SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2303.14443 · 2023

No more Reviewer #2: Subverting Automatic Paper-Reviewer Assignment using Adversarial Learning

arXiv · PDF · Open in the Atlas

Code that ran

7 of 11 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 7 have no contract verdict; 0 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/2303.14443.json · badge: /api/ran/2303.14443.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
rub-syssec/adversarial-papers canonical (classifier label)
classifier basis not recorded on the link
7 ran · 4 no recorded run
FunctionStatusWhere it lives
clean_dict Ran
ran on a synthesized input
rub-syssec/adversarial-papers/src/utils/attack_utils.py
code served (permissive licence) · get_code("0937897fc19205b4")
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 logging
import copy

# Set up logging
logger = logging.getLogger("test_logger")
logger.setLevel(logging.DEBUG)

# Create a simple requested_changes dictionary with some words
# Some words will have characters with high ord values (like unicode characters)
# and some will have normal ASCII characters
requested_changes = {
    "hello": 1,
    "world": 2,
    "café": 3,  # 'é' has ord value 233
    "münchen": 4,  # 'ü' has ord value 252
    "simple": 5,
    "test": 6,
}

# Set max_ord_value to filter out words with characters having ord > 200
# This will keep "hello", "world", "simple", "test" but remove "café" and "münchen"
max_ord_value = 200

ARGS = {
    "requested_changes": requested_changes,
    "logger": logger,
    "max_ord_value": max_ord_value,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
clean_text Ran
ran on a synthesized input
rub-syssec/adversarial-papers/src/utils/attack_utils.py
code served (permissive licence) · get_code("0f71c17baa894982")
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 logging

# Create a logger for testing
logger = logging.getLogger("test_logger")
logger.setLevel(logging.DEBUG)

# Define arguments for clean_text function
ARGS = {
    "input_text": "Hello\u4e2d\u6587World\uffff",  # Text with mixed ASCII and high unicode characters
    "logger": logger,
    "max_ord_value": 127  # ASCII range, so unicode chars will be removed
}
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
create_candidate_grid Ran
ran on a synthesized input
rub-syssec/adversarial-papers/src/featurespace/grid.py
code served (permissive licence) · get_code("8e279b7552181c6b")
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 random
import torch
import torch.nn as nn
from collections import defaultdict

class MockSurrogateModel:
    def __init__(self):
        self.reviewers_list = ['reviewer_a', 'reviewer_b', 'reviewer_c', 'reviewer_d']
    
    def get_ranking(self, words):
        # Returns list of (reviewer_name, score) tuples
        return [
            ('reviewer_a', 0.9),
            ('reviewer_b', 0.8),
            ('reviewer_c', 0.7),
            ('reviewer_d', 0.6),
        ]
    
    def get_reviewer(self, reviewer_name):
        # Returns (name, topic_probs)
        topic_probs = [0.1 if i in [0, 1, 2] else 0.0 for i in range(50)]
        return (reviewer_name, topic_probs)
    
    def get_topics(self, words):
        # Returns topic probabilities for submission
        topic_probs = [0.2 if i in [0, 1] else 0.0 for i in range(50)]
        return topic_probs


class MockStrategyAggregated:
    def create_topic_list(self, topics):
        return list(topics)[:5]
    
    def init_words(self, key):
        return ['word1', 'word2']


class MockStrategyWordBased:
    def create_topic_list(self, topics):
        return list(topics)[:5]
    
    def init_words(self, key):
        return ['word1', 'word2'] if random.random() > 0.5 else []


class MockSubmission:
    def __init__(self):
        self.words = ['machine', 'learning', 'neural', 'network']
        self.strategies = {
            'aggregated': MockStrategyAggregated(),
            'word_based': MockStrategyWordBased(),
        }


class MockLogger:
    def info(self, msg):
        pass
    
    def debug(self, msg):
        pass


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

surrogate_model = MockSurrogateModel()
target = {
    'request': ['reviewer_a', 'reviewer_b'],
    'reject': ['reviewer_c'],
}
submission = MockSubmission()

config_aggregated = {
    'strategy': 'aggregated',
    'step': 2,
    'all_topics': False,
    'no_successors': 10,
}

ARGS = {
    'surrogate_models': [surrogate_model],
    'target': target,
    'submission': submission,
    'config': config_aggregated,
    'grid_size': 50,
    'logger': MockLogger(),
}
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, 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
find_words Ran
ran on a synthesized input
rub-syssec/adversarial-papers/src/utils/pdf_utils.py
code served (permissive licence) · get_code("055f094867896073")
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 re

def is_digit(word):
    try:
        int(word)
        return True
    except ValueError:
        return False

def find_words(string):
    words = re.findall('(\\w+)', string.lower())
    new_words = []
    for w in words:
        if not is_digit(w) and len(w) > 1:
            new_words.append(w)
    return new_words

ARGS = {
    "string": "Hello World 123 test example_42 a 99 programming"
}
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 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_topics_to_words Ran
ran on a synthesized input
rub-syssec/adversarial-papers/src/utils/lda.py
code served (permissive licence) · get_code("4e6dbc7d248497e2")
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 logging
from collections import defaultdict
import numpy as np

logging.basicConfig(level=logging.INFO)

class MockId2Word:
    def __init__(self):
        self.id2token = {
            0: "word0",
            1: "word1",
            2: "word2",
            3: "word3",
            4: "word4",
        }

class MockLDAModel:
    def __init__(self):
        self.id2word = MockId2Word()
    
    def get_term_topics(self, word_id, minimum_probability=0.0):
        """
        Return a list of (topic_id, probability) tuples for a given word.
        """
        if word_id == 0:
            return [(0, 0.8), (1, 0.15), (2, 0.05)]
        elif word_id == 1:
            return [(1, 0.7), (2, 0.2), (0, 0.1)]
        elif word_id == 2:
            return [(2, 0.9)]
        elif word_id == 3:
            return [(0, 0.5), (1, 0.3)]
        elif word_id == 4:
            return [(1, 0.6), (0, 0.25), (2, 0.15)]
        return []

class MockAutobid:
    def __init__(self):
        self.model = MockLDAModel()

ARGS = {
    "autobid": MockAutobid(),
    "lmbd": 0.5,
    "omega": 0.01,
}
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 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
is_digit Ran
ran on a synthesized input
rub-syssec/adversarial-papers/src/utils/pdf_utils.py
code served (permissive licence) · get_code("a6a3554de7bb5b0f")
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

ARGS = {
    "word": "42"
}
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
scrape_via_pdftotext Ran
ran on a synthesized input
rub-syssec/adversarial-papers/src/utils/pdf_utils.py
code served (permissive licence) · get_code("9ab978c5f5b548e3")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import tempfile
import os

# Create a minimal PDF file for testing
# Since we can't actually create a valid PDF without dependencies,
# we'll create a temporary file that pdftotext can attempt to process
def create_test_pdf():
    # Create a simple text file that we'll use as a stand-in
    fd, path = tempfile.mkstemp(suffix='.pdf')
    try:
        # Write some minimal PDF-like content
        os.write(fd, b'%PDF-1.4\n%fake pdf content\n')
        os.close(fd)
    except:
        os.close(fd)
    return path

pdf_file = create_test_pdf()

ARGS = {
    'pdf_file': pdf_file
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
append_to_dataframe Not yet run
unverified (harvested, no recorded run)
rub-syssec/adversarial-papers/src/hypersearch.py
code served (permissive licence) · get_code("2c8daba027e681a1")
How it ran

no invocation recorded.

get_lda_topics Not yet run
unverified (harvested, no recorded run)
rub-syssec/adversarial-papers/src/utils/lda.py
code served (permissive licence) · get_code("a7ed4609df93b63f")
How it ran

no invocation recorded.

loss Not yet run
unverified (harvested, no recorded run)
rub-syssec/adversarial-papers/src/featurespace/loss.py
code served (permissive licence) · get_code("dad61bce0b70cf63")
How it ran

no invocation recorded.

surrogate_loss Not yet run
unverified (harvested, no recorded run)
rub-syssec/adversarial-papers/src/featurespace/attack.py
code served (permissive licence) · get_code("0277b131c214088f")
How it ran

no invocation recorded.

Also present in other papers' code

Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 11 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.

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

The number of papers submitted to academic conferences is steadily rising in many scientific disciplines. To handle this growth, systems for automatic paper-reviewer assignments are increasingly used during the reviewing process. These systems use statistical topic models to characterize the content of submissions and automate the assignment to reviewers. In this paper, we show that this automation can be manipulated using adversarial learning. We propose an attack that adapts a given paper so that it misleads the assignment and selects its own reviewers. Our attack is based on a novel optimization strategy that alternates between the feature space and problem space to realize unobtrusive changes to the paper. To evaluate the feasibility of our attack, we simulate the paper-reviewer assignment of an actual security conference (IEEE S&P) with 165 reviewers on the program committee. Our results show that we can successfully select and remove reviewers without access to the assignment system. Moreover, we demonstrate that the manipulated papers remain plausible and are often indistinguishable from benign submissions.

For agents

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

get_harvested_code_for_paper("2303.14443")
get_code_for_paper("2303.14443")
have("2303.14443")

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

curl https://syntology.ai/api/ran/2303.14443.json

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

[![Syntology run record](https://syntology.ai/api/ran/2303.14443.svg)](https://syntology.ai/paper/2303.14443)

Connect an agent — have() is free.