Jeesup/svd-safety-l3_remove40_swapgapiter_evfrontgap_b010
Jeesup/svd-safety-l3_remove40_swapgapiter_evfrontgap_b010 is an 8 billion parameter Llama-3-8B-Instruct checkpoint, compressed using SVD-LLM to 60% of its original dense parameters. This model has undergone 10 rounds of iterative parameter-neutral swap selected by the 'gap_iter' rule, restoring 1.0% of dense parameters. It serves as a research artifact for studying safety behavior under compression and component selection rules, rather than a general-purpose chat model.
Loading preview...
Overview
This model, svd-safety-l3_remove40_swapgapiter_evfrontgap_b010, is an 8 billion parameter Llama-3-8B-Instruct checkpoint that has been significantly modified through a research process. It was initially compressed using SVD-LLM, removing 40.02% of its parameters, resulting in a model that retains 59.98% of the original dense parameters. Subsequently, it underwent 10 rounds of iterative parameter-neutral swapping, guided by the gap_iter selection rule, which restored 1.0% of dense parameters.
Key Characteristics
- Base Model:
meta-llama/Meta-Llama-3-8B-Instruct - Compression Method: SVD-LLM, reducing parameters by 40.02%
- Parameter Restoration: 1.0% of dense parameters restored over 10 iterative rounds using the
gap_iterselection rule. - Resulting Size: Approximately 60% of the original dense parameters.
- Measured Metrics:
- AdvBench ASR (HarmBench judge): 0.0000
- StrongREJECT ASR (HarmBench judge): 0.0319
- Macro over-refusal (WildGuard): 0.4478
- WikiText-2 perplexity: 38.2635
Intended Use and Limitations
This model is explicitly a research artifact designed to measure safety/utility trade-offs under compression. It is part of a study to quantify how SVD compression impacts safety behavior and to test recovery mechanisms. Users should be aware that several arms in this research grid, including this model, are deliberately safety-degraded relative to the base Llama-3-8B-Instruct. It is not intended as a general-purpose deployable assistant and requires independent evaluation before drawing conclusions from its behavior.