A Safe Prototype Is Not a Safety Direction: Reference Dependence and Prompt Confounds in Response-Safety Embeddings
Reported by arXivBharatHunt Trend Score37
Source: arXiv
Read original articleSummary
Reported by arXiv.
Key points
- First reported by arXiv on 1 Oct 2026.
- Reported so far by arXiv alone — worth checking the original before relying on it.
- Filed under AI Models.
Summary assembled automatically by BharatHunt from the headline and the coverage listed below — not AI-generated, and not a reproduction of the article. The original reporting is the source of truth.