Jeesup/svd-safety-llama3_8b_instruct_remove_30_seed3_jbbmixsft
Jeesup/svd-safety-llama3_8b_instruct_remove_30_seed3_jbbmixsft is an 8 billion parameter Llama-3-8B-Instruct checkpoint that has been compressed using SVD-LLM, removing 30% of its dense parameters. This model is a research artifact designed to study how SVD compression impacts safety behavior and to test component-selection rules for repair. It is specifically intended for experimental evaluation of safety/utility trade-offs under compression, rather than as a general-purpose chat model.
Loading preview...
Overview
This model, svd-safety-llama3_8b_instruct_remove_30_seed3_jbbmixsft, is an 8 billion parameter Llama-3-8B-Instruct checkpoint. It has undergone SVD-LLM compression, resulting in the removal of 30% of its dense parameters, with a 0% budget for restoring SVD components using an 'unknown' selection rule. This specific configuration is a research artifact from a study investigating the effects of SVD compression on model safety and the efficacy of various component-selection rules for recovery.
Key Characteristics
- Base Model: Derived from
meta-llama/Meta-Llama-3-8B-Instruct. - Compression Method: SVD-LLM, with 30.00% of parameters removed.
- Parameter Fraction: The resulting model retains approximately 69.99% of the original dense parameters.
- Measured Metrics:
- AdvBench ASR (HarmBench judge): 0.0250
- StrongREJECT ASR (HarmBench judge): 0.0256
- Macro over-refusal (WildGuard): 0.4777
- WikiText-2 perplexity: 17.0613
Intended Use and Limitations
This model is not a general-purpose chat model. Its primary purpose is to serve as an experimental subject for measuring safety/utility trade-offs under compression. The study aims to quantify how compression alone can increase attack-success rates and to test recovery mechanisms. Users should treat this checkpoint as an experimental artifact, acknowledging that it may be deliberately safety-degraded relative to the original Llama-3-8B-Instruct. Independent evaluation is strongly recommended before drawing any conclusions or considering deployment.