microsoft / TaskTrackerLinks
TaskTracker is an approach to detecting task drift in Large Language Models (LLMs) by analysing their internal activations. It provides a simple linear probe-based method and a more sophisticated metric learning method to achieve this. The project also releases the computationally expensive activation data to stimulate further AI safety research…
☆64Updated 3 weeks ago
Alternatives and similar repositories for TaskTracker
Users that are interested in TaskTracker are comparing it to the libraries listed below
Sorting:
- A benchmark for evaluating the robustness of LLMs and defenses to indirect prompt injection attacks.☆77Updated last year
- Repo for the research paper "SecAlign: Defending Against Prompt Injection with Preference Optimization"☆66Updated last month
- [NDSS'25 Best Technical Poster] A collection of automated evaluators for assessing jailbreak attempts.☆167Updated 5 months ago
- PAL: Proxy-Guided Black-Box Attack on Large Language Models☆54Updated last year
- A Dynamic Environment to Evaluate Attacks and Defenses for LLM Agents.☆244Updated 3 weeks ago
- [ICML 2024] COLD-Attack: Jailbreaking LLMs with Stealthiness and Controllability☆165Updated 8 months ago
- Jailbreaking Leading Safety-Aligned LLMs with Simple Adaptive Attacks [ICLR 2025]☆336Updated 7 months ago
- A repository of Language Model Vulnerabilities and Exposures (LVEs).☆113Updated last year
- ☆71Updated last year
- Improving Alignment and Robustness with Circuit Breakers☆228Updated 11 months ago
- The official implementation of our pre-print paper "Automatic and Universal Prompt Injection Attacks against Large Language Models".☆54Updated 10 months ago
- ☆34Updated 9 months ago
- Code to generate NeuralExecs (prompt injection for LLMs)☆22Updated 9 months ago
- Package to optimize Adversarial Attacks against (Large) Language Models with Varied Objectives☆70Updated last year
- TAP: An automated jailbreaking method for black-box LLMs☆182Updated 8 months ago
- [NeurIPS 2024] Official implementation for "AgentPoison: Red-teaming LLM Agents via Memory or Knowledge Base Backdoor Poisoning"☆141Updated 4 months ago
- ☆145Updated 2 months ago
- Open One-Stop Moderation Tools for Safety Risks, Jailbreaks, and Refusals of LLMs☆89Updated 9 months ago
- Code&Data for the paper "Watch Out for Your Agents! Investigating Backdoor Threats to LLM-Based Agents" [NeurIPS 2024]☆89Updated 11 months ago
- Fluent student-teacher redteaming☆22Updated last year
- ☆61Updated 8 months ago
- This repository provides a benchmark for prompt Injection attacks and defenses☆267Updated last month
- A fast + lightweight implementation of the GCG algorithm in PyTorch☆274Updated 3 months ago
- Bag of Tricks: Benchmarking of Jailbreak Attacks on LLMs. Empirical tricks for LLM Jailbreaking. (NeurIPS 2024)☆144Updated 9 months ago
- A Comprehensive Assessment of Trustworthiness in GPT Models☆302Updated 11 months ago
- Finding trojans in aligned LLMs. Official repository for the competition hosted at SaTML 2024.☆114Updated last year
- AutoDefense: Multi-Agent LLM Defense against Jailbreak Attacks☆53Updated 3 months ago
- Dataset for the Tensor Trust project☆45Updated last year
- ☆105Updated last year
- PromptInject is a framework that assembles prompts in a modular fashion to provide a quantitative analysis of the robustness of LLMs to a…☆410Updated last year