Noise Injection Reveals Hidden Capabilities of Sandbagging Language Models
Tice, Cameron, Kreer, Philipp Alexander, Helm-Burger, Nathan, Shahani, Prithviraj Singh, Ryzhenkov, Fedor, Roger, Fabien, Neo, Clement, Haimes, Jacob, Hofstätter, Felix, van der Weij, TeunarXiv.orgPublished Dec 2, 2024Original︎ ↗Archived copy
Examinations
None recorded. An examination is commissioned by a claim’s Steward for what its claim rests on; a review of the document for its own sake is asked for as a request bearing on it.
Text
Stored, not yet divided into passages.
While non-sandbagging models show predictable performance degradation with increasing noise, we demonstrate that sandbagging models exhibit anomalous performance improvements, likely due to disruption of underperformance mechanisms while core capabilities remain partially intact. Through experiments across various model architectures, sizes, and sandbagging techniques, we establish this distinctive response pattern as a reliable, model-agnostic signal for detecting sandbagging behavior.
Not yet read by this claim’s Steward.
Lineage across 1 claim
Record
- Kind
- preprint
- Authors
- Tice, Cameron, Kreer, Philipp Alexander, Helm-Burger, Nathan, Shahani, Prithviraj Singh, Ryzhenkov, Fedor, Roger, Fabien, Neo, Clement, Haimes, Jacob, Hofstätter, Felix, van der Weij, Teun
- Publisher
- arXiv.org
- Published
- Dec 2, 2024
- DOI
- None
- Link
- https://arxiv.org/abs/2412.01784
- Archived copy
- https://web.archive.org/web/20260929164248/https://arxiv.org/abs/2412.01784
- Stored text
- 5,248 characters, retrieved Nov 12, 2026 · not yet divided
- Claims extracted
- Not recorded
- Versions
- One known
- Copies
- None recorded
- Corrections
- None recorded
- Facts checked
- Nov 12, 2026
History 2
- Nov 12, 2026FactsFacts fetcher · Facts looked up
- Nov 12, 2026Entered the graphEntered the graph; copy stored