SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2407.15847 · 2024

LLMmap: Fingerprinting For Large Language Models

arXiv · PDF · Open in the Atlas

Code that ran

9 of 16 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; 2 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 7 failures of our instrument; 7 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/2407.15847.json · badge: /api/ran/2407.15847.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
pasquini-dario/LLMmap canonical (classifier label)
classifier basis: census
9 ran · 7 no recorded run
FunctionStatusWhere it lives
build_templates Ran
ran, honoured the contract we drafted
pasquini-dario/LLMmap/LLMmap/templates.py
code served (permissive licence) · get_code("4d3e058045786b53")
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, through the batch API) and returned ndarray, ndim 2, dtype family float, shape (3, 4) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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

# Construct valid arguments for build_templates
# The function takes labels (1D array of class IDs) and feats (2D array of feature vectors)
# and returns mean feature vectors for each class

# Create sample labels: 5 samples with class IDs 0, 1, 2
labels = np.array([0, 0, 1, 1, 2], dtype=np.int32)

# Create sample features: 5 samples, each with 4 features
feats = np.array([
    [1.0, 2.0, 3.0, 4.0],
    [1.5, 2.5, 3.5, 4.5],
    [5.0, 6.0, 7.0, 8.0],
    [5.5, 6.5, 7.5, 8.5],
    [9.0, 10.0, 11.0, 12.0],
], dtype=np.float32)

ARGS = {
    "labels": labels,
    "feats": feats,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 2, dtype family float, shape (3, 4) — shape (3, 4) as a later run recorded it
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
euclidean_distance Ran
ran; our contract draft was wrong, not the code
pasquini-dario/LLMmap/LLMmap/inference_model_archs.py
code served (permissive licence) · get_code("47203ac597d1cdaa")
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, through the batch API) and returned Tensor, ndim 2, dtype family float, shape (2, 1) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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

torch.manual_seed(0)

# Create two small feature tensors: (B=2, F=4)
a = torch.randn(2, 4)
b = torch.randn(2, 4)
eps = 1e-08

ARGS = {
    "a": a,
    "b": b,
    "eps": eps,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Tensor, ndim 2, dtype family float, shape (2, 1) — shape (2, 1) as a later run recorded it
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
get_activation Ran
ran; our contract draft was wrong, not the code
pasquini-dario/LLMmap/LLMmap/inference_model_archs.py
code served (permissive licence) · get_code("82f8654112f827f6")
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, through the batch API) and returned GELU. 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 torch.nn as nn

torch.manual_seed(0)

ARGS = {
    'name': 'gelu'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type GELU — a GELU output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a GELU 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
predict_by_templates Ran
ran, honoured the contract we drafted
pasquini-dario/LLMmap/LLMmap/templates.py
code served (permissive licence) · get_code("63abcdbc162b0dc0")
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, through the batch API) and returned ndarray, ndim 1, dtype family int, shape (2,) as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape 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

# Set random seed for reproducibility
np.random.seed(0)

# Create small feature vectors: N=2 samples, feature_dim=4
feats = np.random.randn(2, 4).astype(np.float32)

# Create template vectors: num_classes=3, feature_dim=4
templates = np.random.randn(3, 4).astype(np.float32)

ARGS = {
    "feats": feats,
    "templates": templates,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (2,) — shape (2,) as a later run recorded it
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
read_conf_file Ran
ran; our contract draft was wrong, not the code
pasquini-dario/LLMmap/LLMmap/utility.py
code served (permissive licence) · get_code("145223e448267f7f")
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, through the batch API) 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 json
import tempfile
import os

# Create a temporary JSON configuration file
temp_dir = tempfile.mkdtemp()
config_file_path = os.path.join(temp_dir, "test_config.json")

# Write a sample configuration to the file
sample_config = {
    "model_name": "test_model",
    "learning_rate": 0.001,
    "batch_size": 32,
    "epochs": 10,
    "layers": [128, 64, 32],
    "activation": "relu"
}

with open(config_file_path, 'w') as f:
    json.dump(sample_config, f)

# Define the ARGS dictionary
ARGS = {
    "file_path": config_file_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 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
read_dataset Ran
ran; our contract draft was wrong, not the code
pasquini-dario/LLMmap/LLMmap/input_pipeline.py
code served (permissive licence) · get_code("6e7bb2069a548786")
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, through the batch API) and returned tuple, ndim 2, 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 json
import random
import tempfile
import os

# Create a temporary file with sample dataset entries
temp_file = tempfile.NamedTemporaryFile(mode='w', delete=False, suffix='.jsonl', encoding='utf-8')
try:
    # Write sample training entries
    temp_file.write(json.dumps({'dataset': 'train', 'text': 'hello world', 'label': 0}) + '\n')
    temp_file.write(json.dumps({'dataset': 'train', 'text': 'foo bar', 'label': 1}) + '\n')
    
    # Write sample test entries
    temp_file.write(json.dumps({'dataset': 'test', 'text': 'baz qux', 'label': 0}) + '\n')
    temp_file.write(json.dumps({'dataset': 'test', 'text': 'alpha beta', 'label': 1}) + '\n')
    
    temp_file.flush()
    temp_path = temp_file.name
finally:
    temp_file.close()

ARGS = {
    'path': temp_path,
    'encoding': 'utf-8',
    'shuffle': True,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 tuple, ndim 2, dtype family float — a tuple 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
read_pickle Ran
ran; our contract draft was wrong, not the code
pasquini-dario/LLMmap/LLMmap/utility.py
code served (permissive licence) · get_code("c7497cceeb28ffef")
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, through the batch API) 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 pickle
import tempfile
import os

# Create a temporary directory and file for testing
temp_dir = tempfile.mkdtemp()
pickle_path = os.path.join(temp_dir, "test_data.pkl")

# Create some sample data and pickle it
sample_data = {
    "key1": "value1",
    "key2": [1, 2, 3],
    "key3": {"nested": "dict"}
}

with open(pickle_path, 'wb') as f:
    pickle.dump(sample_data, f)

# Define ARGS with the path to the pickled file
ARGS = {
    "path": pickle_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 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
read_templates Ran
ran; our contract draft was wrong, not the code
pasquini-dario/LLMmap/LLMmap/inference.py
code served (permissive licence) · get_code("adc09759f8d65697")
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, through the batch API) 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 json
import tempfile
import os
import numpy as np

# Create a temporary JSON file with template data
temp_dir = tempfile.mkdtemp()
templates_path = os.path.join(temp_dir, "templates.json")

# Create sample template data
template_data = {
    "template_1": [1.0, 2.0, 3.0, 4.0],
    "template_2": [5.0, 6.0, 7.0, 8.0],
    "template_3": [9.0, 10.0, 11.0, 12.0]
}

# Write the JSON file
with open(templates_path, 'w') as f:
    json.dump(template_data, f)

ARGS = {
    "templates_path": templates_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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
sample_from_multi_universe Ran
ran; our contract draft was wrong, not the code
pasquini-dario/LLMmap/LLMmap/prompt_configuration.py
code served (permissive licence) · get_code("b448e7d58494a339")
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, through the batch API) 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 random

def sample_from_multi_universe(universe):
    sample = {}
    for k, u in universe.items():
        sample[k] = random.sample(u, 1)[0]
    return sample

# The function iterates over universe.items(), so universe must be a dict
# For each value u, it calls random.sample(u, 1), so each u must be a sequence
# The function returns a dict where each key maps to one randomly sampled element

ARGS = {
    "universe": {
        "world_1": [1, 2, 3, 4, 5],
        "world_2": ["a", "b", "c"],
        "world_3": [10.5, 20.5, 30.5, 40.5],
        "world_4": [True, False],
    }
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
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 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
infer_features Not yet run
unverified (harvested, no recorded run)
pasquini-dario/LLMmap/LLMmap/templates.py
code served (permissive licence) · get_code("a5c7254402cb328d")
How it ran

no invocation recorded.

load_LLMmap Not yet run
unverified (harvested, no recorded run)
pasquini-dario/LLMmap/LLMmap/inference.py
code served (permissive licence) · get_code("a0c4036c783edf32")
How it ran

no invocation recorded.

load_datasets Not yet run
unverified (harvested, no recorded run)
pasquini-dario/LLMmap/LLMmap/dataset.py
code served (permissive licence) · get_code("fe509b80ffd62e01")
How it ran

no invocation recorded.

load_model Not yet run
unverified (harvested, no recorded run)
pasquini-dario/LLMmap/LLMmap/embedding_model.py
code served (permissive licence) · get_code("9ee90e0c985481bb")
How it ran

no invocation recorded.

make_norm Not yet run
unverified (harvested, no recorded run)
pasquini-dario/LLMmap/LLMmap/inference_model_archs.py
code served (permissive licence) · get_code("9cb6cf3f23bddfa5")
How it ran

no invocation recorded.

read_dataset Not yet run
unverified (harvested, no recorded run)
pasquini-dario/LLMmap/LLMmap/dataset_maker.py
code served (permissive licence) · get_code("5c38d3eecf731383")
How it ran

no invocation recorded.

train_model Not yet run
unverified (harvested, no recorded run)
pasquini-dario/LLMmap/LLMmap/trainer.py
code served (permissive licence) · get_code("3ce9e1b27efcd170")
How it ran

no invocation recorded.

Also present in other papers' code

Of 16 samples lifted from this paper's linked repositories, 16 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 15 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.

FunctionAlso present inEarliest-datedOthers named
read_pickle 1 other paper
1 dated earlier
2305.18334 (2023-05)
PQA: Exploring the Potential of Product Quantization in DNN Hardware Acceleration

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

We introduce LLMmap, a first-generation fingerprinting technique targeted at LLM-integrated applications. LLMmap employs an active fingerprinting approach, sending carefully crafted queries to the application and analyzing the responses to identify the specific LLM version in use. Our query selection is informed by domain expertise on how LLMs generate uniquely identifiable responses to thematically varied prompts. With as few as 8 interactions, LLMmap can accurately identify 42 different LLM versions with over 95% accuracy. More importantly, LLMmap is designed to be robust across different application layers, allowing it to identify LLM versions--whether open-source or proprietary--from various vendors, operating under various unknown system prompts, stochastic sampling hyperparameters, and even complex generation frameworks such as RAG or Chain-of-Thought. We discuss potential mitigations and demonstrate that, against resourceful adversaries, effective countermeasures may be challenging or even unrealizable.

For agents

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

get_harvested_code_for_paper("2407.15847")
get_code_for_paper("2407.15847")
have("2407.15847")

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

curl https://syntology.ai/api/ran/2407.15847.json

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

[![Syntology run record](https://syntology.ai/api/ran/2407.15847.svg)](https://syntology.ai/paper/2407.15847)

Connect an agent — have() is free.