access before sharing any private material.
The project and toolkit are free; code is MIT licensed. External model services can charge for inference. Third-party documents and models retain their own terms.
No. Andon Labs evaluates spatial reconstruction. PALO studies business development specifications. The projects have no affiliation.
Use CITATION.cff, name the exact release or commit, and disclose whether the cited evidence is synthetic or developmental. No DOI or peer-reviewed validation is claimed.
Research pilot | Zero qualified held-out results | Versioned wiki source | Corrections