SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2601.08467 · 2026

Zero-Shot Distracted Driver Detection via Vision Language Models with Double Decoupling

Takamichi Miyata, Sumiko Miyata, Andrew Morris

arXiv · PDF · Open in the Atlas

Code that ran

8 of 12 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; 1 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 7 failures of our instrument; 4 no recorded run), measured 2026-11-22; 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/2601.08467.json · badge: /api/ran/2601.08467.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
mtakamichi/ZVL-DDD canonical (classifier label)
classifier basis: census
8 ran · 4 no recorded run
FunctionStatusWhere it lives
apply_dad Ran
ran; our contract draft was wrong, not the code
mtakamichi/ZVL-DDD/src/method.py
code served (permissive licence) · get_code("85ab83623d17f6b8")
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 (5, 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

np.random.seed(0)

# Create image embeddings: [N, D] where N=5 images, D=4 embedding dimensions
image_embeddings = np.random.randn(5, 4).astype(np.float32)

# Create subject IDs: 5 images belonging to 3 different drivers
# Use distinct values to make testing easier
subjects = [0, 1, 0, 1, 2]

ARGS = {
    "image_embeddings": image_embeddings,
    "subjects": subjects,
}
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 (5, 4) — shape (5, 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
apply_teo Ran
ran, honoured the contract we drafted
mtakamichi/ZVL-DDD/src/method.py
code served (permissive licence) · get_code("2c7f2f6679de6413")
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 (4, 3) 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

# Create a simple text embeddings matrix
# Shape: [C, D] where C = number of classes, D = embedding dimension
# Using 4 classes and 3-dimensional embeddings
np.random.seed(0)
text_embeddings = np.random.randn(4, 3).astype(np.float32)

ARGS = {
    "text_embeddings": text_embeddings,
}
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 (4, 3) — shape (4, 3) 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
compute_metrics Ran
ran; our contract draft was wrong, not the code
mtakamichi/ZVL-DDD/src/metrics.py
code served (permissive licence) · get_code("4277c568ca217e4e")
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 numpy as np

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

# Create sample data
num_classes = 10
num_samples = 20

# true_labels: class indices from 0 to num_classes-1
true_labels = np.random.randint(0, num_classes, size=num_samples)

# pred_labels: predicted class indices
pred_labels = np.random.randint(0, num_classes, size=num_samples)

# logits: raw scores for each class (num_samples x num_classes)
logits = np.random.randn(num_samples, num_classes).astype(np.float32)

# num_classes: number of output classes
num_classes_val = num_classes

# class_names: optional names for each class
class_names = [f"class_{i}" for i in range(num_classes)]

ARGS = {
    'true_labels': true_labels,
    'pred_labels': pred_labels,
    'logits': logits,
    'num_classes': num_classes_val,
    'class_names': class_names
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — 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
ensure_dir Ran
ran; our contract draft was wrong, not the code
mtakamichi/ZVL-DDD/src/utils.py
code served (permissive licence) · get_code("fc55a1baf2a1587a")
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 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 os
import tempfile

# Create a temporary directory path for testing
temp_dir = tempfile.gettempdir()
test_path = os.path.join(temp_dir, "test_ensure_dir", "nested", "path")

ARGS = {
    "path": test_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 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
format_summary Ran
ran; our contract draft was wrong, not the code
mtakamichi/ZVL-DDD/src/metrics.py
code served (permissive licence) · get_code("43c6e17e238ca615")
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 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 torch

torch.manual_seed(0)

# The function expects a dict with specific metric keys
# All values should be floats in the range [0, 1] since they're multiplied by 100 for percentages

metrics = {
    'top1_accuracy': 0.87,
    'top3_accuracy': 0.95,
    'macro_recall': 0.82,
    'macro_precision': 0.79,
    'binary_auprc': 0.91,
    'binary_fnr': 0.08
}

ARGS = {
    'metrics': metrics
}
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 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
get_device Ran
ran; our contract draft was wrong, not the code
mtakamichi/ZVL-DDD/src/utils.py
code served (permissive licence) · get_code("6caa40890c7343cc")
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 str. 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

torch.manual_seed(0)

# The function get_device takes one parameter: requested
# It should be a string with value 'auto', 'cuda', or 'cpu'
# We'll test with 'auto' as the default

ARGS = {
    'requested': 'auto'
}
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 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
get_display_names Ran
ran; our contract draft was wrong, not the code
mtakamichi/ZVL-DDD/src/prompts.py
code served (permissive licence) · get_code("4c3d2b4c92822d56")
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 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 torch
import numpy as np

# Looking at the function `get_display_names`, it takes a single parameter:
# - dataset: str

# The function normalizes the dataset string using _norm() and looks it up in _REGISTRY
# to return the display_names list. We just need to pass a valid dataset string.

ARGS = {
    'dataset': 'sam-dd'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — 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
get_prompts Ran
ran; our contract draft was wrong, not the code
mtakamichi/ZVL-DDD/src/prompts.py
code served (permissive licence) · get_code("ba3eecb3855cd5b6")
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 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 torch
import numpy as np

# Define arguments for get_prompts function
ARGS = {
    'dataset': 'sam-dd',
    'enable_pe': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — 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
load_clip Not yet run
unverified (harvested, no recorded run)
mtakamichi/ZVL-DDD/src/utils.py
code served (permissive licence) · get_code("207173e78852a2a1")
How it ran

no invocation recorded.

load_dataset Not yet run
unverified (harvested, no recorded run)
mtakamichi/ZVL-DDD/src/datasets.py
code served (permissive licence) · get_code("8869a46fdc28aec6")
How it ran

no invocation recorded.

load_sam_dd Not yet run
unverified (harvested, no recorded run)
mtakamichi/ZVL-DDD/src/datasets.py
code served (permissive licence) · get_code("6e0fb8560e5f6952")
How it ran

no invocation recorded.

load_statefarm Not yet run
unverified (harvested, no recorded run)
mtakamichi/ZVL-DDD/src/datasets.py
code served (permissive licence) · get_code("1a6df3a369011a56")
How it ran

no invocation recorded.

Also present in other papers' code

Of 12 samples lifted from this paper's linked repositories, 12 carry a function identity and 0 do not. Of the 12 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 12 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

Distracted driving is a major cause of traffic collisions, calling for robust and scalable detection methods. Visionlanguage models (VLMs) enable strong zero-shot image classification, but existing VLM-based distracted driver detectors often underperform in real-world conditions. We identify subjectspecific appearance variations (e.g., clothing, age, and gender) as a key bottleneck: VLMs entangle these factors with behavior cues, leading to decisions driven by who the driver is rather than what the driver is doing. To address this, we propose a subject decoupling framework that extracts a driver appearance embedding and removes its influence from the image embedding prior to zero-shot classification, thereby emphasizing distractionrelevant evidence. We further orthogonalize text embeddings via metric projection onto Stiefel manifold to improve separability while staying close to the original semantics. Experiments demonstrate consistent gains over prior baselines, indicating the promise of our approach for practical road-safety applications. Code is available at https://github.com/mtakamichi/ZVL-DDD.

For agents

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

get_harvested_code_for_paper("2601.08467")
get_code_for_paper("2601.08467")
have("2601.08467")

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

curl https://syntology.ai/api/ran/2601.08467.json

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

[![Syntology run record](https://syntology.ai/api/ran/2601.08467.svg)](https://syntology.ai/paper/2601.08467)

Connect an agent — have() is free.