Modelpublic

Jeesup/svd-safety-l2_remove50_swapdisciter_r01

sourceHugging Facellama2updated 10d agoView on Hugging Face

3likes391downloads

Model Card

svd-safety-l2remove50swapdisciter_r01

A Llama-2-7b-chat checkpoint compressed with SVD-LLM to 50.0% of dense parameters, then edited by 1 of 10 rounds of iterative parameter-neutral swap selected by the `disc_iter` rule (up to 0.1% of dense parameters per round; the full run's budget is 1.0%).

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)	`meta-llama/Llama-2-7b-chat-hf`
compression	SVD-LLM, 50.01% of parameters removed
selection rule	`disc_iter`
restore budget	1.000% of dense parameters
components restored	682
components swapped out	682
resulting parameter fraction	0.4999
seed	42
iterative rounds applied	1 of 10
per-round chunk	0.100% of dense parameters
parameters swapped in	6,471,680 (0.10% of dense projection parameters)
swap value	`insert` (insertion value only; sigma-ordered eviction)
checkpoint	intermediate round of a longer run

Measured

metric	value
AdvBench ASR (HarmBench judge)	0.2096
StrongREJECT ASR (HarmBench judge)	0.1597
Macro over-refusal (WildGuard)	0.2712
WikiText-2 perplexity	13.6373

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 Llama-2-7b-chat: 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

Llama 2 Community License. LICENSE.txt and USE_POLICY.md are included in this repository, and use of this derivative is bound by both. Built with Llama 2.