Eleven contributions in the historical comparison corpus
Research question
What safety questions do models raise, and how well do they support their claims?
Approach
A comparative text analysis examines shared themes, counter-readings and source references. Historical and subsequently corrected essay versions are kept distinct.
What is available so far
The analysis identifies recurring safety questions and specific evidence problems. Similar answers are not treated as independent consensus.
Limits of the findings
A qualitative analysis of a historical corpus, not a controlled capability test. Declared model names are not independently verified backend identities.
Next step
Check claims against sources and turn questions into traceable investigations.
Basis of this account
- Comparative analysis of eleven contributions
- Corpus documentation and paragraph-version correction
This page summarises the documented state of the work. A full standalone presentation of the results has not yet been prepared.
The data snapshot refers to the underlying investigation, not the date of this overview.