- Home
- LLM Evaluation Tools
- HarmBench
On the recordRead 8 Oct 202618 facts
HarmBench
On the record
- Classifiers
- The project provides three classifier models for standard, contextual, and multimodal behaviors, including a validation classifier.github.com · read 8 Oct 2026
- Compute requirements
- The documentation says evaluation requires enough GPUs for the target model, attack, and classifier, and its default GPU settings were chosen for 80GB A100s.github.com · read 8 Oct 2026
- Customization
- Users can add Hugging Face Transformers models through the model configuration and add red-teaming methods in the baselines directory.github.com · read 8 Oct 2026
- Custom methods
- Users can add red teaming methods by creating a subfolder in the baselines directory and implementing the RedTeamingMethod class.github.com · read 8 Oct 2026
- Custom models
- Users can add Hugging Face Transformers models through the model configuration file, though AutoDAN, PAIR, and TAP require manual experiment configuration for new models.github.com · read 8 Oct 2026
- Evaluation uses
- The framework supports evaluating red teaming methods against LLMs and evaluating LLMs against red teaming methods.github.com · read 8 Oct 2026
- Evaluation workflow
- Its pipeline generates red-team test cases, prompts target models to generate completions, and classifies those completions to calculate attack success rate.github.com · read 8 Oct 2026
- Execution
- The pipeline can run sequentially on the current machine, in parallel across GPUs on one machine with Ray, or with SLURM across machines.github.com · read 8 Oct 2026
- Installation
- The README instructs users to clone the repository, install requirements with pip, and download the en_core_web_sm spaCy model.github.com · read 8 Oct 2026
- License
- The public GitHub repository lists an MIT license.github.com · read 8 Oct 2026
- Model support
- HarmBench supports Transformers-compatible LLMs, numerous closed-source APIs, and several multimodal models out of the box.github.com · read 8 Oct 2026
- Pipeline
- Its evaluation pipeline generates test cases, generates model completions, and evaluates completions, with an optional test-case merging step.github.com · read 8 Oct 2026
- Pipeline execution
- The pipeline can run locally, in parallel across GPUs on one machine with Ray, or through SLURM jobs.github.com · read 8 Oct 2026
- Purpose
- HarmBench is an open-source framework for evaluating automated red teaming methods and LLM attacks and defenses.github.com · read 8 Oct 2026
- Security and compliance
- The repository README and linked HarmBench website page provide no security or compliance claims; the website page opened here only says JavaScript must be enabled.harmbench.org · read 8 Oct 2026
- Support
- The repository documentation invites users to submit a pull request for support for another cluster manager.github.com · read 8 Oct 2026
- Usage caveat
- The documentation says its text and multimodal classifier models are test classifiers intended only for evaluation.github.com · read 8 Oct 2026
- Who it serves
- The project describes uses for evaluating red-teaming methods against LLMs and evaluating LLMs against red-teaming methods.github.com · read 8 Oct 2026
