CoolFace
Modelpublic

Jeesup/svd-safety-mis7_swift_remove40

sourceHugging Faceapache-2.0updated 5d agoView on Hugging Face
0likes279downloads
Model Card

svd-safety-mis7swiftremove40

A Mistral-7B-Instruct-v0.2 checkpoint compressed with SVD-LLM to 60.0% of dense parameters, then given a 0.0% parameter budget of restored SVD components selected by the `unknown` rule.

This is a research artifact from a study of how SVD compression damages safety behaviour and which component-selection rule best repairs it. It is one cell of a grid over selection rules and budgets; it is not a general-purpose chat model.

Provenance

fieldvalue
base (uncompressed)mistralai/Mistral-7B-Instruct-v0.2
compressionSVD-LLM, 40.00% of parameters removed
selection ruleunknown
restore budget0.000% of dense parameters
components restored0
components swapped out0
resulting parameter fraction0.6003
seed42

Measured

metricvalue
AdvBench ASR (HarmBench judge)0.6558
StrongREJECT ASR (HarmBench judge)0.5495
Macro over-refusal (WildGuard)0.0663
WikiText-2 perplexity10.7309

Intended use and limitations

This checkpoint exists to measure safety/utility trade-offs under compression. Several arms in the grid are deliberately safety-degraded relative to Mistral-7B-Instruct-v0.2: compression alone raises attack-success rate, and the point of the study is to quantify that and test recovery. Treat any given cell as an experimental subject, not as a deployable assistant, and evaluate it yourself before drawing conclusions from it.

Licence

Apache License 2.0. The base model's repository ships no licence file to redistribute; the licence above governs this derivative.