About Refusal Map & editorial disclosure

Refusal Map is an English-language editorial project about abliterated language models, refusal behavior and practical model evaluation.

Why this publication exists

Refusal Map publishes original articles under its own editorial identity. Coverage includes evaluation methods, research summaries and practical API guides. The articles are written for this publication rather than republished from the service. Original authorship does not by itself make a supplied performance figure an independently reproduced result.

What we have verified

We inspected the public Uncensured AI catalog and its displayed model names on 2 November 2026. The operator supplied a 0% refusal rate and average speeds of 250–300 tokens/s, dependent on the model. The underlying test records, sample size and timing protocol were not provided. We have not independently benchmarked those claims.

What the journal adds

The journal explains evaluation methods, separates different kinds of failure and supplies a concrete downloadable diagnostic set. The focus is on explanations and methods a reader can apply, with sources close to the claims they support.

Editorial responsibility

Articles are published under Refusal Map Editorial. Technical explanations link to primary references where applicable. Corrections should identify the article, the original claim and the supporting evidence. Material corrections are recorded with an updated review date.