SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2310.18685 · NeurIPS · 2023

When Reviewers Lock Horn: Finding Disagreement in Scientific Peer Reviews

Sandeep Kumar, Asif Ekbal

arXiv · PDF · Open in the Atlas

Code that ran

7 of 10 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, 1 violated it and 0 have no contract verdict; 6 failures of our instrument; 3 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/2310.18685.json · badge: /api/ran/2310.18685.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
sandeep82945/contradiction-in-peer-review canonical (classifier label)
classifier basis: census
7 ran · 3 no recorded run
FunctionStatusWhere it lives
append_item_from_dict_to_list Ran
ran; our contract draft was wrong, not the code
sandeep82945/contradiction-in-peer-review/src/utils/list_dict_data_tool.py
code served (permissive licence) · get_code("0c0742a4ea1d9582")
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

torch.manual_seed(0)

# d_list: list of dicts, each with a key_fieldname and append_fieldnames
d_list = [
    {'id': 'key1', 'score': None, 'label': None},
    {'id': 'key2', 'score': None, 'label': None},
    {'id': 'key3', 'score': None, 'label': None},
]

# d_dict: dict mapping keys to dicts containing the fields to append
d_dict = {
    'key1': {'score': [0.9, 0.1], 'label': ['A', 'B']},
    'key2': {'score': [0.5, 0.5], 'label': ['C', 'D']},
}

# key_fieldname: the field name in d_list items that serves as the lookup key
key_fieldname = 'id'

# append_fieldnames: field names to append from d_dict to d_list items
append_fieldnames = ['score', 'label']

ARGS = {
    'd_list': d_list,
    'd_dict': d_dict,
    'key_fieldname': key_fieldname,
    'append_fieldnames': append_fieldnames,
}
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 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
calculate_metrics Ran
ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT)
sandeep82945/contradiction-in-peer-review/src/training_scratch3.py
code served (permissive licence) · get_code("8379fdb8173a8157")
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
from sklearn.metrics import precision_score, recall_score, f1_score, accuracy_score

# Create simple binary classification data
# true_labels: ground truth labels
true_labels = [0, 1, 1, 0, 1, 0, 1, 1]

# predicted_labels: model predictions
predicted_labels = [0, 1, 1, 0, 0, 0, 1, 1]

# Define ARGS dict with all parameters
ARGS = {
    'true_labels': true_labels,
    'predicted_labels': predicted_labels,
    'average': 'binary'
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, sklearn — 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
count_F1_manual Ran
ran; our contract draft was wrong, not the code
sandeep82945/contradiction-in-peer-review/src/training_scratch3.py
code served (permissive licence) · get_code("96627f234d4de785")
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.

from sklearn.metrics import precision_recall_fscore_support

# Create ground truth list with uid and label
gt_list = [
    {'uid': 'sample_0', 'label': 0},
    {'uid': 'sample_1', 'label': 1},
    {'uid': 'sample_2', 'label': 0},
    {'uid': 'sample_3', 'label': 1},
]

# Create predictions list with uid and predicted_label
pred_list = [
    {'uid': 'sample_0', 'predicted_label': 0},
    {'uid': 'sample_1', 'predicted_label': 1},
    {'uid': 'sample_2', 'predicted_label': 1},
    {'uid': 'sample_3', 'predicted_label': 1},
]

ARGS = {
    'gt_list': gt_list,
    'pred_list': pred_list,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
sklearn — 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
dict_to_list Ran
ran; our contract draft was wrong, not the code
sandeep82945/contradiction-in-peer-review/src/utils/list_dict_data_tool.py
code served (permissive licence) · get_code("5c0d2b710a43c732")
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. 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)

# The function dict_to_list takes a dictionary and iterates through its items,
# appending values to a list. It requires a dict parameter.

d_dict = {
    'key1': 42,
    'key2': 'value2',
    'key3': [1, 2, 3],
    'key4': {'nested': 'dict'}
}

ARGS = {
    'd_dict': d_dict
}
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 list — a list 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
has_tensor Ran
ran, violated the contract we drafted
sandeep82945/contradiction-in-peer-review/src/flint/data_utils/batchbuilder.py
code served (permissive licence) · get_code("c71aafcb215c5122")
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)

# Test case 1: A simple tensor
obj1 = torch.randn(2, 4)

# Test case 2: A dict containing a tensor
obj2 = {
    "tensor": torch.randn(2, 4),
    "value": 42
}

# Test case 3: A list containing tensors
obj3 = [torch.randn(2, 4), "string", 123]

# Test case 4: A tuple containing tensors
obj4 = (torch.randn(2, 4), "data")

# Test case 5: Nested structure with tensors deep inside
obj5 = {
    "nested": {
        "list": [torch.randn(2, 4), 100],
        "value": "test"
    },
    "other": 999
}

# Test case 6: No tensors at all
obj6 = {
    "a": 1,
    "b": [2, 3, "four"],
    "c": (5, 6)
}

# Test case 7: Empty structures
obj7 = []

# Test case 8: Complex nested without tensors
obj8 = {"x": [1, 2], "y": (3, 4), "z": {"nested": "value"}}

ARGS = {
    "obj": obj1
}
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
list_to_dict Ran
ran; our contract draft was wrong, not the code
sandeep82945/contradiction-in-peer-review/src/utils/list_dict_data_tool.py
code served (permissive licence) · get_code("66cbfd6d3a76a33d")
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
import numpy as np

torch.manual_seed(0)

# Construct arguments for list_to_dict function
# The function takes a list of dictionaries and a key field name
# It creates a new dict using the value of key_fields as the key for each item

# Create a list of dictionaries
d_list = [
    {"id": "a", "name": "Alice", "value": 10},
    {"id": "b", "name": "Bob", "value": 20},
    {"id": "c", "name": "Charlie", "value": 30},
]

# The key_fields parameter should be a string that exists as a key in each dictionary item
key_fields = "id"

ARGS = {
    "d_list": d_list,
    "key_fields": key_fields,
}
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 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
move_to_device Ran
ran; our contract draft was wrong, not the code
sandeep82945/contradiction-in-peer-review/src/flint/data_utils/batchbuilder.py
code served (permissive licence) · get_code("7ebeb1ae8084fb58")
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. 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

torch.manual_seed(0)

# Test case 1: obj with tensors, cuda_device < 0 (should return as-is)
obj_with_tensors = {
    "tensor": torch.randn(2, 4),
    "nested": {
        "list": [torch.randn(3, 5), "string"],
        "tuple": (torch.randn(1, 2), 42)
    }
}

# Test case 2: obj without tensors
obj_without_tensors = {
    "string": "hello",
    "number": 123,
    "list": [1, 2, 3],
    "tuple": ("a", "b")
}

# Test case 3: plain tensor
plain_tensor = torch.randn(2, 4)

# Test case 4: list of tensors and scalars
mixed_list = [torch.randn(2, 3), 10, "text", torch.randn(4, 5)]

# Test case 5: tuple of tensors
mixed_tuple = (torch.randn(1, 4), torch.randn(2, 3), "constant")

ARGS = {
    "obj": obj_with_tensors,
    "cuda_device": -1
}
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
json_dumps Not yet run
unverified (harvested, no recorded run)
sandeep82945/contradiction-in-peer-review/src/utils/common.py
code served (permissive licence) · get_code("0711b3689395f3a7")
How it ran

no invocation recorded.

json_loads Not yet run
unverified (harvested, no recorded run)
sandeep82945/contradiction-in-peer-review/src/utils/common.py
code served (permissive licence) · get_code("821813f4d585524f")
How it ran

no invocation recorded.

unserialize_JsonableObject Not yet run
unverified (harvested, no recorded run)
sandeep82945/contradiction-in-peer-review/src/utils/common.py
code served (permissive licence) · get_code("f0ff1b43a0f5291d")
How it ran

no invocation recorded.

Also present in other papers' code

Of 10 samples lifted from this paper's linked repositories, 10 carry a function identity and 0 do not. Of the 10 distinct functions with an identity, 1 are also present in other papers whose code we hold (1 with at least one paper dated earlier, 0 the same month, 0 later) and 9 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
has_tensor 1 other paper
1 dated earlier
2112.01640 (2021-12)
MultiVerS: Improving scientific claim verification with weak supervision and full-document context

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

To this date, the efficacy of the scientific publishing enterprise fundamentally rests on the strength of the peer review process. The journal editor or the conference chair primarily relies on the expert reviewers' assessment, identify points of agreement and disagreement and try to reach a consensus to make a fair and informed decision on whether to accept or reject a paper. However, with the escalating number of submissions requiring review, especially in toptier Artificial Intelligence (AI) conferences, the editor/chair, among many other works, invests a significant, sometimes stressful effort to mitigate reviewer disagreements. Here in this work, we introduce a novel task of automatically identifying contradictions among reviewers on a given article. To this end, we introduce Con-traSciView, a comprehensive review-pair contradiction dataset on around 8.5k papers (with around 28k review pairs containing nearly 50k review pair comments) from the open reviewbased ICLR and NeurIPS conferences. We further propose a baseline model that detects contradictory statements from the review pairs. To the best of our knowledge, we make the first attempt to identify disagreements among peer reviewers automatically. We make our dataset and code public for further investigations 1 .

For agents

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

get_harvested_code_for_paper("2310.18685")
get_code_for_paper("2310.18685")
have("2310.18685")

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

curl https://syntology.ai/api/ran/2310.18685.json

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

[![Syntology run record](https://syntology.ai/api/ran/2310.18685.svg)](https://syntology.ai/paper/2310.18685)

Connect an agent — have() is free.