SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2212.11696 · ICLR · 2023

Reversible Column Networks

Jun Li, Xiangyu Zhang, Qi Han, Jianjian Sun, Yuxuan Cai, Yizhuang Zhou, Xiangwen Kong

arXiv · PDF · Open in the Atlas

Code that ran

8 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (8 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 8 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/2212.11696.json · badge: /api/ran/2212.11696.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
megvii-research/revcol community
Papers with Code archive
8 ran · 3 no recorded run
FunctionStatusWhere it lives
Classifier Ran
ran on a synthesized input · constructed an object; no result was computed
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("62fc75450a52fc56")
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, Classifier: 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)

# The Classifier takes two parameters: in_channels and num_classes
# in_channels: the number of input channels (used in LayerNorm and Linear)
# num_classes: the number of output classes (used in Linear output dimension)

in_channels = 16
num_classes = 10

ARGS = {
    'in_channels': in_channels,
    'num_classes': num_classes
}
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 Classifier — 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 Classifier 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
ConvNextBlock Ran
ran on a synthesized input · constructed an object; no result was computed
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("008b0e076048797d")
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, ConvNextBlock: 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 timm.models.layers import DropPath

torch.manual_seed(0)

ARGS = {
    "in_channel": 4,
    "hidden_dim": 8,
    "out_channel": 4,
    "kernel_size": 3,
    "layer_scale_init_value": 1e-6,
    "drop_path": 0.0,
}
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 ConvNextBlock — 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 ConvNextBlock 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
Decoder Ran
ran on a synthesized input · constructed an object; no result was computed
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("0aa0f3056567a82f")
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, Decoder: 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
import torch.nn.functional as F

torch.manual_seed(0)

class DummyBlock(nn.Module):
    """Minimal block that satisfies the block_type interface"""
    def __init__(self, in_channels, mid_channels, out_channels, kernel_size):
        super().__init__()
        self.conv = nn.Conv2d(in_channels, out_channels, kernel_size, padding=kernel_size//2)
    
    def forward(self, x):
        return self.conv(x)

ARGS = {
    'depth': [2, 2, 2, 2],
    'dim': [112, 72, 40, 24],
    'block_type': DummyBlock,
    'kernel_size': 3
}
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 Decoder — 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 Decoder 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
Fusion Ran
ran on a synthesized input · constructed an object; no result was computed
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("7bc303c0be07360e")
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, Fusion: 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 channels list - needs to have at least level+2 elements
# Using distinct sizes: 4, 8, 16, 32, 64
channels = [4, 8, 16, 32, 64]

# Test case 1: level=0, first_col=True
ARGS = {
    'level': 0,
    'channels': channels,
    'first_col': True
}
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 Fusion — 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 Fusion 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
LayerNorm Ran
ran on a synthesized input · constructed an object; no result was computed
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("3d31e56894e46115")
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, LayerNorm: 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)

ARGS = {
    'normalized_shape': 4,
    'eps': 1e-06,
    'data_format': 'channels_first',
    'elementwise_affine': True
}
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 LayerNorm — 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 LayerNorm 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
ReverseFunction Ran
ran on a synthesized input · constructed an object; no result was computed
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("1939d987f30fd46b")
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, ReverseFunction: the run constructed the object and did not call it, so no result was computed. 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
import torch.nn as nn

torch.manual_seed(0)

# Create simple layer functions that accept two inputs
class SimpleLayer(nn.Module):
    def __init__(self, in_dim, out_dim):
        super().__init__()
        self.linear = nn.Linear(in_dim, out_dim)
    
    def forward(self, x, y=None):
        out = self.linear(x)
        if y is not None:
            out = out + y
        return out

# Create layer instances
l0 = SimpleLayer(4, 4)
l1 = SimpleLayer(4, 4)
l2 = SimpleLayer(4, 4)
l3 = SimpleLayer(4, 4)

# Wrap them in a way that makes them callable
def make_layer_fn(layer):
    def fn(x, y=None):
        return layer(x, y)
    return fn

layer_fn_0 = make_layer_fn(l0)
layer_fn_1 = make_layer_fn(l1)
layer_fn_2 = make_layer_fn(l2)
layer_fn_3 = make_layer_fn(l3)

run_functions = (layer_fn_0, layer_fn_1, layer_fn_2, layer_fn_3)

# Alpha values for weighting
alpha = (0.5, 0.5, 0.5, 0.5)

# Create input tensors
x = torch.randn(2, 4, requires_grad=True)
c0 = 0  # Can be int (first_col case)
c1 = torch.randn(2, 4, requires_grad=True)
c2 = torch.randn(2, 4, requires_grad=True)
c3 = torch.randn(2, 4, requires_grad=True)

args = (x, c0, c1, c2, c3)

ARGS = {
    'run_functions': run_functions,
    'alpha': alpha,
    'args': args,
}
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 ReverseFunction — 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 ReverseFunction 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
SimDecoder Ran
ran on a synthesized input · constructed an object; no result was computed
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("9da4f6ea726fdfb1")
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, SimDecoder: 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)

ARGS = {
    'in_channel': 4,
    'encoder_stride': 2,
}
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 SimDecoder — 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 SimDecoder 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
UpSampleConvnext Ran
ran on a synthesized input · constructed an object; no result was computed
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("a214203738550dde")
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, UpSampleConvnext: 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)

# Parameters for UpSampleConvnext
ratio = 2
inchannel = 4
outchannel = 8

ARGS = {
    'ratio': ratio,
    'inchannel': inchannel,
    'outchannel': outchannel
}
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 UpSampleConvnext — 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 UpSampleConvnext 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
FullNet Not yet run
unverified (harvested, no recorded run)
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("c57688066cd3ac52")
How it ran

no invocation recorded.

Level Not yet run
unverified (harvested, no recorded run)
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("d6bc6096b81fd178")
How it ran

no invocation recorded.

SubNet Not yet run
unverified (harvested, no recorded run)
megvii-research/revcol/models/revcol.py
code served (permissive licence) · get_code("6d63037ed95088f7")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 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 11 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

We propose a new neural network design paradigm Reversible Column Network (RevCol). The main body of RevCol is composed of multiple copies of subnetworks, named columns respectively, between which multi-level reversible connections are employed. Such architectural scheme attributes RevCol very different behavior from conventional networks: during forward propagation, features in RevCol are learned to be gradually disentangled when passing through each column, whose total information is maintained rather than compressed or discarded as other network does. Our experiments suggest that CNN-style RevCol models can achieve very competitive performances on multiple computer vision tasks such as image classification, object detection and semantic segmentation, especially with large parameter budget and large dataset. For example, after ImageNet-22K pre-training, RevCol-XL obtains 88.2% ImageNet-1K accuracy. Given more pre-training data, our largest model RevCol-H reaches 90.0% on ImageNet-1K, 63.8% AP box on COCO detection minival set, 61.0% mIoU on ADE20k segmentation. To our knowledge, it is the best COCO detection and ADE20k segmentation result among pure (static) CNN models. Moreover, as a general macro architecture fashion, RevCol can also be introduced into transformers or other neural networks, which is demonstrated to improve the performances in both computer vision and NLP tasks. We release code and models at https://github.com/megvii-research/RevCol

For agents

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

get_harvested_code_for_paper("2212.11696")
get_code_for_paper("2212.11696")
have("2212.11696")

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

curl https://syntology.ai/api/ran/2212.11696.json

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

[![Syntology run record](https://syntology.ai/api/ran/2212.11696.svg)](https://syntology.ai/paper/2212.11696)

Connect an agent — have() is free.