virtual-prompt-injection
Unofficial implementation of Virtual Prompt Injection attack on instruction-tuned LLMs
GraphCanon updated Oct 5, 2026 · GitHub synced Oct 5, 2026
33views this month
Decision brief
Virtual Prompt Injection provides an unofficial implementation for backdooring instruction-tuned LLMs with virtual prompt injection, offering tools for data poisoning and evaluation specific to this technique.
Good fit when
- If needing to simulate or study backdoor attacks specifically targeting the behavior of trained language models under certain scenarios without modifying input directly at inference time.
- For researchers aiming to evaluate the robustness of LLMs against sophisticated, hidden prompts that can alter model outputs in specific contexts.
Avoid when
- Not applicable for general training or serving tasks if backdoor insertion is not within scope as it focuses solely on simulating attacks.
- In a production environment where tampering with AI models' integrity and security is strictly prohibited due to ethical considerations.
Observed Jul 12, 2026 · Source: enrich:decision_facts
Verify the decision
Maintenance and security
Full trust report- Maintenance
- Dormant (820d since push)
- As of Oct 5, 2026
- Provenance
- Not a fork · Personal account
- As of Oct 5, 2026
- Security (OSV)
- No lockfile
- As of Jul 11, 2026
Public GitHub metadata and optional OSV scans. Signals, not a guarantee. Trust methodology.
Install
pip install virtual-prompt-injectionSimilar tools
Same-category neighbours. No typed graph edges are catalogued for this tool yet.
Evidence and technical details
Sourced facts, taxonomy, compatibility claims, README excerpt, and machine-readable endpoints.
Overview
This repository contains an unofficial implementation of the paper 'Backdooring Instruction-Tuned Large Language Models with Virtual Prompt Injection', focusing on data poisoning and evaluation for virtual prompt injection, as well as Alpaca training and inference.
Capability facts
- Languages
- python
Source: github.language · Oct 5, 2026
Categories
Compatibility
Sourced claims from the README excerpt - not unsourced marketing copy.
Source: README excerpt (regex_v1, Oct 5, 2026)
You also need to set your OpenAI API Key in `./utils.py` (Line 13).Source link
Source: README excerpt (regex_v1, Oct 5, 2026)
conda create -n vpi python=3.10Source link
Tags
README
Virtual Prompt Injection Virtual Prompt Injection (VPI) is a backdoor attack for instruction tuned large language models (LLMs). It was proposed in the paper "Backdooring Instruction Tuned Large Language Models with Virtual Prompt Injection" [[project website]](https://poison llm...
For agents
This page has a .md twin and JSON over the API.