SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2502.15427 · NeurIPS · 2025

Adversarial Prompt Evaluation: Systematic Benchmarking of Guardrails Against Prompt Input Attacks on LLMs

Pin-Yu Chen, Prasanna Sattigeri, Giulio Zizzo, Kush Varshney, Giandomenico Cornacchia, Kieran Fraser, Muhammad Zaid, Hameed Ambrish, Rawat Beat Buesser, Mark Purcell

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 1 has no contract verdict; 1 failure 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/2502.15427.json · badge: /api/ran/2502.15427.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
IBM/Adversarial-Prompt-Evaluation canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
FunctionStatusWhere it lives
get_plot Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
IBM/Adversarial-Prompt-Evaluation/scripts/main_evaluate.py
code served (permissive licence) · get_code("ff7ed3a671bd2838")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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 numpy as np

# Create a small binary classification example with 10 samples
np.random.seed(42)

# Ground truth labels (0 or 1)
true_label = [0, 0, 0, 0, 0, 1, 1, 1, 1, 1]

# Predictions (0 or 1)
prediction = [0, 0, 1, 0, 0, 1, 1, 0, 1, 1]

# Prediction probabilities (continuous values between 0 and 1)
pred_proba = [0.1, 0.2, 0.6, 0.3, 0.15, 0.8, 0.9, 0.4, 0.75, 0.85]

# Whether to plot (set to False to avoid display issues)
plot = False

ARGS = {
    'prediction': prediction,
    'pred_proba': pred_proba,
    'true_label': true_label,
    'plot': plot
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy, sklearn — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
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
handle_prediction Ran
ran on a synthesized input
IBM/Adversarial-Prompt-Evaluation/scripts/main_evaluate.py
code served (permissive licence) · get_code("4d0b96e1cdd1c174")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Sonnet 4.5, prompt python_fixture_v1) and returned dict. The image it ran in was not recorded. The code mentions cuda, and the harness aliases cuda to the CPU for every sample; whether this run needed that is 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

torch.manual_seed(0)

class MockModel:
    def predict(self, samples, threshold=None):
        results = []
        for sample in samples:
            # Alternate between safe and unsafe predictions
            if len(sample) % 2 == 0:
                results.append({
                    'label': 'safe',
                    'confidence': 0.8
                })
            else:
                results.append({
                    'label': 'unsafe',
                    'confidence': 0.9
                })
        return results

data_name = "test_data"
model_name = "mock_model"
model = MockModel()
x_test = ["This is a test prompt", "Another test", "Third sample"]
y_test = [0, 1, 0]
source_test = ["dataset1", "dataset2", "dataset1"]
threshold = None

ARGS = {
    'data_name': data_name,
    'model_name': model_name,
    'model': model,
    'x_test': x_test,
    'y_test': y_test,
    'source_test': source_test,
    'threshold': threshold
}
Setup block drafted by
a language model, Claude Sonnet 4.5, prompt python_fixture_v1
Imports
numpy, torch, tqdm — 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

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

As large language models (LLMs) become integrated into everyday applications, ensuring their robustness and security is increasingly critical. In particular, LLMs can be manipulated into unsafe behaviour by prompts known as jailbreaks. The variety of jailbreak styles is growing, necessitating the use of external defences known as guardrails. While many jailbreak defences have been proposed, not all defences are able to handle new out-of-distribution attacks due to the narrow segment of jailbreaks used to align them. Moreover, the lack of systematisation around defences has created significant gaps in their practical application. In this work, we perform systematic benchmarking across 15 different defences, considering a broad swathe of malicious and benign datasets. We find that there is significant performance variation depending on the style of jailbreak a defence is subject to. Additionally, we show that based on current datasets available for evaluation, simple baselines can display competitive out-of-distribution performance compared to many state-of-the-art defences. Code is available at https://github.com/IBM/Adversarial-Prompt-Evaluation.

For agents

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

get_harvested_code_for_paper("2502.15427")
get_code_for_paper("2502.15427")
have("2502.15427")

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

curl https://syntology.ai/api/ran/2502.15427.json

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

[![Syntology run record](https://syntology.ai/api/ran/2502.15427.svg)](https://syntology.ai/paper/2502.15427)

Connect an agent — have() is free.