SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2304.14636 · ICML · 2023

PreNAS: Preferred One-Shot Learning Towards Efficient Neural Architecture Search

Xiuyu Sun, Haibin Wang, Ce Ge, Hesen Chen

arXiv · PDF · Open in the Atlas

Code that ran

16 of 23 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (3 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 1 violated it and 14 have no contract verdict; 1 failure of our instrument; 7 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/2304.14636.json · badge: /api/ran/2304.14636.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
tinyvision/PreNAS canonical (classifier label)
classifier basis: census
8 ran · 3 no recorded run
alibaba/lightweight-neural-architecture-search canonical (classifier label)
classifier basis: census
5 ran · 1 no recorded run
tinyvision/prenas community
Papers with Code archive
3 ran · 3 no recorded run
FunctionStatusWhere it lives
PatchembedSuper Ran
ran on a synthesized input · constructed an object; no result was computed
tinyvision/prenas/model/supernet_transformer.py
code served (permissive licence) · get_code("5e87d9096209e29f")
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, PatchembedSuper: 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 numpy as np

torch.manual_seed(0)

# Construct arguments for PatchembedSuper
ARGS = {
    "img_size": 224,
    "patch_size": 16,
    "in_chans": 3,
    "embed_dim": 768,
    "scale": False,
}
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 PatchembedSuper — 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 PatchembedSuper 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
ScalingSuper Ran
ran on a synthesized input · constructed an object; no result was computed
tinyvision/prenas/model/supernet_transformer.py
code served (permissive licence) · get_code("026685c04c506554")
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, ScalingSuper: 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)

# ScalingSuper.__init__ takes embed_dim_list
# ScalingSuper.set_sample_config takes sample_embed_dim (must be in embed_dim_list)
# ScalingSuper.forward takes x (a tensor)
# The forward method accesses self.scalings[self.sample_idx][:self.sample_embed_dim]
# So x should be compatible with scaling operations

# Create a small embed_dim_list with distinct values
embed_dim_list = [2, 4, 6]

# Create a sample input tensor
# The forward expects x * self.scalings[self.sample_idx][:self.sample_embed_dim]
# If sample_embed_dim is 4, then we scale by a vector of length 4
# So x should have shape compatible with broadcasting
x = torch.randn(2, 4)  # batch of 2, embedding dim 4

# sample_embed_dim must be one of the values in embed_dim_list
sample_embed_dim = 4

ARGS = {
    'embed_dim_list': embed_dim_list,
}
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 ScalingSuper — 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 ScalingSuper 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
SwitchableLayerNormSuper Ran
ran on a synthesized input · constructed an object; no result was computed
tinyvision/prenas/model/supernet_transformer.py
code served (permissive licence) · get_code("19eb86babba84eec")
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, SwitchableLayerNormSuper: 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 SwitchableLayerNormSuper class takes embed_dim_list as its only parameter
# It creates LayerNorm modules for each dimension in the list
# The forward method uses set_sample_config to select which LayerNorm to use
# So we need a list of embedding dimensions

ARGS = {
    "embed_dim_list": [4, 8, 12]
}
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 SwitchableLayerNormSuper — 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 SwitchableLayerNormSuper 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
build_transform Ran
ran on a synthesized input
tinyvision/PreNAS/lib/datasets.py
pointer only (licence: Apache-2.0) · get_code("97eb30cd8a8f60a9")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

candidate_to_choices Ran
ran on a synthesized input
tinyvision/PreNAS/candidates_to_choices.py
pointer only (licence: Apache-2.0) · get_code("33be0fe904fcd4bb")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

decode_cand_tuple Ran
ran; our contract draft was wrong, not the code
tinyvision/PreNAS/evolution_pre_train.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from liangyn22/MCUFormer, is pointer only (licence: NONE) · get_code("b328ce4411979624")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (Apache-2.0), but get_code serves the stored copy, first harvested from liangyn22/MCUFormer, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

func_receive_dict_inputs Ran
ran on a synthesized input
alibaba/lightweight-neural-architecture-search/modelscope/utils/file_utils.py
code served (permissive licence) · get_code("6f8b42ebbb59d908")
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 bool, ndim 0, dtype family float, shape () 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 inspect

def func_receive_dict_inputs(func):
    """to decide if a func could recieve dict inputs or not

    Args:
        func (class): the target function to be inspected

    Returns:
        bool: if func only has one arg ``input`` or ``inputs``, return True, else return False
    """
    full_args_spec = inspect.getfullargspec(func)
    varargs = full_args_spec.varargs
    varkw = full_args_spec.varkw
    if not (varargs is None and varkw is None):
        return False
    args = [] if not full_args_spec.args else full_args_spec.args
    args.pop(0) if args and args[0] in ['self', 'cls'] else args
    if len(args) == 1 and args[0] in ['input', 'inputs']:
        return True
    return False


# Test case 1: A function with single 'input' parameter - should return True
def test_func_single_input(input):
    return input

# Test case 2: A function with single 'inputs' parameter - should return True
def test_func_single_inputs(inputs):
    return inputs

# Test case 3: A function with multiple parameters - should return False
def test_func_multiple_args(input, output):
    return input + output

# Test case 4: A function with *args - should return False
def test_func_varargs(input, *args):
    return input

# Test case 5: A function with **kwargs - should return False
def test_func_kwargs(input, **kwargs):
    return input

# Test case 6: A method with self - should return True for 'input' parameter
class TestClass:
    def method_with_input(self, input):
        return input

ARGS = {
    "func": test_func_single_input
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () 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
gen_key Ran
ran on a synthesized input
tinyvision/PreNAS/lib/score_maker.py
pointer only (licence: Apache-2.0) · get_code("3c63fe350b1181ce")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

get_max_min_model Ran
ran on a synthesized input
tinyvision/PreNAS/evolution_pre_train.py
pointer only (licence: Apache-2.0) · get_code("59df9b4c9ddc23e1")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

has_method Ran
ran, violated the contract we drafted
alibaba/lightweight-neural-architecture-search/modelscope/utils/import_utils.py
code served (permissive licence) · get_code("ecaa7193a5bfc39e")
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 bool, ndim 0, dtype family float, shape () 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)

# Create a simple class with a method
class SimpleObject:
    def my_method(self):
        return "test"
    
    def another_method(self):
        pass

# Create an instance
obj = SimpleObject()

# Create the ARGS dictionary with both parameters
ARGS = {
    "obj": obj,
    "method": "my_method"
}
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 bool, ndim 0, dtype family float, shape () — shape () 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
import_modules_from_file Ran
ran on a synthesized input
alibaba/lightweight-neural-architecture-search/modelscope/utils/import_utils.py
code served (permissive licence) · get_code("2eeb911987d072eb")
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 tuple, ndim 1, 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 tempfile
import os

# Create a temporary Python file with valid syntax
temp_dir = tempfile.mkdtemp()
temp_py_file = os.path.join(temp_dir, "test_module.py")

with open(temp_py_file, 'w', encoding='utf-8') as f:
    f.write("# Valid Python module\nVAR = 42\n")

ARGS = {
    'py_file': temp_py_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
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
is_method_overridden Ran
ran on a synthesized input
alibaba/lightweight-neural-architecture-search/modelscope/utils/import_utils.py
code served (permissive licence) · get_code("f22f6715f6935b8b")
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 bool, ndim 0, dtype family float, shape () 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.

# Define base and derived classes for testing
class BaseClass:
    def method_to_check(self):
        return "base"
    
    def other_method(self):
        return "other"


class DerivedClassOverride(BaseClass):
    def method_to_check(self):
        return "derived"


class DerivedClassNoOverride(BaseClass):
    pass


# Construct ARGS dictionary with valid arguments for is_method_overridden
ARGS = {
    "method": "method_to_check",
    "base_class": BaseClass,
    "derived_class": DerivedClassOverride
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () 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
read_file Ran
ran on a synthesized input
alibaba/lightweight-neural-architecture-search/modelscope/utils/file_utils.py
code served (permissive licence) · get_code("0531c04dd4901504")
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 str. 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 tempfile
import os

# Create a temporary file with some text content
temp_dir = tempfile.mkdtemp()
temp_file = os.path.join(temp_dir, "test.txt")
with open(temp_file, 'w') as f:
    f.write("Hello, this is test content for the read_file function.")

ARGS = {
    "path": temp_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str 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
sample_a_cand Ran
ran on a synthesized input
tinyvision/PreNAS/supernet_engine.py
pointer only (licence: Apache-2.0) · get_code("3b90681597c539f9")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

sample_a_config Ran
ran on a synthesized input
tinyvision/PreNAS/supernet_engine.py
pointer only (licence: Apache-2.0) · get_code("bc914e97a3d6f3c5")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

sample_candidates Ran
ran on a synthesized input
tinyvision/PreNAS/supernet_engine.py
pointer only (licence: Apache-2.0) · get_code("e74419e5b0847287")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

AttentionSuper Not yet run
unverified (harvested, no recorded run)
tinyvision/prenas/model/supernet_transformer.py
code served (permissive licence) · get_code("707554979346fab5")
How it ran

no invocation recorded.

TransformerEncoderLayer Not yet run
unverified (harvested, no recorded run)
tinyvision/prenas/model/supernet_transformer.py
code served (permissive licence) · get_code("a17eb310e7b4dff6")
How it ran

no invocation recorded.

Vision_TransformerSuper Not yet run
unverified (harvested, no recorded run)
tinyvision/prenas/model/supernet_transformer.py
code served (permissive licence) · get_code("f1400f07fc9772f7")
How it ran

no invocation recorded.

build_dataset Not yet run
unverified (harvested, no recorded run)
tinyvision/PreNAS/lib/datasets.py
pointer only (licence: Apache-2.0) · get_code("afa7efdfbea334df")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

get_logger Not yet run
unverified (harvested, no recorded run)
alibaba/lightweight-neural-architecture-search/modelscope/utils/logger.py
code served (permissive licence) · get_code("21b93c8efc0f5b5c")
How it ran

no invocation recorded.

is_master Not yet run
unverified (harvested, no recorded run)
tinyvision/PreNAS/lib/score_maker.py
pointer only (licence: Apache-2.0) · get_code("9680c55c004e3572")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

unwrap_model Not yet run
unverified (harvested, no recorded run)
tinyvision/PreNAS/lib/score_maker.py
pointer only (licence: Apache-2.0) · get_code("2bf62c0287050bdd")
How it ran

The body of this sample is withheld: Apache-2.0, as GitHub detects it on this paper's repository, tinyvision/PreNAS, is in the permissive set, but a licence decision keeps the body a pointer: the reason for that decision is not recorded with it. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. No invocation recorded.

Also present in other papers' code

Of 23 samples lifted from this paper's linked repositories, 23 carry a function identity and 0 do not. Of the 23 distinct functions with an identity, 4 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 4 later) and 19 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
is_master 9 other papers
1 dated earlier, 8 later
2302.14042 (2023-02)
Knowledge-enhanced Visual-Language Pre-training on Chest Radiology Images
2308.05734 (2023-08)
AudioLDM 2: Learning Holistic Audio Generation with Self-supervised Pretraining
2311.17048 (2023-11)
Zero-shot Referring Expression Comprehension via Structural Similarity Between Images and Captions
and 6 more not named here
unwrap_model 4 other papers
3 later, 1 same year, order not determinable
Yang_Attentive_Mask_CLIP_ICCV_2023_paper (2023 (year only))
Attentive Mask CLIP
2308.05734 (2023-08)
AudioLDM 2: Learning Holistic Audio Generation with Self-supervised Pretraining
2403.16831 (2024-03)
UrbanVLP: Multi-Granularity Vision-Language Pretraining for Urban Socioeconomic Indicator Prediction
and 1 more not named here
decode_cand_tuple 2 other papers
2 later
2310.16898 (2023-10)
MCUFormer: Deploying Vision Transformers on Microcontrollers with Limited Memory
2501.03782 (2025-01)
Vision Transformer Neural Architecture Search for Out-of-Distribution Generalization: Benchmark and Insights
has_method 1 other paper
1 later
2407.09787 (2024-07)
Semi-supervised 3D Object Detection with PatchTeacher and PillarMix

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 wide application of pre-trained models is driving the trend of once-for-all training in oneshot neural architecture search (NAS). However, training within a huge sample space damages the performance of individual subnets and requires much computation to search for an optimal model. In this paper, we present Pre-NAS, a search-free NAS approach that accentuates target models in one-shot training. Specifically, the sample space is dramatically reduced in advance by a zero-cost selector, and weightsharing one-shot training is performed on the preferred architectures to alleviate update conflicts. Extensive experiments have demonstrated that PreNAS consistently outperforms state-ofthe-art one-shot NAS competitors for both Vision Transformer and convolutional architectures, and importantly, enables instant specialization with zero search cost. Our code is available at https://github.com/tinyvision/PreNAS.

For agents

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

get_harvested_code_for_paper("2304.14636")
get_code_for_paper("2304.14636")
have("2304.14636")

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

curl https://syntology.ai/api/ran/2304.14636.json

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

[![Syntology run record](https://syntology.ai/api/ran/2304.14636.svg)](https://syntology.ai/paper/2304.14636)

Connect an agent — have() is free.