SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2203.08734 · ECCV · 2022

Learning Where To Look -Generative NAS is Surprisingly Efficient

Margret Keuper, Steffen Jung, Jovita Lukasik

arXiv · PDF · Open in the Atlas

Code that ran

8 of 15 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 6 have no contract verdict; 2 failures of our instrument; 7 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/2203.08734.json · badge: /api/ran/2203.08734.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
jovitalukasik/AG-Net marked official in Papers with Code
Papers with Code archive
8 ran · 7 no recorded run
FunctionStatusWhere it lives
Embedding_Network Ran
ran on a synthesized input · constructed an object; no result was computed
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("582d38d6fe781c5d")
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, Embedding_Network: 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_channels": 5,
    "hidden_channels": 32,
    "out_channels": 32,
    "num_layers": 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 Embedding_Network — 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 Embedding_Network 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
GraphAggr Ran
ran on a synthesized input · constructed an object; no result was computed
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("ce4b703f0ca4ac1d")
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, GraphAggr: 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 = {
    'ndim': 4,
    'gdim': 3,
    'aggr': 'gsum'
}
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 GraphAggr — 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 GraphAggr 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_predictor Ran
ran on a synthesized input · constructed an object; no result was computed
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("6108502415b404b5")
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_predictor: 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 = {
    'layers': 3,
    'in_channels': 4,
    'hidden_channels': 8,
    'out_channels': 1,
    'condition': 0
}
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_predictor — 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_predictor 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
NodeAdd Ran
ran on a synthesized input · constructed an object; no result was computed
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("2b9fe11164cf5625")
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, NodeAdd: 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)

# NodeAdd.__init__ takes gdim and num_node_atts
# NodeAdd.forward takes h_G and c tensors

# From the forward method:
# - h_G and c are concatenated along dimension 1, so they must have same batch size and compatible dims
# - After cat([h_G, c], 1), the result goes to f_an which expects (gdim * 2) input features
# - This means h_G has gdim features and c has gdim features
# - h_G and c should have shape (batch_size, gdim)

gdim = 4
num_node_atts = 3
batch_size = 2

# Create sample tensors for forward pass
h_G = torch.randn(batch_size, gdim)
c = torch.randn(batch_size, gdim)

ARGS = {
    'gdim': gdim,
    'num_node_atts': num_node_atts,
}
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 NodeAdd — 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 NodeAdd 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
NodeInit Ran
ran on a synthesized input · constructed an object; no result was computed
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("dad4798b4ede036f")
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, NodeInit: 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 the arguments for NodeInit
ndim = 8
gdim = 4
num_node_atts = 5

ARGS = {
    'ndim': ndim,
    'gdim': gdim,
    'num_node_atts': num_node_atts
}
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 NodeInit — 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 NodeInit 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
Nodes Ran
ran on a synthesized input
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("775bc58aa0b13b2c")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, class construct-then-call prompt, model not recorded) and returned Tensor, ndim 2, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim 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

INIT_ARGS = {
    'ndim': 4,
    'gdim': 3
}

# Create input tensors matching the expected shapes
batch_size = 2
num_nodes = 5  # idx in forward

h = torch.randn(batch_size, num_nodes, INIT_ARGS['ndim'])
h_v = torch.randn(batch_size, INIT_ARGS['ndim'])
h_G = torch.randn(batch_size, INIT_ARGS['gdim'])
c = torch.randn(batch_size, INIT_ARGS['gdim'])

ARGS = {
    'h': h,
    'h_v': h_v,
    'h_G': h_G,
    'c': c
}
Setup block drafted by
a language model, class construct-then-call prompt, model not recorded
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 2, dtype family float — no output shape is stated: observe_signature.py observed this class with its constructor sizes canonicalised where it could, which the node does not record, so its shape is not stated as this setup block's
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: torch and numpy seeded to 0 by the harness immediately before the class was constructed from INIT_ARGS (fixture_check.py)
adj_to_scores Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("79e9ac44d05a5e89")
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 1, dtype family float, shape (21,) 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
import torch.nn.functional as F

torch.manual_seed(0)

# Create a small adjacency matrix
# The function transposes adj and indexes it with row_idx, col_idx from tril_indices
# adj.T.shape[1] is used in tril_indices, so adj needs to have at least 2 dimensions
# Let's make adj a 2D tensor (4, 5) so adj.T is (5, 4)
adj = torch.randn(4, 5)

# The function uses adj.device to move tensors, and indexing with row_idx, col_idx
# The default l=21 means scores should be padded to length 21 if shorter
# tril_indices(5, k=-1) gives lower triangular indices, which for a 5x5 matrix
# gives us (5*4)/2 = 10 indices
# So we get 10 scores, which is less than 21, so padding will occur

ARGS = {
    "adj": adj,
    "l": 21
}
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 1, dtype family float, shape (21,) — shape (21,) 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
scores_to_adj Ran
ran; our contract draft was wrong, not the code
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("07bfd504987b0c00")
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 2, dtype family float, shape (7, 7) 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

torch.manual_seed(0)

# The function expects:
# - scores: a 1D tensor of shape (num_nodes * (num_nodes - 1) / 2,) 
#   since tril_indices with offset=-1 gives lower triangular indices
# - num_nodes: an integer (default 7)

# For num_nodes=7, the lower triangular part (excluding diagonal) has:
# 6 + 5 + 4 + 3 + 2 + 1 = 21 elements
num_nodes = 7
num_scores = num_nodes * (num_nodes - 1) // 2

# Create a scores tensor with the correct size
scores = torch.randn(num_scores)

ARGS = {
    "scores": scores,
    "num_nodes": num_nodes,
}
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 2, dtype family float, shape (7, 7) — shape (7, 7) 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
GNNDecoder Not yet run
unverified (harvested, no recorded run)
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("7a4c97aeb50377dc")
How it ran

no invocation recorded.

GNNLayer_backward Not yet run
unverified (harvested, no recorded run)
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("2d556bf1eff82766")
How it ran

no invocation recorded.

GNNLayer_forward Not yet run
unverified (harvested, no recorded run)
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("010094ca34994468")
How it ran

no invocation recorded.

Generator Not yet run
unverified (harvested, no recorded run)
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("011cde17ef45ee9a")
How it ran

no invocation recorded.

Generator_Decoder Not yet run
unverified (harvested, no recorded run)
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("5d7d83be99daf869")
How it ran

no invocation recorded.

GraphEmbed Not yet run
unverified (harvested, no recorded run)
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("57beeabace709277")
How it ran

no invocation recorded.

NodeEmbUpd Not yet run
unverified (harvested, no recorded run)
jovitalukasik/AG-Net/Generator.py
code served (permissive licence) · get_code("bfe2ce35559e1337")
How it ran

no invocation recorded.

Also present in other papers' code

Of 15 samples lifted from this paper's linked repositories, 15 carry a function identity and 0 do not. Of the 15 distinct functions with an identity, 2 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 2 later) and 13 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
adj_to_scores
brick of this paper's assembly: helper
1 other paper
1 later
2403.06020 (2024-03)
Multi-conditioned Graph Diffusion for Neural Architecture Search
scores_to_adj
brick of this paper's assembly: helper
1 other paper
1 later
2403.06020 (2024-03)
Multi-conditioned Graph Diffusion for Neural Architecture Search

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

The efficient, automated search for well-performing neural architectures (NAS) has drawn increasing attention in the recent past. Thereby, the predominant research objective is to reduce the necessity of costly evaluations of neural architectures while efficiently exploring large search spaces. To this aim, surrogate models embed architectures in a latent space and predict their performance, while generative models for neural architectures enable optimization-based search within the latent space the generator draws from. Both, surrogate and generative models, have the aim of facilitating query-efficient search in a well-structured latent space. In this paper, we further improve the trade-off between query-efficiency and promising architecture generation by leveraging advantages from both, efficient surrogate models and generative design. To this end, we propose a generative model, paired with a surrogate predictor, that iteratively learns to generate samples from increasingly promising latent subspaces. This approach leads to very effective and efficient architecture search, while keeping the query amount low. In addition, our approach allows in a straightforward manner to jointly optimize for multiple objectives such as accuracy and hardware latency. We show the benefit of this approach not only w.r.t. the optimization of architectures for highest classification accuracy but also in the context of hardware constraints and outperform state-of-the-art methods on several NAS benchmarks for single and multiple objectives. We also achieve state-of-the-art performance on ImageNet. The code is available at https://github.com/jovitalukasik/AG-Net.

For agents

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

get_harvested_code_for_paper("2203.08734")
get_code_for_paper("2203.08734")
have("2203.08734")

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

curl https://syntology.ai/api/ran/2203.08734.json

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

[![Syntology run record](https://syntology.ai/api/ran/2203.08734.svg)](https://syntology.ai/paper/2203.08734)

Connect an agent — have() is free.