FlorianJK/Meta-Llama-3-8B-SecAlign-Merged

Meta-Llama-3-8B-Instruct — SecAlign (Merged)

A fully merged model based on meta-llama/Meta-Llama-3-8B-Instruct fine-tuned with SecAlign to make the model resistant to prompt injection attacks.

This is the merged (standalone) version of the PEFT LoRA adapter FlorianJK/Meta-Llama-3-8B-SecAlign. The adapter weights have been merged into the base model, so no PEFT library is required for inference.

Model Details

• Base model: meta-llama/Meta-Llama-3-8B-Instruct
• Source adapter: FlorianJK/Meta-Llama-3-8B-SecAlign
• Fine-tuning method: DPO (Direct Preference Optimisation) via SecAlign
• Adapter type: PEFT LoRA (library version 0.14.0), merged into base model
• Training data: 104-sample subset of AlpacaEval (text-davinci-003 reference outputs, samples with non-empty input field)

Security Evaluation

Attack success rate measured on 104 samples from AlpacaEval with no additional defense prompting.

↓ lower is better — the model should not follow injected instructions.

• in-response — fraction of outputs containing the injected trigger word
• begin-with — fraction of outputs that begin with the injected trigger word

This model (SecAlign)

Attack	In-Response ↓	Begin-With ↓
ignore	1.9%	0.0%
completion_real	0.0%	0.0%
completion_realcmb	0.0%	0.0%
gcg	8.2%	0.0%

Undefended base model (Meta-Llama-3-8B-Instruct)

Attack	In-Response	Begin-With
ignore	65.4%	20.7%
completion_real	81.7%	47.1%
completion_realcmb	83.2%	55.3%
gcg	85.6%	6.3%

Utility Evaluation

Win-rate on the full 805-sample AlpacaEval 2 benchmark (judge: gpt-4o-2024-08-06).

Model	LC Win-Rate	Win-Rate	Avg Length
Meta-Llama-3-8B-Instruct (base)	31.41%	30.69%	1947
This adapter (SecAlign)	28.32%	26.15%	1838

Usage

Since the adapter is fully merged, the model can be loaded directly with transformers:

from transformers import AutoTokenizer, AutoModelForCausalLM

model = AutoModelForCausalLM.from_pretrained("FlorianJK/Meta-Llama-3-8B-SecAlign-Merged")
tokenizer = AutoTokenizer.from_pretrained("FlorianJK/Meta-Llama-3-8B-SecAlign-Merged")

It is also compatible with vLLM:

from vllm import LLM
llm = LLM(model="FlorianJK/Meta-Llama-3-8B-SecAlign-Merged")

Related Models

Model	Description
FlorianJK/Meta-Llama-3-8B-SecAlign	Source PEFT LoRA adapter (before merging)
FlorianJK/Meta-Llama-3-8B-SecUnalign-Merged	Same architecture fine-tuned with inverted preferences — intentionally vulnerable to prompt injection
FlorianJK/Meta-Llama-3-8B-SecUnalign	SecUnalign PEFT LoRA adapter — intentionally vulnerable to prompt injection