SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2406.18451 · NeurIPS · 2024

Detecting Brittle Decisions for Free: Leveraging Margin Consistency in Deep Robust Classifiers

Christian Gagné, Jonas Ngnawé, Frédéric Precioso, Sabyasachi Sahoo, Yann Pequignot

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/2406.18451.json · badge: /api/ran/2406.18451.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
ngnawejonas/margin-consistency canonical (classifier label)
classifier basis: classifier_2025
2 ran · 0 no recorded run
FunctionStatusWhere it lives
PM Ran
ran; our contract draft was wrong, not the code
ngnawejonas/margin-consistency/train/pkg/gairat.py
code served (permissive licence) · get_code("776602b4e9f0a9e6")
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 1, dtype family float, 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 torch

torch.manual_seed(0)

# The function PM expects:
# - logit: a tensor with shape (batch_size, num_classes) where num_classes=10
# - target: a tensor with shape (batch_size,) containing class indices (0-9)
# - gpu_id: device id (None for CPU)

batch_size = 2
num_classes = 10

# Create logit tensor (batch_size x num_classes)
logit = torch.randn(batch_size, num_classes)

# Create target tensor with valid class indices (0-9)
target = torch.tensor([3, 7])

# gpu_id should be None for CPU
gpu_id = None

ARGS = {
    "logit": logit,
    "target": target,
    "gpu_id": gpu_id
}
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 1, dtype family float, 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
weight_assign Ran
ran; our contract draft was wrong, not the code
ngnawejonas/margin-consistency/train/pkg/gairat.py
code served (permissive licence) · get_code("55d3ff95b7be99b2")
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 Tensor, ndim 1, dtype family float, 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 torch

torch.manual_seed(0)

# The function weight_assign calls PM which:
# - Takes logit (batch_size, 10) and target (batch_size,)
# - Uses torch.eye(10, device=gpu_id)
# - Calls logit.softmax(1) and logit.softmax(1).topk(2, largest=True)
# - Indexes with target and eye[target]
# So logit needs shape (batch_size, 10) and target needs to be class indices in [0, 9]

batch_size = 2
num_classes = 10

# Create logits: (batch_size, num_classes)
logit = torch.randn(batch_size, num_classes)

# Create target: class indices in range [0, 9]
target = torch.tensor([3, 7])

# bias and slope are floats
bias = -1.5
slope = 1.0

# gpu_id is None for CPU
gpu_id = None

ARGS = {
    "logit": logit,
    "target": target,
    "bias": bias,
    "slope": slope,
    "gpu_id": gpu_id,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 1, dtype family float, 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

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

Despite extensive research on adversarial training strategies to improve robustness, the decisions of even the most robust deep learning models can still be quite sensitive to imperceptible perturbations, creating serious risks when deploying them for high-stakes real-world applications. While detecting such cases may be critical, evaluating a model's vulnerability at a per-instance level using adversarial attacks is computationally too intensive and unsuitable for real-time deployment scenarios. The input space margin is the exact score to detect non-robust samples and is intractable for deep neural networks. This paper introduces the concept of margin consistency -a property that links the input space margins and the logit margins in robust models -for efficient detection of vulnerable samples. First, we establish that margin consistency is a necessary and sufficient condition to use a model's logit margin as a score for identifying non-robust samples. Next, through comprehensive empirical analysis of various robustly trained models on CIFAR10 and CIFAR100 datasets, we show that they indicate high margin consistency with a strong correlation between their input space margins and the logit margins. Then, we show that we can effectively and confidently use the logit margin to detect brittle decisions with such models. Finally, we address cases where the model is not sufficiently margin-consistent by learning a pseudo-margin from the feature representation. Our findings highlight the potential of leveraging deep representations to assess adversarial vulnerability in deployment scenarios efficiently.

For agents

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

get_harvested_code_for_paper("2406.18451")
get_code_for_paper("2406.18451")
have("2406.18451")

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

curl https://syntology.ai/api/ran/2406.18451.json

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

[![Syntology run record](https://syntology.ai/api/ran/2406.18451.svg)](https://syntology.ai/paper/2406.18451)

Connect an agent — have() is free.