Home/Evaluation & Observability/llm-self-defense

llm-self-defense

poloclub/llm-self-defense

LLM Self Defense: By Self Examination, LLMs know they are being tricked

GitHub

GraphCanon updated Oct 5, 2026 · GitHub synced Oct 5, 2026

28views this month

53 stars7 forksLast push May 21, 2024 Python BSD-3-Clause

Decision brief

Mitigates harmful content generation via self-examination by LLM outputs without fine-tuning.

Good fit when

When you need to reduce the success rate of adversarial attacks on text generation.
For environments requiring high safety standards from generated text, like payroll systems.

Avoid when

If real-time performance is critical and additional latency cannot be tolerated.
In scenarios where API access to both GPT 3.5 and Llama models is not feasible.

Observed Jul 12, 2026 · Source: enrich:decision_facts

llm-self-defense alternatives Trust report

Verify the decision

Maintenance and security

Full trust report

Maintenance: Dormant (866d since push); As of Oct 5, 2026
Provenance: Not a fork · Organization account; As of Oct 5, 2026
Security (OSV): 157 low (157 low); As of Jul 11, 2026

Public GitHub metadata and optional OSV scans. Signals, not a guarantee. Trust methodology.

Install

pip install llm-self-defense

PyPI

Similar tools

Same-category neighbours. No typed graph edges are catalogued for this tool yet.

llm-attacks

Universal and Transferable Attacks on Aligned Language Models

4.8k★

PurpleLlama

Set of tools to assess and improve LLM security

4.4k★

awesome-llm-security

A curation of tools, documents and projects about LLM Security

1.7k★

Evidence and technical details

Sourced facts, taxonomy, compatibility claims, README excerpt, and machine-readable endpoints.

Overview

Introduces a method to mitigate harmful content generation from LLMs by analyzing the outputs with another instance of an LLM without requiring fine-tuning or iterative output generation.

Capability facts

Languages: python
Source: github.language · Oct 5, 2026

Compatibility

Sourced claims from the README excerpt - not unsourced marketing copy.

Python runtimePython

Source: README excerpt (regex_v1, Oct 5, 2026)

```conda create --name llmdefense python=3.9 --file requirements.txt```

Source link

README

LLM Self Defense LLM Self Defense: By Self Examination, LLMs know they are being tricked. Mansi Phute, Alec Heibling, Matthew Hull, ShengYun Peng, Sebastian Szyller, Cory Cornelius, Duen Horng Chau. In ICLR 2024 TinyPaper , 2024. 📄 <a href="https://arxiv.org/abs/2308.07308" Rese...

For agents

This page has a .md twin and JSON over the API.

/tools/poloclub-llm-self-defense.md GET /api/graphcanon/tools/poloclub-llm-self-defense GET /api/graphcanon/graph?tool=poloclub-llm-self-defense