Downloads · 30 days
381
75% of all-time downloads
svd-safety-l2_remove40_random_b010
Jeesup/svd-safety-l2_remove40_random_b010
svd-safety-l2_remove40_random_b010 is a text generation model from Jeesup. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as llama2.
A Llama-2-7b-chat checkpoint compressed with SVD-LLM to 61.0% of dense parameters, then given a 1.0% parameter budget of restored SVD components selected by the random rule.
All-time downloads
509
Public
Parameters
6.7B
13.5 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors13.5 GB · 100%
At a glance
- Task
- Text Generation
- Library
- transformers
- License
- llama2
- Model type
- llama
- Access
- Public
- Created
- Sep 9, 2026
- Updated
- Sep 10, 2026
- SHA
- 2928760e
Try a prompt
Base models
Model card
From the Hugging Face model README
svd-safety-l2_remove40_random_b010
A Llama-2-7b-chat checkpoint compressed with SVD-LLM to 61.0% of dense
parameters, then given a 1.0% parameter budget of restored SVD
components selected by the random 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) | meta-llama/Llama-2-7b-chat-hf |
| compression | SVD-LLM, 39.03% of parameters removed |
| selection rule | random |
| restore budget | 1.000% of dense parameters |
| components restored | 5901 |
| components swapped out | 0 |
| resulting parameter fraction | 0.6097 |
| seed | 42 |
Measured
| metric | value |
|---|---|
| AdvBench ASR (HarmBench judge) | 0.3346 |
| StrongREJECT ASR (HarmBench judge) | 0.1725 |
| Macro over-refusal (WildGuard) | 0.1004 |
| WikiText-2 perplexity | 11.2594 |
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.
- Task
- Text Generation
- Library
- transformers
- Type
- llama
- License
- llama2
- Created
- Sep 9, 2026
- Updated
- Sep 10, 2026