SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2106.07411 · NeurIPS · 2021

Partial success in closing the gap between human and machine vision

Matthias Bethge, Robert Geirhos, Felix Wichmann, Wieland Brendel, Kantharaju Narayanappa, Benjamin Mitzkus, Tizian Thieringer

arXiv · PDF · Open in the Atlas

Code that ran

2 of 3 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 2 have no contract verdict; 0 failures of our instrument; 1 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/2106.07411.json · badge: /api/ran/2106.07411.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
kantharajucn/CLIP-imagenet-evaluation canonical (classifier label)
classifier basis: census
2 ran · 0 no recorded run
bethgelab/model-vs-human canonical (classifier label)
classifier basis: classifier_2025
0 ran · 1 no recorded run
FunctionStatusWhere it lives
data_loader Ran
ran on a synthesized input
kantharajucn/CLIP-imagenet-evaluation/data_loader.py
code served (permissive licence) · get_code("68a72b229739b428")
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 DataLoader. 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
from torchvision import transforms
import tempfile
import os
from PIL import Image

torch.manual_seed(0)

# Create a temporary directory with mock ImageNet-like structure
temp_dir = tempfile.mkdtemp()
category_dir = os.path.join(temp_dir, "cat1")
os.makedirs(category_dir, exist_ok=True)

# Create a small dummy image
dummy_img = Image.new('RGB', (32, 32), color='red')
dummy_img.save(os.path.join(category_dir, "image1.jpg"))

dummy_img2 = Image.new('RGB', (32, 32), color='blue')
dummy_img2.save(os.path.join(category_dir, "image2.jpg"))

# Create a simple args object
class Args:
    def __init__(self):
        self.data_dir = temp_dir
        self.batch_size = 2
        self.num_workers = 0

args = Args()

# Create transform
transform = transforms.Compose([
    transforms.Resize((32, 32)),
    transforms.ToTensor(),
])

ARGS = {
    "transform": transform,
    "args": args,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
PIL, torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type DataLoader — a DataLoader output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a DataLoader 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_numpy_array Ran
ran on a synthesized input
kantharajucn/CLIP-imagenet-evaluation/save_predictions.py
code served (permissive licence) · get_code("1b88c3dbe095f687")
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 (2, 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 torch
import numpy as np

torch.manual_seed(0)

# The function expects a tensor x with is_cuda attribute and methods detach(), cpu(), numpy()
# We'll create a simple CPU tensor
x = torch.randn(2, 4)

ARGS = {
    "x": x
}
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 ndarray, ndim 2, dtype family float, shape (2, 4) — shape (2, 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
model_pytorch Not yet run
unverified (harvested, no recorded run)
bethgelab/model-vs-human/modelvshuman/models/pytorch/model_zoo.py
pointer only (licence: NONE) · get_code("b6e1f6acd97f93c4")
How it ran

The body of this sample is withheld: GitHub detected no licence in this paper's repository, bethgelab/model-vs-human when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 3 samples lifted from this paper's linked repositories, 3 carry a function identity and 0 do not. Of the 3 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 3 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

A few years ago, the first CNN surpassed human performance on ImageNet. However, it soon became clear that machines lack robustness on more challenging test cases, a major obstacle towards deploying machines "in the wild" and towards obtaining better computational models of human visual perception. Here we ask: Are we making progress in closing the gap between human and machine vision? To answer this question, we tested human observers on a broad range of out-ofdistribution (OOD) datasets, recording 85,120 psychophysical trials across 90 participants. We then investigated a range of promising machine learning developments that crucially deviate from standard supervised CNNs along three axes: objective function (self-supervised, adversarially trained, CLIP language-image training), architecture (e.g. vision transformers), and dataset size (ranging from 1M to 1B). Our findings are threefold. (1.) The longstanding distortion robustness gap between humans and CNNs is closing, with the best models now exceeding human feedforward performance on most of the investigated OOD datasets. (2.) There is still a substantial image-level consistency gap, meaning that humans make different errors than models. In contrast, most models systematically agree in their categorisation errors, even substantially different ones like contrastive self-supervised vs. standard supervised models. (3.) In many cases, human-to-model consistency improves when training dataset size is increased by one to three orders of magnitude. Our results give reason for cautious optimism: While there is still much room for improvement, the behavioural difference between human and machine vision is narrowing. In order to measure future progress, 17 OOD datasets with image-level human behavioural data and evaluation code are provided as a toolbox and benchmark at https://github.com/bethgelab/model-vs-human/.

For agents

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

get_harvested_code_for_paper("2106.07411")
get_code_for_paper("2106.07411")
have("2106.07411")

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

curl https://syntology.ai/api/ran/2106.07411.json

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

[![Syntology run record](https://syntology.ai/api/ran/2106.07411.svg)](https://syntology.ai/paper/2106.07411)

Connect an agent — have() is free.