SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2009.09405 · 2020

Scale-Localized Abstract Reasoning

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 (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 0 violated it and 11 have no contract verdict; 2 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/2009.09405.json · badge: /api/ran/2009.09405.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
yanivbenny/MRNet canonical (classifier label)
classifier basis: census
7 ran · 2 no recorded run
yanivbenny/RAVEN_FAIR canonical (classifier label)
classifier basis: census
6 ran · 2 no recorded run
FunctionStatusWhere it lives
calculate_acc Ran
ran on a synthesized input
yanivbenny/MRNet/src/criteria.py
code served (permissive licence) · get_code("7c4fa50082da02b4")
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 float64, ndim 0, 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

torch.manual_seed(0)

# Create output tensor: logits with shape (batch_size=2, num_classes=4)
# The function calls output.data.max(1)[1] to get predictions
output = torch.randn(2, 4)

# Create target tensor: class indices with shape (batch_size=2,)
# The function calls target.data and target.size()[0]
target = torch.tensor([1, 3])

ARGS = {
    "output": output,
    "target": target
}
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 float64, ndim 0, dtype family float — no output shape is stated: observe_signature.py recorded float ndim 0, which disagrees with the fixture run
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
calculate_correct Ran
ran on a synthesized input
yanivbenny/MRNet/src/criteria.py
code served (permissive licence) · get_code("c673f412e11d09bb")
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 ndarray, ndim 0, dtype family int, 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
import torch.nn as nn

torch.manual_seed(0)

# The function expects:
# - output: a tensor with .data attribute and max() method that returns (values, indices)
# - target: a tensor with .data attribute
# Both need to support .eq(), .cpu(), .sum(), and .numpy()

# Create output tensor: batch_size=2, num_classes=4
# max(1) returns top value and index along dimension 1
output = torch.randn(2, 4)

# Create target tensor: batch_size=2 with class indices
target = torch.tensor([0, 2])

ARGS = {
    "output": output,
    "target": target
}
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 ndarray, ndim 0, dtype family int, 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
check_consistency Ran
ran on a synthesized input
yanivbenny/RAVEN_FAIR/src/solver.py
code served (permissive licence) · get_code("dd4d64202796663f")
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
import torch.nn as nn

torch.manual_seed(0)

# Create a small stand-in class for attribute objects with get_value_level() method
class Attribute:
    def __init__(self, value_level):
        self._value_level = value_level
    
    def get_value_level(self):
        return self._value_level

# Create entity objects with attributes
class Entity:
    def __init__(self, attr_value):
        self.color = Attribute(attr_value)
        self.size = Attribute(attr_value)
        self.shape = Attribute(attr_value)

# Build the nested structure: candidate -> children[0] -> children[component_idx] -> children[0] (layout) -> children (entities)
class Container:
    def __init__(self, children):
        self.children = children

# Create entities with consistent attribute values
entity_0 = Entity(attr_value=1)
entity_1 = Entity(attr_value=1)
entity_2 = Entity(attr_value=1)

# Create layout (container of entities)
layout = Container(children=[entity_0, entity_1, entity_2])

# Create component (container with layout as first child)
component = Container(children=[layout])

# Create the main structure with multiple components
component_0 = Container(children=[layout])
component_1 = Container(children=[layout])

# Create candidate with components nested inside another layer
candidate = Container(children=[Container(children=[component_0, component_1])])

ARGS = {
    "candidate": candidate,
    "attr": "color",
    "component_idx": 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 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
check_num_pos Ran
ran on a synthesized input
yanivbenny/RAVEN_FAIR/src/solver.py
code served (permissive licence) · get_code("1dd8a5c97d899a22")
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 int, ndim 0, dtype family int, 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
import numpy as np

# Stand-in classes to represent the AoT (Analogy on Tests) node structure
class ValueAttribute:
    """Mock attribute that can return indices or levels"""
    def __init__(self, indices, level=None):
        self.indices = indices
        self.level = level
        self.values = list(range(max(indices) + 1)) if indices else []
    
    def get_value_idx(self):
        return self.indices
    
    def get_value_level(self):
        return self.level
    
    def get_value(self):
        return sum(self.indices) if self.indices else 0

class Layout:
    """Mock layout node with position and number attributes"""
    def __init__(self, position_indices, position_level, number_indices, number_level):
        self.position = ValueAttribute(position_indices, position_level)
        self.number = ValueAttribute(number_indices, number_level)
    
    def __repr__(self):
        return f"Layout(pos={self.position.get_value_idx()}, num={self.number.get_value_idx()})"

class Component:
    """Mock component node"""
    def __init__(self, layout):
        self.children = [layout]
    
    def __getitem__(self, idx):
        return self.children[idx]

class Figure:
    """Mock figure node with components"""
    def __init__(self, components):
        self.children = [components]
    
    def __getitem__(self, idx):
        return self.children[idx]

class AoTNode:
    """Mock AoT node with children"""
    def __init__(self, figure):
        self.children = [figure]
    
    def __getitem__(self, idx):
        return self.children[idx]

class Rule:
    """Mock rule object"""
    def __init__(self, name, component_idx=0, attr='Number', value=1, value_levels=None):
        self.name = name
        self.component_idx = component_idx
        self.attr = attr
        self.value = value
        self.value_levels = value_levels or {2: [[1], [2], [3]]}

# Build the context list with 8 AoTNode objects
context = []
for i in range(8):
    # Create layouts with varying position and number values
    pos_indices = [i % 3, (i + 1) % 3]
    pos_level = i % 2
    num_indices = [i % 2]
    num_level = i
    
    layout = Layout(pos_indices, pos_level, num_indices, num_level)
    component = Component(layout)
    figure = Figure(component)
    node = AoTNode(figure)
    context.append(node)

# Build candidate AoTNode
candidate_pos_indices = [1, 2]
candidate_pos_level = 0
candidate_num_indices = [0]
candidate_num_level = 3

candidate_layout = Layout(candidate_pos_indices, candidate_pos_level, 
                          candidate_num_indices, candidate_num_level)
candidate_component = Component(candidate_layout)
candidate_figure = Figure(candidate_component)
candidate = AoTNode(candidate_figure)

# Create a rule (using Constant rule type for simplicity)
rule_num_pos = Rule(name='Constant', component_idx=0, attr='Number')

ARGS = {
    'rule_num_pos': rule_num_pos,
    'context': context,
    'candidate': candidate
}
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 int, ndim 0, dtype family int, 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
contrast_loss Ran
ran on a synthesized input
yanivbenny/MRNet/src/criteria.py
code served (permissive licence) · get_code("c6825a5c045307ea")
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 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
import torch.nn.functional as F

torch.manual_seed(0)

# Create small tensors for testing
batch_size = 2
num_classes = 4

# output: logits tensor, shape (batch_size, num_classes)
output = torch.randn(batch_size, num_classes)

# target: class indices tensor, shape (batch_size,)
target = torch.tensor([0, 2])

# reduction parameter
reduction = 'mean'

# weighted_loss parameter
weighted_loss = False

ARGS = {
    'output': output,
    'target': target,
    'reduction': reduction,
    'weighted_loss': weighted_loss
}
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 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
conv1x1 Ran
ran; our contract draft was wrong, not the code
yanivbenny/MRNet/src/networks/blocks.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from visiontao/dcnet, is pointer only (licence: NONE) · get_code("6c693fb7c03c87bd")
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 visiontao/dcnet, 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.

conv3x3 Ran
ran; our contract draft was wrong, not the code
yanivbenny/MRNet/src/networks/blocks.py
pointer only (licence: Apache-2.0); the stored copy get_code serves, from visiontao/dcnet, is pointer only (licence: NONE) · get_code("52601677ce5d1634")
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 visiontao/dcnet, 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.

init_acc_regime Ran
ran on a synthesized input
yanivbenny/MRNet/src/report_acc_regime.py
code served (permissive licence) · get_code("10ed5d2031a9bc68")
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 dict. 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)

ARGS = {
    'dataname': 'RAVEN'
}
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 dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict 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
n_tree_serialize Ran
ran on a synthesized input
yanivbenny/RAVEN_FAIR/src/serialize.py
code served (permissive licence) · get_code("8db0b273506cda21")
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 torch
import torch.nn as nn

class TreeNode:
    def __init__(self, name, level='Node', is_pg=True, children=None):
        self.name = name
        self.level = level
        self.is_pg = is_pg
        self.children = children if children is not None else []

# Create a simple tree structure
leaf1 = TreeNode(name='leaf1', level='Leaf', is_pg=True, children=[])
leaf2 = TreeNode(name='leaf2', level='Leaf', is_pg=True, children=[])
intermediate = TreeNode(name='inter', level='Node', is_pg=True, children=[leaf1, leaf2])
root = TreeNode(name='root', level='Layout', is_pg=True, children=[intermediate])

ARGS = {
    'aot': root
}
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 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
renormalize Ran
ran on a synthesized input
yanivbenny/MRNet/src/trainer.py
code served (permissive licence) · get_code("00bd265c90d10459")
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 (2, 4) 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)

# The function takes images and performs normalization:
# 1. Divides by 255 (converts from [0, 255] range to [0, 1])
# 2. Subtracts 0.5 (shifts to [-0.5, 0.5])
# 3. Multiplies by 2 (scales to [-1, 1])
# 
# The input should be a tensor of images with values in range [0, 255]

images = torch.randint(0, 256, (2, 4), dtype=torch.float32)

ARGS = {
    "images": images
}
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 2, dtype family float, shape (2, 4) — shape (2, 4) 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
rle_encode Ran
ran on a synthesized input
yanivbenny/RAVEN_FAIR/src/api.py
code served (permissive licence) · get_code("80366123e633d920")
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 numpy as np

# Create a simple binary mask image (numpy array with 1s and 0s)
# The function expects a 2D or flattened array where 1 = mask, 0 = background
img = np.array([
    [0, 1, 1, 0],
    [0, 1, 1, 0],
    [1, 0, 0, 1]
], dtype=np.uint8)

ARGS = {
    'img': img
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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_attr Ran
ran on a synthesized input
yanivbenny/RAVEN_FAIR/src/sampling.py
code served (permissive licence) · get_code("f07118b01c265872")
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 numpy as np

np.random.seed(0)

# Construct a valid attrs_list for the sample_attr function
# Each element should be: [component_idx, attr_name, available_times, min_level, max_level]
attrs_list = [
    [0, "color", 3, 0, 255],
    [1, "size", 2, 10, 100],
    [2, "opacity", 1, 0, 1],
    [3, "rotation", 5, 0, 360]
]

ARGS = {
    "attrs_list": attrs_list
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — 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
serialize_rules Ran
ran on a synthesized input
yanivbenny/RAVEN_FAIR/src/serialize.py
code served (permissive licence) · get_code("9be696f331d13ff4")
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. 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 numpy as np

class Rule:
    def __init__(self, name, attr):
        self.name = name
        self.attr = attr

# Create rule groups with different rule types and attributes
rule_group_1 = [
    Rule('Constant', 'Number'),
    Rule('Progression', 'Position'),
]

rule_group_2 = [
    Rule('Arithmetic', 'Type'),
    Rule('Distribute_Three', 'Size'),
    Rule('Constant', 'Color'),
    Rule('Progression', 'Number/Position'),
]

rule_group_3 = [
    Rule('Arithmetic', 'Size'),
    Rule('Constant', 'Number'),
]

rule_groups = [rule_group_1, rule_group_2, rule_group_3]

ARGS = {
    'rule_groups': rule_groups
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple — a tuple output has no array shape of its own; none is 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
conv1x1 Not yet run
unverified (harvested, no recorded run)
yanivbenny/MRNet/src/networks/mrnet.py
code served (permissive licence) · get_code("08d63b2cfd3e3139")
How it ran

no invocation recorded.

conv3x3 Not yet run
unverified (harvested, no recorded run)
yanivbenny/MRNet/src/networks/mrnet.py
code served (permissive licence) · get_code("11fda3e1d6f755ae")
How it ran

no invocation recorded.

sample_attr_avail Not yet run
unverified (harvested, no recorded run)
yanivbenny/RAVEN_FAIR/src/sampling.py
code served (permissive licence) · get_code("4f3104a61ba3df12")
How it ran

no invocation recorded.

solve Not yet run
unverified (harvested, no recorded run)
yanivbenny/RAVEN_FAIR/src/solver.py
code served (permissive licence) · get_code("b8c75f90a1932121")
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 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
conv1x1 2 other papers
2 later
2205.13720 (2022-05)
Effective Abstract Reasoning with Dual-Contrast Network
2312.09997 (2023-12)
One Self-Configurable Model to Solve Many Abstract Visual Reasoning Problems
conv3x3 2 other papers
2 later
2205.13720 (2022-05)
Effective Abstract Reasoning with Dual-Contrast Network
2312.09997 (2023-12)
One Self-Configurable Model to Solve Many Abstract Visual Reasoning Problems

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 consider the abstract relational reasoning task, which is commonly used as an intelligence test. Since some patterns have spatial rationales, while others are only semantic, we propose a multi-scale architecture that processes each query in multiple resolutions. We show that indeed different rules are solved by different resolutions and a combined multi-scale approach outperforms the existing state of the art in this task on all benchmarks by 5-54%. The success of our method is shown to arise from multiple novelties. First, it searches for relational patterns in multiple resolutions, which allows it to readily detect visual relations, such as location, in higher resolution, while allowing the lower resolution module to focus on semantic relations, such as shape type. Second, we optimize the reasoning network of each resolution proportionally to its performance, hereby we motivate each resolution to specialize on the rules for which it performs better than the others and ignore cases that are already solved by the other resolutions. Third, we propose a new way to pool information along the rows and the columns of the illustration-grid of the query. Our work also analyses the existing benchmarks, demonstrating that the RAVEN dataset selects the negative examples in a way that is easily exploited. We, therefore, propose a modified version of the RAVEN dataset, named RAVEN-FAIR. Our code and pretrained models are available at https://github.com/yanivbenny/MRNet.

For agents

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

get_harvested_code_for_paper("2009.09405")
get_code_for_paper("2009.09405")
have("2009.09405")

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

curl https://syntology.ai/api/ran/2009.09405.json

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

[![Syntology run record](https://syntology.ai/api/ran/2009.09405.svg)](https://syntology.ai/paper/2009.09405)

Connect an agent — have() is free.