SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2402.04249 · 2024

HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusal

Bo Li, Dan Hendrycks, Steven Basart, Andy Zou, Mantas Mazeika, Zifan Wang, David Forsyth, Long Phan, Nathaniel Li, Xuwang Yin, Norman Mu, Elham Sakhaee, and 2 more

arXiv · PDF · Open in the Atlas

Code that ran

2 of 2 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 2 failures of our instrument; 0 no recorded run), measured 2026-11-22; repository links from the Papers with Code archive attributed under CC BY-SA, no affiliation; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2402.04249.json · badge: /api/ran/2402.04249.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
centerforaisafety/harmbench canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
FunctionStatusWhere it lives
get_expanded_pipeline_configs Ran
ran; our contract draft was wrong, not the code
centerforaisafety/harmbench/scripts/run_pipeline.py
code served (permissive licence) · get_code("899b390c029421da")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# Build minimal valid arguments for get_expanded_pipeline_configs

# pipeline_config: dict mapping proper method names to their configs
pipeline_config = {
    'method_gcg': {
        'experiment_name_template': 'gcg_<model_name>_exp',
        'allowed_target_model_types': ['llm', 'transformer'],
        'behavior_chunk_size': 10
    },
    'method_transfer': {
        'experiment_name_template': 'transfer_ensemble_exp',
        'allowed_target_model_types': ['llm'],
        'behavior_chunk_size': 5
    }
}

# model_configs: dict mapping model names to their configs
model_configs = {
    'gpt2': {
        'model_type': 'transformer',
        'num_gpus': 1
    },
    'llama7b': {
        'model_type': 'llm',
        'num_gpus': 2
    },
    'bert': {
        'model_type': 'transformer',
        'num_gpus': 0
    },
    'mistral': {
        'model_type': 'llm',
        'num_gpus': 1
    }
}

# proper_method_names: list of method names to expand
proper_method_names = ['method_gcg', 'method_transfer']

# models: list of model names
models = ['gpt2', 'llama7b', 'bert', 'mistral']

ARGS = {
    'pipeline_config': pipeline_config,
    'model_configs': model_configs,
    'proper_method_names': proper_method_names,
    'models': models
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
run_subprocess_slurm Ran
ran; our contract draft was wrong, not the code
centerforaisafety/harmbench/scripts/run_pipeline.py
code served (permissive licence) · get_code("11ac5d8b148a7915")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import subprocess
import re

def run_subprocess_slurm(command):
    result = subprocess.run(command, shell=True, capture_output=True, text=True)
    print('command:', command)
    print('stdout:', result.stdout)
    print('stderr:', result.stderr)
    match = re.search('Submitted batch job (\\d+)', result.stdout)
    if match:
        job_id = match.group(1)
    else:
        job_id = None
    return job_id


# The function takes a command string and runs it via subprocess
# It expects the stdout to potentially contain a SLURM job submission message
# We'll create a simple command that produces output matching the expected pattern

ARGS = {
    'command': 'echo "Submitted batch job 12345"'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block

Also present in other papers' code

Of 2 samples lifted from this paper's linked repositories, 2 carry a function identity and 0 do not. Of the 2 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 2 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Automated red teaming holds substantial promise for uncovering and mitigating the risks associated with the malicious use of large language models (LLMs), yet the field lacks a standardized evaluation framework to rigorously assess new methods. To address this issue, we introduce HarmBench, a standardized evaluation framework for automated red teaming. We identify several desirable properties previously unaccounted for in red teaming evaluations and systematically design HarmBench to meet these criteria. Using HarmBench, we conduct a largescale comparison of 18 red teaming methods and 33 target LLMs and defenses, yielding novel insights. We also introduce a highly efficient adversarial training method that greatly enhances LLM robustness across a wide range of attacks, demonstrating how HarmBench enables codevelopment of attacks and defenses. We open source HarmBench at https://github. com/centerforaisafety/HarmBench.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2402.04249")
get_code_for_paper("2402.04249")
have("2402.04249")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/2402.04249.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/2402.04249.svg)](https://syntology.ai/paper/2402.04249)

Connect an agent — have() is free.