SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.03282 · CVPR · 2022

Interpretable part-whole hierarchies and conceptual-semantic relationships in neural networks

Zeno Sambugaro, Nicola Conci, Nicola Garau, Niccoló Bisagno

arXiv · PDF · Open in the Atlas

Code that ran

5 of 7 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (5 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 5 have no contract verdict; 0 failures of our instrument; 2 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/2203.03282.json · badge: /api/ran/2203.03282.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
mmlab-cv/Agglomerator marked official in Papers with Code
Papers with Code archive
5 ran · 2 no recorded run
FunctionStatusWhere it lives
ColumnNet Ran
ran on a synthesized input · constructed an object; no result was computed
mmlab-cv/Agglomerator/src/models.py
code served (permissive licence) · get_code("84e5b05b0dd6824a")
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, ColumnNet: 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 pytorch_lightning as pl
from einops.layers.torch import Rearrange

torch.manual_seed(0)

class FLAGSConfig:
    def __init__(self):
        self.conv_image_size = 32
        self.patch_size = 8

FLAGS = FLAGSConfig()
dim = 4
groups = 3
mult = 4
activation = nn.GELU

ARGS = {
    "FLAGS": FLAGS,
    "dim": dim,
    "groups": groups,
    "mult": mult,
    "activation": activation
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, pytorch_lightning, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ColumnNet — 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 ColumnNet 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
ConsensusAttention Ran
ran on a synthesized input · constructed an object; no result was computed
mmlab-cv/Agglomerator/src/models.py
code served (permissive licence) · get_code("0e38b8d823bd2d76")
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, ConsensusAttention: 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
from torch import einsum
import pytorch_lightning as pl
from einops import rearrange, repeat

torch.manual_seed(0)

TOKEN_ATTEND_SELF_VALUE = -0.0005

class ConsensusAttention(pl.LightningModule):

    def __init__(self, num_patches_side, attend_self=True, local_consensus_radius=0):
        super().__init__()
        self.attend_self = attend_self
        self.local_consensus_radius = local_consensus_radius
        if self.local_consensus_radius > 0:
            coors = torch.stack(torch.meshgrid(torch.arange(num_patches_side), torch.arange(num_patches_side), indexing='ij')).float()
            coors = rearrange(coors, 'c h w -> (h w) c')
            dist = torch.cdist(coors, coors)
            mask_non_local = dist > self.local_consensus_radius
            mask_non_local = rearrange(mask_non_local, 'i j -> () i j')
            self.register_buffer('non_local_mask', mask_non_local)

    def forward(self, levels):
        _, n, _, d, device = (*levels.shape, levels.device)
        q, k, v = (levels, F.normalize(levels, dim=-1), levels)
        sim = einsum('b i l d, b j l d -> b l i j', q, k) * d ** (-0.5)
        if not self.attend_self:
            self_mask = torch.eye(n, device=device, dtype=torch.bool)
            self_mask = rearrange(self_mask, 'i j -> () () i j')
            sim.masked_fill_(self_mask, TOKEN_ATTEND_SELF_VALUE)
        if self.local_consensus_radius > 0:
            max_neg_value = -torch.finfo(sim.dtype).max
            sim.masked_fill_(self.non_local_mask, max_neg_value)
        attn = sim.softmax(dim=-1)
        out = einsum('b l i j, b j l d -> b i l d', attn, levels)
        return out


ARGS = {
    "num_patches_side": 3,
    "attend_self": True,
    "local_consensus_radius": 0,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
einops, pytorch_lightning, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ConsensusAttention — 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 ConsensusAttention 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
ConvTokenizer Ran
ran on a synthesized input · constructed an object; no result was computed
mmlab-cv/Agglomerator/src/models.py
code served (permissive licence) · get_code("cee68d0097065965")
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, ConvTokenizer: 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 pytorch_lightning as pl

torch.manual_seed(0)

ARGS = {
    "in_channels": 3,
    "embedding_dim": 64,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pytorch_lightning, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ConvTokenizer — 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 ConvTokenizer 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
Siren Ran
ran on a synthesized input · constructed an object; no result was computed
mmlab-cv/Agglomerator/src/models.py
code served (permissive licence) · get_code("123e34158db9e6fd")
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, Siren: 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 pytorch_lightning as pl
import torch.nn as nn

torch.manual_seed(0)

ARGS = {
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pytorch_lightning, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type Siren — 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 Siren 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
SupConLoss Ran
ran on a synthesized input · constructed an object; no result was computed
mmlab-cv/Agglomerator/src/models.py
code served (permissive licence) · get_code("299961c9af1dd5a2")
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, SupConLoss: 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)

# Construct tiny tensors for supervised contrastive learning
# features shape: [bsz, n_views, feature_dim]
batch_size = 2
n_views = 2
feature_dim = 4

# Create features tensor: [2, 2, 4]
features = torch.randn(batch_size, n_views, feature_dim)

# Create labels tensor: [2] with class assignments
labels = torch.tensor([0, 1])

# Create a mask tensor: [2, 2] for explicit contrastive mask
mask = torch.tensor([[1.0, 0.0], [0.0, 1.0]])

ARGS = {
    'temperature': 0.07,
    'contrast_mode': 'all',
    'base_temperature': 0.07,
}
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 SupConLoss — 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 SupConLoss 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
Agglomerator Not yet run
unverified (harvested, no recorded run)
mmlab-cv/Agglomerator/src/models.py
code served (permissive licence) · get_code("bfe0064764913784")
How it ran

no invocation recorded.

plot_islands_agreement Not yet run
unverified (harvested, no recorded run)
mmlab-cv/Agglomerator/src/models.py
code served (permissive licence) · get_code("9e135befb1bec51f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 7 samples lifted from this paper's linked repositories, 7 carry a function identity and 0 do not. Of the 7 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 1 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.

FunctionAlso present inEarliest-datedOthers named
SupConLoss
brick of this paper's assembly: helper
6 other papers
4 dated earlier, 2 later · 2 only as an assembly brick
2002.05709 (2020-02)
A Simple Framework for Contrastive Learning of Visual Representations
2104.08821 (2021-04)
SimCSE: Simple Contrastive Learning of Sentence Embeddings
2109.06349 (2021-09)
Few-Shot Intent Detection via Contrastive Pre-Training and Fine-Tuning
and 3 more not named here

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

Figure 1. [Better seen in color]. Overview of the proposed solution. Our Agglomerator is a novel architecture for vision applications, in which column structure (c) mimics hyper-columns typical of the human visual cortex [14]. The input data (a) is fed to the columns using a patch-based embedding (b). The Agglomerator architecture iteratively routes the information across its structure, creating a neural representation of each image, similar to neural fields [37]. In the neural representation, part-whole hierarchies (d) emerge at different levels of the columns. The same column can represent the same patch of the image with different levels of abstraction (e.g., the ears, the head, and the dog) corresponding to each level in the column. Neighbor columns agree on a part representation (e.g ears, head) at lower levels, ideally representing the same whole (e.g. dog) at the top level. The resulting feature space represents the conceptual-semantic relationships between data (e) resembling the human hierarchical organization [38]. Samples belonging to the same super-class (e.g., animals, vehicles) are clustered together, with conceptually close categories (e.g., birds and airplanes) represented on the edge of the super-classes.

For agents

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

get_harvested_code_for_paper("2203.03282")
get_code_for_paper("2203.03282")
have("2203.03282")

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

curl https://syntology.ai/api/ran/2203.03282.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.03282.svg)](https://syntology.ai/paper/2203.03282)

Connect an agent — have() is free.