SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 1907.09701 · 2019

Benchmarking Attribution Methods with Relative Feature Importance

arXiv · PDF · Open in the Atlas

Code that ran

3 of 6 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 2 have no contract verdict; 0 failures of our instrument; 3 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/1907.09701.json · badge: /api/ran/1907.09701.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
google-research-datasets/bim canonical (classifier label)
classifier basis: census
2 ran · 3 no recorded run
google-research-datasets/bam canonical (classifier label)
classifier basis: classifier_2025
1 ran · 0 no recorded run
FunctionStatusWhere it lives
get_global_indices Ran
ran, honoured the contract we drafted
google-research-datasets/bam/bam/metrics.py
code served (permissive licence) · get_code("2965857b8000dac0")
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 range. 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
from absl import flags
from unittest.mock import MagicMock

# Initialize flags
FLAGS = flags.FLAGS

# Define the flags that the function uses
flags.DEFINE_boolean('scratch', False, 'Whether to generate from scratch')
flags.DEFINE_integer('seed', 42, 'Random seed')
flags.DEFINE_integer('num_imgs', 50, 'Number of images to sample')

# Mark flags as parsed to avoid errors
FLAGS.mark_as_parsed()

# Define ARGS dict with the metric parameter
ARGS = {
    'metric': 'MCS'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
Imports
absl — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type range — a range output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a range output the recorded ndim 1 and dtype family int 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
single_attr Ran
ran on a synthesized input
google-research-datasets/bim/bam/attributions.py
code served (permissive licence) · get_code("4afd748b2a52289d")
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 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 numpy as np
import PIL.Image

# Create a small 2D saliency map with batch dimension
# Shape: (batch=2, height=5, width=4)
map_2d = np.random.rand(2, 5, 4).astype(np.float32)

# Create location tuple [row_min, col_min, row_max, col_max]
# This defines a bounding box: row from 1 to 4, col from 1 to 3
loc = [1, 1, 4, 3]

# Create a binary object mask
# Size should match the original object region before resizing
# The function will resize it to (col_max - col_min, row_max - row_min) = (2, 3)
obj_mask = np.array([[1, 0, 1],
                     [1, 1, 0],
                     [0, 1, 1]], dtype=np.uint8)

ARGS = {
    "map_2d": map_2d,
    "loc": loc,
    "obj_mask": obj_mask,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, 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 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
visualize_pos_attr Ran
ran on a synthesized input
google-research-datasets/bim/bam/attributions.py
code served (permissive licence) · get_code("96c5223d1f404db5")
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

# Construct a 3D numpy array with shape (height, width, channels)
# Using distinct dimensions: 4 for height, 3 for width, 2 for channels
image_3d = np.random.randn(4, 3, 2)

# Add some positive and negative values to make clipping meaningful
image_3d[0, 0, 0] = 5.0
image_3d[1, 1, 1] = -3.0
image_3d[2, 2, 0] = 10.0

ARGS = {
    "image_3d": image_3d,
    "percentile": 99,
}
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
corr_indices Not yet run
unverified (harvested, no recorded run)
google-research-datasets/bim/bam/metrics.py
code served (permissive licence) · get_code("8db06832e6dbf0b6")
How it ran

no invocation recorded.

load_imgs Not yet run
unverified (harvested, no recorded run)
google-research-datasets/bim/bam/attributions.py
code served (permissive licence) · get_code("a539b0fef0ed651a")
How it ran

no invocation recorded.

load_pos_neg_attr Not yet run
unverified (harvested, no recorded run)
google-research-datasets/bim/bam/metrics.py
code served (permissive licence) · get_code("0ac8c7eece2eb98d")
How it ran

no invocation recorded.

Also present in other papers' code

Of 6 samples lifted from this paper's linked repositories, 6 carry a function identity and 0 do not. Of the 6 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 6 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

Interpretability is an important area of research for safe deployment of machine learning systems. One particular type of interpretability method attributes model decisions to input features. Despite active development, quantitative evaluation of feature attribution methods remains difficult due to the lack of ground truth: we do not know which input features are in fact important to a model. In this work, we propose a framework for Benchmarking Attribution Methods (BAM) with a priori knowledge of relative feature importance. BAM includes 1) a carefully crafted dataset and models trained with known relative feature importance and 2) three complementary metrics to quantitatively evaluate attribution methods by comparing feature attributions between pairs of models and pairs of inputs. Our evaluation on several widely-used attribution methods suggests that certain methods are more likely to produce false positive explanations---features that are incorrectly attributed as more important to model prediction. We open source our dataset, models, and metrics.

For agents

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

get_harvested_code_for_paper("1907.09701")
get_code_for_paper("1907.09701")
have("1907.09701")

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

curl https://syntology.ai/api/ran/1907.09701.json

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

[![Syntology run record](https://syntology.ai/api/ran/1907.09701.svg)](https://syntology.ai/paper/1907.09701)

Connect an agent — have() is free.