svd-safety-mistral_keep50_gap_b001
A Mistral-7B-Instruct-v0.2 checkpoint compressed with SVD-LLM to 50.1% of dense
parameters, then given a 0.1% parameter budget of restored SVD
components selected by the gap 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
| field | value |
|---|---|
| base (uncompressed) | mistralai/Mistral-7B-Instruct-v0.2 |
| compression | SVD-LLM, 49.90% of parameters removed |
| selection rule | gap |
| restore budget | 0.100% of dense parameters |
| components restored | 1582 |
| components swapped out | 0 |
| resulting parameter fraction | 0.5010 |
| seed | 42 |
Measured
| metric | value |
|---|---|
| AdvBench ASR (HarmBench judge) | 0.2827 |
| StrongREJECT ASR (HarmBench judge) | 0.3163 |
| Macro over-refusal (WildGuard) | 0.1484 |
| WikiText-2 perplexity | 12.7494 |
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.