SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 136960462 · ECCV · 2022

Classification-Regression for Chart Comprehension

Matan Levy, Rami Ben-Ari, Dani Lischinski

Code that ran

20 of 23 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; 2 honoured the contract we drafted, 0 violated it and 17 have no contract verdict; 1 failure of our instrument; 3 no recorded run); 1 of them was also run on the host, not in the jail, by the peer-agreement lane that set its level, 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/136960462.json · badge: /api/ran/136960462.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
levymsn/CQA-CRCT canonical (classifier label)
classifier basis: census
20 ran · 3 no recorded run
FunctionStatusWhere it lives
swish Ran
ran; our contract draft was wrong, not the code
levymsn/CQA-CRCT/CRCT/backbone/vilbert.py
code served (permissive licence) · get_code("0f786c407fb1ee4c")
How it ran

Ran in Syntology's jail on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1) 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. The comparison behind this sample's level is a different run: the peer-agreement lane ran it again on the host, not in the jail (report dated 2026-09-04T22:33:52Z). 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 swish takes a single parameter 'x' which is used with:
# - torch.sigmoid(x) - requires a tensor
# - x * ... - requires a tensor that supports multiplication
# So x should be a tensor

x = torch.randn(2, 4)

ARGS = {
    "x": x
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1
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 (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
OCR_oracle Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/feature_extraction/bbox_match.py
code served (permissive licence) · get_code("bdbd56ae5703372f")
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 numpy as np
import scipy.optimize

# Create sample bounding boxes in format [x1, y1, x2, y2]
# Ground truth bounding boxes (3 boxes)
gt_bboxes = np.array([
    [10, 10, 50, 50],
    [100, 100, 150, 150],
    [200, 200, 250, 250]
], dtype=np.float32)

# Predicted bounding boxes (2 boxes)
pred_bboxes = np.array([
    [12, 12, 48, 48],
    [105, 105, 145, 145]
], dtype=np.float32)

# Predicted class labels (matching number of predicted boxes)
pred_class = np.array([1, 2], dtype=np.int32)

# Ground truth text labels (matching number of ground truth boxes)
gt_text = np.array(['hello', 'world', 'test'], dtype=object)

# Create feature dictionaries
pred_feats = {
    'bboxes': pred_bboxes,
    'class': pred_class
}

gt_feats = {
    'bboxes': gt_bboxes,
    'text': gt_text
}

ARGS = {
    'pred_feats': pred_feats,
    'gt_feats': gt_feats
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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
bbox_iou Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/feature_extraction/bbox_match.py
code served (permissive licence) · get_code("a9517dd6b8183083")
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 float, 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)

# bbox_iou expects two bounding boxes in format [x1, y1, x2, y2]
# where (x1, y1) is top-left and (x2, y2) is bottom-right

boxA = [10, 20, 50, 60]  # x1=10, y1=20, x2=50, y2=60
boxB = [30, 40, 70, 80]  # x1=30, y1=40, x2=70, y2=80

ARGS = {
    "boxA": boxA,
    "boxB": boxB
}
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 float, 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
color Ran
ran on a synthesized input
levymsn/CQA-CRCT/CRCT/Interactive_demo.py
code served (permissive licence) · get_code("d801abdee9495f30")
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.

from colorama import Fore, Back, Style

# The function takes three string parameters: text, fore, and back
# fore and back are color codes (likely from colorama.Fore and colorama.Back)
# text is the string to be colored

ARGS = {
    'text': 'Hello, World!',
    'fore': Fore.RED,
    'back': Back.YELLOW
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
colorama — 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
encode_input Ran
ran on a synthesized input
levymsn/CQA-CRCT/CRCT/utils.py
code served (permissive licence) · get_code("75cd536b57d434c1")
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 torch
import random

torch.manual_seed(0)
random.seed(0)

# Analyze the function:
# - utterances: list of lists of token ids (integers)
# - start_segment: integer (0 or 1)
# - CLS, SEP, MASK: integers (special token ids)
# - max_seq_len: integer, default 256
# - max_sep_len: integer, default 25
# - mask_prob: float, default 0.2

# The function processes utterances as lists of token ids
# Each utterance is a list of integers
# We need small utterances to keep tensors tiny

ARGS = {
    "utterances": [
        [101, 102, 103],           # First utterance: 3 tokens
        [104, 105],                # Second utterance: 2 tokens
    ],
    "start_segment": 0,
    "CLS": 100,
    "SEP": 102,
    "MASK": 103,
    "max_seq_len": 64,
    "max_sep_len": 10,
    "mask_prob": 0.2,
}
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 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
encode_text_input Ran
ran on a synthesized input
levymsn/CQA-CRCT/CRCT/utils.py
code served (permissive licence) · get_code("db3e6aedce62ff71")
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 torch
import random
import numpy as np

torch.manual_seed(0)
random.seed(0)

# Parameters for encode_text_input
utterances = [
    [101, 2054, 2003, 102],  # First utterance with 4 tokens
    [1045, 2572, 102]         # Second utterance with 3 tokens
]

locations = [
    [10.5, 20.3, 30.1, 40.2],  # Single location for first utterance
    [[5.1, 15.2, 25.3, 35.4], [6.2, 16.3, 26.4, 36.5], [7.3, 17.4, 27.5, 37.6]]  # List of locations for second utterance
]

token_types = [0, -1]  # Segment IDs for each utterance

CLS = 101  # BERT CLS token ID
SEP = 102  # BERT SEP token ID
MASK = 103  # BERT MASK token ID

max_seq_len = 256
max_sep_len = 50
mask_prob = 0.15

ARGS = {
    'utterances': utterances,
    'locations': locations,
    'token_types': token_types,
    'CLS': CLS,
    'SEP': SEP,
    'MASK': MASK,
    'max_seq_len': max_seq_len,
    'max_sep_len': max_sep_len,
    'mask_prob': mask_prob
}
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 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
est Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/extract_features.py
code served (permissive licence) · get_code("1afc750dfbf4c84d")
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 time

def sec_to_hhmmss(secs):
    return time.strftime('[%H:%M:%S]', time.gmtime(secs))

def est(num_ready, total, total_time):
    """  Returns estimated time for the process"""
    avg_time = total_time / num_ready
    return sec_to_hhmmss((total - num_ready) * avg_time)

ARGS = {
    'num_ready': 5,
    'total': 20,
    'total_time': 120.0
}
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
fig_type_to_id Ran
ran on a synthesized input
levymsn/CQA-CRCT/CRCT/fig_dataloader.py
code served (permissive licence) · get_code("2bda61dcb74571d3")
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 torch.nn as nn

ARGS = {
    'str_type': 'line'
}
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 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
forward Ran
ran on a synthesized input
levymsn/CQA-CRCT/CRCT/backbone/encoder_decorator.py
code served (permissive licence) · get_code("c18bcab3fca0524e")
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. 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
import numpy as np
from torch.autograd import Variable

torch.manual_seed(0)

# Create a minimal dialog_encoder that mimics the expected interface
class MockDialogEncoder(nn.Module):
    def __init__(self):
        super().__init__()
        self.linear = nn.Linear(10, 5)
    
    def forward(self, tokens, txt_loc, features, image_loc, sep_indices=None, sep_len=None, 
                token_type_ids=None, masked_lm_labels=None, attention_mask=None, 
                next_sentence_label=None, output_nsp_scores=False, output_lm_scores=False,
                image_attention_mask=None, image_label=None, image_target=None, 
                gt_reg=None, areas=None):
        batch_size = tokens.shape[0]
        # Return tuple: (lm_loss, img_loss, nsp_loss, nsp_scores, regression, legend_loss)
        lm_loss = torch.tensor(0.5)
        img_loss = torch.tensor(0.3)
        nsp_loss = torch.tensor(0.2)
        nsp_scores = torch.randn(batch_size, 2)
        # regression should be a tuple with at least: (output, loss, something, right_count)
        regression = (torch.randn(batch_size, 4), torch.randn(batch_size), torch.tensor(0.1), torch.tensor(5))
        legend_loss = torch.tensor(0.1)
        return (lm_loss, img_loss, nsp_loss, nsp_scores, regression, legend_loss)

# Create batch dict with all required keys
batch_size = 2
seq_len = 8
img_feat_dim = 4
num_images = 3

batch = {
    'tokens': torch.randint(0, 100, (batch_size, seq_len)),
    'loc': torch.randint(0, 10, (batch_size, seq_len)),
    'segments': torch.randint(0, 2, (batch_size, seq_len)),
    'sep_indices': torch.randint(0, seq_len, (batch_size, 5)),
    'mask': torch.randint(0, 2, (batch_size, seq_len)),
    'hist_len': torch.randint(0, 5, (batch_size,)),
    'image_feat': torch.randn(batch_size, num_images, img_feat_dim),
    'image_loc': torch.randn(batch_size, num_images, 4),
    'image_mask': torch.randint(0, 2, (batch_size, num_images)),
    'R': torch.randn(batch_size),
    'areas': torch.randn(batch_size, num_images),
    'next_sentence_labels': torch.randint(0, 2, (batch_size,)),
    'image_label': torch.randint(0, 2, (batch_size, num_images)),
    'image_target': torch.randn(batch_size, num_images, 5),
}

# Create params dict
params = {
    'device': 'cpu',
    'nsp_loss_coeff': 1.0,
    'reg_loss_coeff': 1.0,
}

# Create the dialog_encoder model
dialog_encoder = MockDialogEncoder()

ARGS = {
    'dialog_encoder': dialog_encoder,
    'batch': batch,
    'params': params,
    'output_nsp_scores': False,
    'output_lm_scores': False,
    'evaluation': False,
    'sample_ids': None,
}
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 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
gelu Ran
ran, honoured the contract we drafted
levymsn/CQA-CRCT/CRCT/backbone/vilbert.py
pointer only (licence: MIT); the stored copy get_code serves, from tanmay1618/text_to_sql_bert, is pointer only (licence: NONE) · get_code("fdc64f4c72036ae4")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from tanmay1618/text_to_sql_bert, 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.

get_box_area Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/feature_extraction/pie_area.py
code served (permissive licence) · get_code("6466a4bda3a5583c")
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 numpy as np

class Point:
    def __init__(self, x, y):
        self.x = float(x)
        self.y = float(y)

    def __str__(self):
        return f'({self.x}, {self.y})'

def get_minibox_area(x, y, r):
    """Assuming the center of the circle is at point (0,0) - top left."""
    if x <= 0 or y <= 0:
        return 0
    if np.sqrt(x ** 2 + y ** 2) < r:
        return 0
    sqrt = np.sqrt(r ** 2 - y ** 2) if r ** 2 - y ** 2 > 0 else 0
    q1 = Point(sqrt, y)
    sqrt = np.sqrt(r ** 2 - x ** 2) if r ** 2 - x ** 2 > 0 else 0
    q2 = Point(x, sqrt)
    return np.arctan2(q1.y, q1.x) - np.arctan2(q2.y, q2.x)

def get_box_area(p1, p2, r, center):
    n_p1 = Point(-(center.x - p1.x), center.y - p1.y)
    n_p2 = Point(p2.x - center.x, -(p2.y - center.y))
    boxes_xy = [(abs(n_p2.x), abs(n_p2.y)), (abs(n_p1.y), abs(n_p2.x)), (abs(n_p1.x), abs(n_p1.y)), (abs(n_p2.y), abs(n_p1.x))]
    return sum([get_minibox_area(box[0], box[1], r) for box in boxes_xy])

# Construct valid arguments for get_box_area
p1 = Point(x=2.0, y=3.0)
p2 = Point(x=5.0, y=7.0)
r = 10.0
center = Point(x=4.0, y=4.0)

ARGS = {
    'p1': p1,
    'p2': p2,
    'r': r,
    'center': center
}
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 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
get_data_lst Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/frcnn.py
code served (permissive licence) · get_code("c4bc90aeceaf6d34")
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 list, 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 numpy as np

ARGS = {
    'dataset': 'plotqa'
}
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 list, ndim 2, dtype family float — a list 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
get_input Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/extract_features.py
code served (permissive licence) · get_code("4c2fad57f5285a05")
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 list, 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 torch
import numpy as np
from unittest.mock import Mock, MagicMock
import tempfile
import os
import cv2

torch.manual_seed(0)

# Create a temporary test image
temp_dir = tempfile.mkdtemp()
img_path = os.path.join(temp_dir, "test_image.jpg")
test_image = np.random.randint(0, 256, (480, 640, 3), dtype=np.uint8)
cv2.imwrite(img_path, test_image)

# Create a mock transform object
class MockTransform:
    def apply_image(self, image):
        return image

class MockAugmentor:
    def get_transform(self, image):
        return MockTransform()

class MockPredictor:
    def __init__(self):
        self.aug = MockAugmentor()

# Construct args
args = Mock()
args.predictor = MockPredictor()

ARGS = {
    'args': args,
    'img_path': img_path
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
cv2, numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list 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
get_minibox_area Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/feature_extraction/pie_area.py
code served (permissive licence) · get_code("c4e163630a944c0c")
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 numpy as np

# The function get_minibox_area takes three parameters: x, y, and r
# All are used in arithmetic operations (sqrt, comparison, arctan2)
# They should be numeric values (int or float)

x = 5.0
y = 3.0
r = 7.0

ARGS = {
    'x': x,
    'y': y,
    'r': r
}
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 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
get_prev_csvs Ran
ran, honoured the contract we drafted
levymsn/CQA-CRCT/CRCT/evaluation.py
code served (permissive licence) · get_code("47060fb467fa6fd4")
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 1, dtype family int, shape (6,) 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 tempfile
import os
import pandas as pd
import glob

# Create a temporary directory for test CSV files
temp_dir = tempfile.mkdtemp()

# Create sample CSV files matching the expected pattern
save_path = temp_dir
eval_set = "validation"
start_checkpoint = "/path/to/checkpoint_5"

# Create a few CSV files with the expected naming pattern
csv_pattern = f"{save_path}/eval_results_{eval_set}_{start_checkpoint.split('/')[-1]}_*.csv"

# Create sample CSV files
for i in range(2):
    df = pd.DataFrame({
        'qa_ind': [10 + i, 20 + i, 30 + i],
        'score': [0.8, 0.9, 0.7]
    })
    csv_file = f"{save_path}/eval_results_{eval_set}_{start_checkpoint.split('/')[-1]}_{i}.csv"
    df.to_csv(csv_file, index=False)

ARGS = {
    'params': {
        'save_path': save_path,
        'eval_set': eval_set,
        'start_checkpoint': start_checkpoint
    }
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
pandas — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type ndarray, ndim 1, dtype family int, shape (6,) — shape (6,) 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
list2tensorpad Ran
ran on a synthesized input
levymsn/CQA-CRCT/CRCT/utils.py
code served (permissive licence) · get_code("28aaa97940d13fbd")
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 int, shape (1, 8) 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 takes:
# 1. inp_list: a list that will be converted to a LongTensor
# 2. max_seq_len: an integer specifying the maximum sequence length for padding

# Looking at the function body:
# - inp_list is converted to torch.LongTensor([inp_list]), so it should be a list of integers
# - max_seq_len is used in slicing operations, so it should be an integer

inp_list = [1, 2, 3, 4, 5]
max_seq_len = 8

ARGS = {
    "inp_list": inp_list,
    "max_seq_len": max_seq_len,
}
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 int, shape (1, 8) — shape (1, 8) 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
match_bboxes Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/feature_extraction/bbox_match.py
code served (permissive licence) · get_code("a92ad5f7ea925a5b")
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
import scipy.optimize

# Create small sample bounding boxes
# Format: [x1, y1, x2, y2]
bbox_gt = np.array([
    [10, 10, 50, 50],
    [100, 100, 150, 150]
], dtype=np.float32)

bbox_pred = np.array([
    [12, 12, 48, 48],
    [105, 105, 155, 155],
    [200, 200, 250, 250]
], dtype=np.float32)

IOU_THRESH = 0.5

ARGS = {
    "bbox_gt": bbox_gt,
    "bbox_pred": bbox_pred,
    "IOU_THRESH": IOU_THRESH
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, scipy — 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
num_sides_intersections Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/feature_extraction/pie_area.py
code served (permissive licence) · get_code("f604ba3b7ca1c63f")
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 int64, 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 numpy as np

class Point:
    def __init__(self, x, y):
        self.x = x
        self.y = y

# Create two boxes, each with two points (corner1, corner2)
box1 = [Point(0.0, 0.0), Point(10.0, 10.0)]
box2 = [Point(0.1, 0.0), Point(10.1, 10.0)]

ARGS = {
    "box1": box1,
    "box2": box2
}
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 int64, 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
sec_to_hhmmss Ran
ran on a synthesized input
levymsn/CQA-CRCT/Detector/extract_features.py
code served (permissive licence) · get_code("88f9fb967f836af4")
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 time

def sec_to_hhmmss(secs):
    return time.strftime('[%H:%M:%S]', time.gmtime(secs))

ARGS = {
    'secs': 3661
}
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
sequence_mask Ran
ran on a synthesized input
levymsn/CQA-CRCT/CRCT/backbone/encoder_decorator.py
code served (permissive licence) · get_code("fefc09b926e59cdf")
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, 5) as a later run recorded it. 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, 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
from torch.autograd import Variable

torch.manual_seed(0)

# Create a sequence_length tensor with batch_size=2
# Each element represents the length of a sequence in that batch
sequence_length = torch.tensor([3, 5], dtype=torch.long)

# Create ARGS dict with all parameters
ARGS = {
    "sequence_length": sequence_length,
    "max_len": None
}
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 (2, 5) — shape (2, 5) 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
get_class_list Not yet run
unverified (harvested, no recorded run)
levymsn/CQA-CRCT/Detector/frcnn.py
code served (permissive licence) · get_code("bec4326b26874da0")
How it ran

no invocation recorded.

load_tf_weights_in_bert Not yet run
unverified (harvested, no recorded run)
levymsn/CQA-CRCT/CRCT/backbone/vilbert.py
code served (permissive licence) · get_code("63126a3d0d53cf84")
How it ran

no invocation recorded.

read_command_line Not yet run
unverified (harvested, no recorded run)
levymsn/CQA-CRCT/CRCT/options.py
code served (permissive licence) · get_code("094f5e6b965929f5")
How it ran

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, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 2 later) and 20 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
swish 206 other papers
90 dated earlier, 79 later, 37 same year, order not determinable · 44 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1805.12114 (2018-05)
Deep Reinforcement Learning in a Handful of Trials using Probabilistic Dynamics Models
1809.01357 (2018-09)
Zero Shot Learning for Code Education: Rubric Sampling with Deep Learning Inference
and 203 more not named here
gelu 198 other papers
96 dated earlier, 69 later, 33 same year, order not determinable · 32 only as an assembly brick
1408.5882 (2014-08)
Convolutional Neural Networks for Sentence Classification
1810.04805 (2018-10)
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
1901.10125 (2019-01)
Glyce: Glyph-vectors for Chinese Character Representations
and 195 more not named here
get_prev_csvs 1 other paper
1 dated earlier
2111.14792 (2021-11)
Classification-Regression for Chart Comprehension

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

Chart question answering (CQA) is a task used for assessing chart comprehension, which is fundamentally different from understanding natural images. CQA requires analyzing the relationships between the textual and the visual components of a chart, in order to answer general questions or infer numerical values. Most existing CQA datasets and models are based on simplifying assumptions that often enable surpassing human performance. In this work, we address this outcome and propose a new model that jointly learns classification and regression. Our language-vision setup uses co-attention transformers to capture the complex real-world interactions between the question and the textual elements. We validate our design with extensive experiments on the realistic PlotQA dataset, outperforming previous approaches by a large margin, while showing competitive performance on FigureQA. Our model is particularly well suited for realistic questions with out-of-vocabulary answers that require regression.

For agents

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

get_harvested_code_for_paper("136960462")
get_code_for_paper("136960462")
have("136960462")

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

curl https://syntology.ai/api/ran/136960462.json

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

[![Syntology run record](https://syntology.ai/api/ran/136960462.svg)](https://syntology.ai/paper/136960462)

Connect an agent — have() is free.