Construct · 1 fields · 1 citations

exaggerated safety: a standalone construct

One name. One construct. A landmark on the map.

● anchor · a standalone construct Nearest construct by meaning·not indexed

A single standalone construct; nearby constructs are located against it in the Collider. Tap a field for its definition, unit, and measures.

Natural language processing

a safety-tuned model refusing clearly safe requests because they superficially resemble unsafe ones, such as asking how to kill a computer process.

In natural language processing, exaggerated safety refers to a safety-tuned model refusing clearly safe requests because they superficially resemble unsafe ones, such as asking how to kill a computer process. This meaning comes from Röttger (2023) in arXiv.[1]

Unit
Recovery means
Measured by

Run it through the Collider, or take the citations with you.

See how exaggerated safety sits against neighbouring constructs in the Collider, or export every reference on this page in one click.

Open in Collider Export BibTeX Report this entry

Nearby constructs

model3 fields · 3 citationsjingle · one word, three constructs natural3 fields · 7 citationsjingle · one word, three constructs process3 fields · 3 citationsjingle · one word, three constructs

References

Röttger, P., Kirk, H. R., Vidgen, B., Attanasio, G., Bianchi, F., & Hovy, D. (2023). XSTest: A Test Suite for Identifying Exaggerated Safety Behaviours in Large Language Models. arXiv. https://doi.org/10.48550/arxiv.2308.01263

Cited on this page?

Confirm your construct, or correct us.

If your work is cited behind one of these entries, you know better than we do whether we captured the meaning right. Confirm it, suggest an edit, or dispute it, and your name goes on the entry as an author who has checked it, linked to your ORCID.

Claim your construct