Jeesup/svd-safety-mis7_swift_remove50
Jeesup/svd-safety-mis7_swift_remove50 is a 7 billion parameter Mistral-7B-Instruct-v0.2 checkpoint compressed using Swift-SVD to 50% of its original parameters, then recovered with SVD-LLM's stage-2 LoRA. This model is a research artifact designed to study how SVD compression impacts safety behavior and the effectiveness of different component-selection rules for recovery. 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
Jeesup/svd-safety-mis7_swift_remove50 is a research artifact derived from mistralai/Mistral-7B-Instruct-v0.2. It has been significantly compressed using Swift-SVD, reducing its parameters to 50.0% of the original, and subsequently recovered using SVD-LLM's stage-2 LoRA. This model is part of a larger study investigating the effects of SVD compression on safety behavior and the efficacy of various component-selection rules for repair.
Key Characteristics
- Base Model:
mistralai/Mistral-7B-Instruct-v0.2 - Compression Method: Swift-SVD with dynamic rank allocation (alpha 0.6, WikiText2 calibration), achieving 50.0% parameter reduction.
- Recovery Method: SVD-LLM's stage-2 LoRA (sequential U then V, alpaca-cleaned, r=8, alpha=16).
- Measured Metrics:
- AdvBench ASR (HarmBench judge): 0.6558
- StrongREJECT ASR (HarmBench judge): 0.5623
- Macro over-refusal (WildGuard): 0.0801
- WikiText-2 perplexity: 13.9674
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 compression process has deliberately degraded safety relative to the base Mistral-7B-Instruct-v0.2, and the study aims to quantify this degradation and test recovery mechanisms. Users should treat this checkpoint as an experimental subject and conduct their own evaluations before drawing conclusions or considering deployment.