SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2306.07957 · ICCV · 2023

Hidden Biases of End-to-End Driving Models

Andreas Geiger, Kashyap Chitta, Bernhard Jaeger

arXiv · PDF · Open in the Atlas

Code that ran

13 of 17 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (10 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 10 have no contract verdict; 3 failures of our instrument; 4 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/2306.07957.json · badge: /api/ran/2306.07957.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
autonomousvision/carla_garage marked official in Papers with Code
Papers with Code archive
13 ran · 4 no recorded run
FunctionStatusWhere it lives
Block Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("80dbac57b0bfe077")
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 an instance of its own class, Block: the run constructed the object and did not call it, so no result was computed. 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 torch import nn

torch.manual_seed(0)

# Define the ARGS dict with all parameters for Block.__init__
ARGS = {
    "n_embd": 4,
    "n_head": 2,
    "block_exp": 2,
    "attn_pdrop": 0.1,
    "resid_pdrop": 0.1,
}
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 Block — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Block 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
CustomNorm Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("c883c728a459b83a")
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 an instance of its own class, CustomNorm: the run constructed the object and did not call it, so no result was computed. 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
from einops import rearrange

torch.manual_seed(0)

class CustomNorm(nn.Module):
    """
    Changes the channel dimensions before applying a norm and reverts it back afterwards.
    """

    def __init__(self, norm_layer, num_features):
        super().__init__()
        self.norm = norm_layer(num_features)

    def forward(self, x):
        x = rearrange(x, 'n c d h w -> n d h w c')
        x = self.norm(x)
        x = rearrange(x, 'n d h w c -> n c d h w')
        return x


# Define the norm_layer - a callable that takes num_features and returns a norm module
norm_layer = nn.LayerNorm

# Define num_features
num_features = 4

ARGS = {
    "norm_layer": norm_layer,
    "num_features": num_features
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type CustomNorm — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a CustomNorm 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
GPT Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("d07d06f85b424d92")
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 an instance of its own class, GPT: the run constructed the object and did not call it, so no result was computed. 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)

# Create a minimal config class with all required attributes
class Config:
    def __init__(self):
        self.lidar_seq_len = 1
        self.img_vert_anchors = 2
        self.img_horz_anchors = 3
        self.lidar_vert_anchors = 2
        self.lidar_horz_anchors = 3
        self.n_head = 2
        self.block_exp = 2
        self.attn_pdrop = 0.1
        self.resid_pdrop = 0.1
        self.embd_pdrop = 0.1
        self.n_layer = 1
        self.gpt_linear_layer_init_mean = 0.0
        self.gpt_linear_layer_init_std = 0.02
        self.gpt_layer_norm_init_weight = 1.0

config = Config()

# Parameters for GPT
n_embd = 4
lidar_video = False
lidar_time_frames = 1

# Create minimal image_tensor: B*4*seq_len, C, H, W
# With seq_len=1: B*4*1, C, H, W -> shape (4, 4, 2, 3)
image_tensor = torch.randn(4, 4, 2, 3)

# Create minimal lidar_tensor for lidar_video=False: B*seq_len, C, H, W -> (2, 4, 2, 3)
lidar_tensor = torch.randn(2, 4, 2, 3)

ARGS = {
    "n_embd": n_embd,
    "config": config,
    "lidar_video": lidar_video,
    "lidar_time_frames": lidar_time_frames,
}
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 GPT — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a GPT 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
InfoDummy Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("71fc440ae60083f6")
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 an instance of its own class, InfoDummy: the run constructed the object and did not call it, so no result was computed. 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)

info_dict = {
    'architecture': 'resnet18',
    'pretrained': True,
    'num_classes': 1000,
    'input_size': (3, 224, 224),
    'pool_size': (7, 7),
    'crop_pct': 0.875,
    'interpolation': 'bicubic',
    'mean': [0.485, 0.456, 0.406],
    'std': [0.229, 0.224, 0.225],
}

ARGS = {
    'info': info_dict,
}
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 InfoDummy — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a InfoDummy 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
Mlp Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("3d4489efff0d96f9")
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 an instance of its own class, Mlp: the run constructed the object and did not call it, so no result was computed. 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)

# Create a sample input tensor for forward pass testing
sample_input = torch.randn(2, 4)  # batch_size=2, in_features=4

# Define ARGS dict with all parameters for Mlp.__init__
ARGS = {
    'in_features': 4,
    'hidden_features': 8,
    'out_features': 3,
    'act_layer': nn.GELU,
    'drop': 0.1,
}
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 Mlp — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a Mlp 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
PatchEmbed3D Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("fdf0d85afa378d45")
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 an instance of its own class, PatchEmbed3D: the run constructed the object and did not call it, so no result was computed. 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)

# Define a minimal normalization layer that matches the expected interface
class MinimalNorm(nn.Module):
    def __init__(self, num_features):
        super().__init__()
        self.num_features = num_features
    
    def forward(self, x):
        # x shape: (batch, seq_len, embed_dim)
        return x

# Create test input tensor with shape (batch, channels, depth, height, width)
# Using distinct sizes: batch=2, channels=3, depth=4, height=5, width=6
test_input = torch.randn(2, 3, 4, 5, 6)

# Define ARGS with all parameters
ARGS = {
    "patch_size": (2, 4, 4),
    "in_chans": 3,
    "embed_dim": 96,
    "norm_layer": MinimalNorm
}
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 PatchEmbed3D — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a PatchEmbed3D 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
PatchMerging Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("aa4f65952e82e439")
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 an instance of its own class, PatchMerging: the run constructed the object and did not call it, so no result was computed. 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 torch import nn

torch.manual_seed(0)

ARGS = {
    "dim": 4,
    "norm_layer": nn.LayerNorm,
}
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 PatchMerging — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a PatchMerging 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
SelfAttention Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("85cfd6ab309be0b8")
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 an instance of its own class, SelfAttention: the run constructed the object and did not call it, so no result was computed. 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 torch import nn

torch.manual_seed(0)

# Define the arguments for SelfAttention
ARGS = {
    'n_embd': 64,
    'n_head': 4,
    'attn_pdrop': 0.1,
    'resid_pdrop': 0.1,
}
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 SelfAttention — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a SelfAttention 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
VideoResNet Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("429af4867aa42dde")
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 an instance of its own class, VideoResNet: the run constructed the object and did not call it, so no result was computed. 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
from torchvision.models.video import r2plus1d_18
import types as _types

torch.manual_seed(0)

class InfoDummy(object):
    """ Info dummy that tries to mimic TIMMs info class"""

    def __init__(self, info):
        super().__init__()
        self.info = info

class VideoResNet(nn.Module):
    """ R2+1D Video ResNet 18 based on torchvision implementation.
      We adapt the code here so that it matches the structure of timm models and we can interchange them more easily.
  """

    def __init__(self, in_channels=1, pretrained=False):
        super().__init__()
        self.model = r2plus1d_18(pretrained=pretrained)
        del self.model.fc
        del self.model.avgpool
        tmp = self.model.stem._modules['0']
        self.model.stem._modules['0'] = nn.Conv3d(in_channels, out_channels=tmp.out_channels, kernel_size=tmp.kernel_size, stride=tmp.stride, padding=tmp.padding, bias=tmp.bias)
        del tmp
        self.return_layers = {}
        self.feature_info = InfoDummy([dict(num_chs=64, reduction=2, module='stem'), dict(num_chs=64, reduction=2, module='layer1'), dict(num_chs=128, reduction=4, module='layer2'), dict(num_chs=256, reduction=8, module='layer3'), dict(num_chs=512, reduction=16, module='layer4')])
        for idx, layer in enumerate(self.model.named_children()):
            self.return_layers[layer[0]] = idx
            setattr(self, layer[0], layer[1])

    def items(self):
        return self.model.named_children()


ARGS = {
    'in_channels': 1,
    'pretrained': False
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch, torchvision — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type VideoResNet — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a VideoResNet 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
WindowAttention3D Ran
ran on a synthesized input · constructed an object; no result was computed
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("e18ea685986aac66")
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 an instance of its own class, WindowAttention3D: the run constructed the object and did not call it, so no result was computed. 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 torch import nn
from timm.layers import DropPath, trunc_normal_

torch.manual_seed(0)

# Define arguments for WindowAttention3D
ARGS = {
    "dim": 64,
    "window_size": (2, 3, 4),  # temporal, height, width - distinct sizes
    "num_heads": 4,
    "qkv_bias": True,
    "qk_scale": None,
    "attn_drop": 0.1,
    "proj_drop": 0.1,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
timm, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type WindowAttention3D — the run constructed an instance of its own class and did not call it; no shape is stated — the instrument coerces any output to an array; for a WindowAttention3D 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
compute_mask Ran
ran; our contract draft was wrong, not the code
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("2a16557c0e01720d")
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 3, dtype family float, shape (8, 24, 24) 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
from functools import reduce
from operator import mul

torch.manual_seed(0)

# The function computes_mask takes:
# - depth, height, width: integers for tensor dimensions
# - window_size: tuple of ints for window partitioning
# - shift_size: tuple of ints for shifted window slicing
# - device: torch device

# We need window_size and shift_size to divide evenly into depth, height, width
# Let's use small values that divide evenly

depth = 4
height = 6
width = 8
window_size = (2, 3, 4)  # Must divide depth, height, width respectively
shift_size = (1, 1, 2)   # Must be less than window_size
device = torch.device('cpu')

ARGS = {
    'depth': depth,
    'height': height,
    'width': width,
    'window_size': window_size,
    'shift_size': shift_size,
    'device': device
}
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 3, dtype family float, shape (8, 24, 24) — shape (8, 24, 24) 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
window_partition Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("b02810537588055a")
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 3, dtype family float, shape (16, 24, 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 torch
from functools import reduce
from operator import mul

torch.manual_seed(0)

# Create a 5D tensor with shape (B, D, H, W, C)
# B=2 (batch), D=4 (depth), H=6 (height), W=8 (width), C=3 (channels)
# Dimensions chosen to be divisible by window sizes
x = torch.randn(2, 4, 6, 8, 3)

# window_size is a tuple of 3 integers for partitioning D, H, W dimensions
window_size = (2, 3, 4)

ARGS = {
    "x": x,
    "window_size": window_size,
}
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 3, dtype family float, shape (16, 24, 3) — shape (16, 24, 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
window_reverse Ran
ran; our contract draft was wrong, not the code
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("c15d21e9e6193af7")
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 5, dtype family float, shape (2, 4, 8, 6, 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 torch
import numpy as np

torch.manual_seed(0)

# Parameters for window_reverse
# windows: (B*num_windows, window_size, window_size, C)
# The function expects windows from a windowed operation on 3D data
# Let's construct a minimal valid case

b = 2  # batch size
d = 4  # depth
h = 8  # height
w = 6  # width
window_size = (2, 2, 2)  # (depth_window, height_window, width_window)
c = 3  # channels

# Calculate number of windows
num_windows_d = d // window_size[0]  # 2
num_windows_h = h // window_size[1]  # 4
num_windows_w = w // window_size[2]  # 3
num_windows = num_windows_d * num_windows_h * num_windows_w  # 24

# windows shape: (b * num_windows, window_size[0], window_size[1], window_size[2], c)
# But the docstring says (B*num_windows, window_size, window_size, C)
# Looking at the view operation, it expects 4D input reshaped to 8D
# The actual input shape should be: (B*num_windows, window_size[0], window_size[1], window_size[2], c)
# But the function treats it as 4D initially, so we construct it as 4D with last dim containing everything

windows = torch.randn(b * num_windows, window_size[0], window_size[1], window_size[2] * c)

ARGS = {
    "windows": windows,
    "window_size": window_size,
    "b": b,
    "d": d,
    "h": h,
    "w": w,
}
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 Tensor, ndim 5, dtype family float, shape (2, 4, 8, 6, 3) — shape (2, 4, 8, 6, 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
BasicLayer Not yet run
unverified (harvested, no recorded run)
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("877ecd68b92683ed")
How it ran

no invocation recorded.

SwinTransformer3D Not yet run
unverified (harvested, no recorded run)
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("b95dccd5c9f74fb4")
How it ran

no invocation recorded.

SwinTransformerBlock3D Not yet run
unverified (harvested, no recorded run)
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("17d362af22ec9f8b")
How it ran

no invocation recorded.

TransfuserBackbone Not yet run
unverified (harvested, no recorded run)
autonomousvision/carla_garage/team_code/transfuser.py
code served (permissive licence) · get_code("8bfa7814e611be75")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 17 distinct functions with an identity, 2 are also present in other papers whose code we hold (2 with at least one paper dated earlier, 1 the same month, 2 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
Mlp
brick of this paper's assembly: helper
130 other papers
77 dated earlier, 1 the same month, 52 later · 114 only as an assembly brick
2003.12039 (2020-03)
RAFT: Recurrent All-Pairs Field Transforms for Optical Flow
2010.11929 (2020-10)
An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale
2012.12877 (2020-12)
Training data-efficient image transformers & distillation through attention
and 127 more not named here
Block
brick of this paper's assembly: model
2 other papers
1 dated earlier, 1 later
2104.09224 (2021-04)
Multi-Modal Fusion Transformer for End-to-End Autonomous Driving
2504.01941 (2025-04)
End-to-End Driving with Online Trajectory Evaluation via BEV World Model

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

End-to-end driving systems have recently made rapid progress, in particular on CARLA. Independent of their major contribution, they introduce changes to minor system components. Consequently, the source of improvements is unclear. We identify two biases that recur in nearly all state-of-the-art methods and are critical for the observed progress on CARLA: (1) lateral recovery via a strong inductive bias towards target point following, and (2) longitudinal averaging of multimodal waypoint predictions for slowing down. We investigate the drawbacks of these biases and identify principled alternatives. By incorporating our insights, we develop TF++, a simple end-to-end method that ranks first on the Longest6 and LAV benchmarks, gaining 11 driving score over the best prior work on Longest6.

For agents

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

get_harvested_code_for_paper("2306.07957")
get_code_for_paper("2306.07957")
have("2306.07957")

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

curl https://syntology.ai/api/ran/2306.07957.json

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

[![Syntology run record](https://syntology.ai/api/ran/2306.07957.svg)](https://syntology.ai/paper/2306.07957)

Connect an agent — have() is free.